EARLY HISTORY of MAGNETIC RESONANCE Norman F. Ramsey

EARLY HISTORY OF MAGNETIC RESONANCE

Norman F. Ramsey

Lyman Laboratory of Physics Harvard University Cambridge, Massachusetts 02138

INTRODUCTION change in the quantum mechanical space quantization when the direction of a magnetic field is In the title of this report, emphasis should be changed. The problem was first posed and par- given to the word early. Some readers may even tially solved in 1927 by C. G. Darwin(2) and his believe that "Pre-History" would be a better title analysis was subsequently improved by P. than early history. The report will cover the Gutinger(3), E. Majorana(4), and L. Mote and period from 1921 to the first nuclear resonance M. Rose(4). absorption experiments of Purcell, Torrey and In the period 1931-33 several experiments in Pound and the first nuclear induction experi- Otto Stern's laboratory in Hamburg successfully ments of Bloch, Hansen and Packard even measured the changes in the space quantization though from some points of view the history of when the direction of the magnetic field was magnetic resonance can be said to begin with the changed. The experiments of Phipps and experiments that end this report. Stern(5) and Frisch and Segre(6) partly agreed My interest in the history of magnetic reso- with the best theory and partially disagreed. I. nance began with preparations for my Ph.D. I. Rabi(7) pointed out that the discrepancy final examination in 1939. Since mine was the between theory and experiment was due to the first Ph.D. thesis based on nuclear magnetic, neglect of nuclear spins in previous theories. resonance, I feared that my examining commit- Although the magnetic moment of the electron is tee would ask searching questions as to the ori- about 2000 times larger than the typical nuclear gins of the ideas of magnetic resonance and of magnetic moment, the angular momenta are the molecular beam technique we used to detect comparable in size and at the low fields used in the resonance transitions. some of the experiments the nuclear spin angular momenta were tightly coupled to the electron EARLIEST SEARCH FOR A DEPENDENCE spin making large effects on the observations. In OF MAGNETIC SUSCEPTIBILITY ON all of these experiments the direction of the field FREQUENCY was changed in space as the atoms went by. Since the atoms had a thermal velocity distribu- The earliest reported search for a dependence tion the frequency components were different for of magnetic susceptibility on frequency was car- different velocities, so on averaging over the ried out by Belz(l) in 1922 for solutions of a velocity distribution, no sharp resonances were variety of paramagnetic salts. No frequency either anticipated or observed. Rabi(8) and dependence was found. Acting on a suggestion of Schwinger(9) in 1937 calculated the transition Lenz and Ehrenfest, G. Breit(2) searched for a probability for molecules that passed through a frequency dependence of the magnetic suscepti- region in which the direction of the field varied bility of various paramagnetic substances but rapidly. found no dependence on frequency. Perhaps this disappointment contributed to Breit's decision to FIRST ATTEMPT TO OBSERVED concentrate in theory, where he later had such a NUCLEAR MAGNETIC RESONANCE IN productive career. CONDENSED MATTER

SPACE QUANTIZATION WHEN In 1936 with calorimetric techniques, C. J. DIRECTION OF MAGNETIC FIELD Gorter(lO) successfully observed a frequency CHANGES dependence of the paramagnetic relaxation of a number of alums. He found that the observed The origins of the molecular beam magnetic effects depended on the frequency, v, as vx resonance method can be traced back to early where x was a number, usually between 1 and 2. theoretical speculations and experiments on the No resonance effects were observed. Gorter(lO)

94 Bulletin of Magnetic Resonance also utilized the same calorimetric method in an attempt to look at 7Li nuclear magnetic reso- MOLECULAR BEAM MAGNETIC nance in LiCl and for an *H resonance in A1K RESONANCE alum but found no such resonance. The following year, Lasarew and Schubnikowt(21) showed While writing his paper on the gyrating field, at low temperature that the nuclear magnetic Rabi discussed with some of his colleagues the moments in solid hydrogen contributed signifi- possibility of using oscillatory rather than space cantly to the observed static magnetic suscepti- varying magnetic fields, but Rabi's laboratory bility of solid hydrogen. had a full .program of important experiments In an experiment reported in 1942 subse- which did not require oscillatory fields, and no quent to the successful molecular beam nuclear experiments utilizing oscillatory fields were magnetic resonance experiments described in the started during the first six months following the next two sections, Gorter and Broer(lO) submission of Rabi's theoretical paper on the attempted to observe nuclear magnetic resonance gyrating magnetic field. In September 1937, C. in powders of LiCl and KF, but no resonance J. Gorter visited Rabi's laboratory(12) and was observed. It is still a mystery as to why described his brilliantly conceived but experi- Gorter did not detect a resonance. In part he mentally unsuccessful efforts to observe nuclear suffered from a poor choice of material since R. magnetic resonance in lithium fluoride, as V. Pound much later showed that pure crystal- described in Gorter's publications of the previous line LiF has an unusually long nuclear spin-lat- year(10). The research efforts in Rabi's labora- tice relaxation time. However, that alone does tory at Columbia University were soon directed not explain the failure of Gorter's inspired primarily toward the construction of molecular experiments since at a much later date N. Blo- beam magnetic resonance experiments with embergen found one of Gorter's original crystals oscillator driven magnetic fields. Two successful and was able to observe an NMR signal with it magnetic resonance devices were soon con- even though the relaxation time was large. The structed by Rabi(13,14), Zacharias(13,14), Mill- most likely explanation for the failure of Gorter's man(13), Kusch(13), Kellogg(14), and Ram- experiments was an unfavorable signal-to-noise sey(14, 15), A schematic view(13) of the method ratio in his apparatus. It is of interest to note is shown in Figure 1. In these experiments the that the first appearance of the phrase "nuclear atoms or molecules were deflected by a first magnetic resonance" in a publication title is in inhomogeneous magnetic field and refocused by a Gorter's 1942 paper, but he attributes the coin- second one. When the resonance transition was ing of this phrase to I. I. Rabi. induced in the region between the two inhomogeneous fields, the occurrence of the transition TRANSITIONS INDUCED BY PASSAGE OF could easily be recognized by the reduction of MOLECULES THROUGH intensity associated with the accompanying failure of refocusing. For transitions induced by the DIFFERENTLY ORIENTATED MAGNETIC radiofrequency oscillatory field, the apparent FIELDS frequency was almost the same for all molecules independent of molecular velocity. As a result, While Gorter was pursuing his unsuccessful when the oscillator freguency was equal to the NMR experiments, I. I. Rabi was independently Larmor angular frequency « o of a nucleus, a studying transitions induced when atoms or mol- sharp resonance was obtained where ecules in a molecular beam traversed a region in space of space in which the directions of the magnetic field change successively. In his bril- (1) liant 1937 theoretical paper entitled "Space Quantization in a Gyrating Magnetic Field", is the angular precession frequency of a classical Rabi(8) assumed for simplicity that the field was magnetized top with the same ratio Yj of mag- oscillatory in time even though the initial appli- netic moment to angular momentum when in a cation was to a field varying along the beam magnetic field Ho. Figure 2 shows the first rather than oscillatory with time. As a conse- reported nuclear magnetic resonance curve; the quence, all the formulae in that paper are appli- curve was obtained with a beam of LiCl mol- cable to the resonance case with oscillatory fields ecules(13). and the paper, without alteration, provides the Kellogg, Rabi, Ramsey, and Zacharias(14, fundamental theory for present molecular beam 15) soon extended the method to the molecules magnetic resonance experiments as well as for H 2 , D2 and HD for which the resonance fre- other experiments with magnetic resonance. quencies depended not only on eqn. 1 but also on

Vol. 7, No. 2/3 95 ffl

Figure 1. Schematic diagram(13) showing the principle of the first molecular beam magnetic resonance apparatus. The two solid curves indicate two paths of molecules having different orientations that are not changed during passage through the apparatus. The two dashed curves in the region of the B magnet indicate two paths of molecules whose orientation has been changed in the C region so the refocusing is lost due to the change in the component along the direction of the magnetic field.

IOO for an allowed transition

rw = Ej - Ef (2)

For the first time the authors described their results as "radiofrequency spectroscopy". The radiofrequency spectrum for H2 is shown in Fig- 75 ure 3. The first molecular beam magnetic resonance experiments were with *2 molecules for which the primary interactions were those of the nuclear magnetic moments in external magnetic fields, but in 1940 Kusch, Millman and Rabi(16, IlIS II2O 17) first extended the method to paramagnetic MAGNET CURRENT IN AMPERES atoms and in particular to AF = ± 1 transitions of atoms where the relative orientation of the Figure 2. Curve showing refocused beam inten- nuclear and electronic magnetic moments were sity at various values of the homogeneous field. changed, in which case the resonance frequencies One ampere corresponds to about 18.4 Gauss. were determined dominantly by fixed internal The frequency of the oscillating field was held properties of the atom rather than by interac- constant at 3.518 X 10* cycles per second. tions with an externally applied magnetic field. In 1949, N. F. Ramsey(18,20) invented the separated oscillatory field method for magnetic resonance experiments. In this new method, the internal interactions within the molecule. The oscillatory field, instead of being distributed transitions in this case occurred whenever the throughout the transition region, was oscillatory field was at a Bohr angular frequency

96 Bulletin of Magnetic Resonance «

H in H2 FREQUENCY 6 9 8 7 MC I, : O 5 AMP

I6OO I6SO M'AGNETIC FIELD IN GAUSS

Figure 3. Radiofrequency spectrum of H2 in the vicinity of the proton resonance frequency(14). The resonance frequencies are primarily determined by the interaction of the proton magnetic moment with the external magnetic but the state of different mj and mj are displaced relative to each other by the different values of the nuclear spin-nuclear spin interaction energies and of the spin-rotational interaction.

concentrated in two coherently driven oscillatory longer than average in the excited state can fields in short regions at the beginning and end reach the second oscillatory field before decaying. of the resonance region. In an alternative ver- Essentially the same magnetic resonance sion of the same method, the coherent oscillatory technique as developed by Rabi for measuring fields are applied in two short pulses - at the nuclear magnetic moments with a molecular beginning and end of the observation time. The beam was used by Alvarez and Bloch(21) to method has the following advantages(20): (1) measure the magnetic moment of the neutron the resonances are 40% narrower than even the with a neutron beam. Since the first publication most favorable Rabi resonances with the same on the neutron magnetic resonance studies was length of apparatus; (2) the resonance are not published about two years after the first molecu- broadened by field inhomogeneities: (3) the lar beam magnetic resonance papers appeared, it length of the transition region can be much is often considered that the neutron studies of longer than the wavelength of the radiation, pro- Alvarez and Bloch were merely adaptations of vided that the two oscillatory field regions are the resonance methods developed by Rabi and his short, whereas there are difficulties with the associates. However, Alvarez recently has told Rabi method due to phase shifts when the length me that Bloch had thought of doing the neutron of the oscillatory region is comparable to the beam magnetic resonance experiment before wave length; (4) the first-order Doppler shift can either Alvarez or Bloch had heard of the molecu- mostly be eliminated when sufficiently short lar beam magnetic resonance experiments of oscillatory field regions are used; (5) the sensi- Rabi and his associates. It must have been a tivity of the resonance can be increased by the bitter disappointment to Bloch and Alvarez to deliberate use of appropriate relative phase learn that their clever idea for magnetic reso- shifts between the two oscillatory fields; and (6) nance had been anticipated by Rabi and his with short lived states the resonance width can associates. It is to their credit that they did not let this disappointment blight their research be narrowed below that expected from the life- careers; instead each went on to win separate time of the state and the Heisenberg uncertainty Nobel Prizes for subsequent research. principle if the separation of the oscillatory fields is sufficiently great that only molecules living Work on both molecular beam and neutron

Vol. 7, No. 2/3 97 beam magnetic resonance experiments were interrupted by World War n. In 1944 Rabi and Ramsey spent one evening together in Cam- bridge, Massachusetts, planning possible post- war research experiments. Two ideas emerged as leading candidates. One was to use the molecular beam magnetic resonance method to measure the hyperfine separation in atomic hydrogen since a presumably exact theoretical calculation of this separation existed. This experiment was eventually carried out and led to the first indication of an anomalous magnetic moment of the electron. The other idea was to detect the existence of nuclear magnetic resonance transitions by their effect on the oscillator. To our pleasant surprise, the signal-to-noise calculations were favorable and we became quite enthusiastic about the possibility. We then real- ized that we were merely reinventing Gorter's nuclear magnetic resonance experiments and that those experiments had failed for unknown reasons. We, therefore, decided that efforts in that direction should be given a low priority compared to the various molecular and atomic beam experiments, including the one on the 0 <,uu 600 bOO WOO 120Q WOO atomic hydrogen hyperfine separation.

ELECTRON PARAMAGNETIC RESONANCE EXPERIMENTS IN CONDENSED MATTER Figure 4. Electron paramagnetic resonance curve In addition to his unsuccessful efforts to obtaine d by Zavoisky(23) with CrCl3. The observe nuclear magnetic resonance, Gorter(lO) microwave radiation wavelength was X = 13.70 successfully observed paramagnetic relaxation in cm and T = 298 K. condensed matter. However, his attempts to observe an electron paramagnetic resonance failed. The first successful paramagnetic resonance experiments in condensed matter were those of Zavoisky(23). His observed paramag- Packard(26) at Stanford University. Each group netic resonance with CrCl3 is shown in Figure 4, had different reasons for being willing to procede was first reported in a 1944 Ph.D. thesis, and with its experiments despite the failure of Gort- several years elapsed before there was wide- er's earlier experiments. spread recognition of his accomplishment. In the case of Purcell, Torrey and Pound(25) Shortly after Zavoisky's pioneering work, obser- they were initially unaware of Gorter's work vations of electron paramagnetic resonances when they first started their experiment. When were made by Cummerow and Halliday (24) and Rabi learned of their plans and pointed out to others. Purcell that Gorter's experiment was similar and had failed, Purcell was disappointed by the news NUCLEAR MAGNETIC RESONANCE but felt that the work on the new experiment EXPERIMENTS IN CONDENSED MATTER had already gone so far that it should be completed, particularly since their extensive theoret- Following World War II, two groups in the ical calculations of relaxation and other feasibil- United States sought to develop nuclear mag- ity requirements appeared favorable. Purcell netic resonance experiments with condensed and his associates observed the absorption in the matter. One was E. M. Purcell, N. G. Torrey resonance circuit and devoted considerable atten- and R.V. Pound(25) at Harvard University and tion to problems of signal size and noise. On the other was F. Bloch, W. Hansen and M. E. December 24, 1945 their letter(25) was received

98 Bulletin of Magnetic Resonance by the Physical Review announcing the success- 3G. Gutinger, Zeits. f. Physik 73, 169 (1931). ful observation of nuclear magnetic resonance 4 E. Maiorana, Nuovo Cimento 9, 43 (1932) absorption of the protons in a paraffin filled 30 and L. Motz and M. E. Rose, Phys. Rev. 50, 348 MHz resonant cavity whose output was balanced (1936). against a portion of the signal generator output. 5 T. E. Phipps and O. Stern, Zeits f. Physik 73, When the magnetic field passed through reso- 185 (1931). nance, an unbalanced signal 20 times noise was •O. Frisch and E. Segre, Zeits f. Physik 80, observed. 610 (1933). When Bloch, Hansen and Packard (26) 7 1.1. Rabi, Phys. Rev. 49, 324 (1936). started their experiments, they were fully aware •I. I. Rabi, Phys. Rev. 51, 652 (1937). of Gorter's experiments but they were encour- ' J. Schwinger, Phys. Rev. 51, 645 (1937). aged to proceed because they thought they knew 1 c C. J. Gorter, Physica 3, 503, 995 and 1006 the source of the previous failure and a means (1936) and C. J. Gorter and I. J. F. Broer, for overcoming it. They believed that Gorter's Physica 9, 591 (1942). experiment had failed because the thermal 1XR. V. Pound, Phys. Rev. 81, 156 (1951).. relaxation time T, was much longer than Goiter 12Rabi(13) refers to the visit of Gorter in a had allowed for. To overcome this difficulty they footnote to the first paper experimentally dem- proposed to put their water sample in a strong onstrating a successful molecular beam magnetic magnetic field for several days to allow the resonance and Gorter 29 years later published nuclear spin system to reach thermal equilib- an article giving his own somewhat different rium. They in fact did so: when their apparatus recollections of the same visit [Physics Today 20, was all ready for a first test they inserted the 76 (Jan. 1967)]. water in the high field and before attempting a 1 3 1 . I. Rabi, J. R. Zacharias, S. Millman and careful search for a resonance Bloch went off on P. Kusch, Phys. Rev. 53, 318 (1938) and 55, a ski trip to allow the system to come to equilib- 526 (1939). rium. When he returned he and his associates 1 4 J . M. B. Kellogg, I. I. Rabi, N. F. Ramsey found the desired resonance after some initial and J. R. Zacharias, Phys. Rev. 55, 729 (1939); searching, but they also found that the relaxa- 57, 728 (1939); and 57, 677 (1940). tion time was short and not long. Instead of l s N . F. Ramsey, Phys. Rev. 58, 226 (1940). waiting several days to begin their observations, 1«;P. Kusch, S. Millman and I. I. Rabi, Phys. a few seconds would have sufficed. The detec- Rev. 57, 765 (1940). tion method of Bloch, Hansen and Packard(26) 17 S. Millman and P. Kusch, Phys. Rev. 57, was rather different from that of Purcell, Torrey 438 (1940). and Pound(25). Instead of observing the absorp- 1 4 N. F. Ramsey, Phys. Rev. 76, 966 (1949). tion signal with a single coil, they used two 1»N. F. Ramsey and H. B. Silsbee, Phys. Rev. orthogonal coils and picked up the signal induced 84, 506 (1951). in the second coil by the coherently precessing 2 ° N. F. Ramsey, Physics Today 33, 25 (July nuclei driven by the first coil. For this reason 1980). they called their experiments nuclear induction. 2 1 L. Alvarez and F. Bloch, Phys. Rev. 51, 11 A letter(2 6) announcing their successful experi- (1940). ment was received by the Physical Review on 2 2 J . E. Nafe. E. B. Nelson and I. I. Rabi, January 29, 1946. Phys. Rev. 71, 914, (1947) and 73, 718 (1948). From the time of these experiments onward, 2 3 E. K. Zavoisky, Ph.D. Thesis 1944 and J. developments in magnetic resonance occurred at Phys. USSR 9, 211 and 245 (1945) and 10, 197 a rapid pace. For this reason, I have chosen that (1946). time to bring to an end this account of the early 2 * R. L. Cummerow and D. H. Halliday, Phys. history of magnetic resonance. Rev. 70, 483 (1946). 2 S E. M. Purcell, H. G. Torrey and R. V. REFERENCES Pound, Phys. Rev. 69, 37 (1946). 2 *F. Bloch, W. Hansen and M. E. Packard, 1M. H. Belz, Phil Mag. 44, 479 (1922) and Phys. Rev. 69, 127 (1946). G. Beit, Comm. K. Lab. 168c (1926). 2 7 B . G. Lasarew and L. W. Schubnikow, 2C. G. Darwin, Proc Roy. Soc. 117, 258 Phys. Zeits. Sowjet 11, 445 (1937). (1927).

Vol. 7, No. 2/3 99 Why Just NMR?

Richard R. Ernst

It must have been sensational to see, for the very first time, the nuclear spin precession on a scope through the resonant absorption of energy and even more directly by means of a free induction decay. The magnetic moments of inconceivably small particles had been put into a coherent motion! This was truly a great discovery by Felix Bloch, Edward M. Purcell, and their coworkers in the mid forties. The Second World War was just over, and the newly developed radio frequency equipment could be put to better usage.

Much has happened since then in the field of nuclear magnetic resonance (NMR). The chemical shift has been discovered by W.D. Knight, and W.G. Proctor and F.C. Yu, and exploited first for chemical analysis by Herbert S. Gutowsky. The scalar spin-spin coupling has been noticed. Pulse techniques have been introduced by Erwin L. Hahn and Henry C. Torrey. Fourier transformation spectroscopy, proposed by Russell Varian and Weston A. Anderson, has eased the sensitivity problem, and two-dimensional spectroscopy, conceived by Jean Jeener, has truly opened up new dimensions in spectroscopy. The structure determination procedures developed by Kurt Wuthrich and his group made NMR into an indispensable tool also for the molecular biologist. On the other hand, John S. Waugh, Alexander Pines, and Peter Mansfield, together with the magic angle spinning techniques of Raymond E. Andrew and coworkers and Irving J. Lowe, made high resolution NMR also feasible in the solid state. Finally, medical imaging by NMR, developed initially by Paul Lauterbur, led to a revolution in medical imaging.

Despite these obvious successes, it took several decades to accept NMR in chemistry as a discipline of a status comparable to that of optical spectroscopy or X-ray diffraction. To be honest, a new generation of scientists, growing up from childhood with NMR, had to take over key positions in the scientific arena.

What are the reasons behind NMR's success? First, nature has generously provided us with three basic physical properties: (1) The nuclear sensors interrogated in NMR experiments are as localized as ever needed, with a diameter as small as 2 fm, allowing for almost unlimited spatial resolution. (2) Interactions with the environment at less than 0.2 J/mol are extremely weak, permitting virtually perturbation-free sensing of the surroundings. Nevertheless, the interactions are highly sensitive to the environmental conditions. (3) Internuclear pair interactions provide accurate distance information and information on bond angles.

The precision of the spatial measurements, compared with the weakness of the relevant interactions, seems to contradict the quantum mechanical uncertainty relations or the resolution limits of a "radiofrequency microscope". Fortunately, no such restriction applies to NMR. All the relevant information is encoded in the spacing of discrete energy levels and the measurements are reduced to the determination of transition frequencies. They can be determined with arbitrary accuracy provided sufficient time is taken. Thus, there is no limiting uncertainty relation between spatial resolution and wave length. However, finite relaxation times may limit the effective observation time, and an uncertainty relation between the accuracy of an energy measurement and the measurement time comes into play.

The energy- or frequency-encoding of the information provides NMR with a great advantage in comparison to scattering techniques, such as X-ray or neutron diffraction, where the high quantum energies required for high spatial resolution can severely damage materials. NMR, on the other hand, is fully non-destructive. The same considerations also apply to magnetic resonance imaging where no wavelength-dependent resolution limit exists as the spatial information is again frequency-encoded.

These are, in a nutshell, the basic reasons for the success of NMR: Very localized and accurate measurements involving extremely low quantum energies. So far these are the gifts of nature. The rest is human skill and sometimes ingenuity: How to elicit and interpret the messages of nature? The solution is known by now: Time-domain experiments, pulse techniques, and Fourier transformation are the key words to success in this field. No further details shall be given here except to say that a large number of the very best scientists have invented an incredible array of tools and procedures to squeeze out even the last drop of information from molecular systems. Teamwork in the best sense of the word has led to the success that the field is experiencing at the present.

If I may try to predict the course of the developments during the remaining years of this decade, I think that certainly unexpected methodological developments are still possible and may further enhance the power of NMR. But more important will be spectacular applications in a variety of fields, ranging from mineralogy to medicine, and providing unprecedented insights into chemical and biological processes. This may initiate unforeseen technological and biomedical possibilities based on a much improved understanding of nature. I am convinced that the next 10 years will be very rewarding in terms of results and also in terms of general and individual recognition of achievements based on NMR.

NMR will not replace other valuable techniques. It should rather be regarded as a complement to more traditional methods, whereby the answers to the question: which is the complement of the other? Could be very well change in due time. I think it is important to stress the unique features of NMR, rather than to emphasize the competitive aspects. With this view, I believe that in the future rewarding activities will be invested in the study of molecular processes and molecular interactions that, after all, are the essentials of chemistry and biology, rather than the static or motionally averaged structures that are merely essential prerequisites for the relvant processes to occur.

Why just NMR? - Because there is hardly another technique that is so informative for so many different types of applications, and because there is no other technique that provides so much fun.

J. Phys. C: Solid State Phys., Vol. 6, 1973. Printed in Great Britain. @ 1973

LETTER TO THE EDITOR

NMR ‘diffraction’ in solids?

P Mansfield and P K Grannellt Department of Physics, University of Nottingham, University Park, Nottingham, NG7 2RD

Received 24 August 1973

Abstract. A new approach to the study of structure in solids by NMR is described. Multi- ple-pulse line-narrowing sequences and an applied magnetic field gradient are used. The theoretical analysis highlights the analogy with x-ray diffraction. Experimental results from a model one-dimensional lattice are presented.

In this letter, we wish to introduce a new method for the determination of spatial structures in solids which relies on NMR ‘diffraction’ effects. The study of internuclear spacings in solids by NMR has traditionally relied upon the dipole-dipole interaction and its effect on lineshape and second moment in order to estimate intramolecular distances. Although all information on the unit cell is contained in the dipolar lattice sums, there is no direct way of obtaining the lattice structure from the free-induction decay (FID) or from the lineshape of a solid. In general, one needs a model of the structure being determined, so that the theoretical predictions for the second moment or for the lineshape may be compared with results obtained experimentally. This all comes about because a particular lattice site is not uniquely determined magnetically and hence is not uniquely identifiable in the frequency spectrum. Identification of lattice sites in the frequency spectrum may be obtained by applying a linear magnetic field gradient to the sample. The usual effect of this is to produce a FID which reflects the bulk shape of the solid, assumed to be a continuous distribution of spins (Carr and Purcell 1954). Of course, in a solid, the spins are actually distributed in a discrete manner at atomic sites. One reason that this discrete nature is not apparent in the observed FID signals is the large dipole-dipole broadening in solids. In mobile liquids, however, this broadening can be very small, so why do we not observe diffraction effects there? There are two reasons: the first is that temporal coherence of the signals from all sites is partially destroyed by random motions, though perhaps one could observe a partially coherent diffraction effect, as for example in a random solid, were it not for the second more important effect of self-diffusion in the applied field gradient. In solids, the self-diffusion may be made arbitrarily small by lowering the sample temperature, but of course, the spin-spin interactions remain. By employing one of the recently developed multiple-pulse sequences (Waugh et al 1968, Mansfield et a2 1973,) or a suitable modification described below, the dipolar and chemical shift interactions t IC1 Postdoctoral Research Fellow. L422 Letter to the Editor L423 may be artificially reduced to a very high degree, while at the same time leaving the spin- field gradient interaction only slightly reduced in value. This can be achieved by applying a modified compensated reflection symmetry cycle, designed to remove dipolar and chemical shift terms, while at the same time appropriately reversing the linear field gradient direction. Such a sequence is, in the pulse-timing representation,

P-y - {T(+) - Px - 7(+) - Py - 2T(+) - Pg - T(+) - P$ - 2T(-)

- px - 7(+) - f'y - 27(-) - py - T(+) - px - .(+)IN where ~(i)indicates the sense of the field gradient applied during the delay 7. For a set of non-interacting spins in a tetragonal lattice with unit-cell dimensions a, b, e, the displacement vector rlmn (k)from the origin to the kth spin site in the I, m, nth unit cell (Z, m, n integers) is given by

rlmn (k)= (I + uk)a + (~2+ uk)b + (a + wk)~ (1) where a = ai etc, and U*, uk, wk are fractions of the primitive-cell dimensions. For a semi-discrete spin distribution in a uniform field gradient G,in which the spins are distributed with a density p(r) over a range Ax, Ay, Az from the position r = rlmn(k) + r', where r' is a continuous variable, we obtain for the FID function of the system at resonance

la+ur+AX mb + v, +Ay nc+ Wk +A2 '(*) = #% 1,&&&/!#2+U& IP?lb+Uk /"c+W, p(r) exp (ip.v)dr' (2) wherep = yGt in which y is the magnetogyric ratio and t the time. For a set of point spins, as in a crystal lattice, equation (2) reduces to

S = 9 C azmn exp [271i(le + mf + ng)l Zfk exp [27i(uke + ukf+ wkg)] (3) Lm,n k where e = ytaGX/2r= apx/2.rretc, and almn = 0, 1. The term involving the summation over k corresponds to signal contributions within the unit cell and is equivalent to the scattering factor sk in electron or neutron scattering. In our case, the scattering cross section fk = 0, 1. Unlike ordinary x-ray scattering, NMR scattering is in principle selective, since only resonant spins contribute to the signal, non-resonant spins havingfk = 0. This is an important point in the study of protons in solids which are effectively transparent to x-rays. It is clear that the dimensionless quantities e,f, g correspond to the lattice Miller indices at appropriate times t. The point-spin formula equation (3) shows that in a cubic lattice with G along the [Ool] axis, observation of first-order diffraction requires g = 1. For protons with c = 3 A and Gz= 103 G cm-1 the diffraction peak would occur at 8 s from the time origin. Thus in order to observe this signal, an intrinsic narrowed linewidth of about 0.1 Hz is required. To date, the best line narrowing achieved in a single crystal of CaF2 is about 20 Hz (Rhim et al 1973). Thus practical realization of NMR crystallography is some way off. In addition, the application of large field gradients degrades the line-narrowing efficiency in the present-day cycles, but this effect may be reduced by using samples of very small diameter so that the total static field variation over the sample is kept within reasonable limits. At this point we may ask what field gradient would be necessary to observe first-order diffraction in a solid excited by a single 90" RF pulse? Again if we take c = 3 A and L424 Letter to the Editor insist that the diffraction peak be observed within about 2T2 - 100 ys, we find that Gz = 10s G cm-1. NMR diffraction could be useful at the macroscopic level for microscopy in bio- physical systems with regular, or approximately regular, macroscopic structures; eg cell membranes and filamentary or fibrous structures. As an approximation to such a system we consider a uniform one-dimensional lattice of lattice constant c, which comprises N + 1 flat slabs of thickness Az containing uniformly distributed spins. For this model, equation (2) gives for the normalized signal

where ,5? = yGzthz/2. This result is similar to the classical diffraction grating formula. The sin /3 /,8 term represents the signal coming from one plate of thickness Az, and corresponds to the rcsults obtained by Carr and Purcell (1954) in the interference limit of a set of continuously distributed spins. As a preliminary experimental test of our result, we have applied the 11, 3,2; 1, 3, 21 multipulse line-narrowing sequence (Mansfield et a1 1973) with 7 I6.4 [AS to model

Au=368 kHz (a)

~~ 0 30 40 50 6.0 70

0 20. 40 60 80 100 120 140 p,(cm-l)

5 layers, Gz=0,77 G cm-1

0 20 40 60 80 100 120 140 pr (cm-1)

Figure 1. (U) The transient nuclear signal from protons in a three-layer sample of synthetic camphor ClOH160 in response to the [[I, 3,2; 1, 211 multiple-pulsesequence, T = 6.4 ps, with zero applied field gradient. (b) The same as (a) but with an applied field gradient of 0.77 G cm-l. A first-order diffraction peak is observed. (c) The transient nuclear signal from protons in a five-layer sample of synthetic camphor in response to the same pulse sequence and the same field gradient as in (b). The abscissae in (6) and (c) were calculated from the measured values of the field gradient and the scaling factor of the multiple-pulse sequence, which was 2.1, Letter to the Editor L425 one-dimensional lattices comprising equally spaced plates of camphor. In these experiments, the field gradient is kept constant so that the chemical shift terms, which are rather small for protons, are retained in the average hamiltonian. However, the spatial resolution obtained is governed by much larger deviations from field gradient uni- formity due to the coil design (Tanner 1965). The important point is not so much the resolution, which at present corresponds to 0.05 cm, but that temporal coherence of the first-order diffraction peak is restored due to removal of the dipole-dipole interaction in a solid. Figure l(u) shows the transient signal from a three-layer camphor sample with zero applied field gradient in response to the 81, 3, 2; 1, 5, 21 multipulse sequence with 7 = 6.4 p. The nuclear signal from camphor in the absence of artificial line narrowing decays with TZ- 44 ps. Figure l(h) shows the transient signal as for l(u) but with an applied field gradient Gz = 0.77 G cm-1. Note the first-order diffraction peak. Figure l(c) shows the narrowed transient signal from a five-layer camphor sample under the same conditions. A first-order diffraction peak is observed here also. All these data were recorded at room temperature, and off-resonance to facilitate Fourier transformation (Mansfield et a1 1973). We see from equation (2) that the inverse Fourier transform of S(p) yields the spatial

Figure 2. (a) The Fourier cosine transform of the transient response figure l(a). A narrowed linewidth of 150 Hz is observed. (6) The Fourier cosine transform of the transient response figure I@). The three camphor layers are clearly resolved. (c) The Fourier cosine transform of the transient response figure I@). The five camphor layers are well resolved. The abscissae in (6) and (c) were calculated from the measured values of the field gradient and the scaling factor of the multiple-pulse sequence, which was 2.1. The peaks observed at the frequency origin arise from the damping and baseline shifts in the transient responses in figure 1. L426 Letter to the Editor transforming the data shown in figures I(b) and l(c) is presented in figures 2(b) and 2(c), and indicates clear resolution of the plate assemblies. The actual spacings of the plates agree with the spacings derived from figures (2b) and 2(c) to within the 10% accuracy of our field gradient calibration. These experiments are to our knowledge the first demonstration of NMR diffraction in a solid. We have obtained similar-looking results for layered liquid samples and liquid- like rubber samples in response to a single 90" pulse. However, these results are not as valuable for reasons stated earlier. Although with improved field gradient coils the best spatial resolution that could be expected with current multipulse line-narrowing sequences is still only 10 pm, we believe that the practical realization of NMR diffraction and microscopy presents new and compelling reasons for continued effort to improve the line-narrowing efficiencies of these sequences.

We are glad to thank A N Garroway and D C Stalker for their assistance in carrying out the experiments described above. We also thank the Science Research Council for an equipment grant, and Imperial Chemical Industries Ltd for a postdoctoral fellowship for PKG

References

Carr H Y and Purcell E M 1954 Plzys. Rev. 94 630-8 Mansfield P, Orchard M J, Stalker D C and Richards K H B 1973 Phys. Rev. B 7 90-105 Rhim W-K, Elleman D D and Vdughan R W 1973 J. chem. Phys. 58 1772-3 Tanner J E 1965 Rev. Sci. Instrum. 36 1086-7 Waugh J S, Huber L M and Haeberlen U 1968 Phys. Rev. Lett. 20 180-2

ORIGINAL RESEARCH n SPECIAL REVIEW 658 For theMRSConsensusGroup Ralph E. Hurd, PhD Petra S. Hüppi, MD Franklyn A. Howe, DPhil P.Hoby Hetherington, PhD Anke Henning, PhD Arend Heerschap, PhD Rakesh K. Gupta, MD Rolf Gruetter, PhD Stephan Gruber, PhD Ramón GilbertoGonzález, MD, PhD Jens Frahm, PhD E.Uzay Emir, PhD Ulrike Dydak, PhD Alp Dinçer, MD Cristina Cudalbu, PhD Kevin M. Brindle, DPhil, J.Patrick Bolan, PhD Chris Boesch, MD, PhD Alberto Bizzi, MD Robert Bartha, PhD Peter B. Barker, MPhil R.Jeffry Alger, MPhil, PhD Gülin Öz, PhD q of thisarticle. 2 correspondence to (NHS). HealthService funding fromtheNational Research (NIHR)BiomedicalCentre received InstituteforHealth land) GrantC1060/A10334;theNational Research Council (MRC) and the Department of Health (Eng Centre receivedsupportfromtheCR-UK, EPSRC, Medical Sciences ResearchCouncil(EPSRC)CancerImaging Cancer ResearchUK(CR-UK)andEngineeringPhysical Leenaards Foundation, andLouis-JeantetFoundations. The (CHUV), EcolePolytechnique Fédérale deLausanne(EPFL), Hospitals (HUG), CentreHospitalierUniversitaire Vaudois University deGeneva(UNIGE), UniversityofGeneva Biomedicale (CIBM)oftheUniversityLausanne(UNIL), FoundationNational (320030_135743), Centred’Imagerie version accepted August 27. SupportedinpartbytheSwiss 9;revisionreceivedJuly30;accepted April August 30;ﬁnal 55455 (G.O.)ReceivedMarch1, 2013;revisionrequested University ofMinnesota, 20216thStSE, Minneapolis, MN 1 theend isat The completelistofauthorsandafﬁliations

From ResonanceResearch, theCenterforMagnetic

RSNA, 2014

G.O. (e-mail: copies for distribution toyourcopies fordistribution colleaguesorclients, contactusat Note: [email protected]

2 This copy is for your personal non-commercial use only. To order presentation-ready Address ). - C Disorders in linical ProtonM C entral clinical units. on advances technical of incorporation including setting, use of robust MR spectroscopy methodology in the clinical the expedite to recommendations offer nally,authors the Fi interpretation. and assessment, quality analysis, and methodology,areprovidedguidelines acquisition data for spectroscopy MR of standardization and acceptance cal clini expanded facilitate To stroke. and epilepsy, eases, to patient management extends to neurodegenerative dis The article documents the impact of impact the documents article The procedures. processing and acquisition data common of consideration critical a with together management, tient disorders in which MR spectroscopy has an impact on pa modality. Herein, the authors present a summary of brain neuroimaging clinical a into tool researchevolved a from system. The clinical usefulness of in the clinical evaluation of disorders of the central nervous gen 1 [ 1 gen A large body of published work shows that proton (hydro Online supplementalmaterialisavailable forthisarticle. q disorders for which for disorders of list growing The lesions. brain infectious and orders, dis demyelinating injury), brain traumatic and diseases, metabolic inherited (hypoxia-ischemia, disorders diatric pe and neonatal neoplasms, brain for established been

RSNA, 2014 1 N

1 ervous ervous H]) magnetic resonance (MR) spectroscopy has spectroscopy (MR) resonance magnetic H]) radiology.rsna.org www.rsna.org/rsnarights. 1 R H MR spectroscopy may contribute mayspectroscopy MR H

S S n

Radiology: pectroscopy ystem 1 H MR spectroscopy has Volume 270: Number 3—March 2014 1 H MR spectroscopyMR H

------SPECIAL REVIEW: Clinical Proton MR Spectroscopy in Central Nervous System Disorders Öz et al

ince the inception of magnetic techniques were developed. These early resulting in improved outcomes. This resonance (MR) imaging in the localization techniques included point- consensus article has been produced by S1980s, its employment in the resolved spectroscopy (PRESS) (1,2) an international group of imaging sci- diagnostic evaluation of the central and stimulated echo acquisition mode entists, neuroradiologists, neurologists, nervous system (CNS) has had a ma- (STEAM) (3), methods that are now oncologists, and clinical neuroscientists jor impact on patient management. widely used in clinical MR spectroscopy from universities and MR vendors to With the advent of 1.5-T whole-body applications. document the impact of 1H MR spec- magnets, imaging of the CNS with un- Preliminary studies revealed large troscopy in the clinical evaluation of precedented detail became possible differences in metabolite levels in acute disorders of the CNS. The MR Spec- by using the proton (hydrogen 1 [1H]) stroke (4), chronic multiple sclerosis troscopy Consensus Group was formed signal of water. Complementary to (5), and brain tumors compared with from October 2011 to April 2012. The structural MR imaging, 1H MR spec- healthy brain (6). Although this work group drafted and finalized the manu- troscopy has become an attractive ap- stimulated a surge of interest in 1H MR script jointly through e-mail correspon- proach with which to assess the levels spectroscopy for diagnosing and assess- dence and teleconferences with the of metabolites in normal and diseased ing CNS disorders during the early days group members and by means of two CNS, especially as image-controlled, of the “Decade of the Brain” (1990– special interest group meetings held in localized MR spectroscopy acquisition 1999), many suboptimal patient studies connection to the 20th Scientific Meet- (7) and the lack of consistent guidelines ing of the International Society for Mag- have led to a situation where, 20 years netic Resonance in Medicine in May Essentials later, MR spectroscopy is still consid- 2012 and the 21st Scientific Meeting of n Hydrogen 1 (1H) MR spectros- ered an “investigational technique” by the International Society for Magnetic copy is complementary to MR some medical professionals and health Resonance in Medicine in April 2013. imaging and adds clinically rele- care organizations. However, the ability vant information about metabo- to make an early, noninvasive diagnosis 1H MR Spectrum of the Brain: lites in common brain or to increase confidence in a suspected Metabolites and Their Biomarker abnormalities. diagnosis is highly valued by patients Potential n MR spectroscopy is clinic-ready and clinicians alike. As a result, an MR spectroscopy provides a very dif- for diagnostic, prognostic, and increasing number of imaging centers ferent basic “readout” than MR imag- treatment assessment of brain are incorporating MR spectroscopy into ing, namely a spectrum rather than an tumors, various neonatal and their clinical protocols for brain exam- pediatric disorders (hypoxia-isch- inations in selected patients. To facili- emia, inherited metabolic dis- tate expanded use of MR spectroscopy eases, and traumatic brain in the clinical setting, this consensus Published online 10.1148/radiol.13130531 Content code: injury), demyelinating disorders, statement encourages standardization and infectious brain lesions; it is of data acquisition, analysis, and re- Radiology 2014; 270:658–679 expected to contribute to patient porting of results. Abbreviations: management in neurodegenera- When assessing the impact of im- CNS = central nervous system tive disorders, epilepsy, and aging techniques on health care (8), Cr = creatine stroke. it is recommended that six criteria be Gln = glutamine n Provided that spectra are evaluated: (a) technical feasibility, (b) Glu = glutamate Lac = lactate acquired reproducibly with a diagnostic accuracy, (c) diagnostic impact, (d) therapeutic impact, (e) impact mIns = myo-inositol protocol that adheres to quality NAA = N-acetylaspartate on outcome, and (f) societal impact (9). standards, clinical MR spectros- PRESS = point-resolved spectroscopy copy can be performed success- Although MR spectroscopy certainly SNR = signal-to-noise ratio fully at either 1.5 or 3.0 T. fulfills the first two criteria, only a few STEAM = stimulated echo acquisition mode studies have demonstrated that it has tCho = total choline n MR spectroscopy data acquisition a wide impact on differential diagno- tCr = total creatine and processing procedures must sis, patient treatment, and outcome TE = echo time tNAA = NAA + N-acetylaspartylglutamate be harmonized across vendors and none have measured the societal for expanded clinical acceptance, impact (ie, cost-benefit analysis) (8). Funding: as lack of standardization and Thus, it remains a challenge and task of Supported in part by the National Center for Research quality assurance of MR spec- high priority for the MR spectroscopy Resources (P41 RR008079), National Institute of Biomed- ical Imaging and Bioengineering (P41 EB015894), and troscopy data acquisition and community to focus on studies that analysis methods is a current National Institute of Neurologic Disorders and Stroke (P30 will quantify the extent to which MR NS076408). impediment to widespread clin- spectroscopy improves diagnosis and ical use. leads to changes in patient treatment Conflicts of interest are listed at the end of this article.

Radiology: Volume 270: Number 3—March 2014 n radiology.rsna.org 659 SPECIAL REVIEW: Clinical Proton MR Spectroscopy in Central Nervous System Disorders Öz et al

Figure 1 quantified at all clinical magnetic field strengths and at almost all practical TEs up to 280 msec (19,20). At 1.5 T and short TEs (25–35 msec for PRESS, 20 msec or shorter for STEAM), mIns and combined Glu and Gln can also be quantified (21). At field strengths of 3.0 T and higher, additional metabolites are detected at short TEs (eg, g-aminobutyric acid and glutathione) and the separation of Glu and Gln is feasible (22,23). Up to 18 metabolites can be quantified at short TEs and field strengths of 7.0 or 9.4 T (23–25). A subset of the metabolites detectable by using MR spectroscopy may serve as biomarkers in the context of Figure 1: 1H MR spectrum acquired at 3.0 T from a volume of interest physiologic and pathologic states. For in occipital lobe (20 3 20 3 20 mm3, T1-weighted axial image) of healthy subject with the STEAM sequence (repetition time msec/echo time [TE] msec at least one MR spectroscopy–detected = 5000/8; 128 repetitions). tNAA = total N-acetylaspartate (NAA), tCr = total metabolite, NAA, evidence from cell creatine (Cr), tCho = total choline, Glu = glutamate, Gln = glutamine, mIns = (26), ex vivo brain (27), and histologic myo-inositol, MM = macromolecules. studies (28) show unequivocally that, in the mature CNS, NAA is present only in neurons, axons, and dendrites—not image (Fig 1). Although MR images are neuronal metabolite NAA, the glial me- in glial cells. Together with 1H MR spec- conventionally displayed as gray-scale tabolite mIns, choline-containing com- troscopy results of human brain ex vivo images that radiologists interpret by pounds such as glycerophosphocholine specimens (29) and in vivo data (30), means of visual inspection of signal in- and phosphocholine, neurotransmitters these observations make a strong case tensities and geometric structures, the Glu and g-aminobutyric acid, antioxi- that NAA is a biomarker for neuronal MR spectrum consists of resonances dants glutathione and ascorbate, and integrity. In addition, NAA levels may or peaks that represent signal inten- other important metabolites such as Cr, reflect mitochondrial (dys)function sities as a function of frequency (com- phosphocreatine, Gln, and lactate (Lac) (31). tNAA (comprised primarily of monly expressed as parts per million, (10,11). Additional metabolites arise in NAA, with a small contribution from a relative, magnetic field–independent specific clinical conditions, such as suc- N-acetylaspartylglutamate) is therefore frequency scale). Spectra are obtained cinate and acetate in abscesses (12), commonly used as a positive or negative either from one selected brain region lipids in various abnormalities (13,14), in vivo biomarker either for the pres- in the case of single-voxel spectroscopy and even exogenous substances that ence of viable neurons or the assess- or from multiple brain regions in the cross the blood-brain barrier, such as ment of parenchymal damage. Elevated case of MR spectroscopic imaging. The propylene glycol after administration of mIns is generally considered a marker spectral data format has no antecedent some parenteral preparations (15) and for gliosis (32,33), and high tCho may in radiology, as MR images do in radio- ethanol after at least moderate alcohol be a marker for cellular proliferation, graphic films, which may be one of the consumption (16). increased membrane turnover, or in- reasons for the relatively slow accep- The number of quantifiable me- flammation (13,29,34,35). Elevated Lac tance of MR spectroscopy in the clinical tabolites depends on the chosen pulse is indicative of anaerobic glycolysis and imaging community. Nevertheless, cur- sequence and parameters, as well as the may be considered an unspecific MR rently available analysis methods can spectral resolution and signal-to-noise spectroscopy biomarker for several ab- help automatically and reliably quantify ratio (SNR), which are affected by many normalities (36,37). MR spectra in the clinical setting. factors including the static magnetic field 1 In vivo H MR spectroscopy focuses strength, quality of B0 field homogeneity, on carbon-bound protons in the 1–5 and radiofrequency coil used (17,18). MR Spectroscopy of CNS Disorders ppm range of the chemical shift scale The major singlet resonances originat- Neurologic diseases affect as many as 1 (Fig 1) and can depict metabolites that ing from total MR spectroscopy–visible billion people worldwide and are a major are present at high enough concentra- NAA (tNAA) (ie, NAA + N-acetylaspar- cause of disability and human suffering. tions (within the micromoles per gram tylglutamate), tCr (ie, Cr + phosphocre- Diagnosis is often complex, and the time range) and mobile on the MR spec- atine), and tCho (ie, primarily phospho- window for effective therapy may be troscopy time scale. These include the choline + glycerophosphocholine) can be limited. MR imaging, with its excellent

660 radiology.rsna.org n Radiology: Volume 270: Number 3—March 2014 SPECIAL REVIEW: Clinical Proton MR Spectroscopy in Central Nervous System Disorders Öz et al

soft-tissue contrast, is commonly the Although it plays a central role in the subtypes of gliomas with isocitrate modality of choice for the detection of clinical management of patients with dehydrogenase mutations, an example brain lesions. The morphologic details brain tumors, MR imaging alone can- of molecular fingerprinting in vivo, on and the sensitivity to changes in content not provide the answer to many impor- the basis of levels of 2-hydroxyglutarate and physical properties of water are ex- tant clinical questions. These include (47). Few studies compared the diag- quisite. However, conventional MR im- differentiating tumor from other focal nostic accuracy of MR spectroscopy aging is not able to depict changes in cell lesions (giant demyelinating plaques, with that of conventional MR imag- density, cell type, or biochemical compo- encephalitis), obtaining a definitive ing, but one study established added sition—all of which can be investigated diagnosis of atypical ring-enhancing value for a decision support system with MR spectroscopy. Furthermore, le- focal lesions (ie, high-grade gliomas, constructed from multicenter data. sions of different underlying pathophys- metastasis, lymphoma, and abscess), Namely, 1H MR spectroscopy data im- iology often manifest with a similar MR identifying the optimal biopsy sites in proved low- and high-grade tumor pre- imaging appearance. Accordingly, MR heterogeneous gliomas, monitoring the diction relative to MR imaging alone; imaging and MR spectroscopy are com- response to treatment, and differentiat- the area under the receiver operating plementary tools for diagnosing disease ing between treatment-induced changes characteristic curve for low-grade tu- and monitoring disease progression and and recurrent tumor. MR spectroscopy mors was 0.93 for MR imaging plus response to therapy. can provide information in all of these MR spectroscopy versus 0.81 for MR In the next sections, we will first key clinical areas, and it is increas- imaging alone, and the area under the report on the clinical impact of 1H MR ingly being used as an adjunct to MR receiver operating characteristic curve spectroscopy in the evaluation of dis- imaging. for high-grade tumors was 0.93 for MR eases in which it has already been dem- The earliest reports in human brain imaging plus MR spectroscopy versus onstrated to be valuable and next on the tumors (6), together with work in ex 0.85 for MR imaging alone (48). Ele- potential clinical utility of MR spectros- vivo specimens (39,40) and cancer cells vated tCho along with decreased tNAA copy in disorders where substantial re- (41), demonstrated that MR spectros- is typically regarded as a diagnostic search activity has occurred in the past copy offers great potential for noninva- feature of brain tumors (13) (Fig 2). 2 decades with consistent results across sive assessment of brain neoplasms. For In addition, the prominent signal at laboratories. The breakdown is based on example, MR spectroscopy in conjunc- 1.3 ppm, which arises from lipids pre- (a) the demonstration of improved diag- tion with perfusion imaging provided a sent in cytoplasmic droplets associated nostic accuracy of MR spectroscopy over sensitivity of 72% and a specificity of with necrosis or hypoxia, is generally other commonly used clinical imaging 92% in the differentiation of tumors associated with higher grade and poor modalities, (b) the presence of disease- from nonneoplastic lesions (42). Simi- survival (49–51) (Fig 2). Conversely, linked specific metabolites in the 1H MR larly, a sensitivity of 93% and a speci- nonneoplastic lesions such as abscesses spectrum, and (c) the demonstration ficity of 60% were achieved when using and tuberculomas often demonstrate of reduced need for invasive diagnostic these two methods for identifying high- elevated amino acids and lipids (52). procedures. In general, the “patient- versus low-grade gliomas, a substantial Other metabolites observed in brain ready” applications involve large disease improvement in sensitivity over that neoplasms include taurine in primitive effects detectable in an individual MR with conventional MR imaging (43). neuroectodermal tumor (53), alanine spectrum, whereas disorders for which Large multicenter studies have de- in meningiomas (13), and glycine in 1H MR spectroscopy is expected to con- termined the accuracy of single-voxel high-grade pediatric tumors (54). If bi- tribute to future patient management MR spectroscopy with pattern recogni- opsy is needed for diagnosis, the tCho/ involve subtle spectroscopic changes tion algorithms for diagnosing brain tu- tNAA ratio can help differentiate areas that are more challenging to detect in mor histology and grade (44–46). Short of solid tumor with the highest cell den- individual cases. Table 1 summarizes the TE MR spectroscopy gives an accuracy sity from edema (55,56). The detection CNS disorder entities that are covered of approximately 90% for all pairwise of an increased tCho/tNAA ratio in the herein and lists metabolites of interest comparisons of the main adult tumor peritumoral region further reflects tu- for these disorders. types (meningiomas, low-grade glioma, mor invasiveness and can thus be used glioblastoma multiforme, metastases) to differentiate high-grade gliomas from 1 Neurologic Diseases in Which H MR except for glioblastoma multiforme ver- brain metastases that exhibit a near- Spectroscopy Is Valuable for Clinical sus metastasis, where the accuracy was normal spectrum in the peritumoral re- Decision Making 78% (44,46). Combining short and long gion (57,58). MR spectroscopy has also TE MR spectroscopy gives a diagnostic been shown to have a decisive role in Brain Tumors accuracy for the main childhood brain the diagnosis of low-grade versus high- Clinical decision making in neuro-on- tumor types (pilocytic astrocytoma, grade tumors, as well as in the diag- cology is achieved by a multidisciplin- medulloblastoma, and ependymoma) nosis of metastasis versus high-grade ary team combining information from of 98% (45). More recently, MR spec- tumors, as part of a diagnostic work-up many sources, including MR imaging. troscopy helped identify molecular that includes conventional MR imaging

Radiology: Volume 270: Number 3—March 2014 n radiology.rsna.org 661 SPECIAL REVIEW: Clinical Proton MR Spectroscopy in Central Nervous System Disorders Öz et al ‡ Metabolites of Interest tNAA, tCho, tCr, Lac, mIns, lipids mIns, Lac, tCr, tCho, tNAA, Lac tCr, tCho, tNAA, lipids Lac, tCr, tCho, tNAA, Val) Isoleu, Leu, amino acids (Ala, lipids, Lac, Suc, Ac, Suc Pyr, Ala, Gly, Lac, tCr, tNAA, Suc Pyr, Ala, Gly, Lac, tCr, tNAA, MM Lac, tCr, tNAA, MM mIns, tCho, tCr, tNAA, tCho tCr, tNAA, MM tCho, tCr, tNAA, tCho tCr, tNAA, Glx mIns, tCho, tCr, tNAA, (mIns) tCr, tCho, tNAA, tCho tCr, tNAA, tCho tCr, tNAA, tCho Lac, tCr, tNAA, tCho Lac, tCr, tNAA,

† necrotic core, and on FLAIR abnormality for nonenhancing tumors necrotic core, and on FLAIR abnormality for nonenhancing tumors necrotic core, and on FLAIR abnormality for nonenhancing tumors necrotic core, white matter in most cases in basal ganglia for Leigh disease, Location of VOI or ROI Location of avoiding VOI or ROI on the contrast-enhancing region of tumor if it exists, avoiding VOI or ROI on the contrast-enhancing region of tumor if it exists, avoiding VOI or ROI on the contrast-enhancing region of tumor if it exists, VOI within the lesion in parietal cortex for Cr deﬁciencies, eg, VOI according to metabolic disorder, Section through parietal cortex/white matter/basal ganglia VOI in basal ganglia T2 hyperintense white matter lesion VOI in T2 hyperintense white matter lesions Section covering VOI within white matter lesion corpus callosum Section covering white matter including VOIs in posterior cingulate and mesial temporal lobes Section angulated along planum temporale and above the lateral ventricles data ROI best deﬁned by clinical planum temporale angulation Bilateral voxels in mesial temporal structures, VOI within reduced diffusion volume Section through reduced diffusion volume 10 mL at gadolinium-enhanced T1-weighted 10 mL at gadolinium-enhanced T1-weighted 10 mL at gadolinium-enhanced . , MR imaging or FLAIR MR imaging MR imaging or FLAIR MR imaging

SVS, STE or LTE PRESS, STE STEAM PRESS, STE or LTE SVS, MRSI LTE MRSI STE or LTE PRESS STE or LTE SVS, STE STEAM PRESS, STE or LTE SVS, MRSI LTE STE STEAM PRESS, STE or LTE SVS, PRESS STE or LTE SVS, MRSI LTE PRESS STE or LTE SVS, MRSI LTE STE STEAM STE PRESS, SVS, MRSI LTE MRSI LTE PRESS STE or LTE SVS, PRESS STE or LTE SVS, MRSI LTE Suspected tumor Suspected tumor Suspected infective focal lesion Suspected metabolic disorder Neonatal hypoxia-ischemia Suspected demyelinating disorder Multiple sclerosis Suspected dementia epilepsy Focal Mesial temporal lobe epilepsy Ischemic lesion ROI = region of interest, VOI = volume of interest. ROI = region of interest, isoleucine, = Isoleu alanine, = Ala acetate, = Ac sequences. TE long with reliably detected is (Gly) glycine whereas sequences, TE short with reliably detected are lipids and (MM), macromolecules (Glx), Gln and Glu of combination a mIns, that Note Leu = leucine, Pyr = pyruvate, Suc = succinate, Val = valine. Val Suc = succinate, Pyr = pyruvate, Leu = leucine, MR Spectroscopy Methods Used to Image Brain Disorders and Metabolites of Interest for Each Disorder for Each and Metabolites of Interest Disorders Methods Used to Image Brain MR Spectroscopy Disorder and MR Spectroscopy Method* Note.—FLAIR = ﬂuid-attenuated inversion recovery. * Clinically available MR spectroscopy methods that have been widely used TE for is the 25–35 abnormalities msec Short indicatedfor TE PRESS is sequence are 135–270 and listed. msec 20 Long (typically msec only for used STEAM sequence. for single-voxel spectroscopy); 144 msec is used for Lac precaution inversion, is however, required about chemical shift displacement errors at 3.0 T = (38). LTE long TE, MRSI = MR spectroscopic imaging, STE = short spectroscopy. TE, SVS = single-voxel † ‡ Table 1 Table

662 radiology.rsna.org n Radiology: Volume 270: Number 3—March 2014 SPECIAL REVIEW: Clinical Proton MR Spectroscopy in Central Nervous System Disorders Öz et al

Figure 2

Figure 2: MR spectroscopy of astrocytomas. Average (solid line) and standard deviation (shaded area) 1H MR spectra (1.5 T, STEAM or PRESS, 2000/30, 128–256 repetitions per spectrum included in average) in World Health Organization (a) grade II (n = 14) and (b) grade IV (n = 42) astrocytomas. Characteristically elevated tCho/tCr ratio and absence of tNAA is apparent in both tumor spectra compared with that from normal brain (see Fig 1). Lac in low-grade tumor may be the result of hypoxia and/or a metabolic shift toward glycolysis, as is commonplace in cancer. In high-grade tumor, large macromolecule (MM) and lipid (Lip) signals (at chemical shifts 2.0, 1.3, and 0.9 ppm) are associated with necrosis. Glx = combination of Glu and Gln. (Reprinted, with permission, from reference 49.) with gadolinium and diffusion-weighted TE MR spectroscopic imaging has been (67), sometimes even despite apparent and perfusion MR imaging (59). used to identify regions of more aggres- homogeneous imaging characteristics. MR spectroscopy may be used to sive phenotype within a heterogeneous It is common to find low-grade oligo- determine prognosis and to guide treat- gliobastoma multiforme to improve dendrogliomas with malignant imag- ment planning in oncology patients gamma knife radiosurgery (64). ing features, nonenhancing high-grade when surgery is not indicated, such as For neurosurgical treatment plan- gliomas with benign imaging features, in diffuse brainstem gliomas and intra- ning, MR spectroscopy plays a role in and focal areas of malignancy in low- medullary tumors in the spinal cord differentiating areas of tumor from grade gliomas. In low-grade gliomas, (60). A tCho/tNAA peak amplitude ra- benign processes and, together with detection of areas with infiltrative tu- tio of at least 2.1 (either at single-voxel other MR imaging methods, in estab- mor cells (close or distant to the main spectroscopy with a TE of 144 or 270 lishing their relationship to key nor- mass) is very important as these can msec or at MR spectroscopic imaging mal brain structures (56), particularly be the primary sites of tumor recur- with a TE of 280 msec) was found prog- in gliomas. Infiltrative gliomas extend rence. Delineation of tumor infiltration nostic of unfavorable outcome in pedi- well beyond the T2-defined main tumor is an essential part of (a) preoperative atric diffuse pontine gliomas (61). Prog- bulk. One study reported that the MR decision making, (b) intraoperative nostic MR spectroscopy markers are spectroscopy–defined abnormal area MR imaging–guided resections, and important for treatment stratification was an average of 24% larger than (c) postoperative follow-up and appli- and can help identify patients who need that delineated by T2 hyperintensity cation of additional therapies (post- more intensive treatment from the out- and confirmed the accuracy of an ele- surgery radiation and/or chemother- set for some tumor types (47,62,63). vated tCho/tNAA ratio with histologic apy). MR spectroscopy was shown to These include the detection of 2-hy- and immunohistochemistry findings spatially correlate with histologic type droxyglutarate in isocitrate dehydroge- for tumor cells (65). Another study and grade and to reflect heterogene- nase-1 mutated gliomas (47), citrate demonstrated increased mIns and Gln ity in brain tumors before surgery: in proliferating pediatric astrocytomas levels in the contralateral hemisphere A tCho/tNAA ratio greater than 2, a (62), and highly MR spectroscopy–visi- of patients with untreated gliobastoma Lac/tNAA ratio greater than 0.25, and ble saturated lipids with elevated scyllo- multiforme, a finding that was indic- the presence of lipid at MR spectro- inositol and low glutamine in high-risk ative of early neoplastic infiltration scopic imaging with a long TE (144 pediatric brain tumors (64). A tCho/ (66). In addition, gliomas of all grades msec) are characteristics of a high- tNAA ratio of more than 2.1 at long may have intratumoral heterogeneity grade tumor, allowing demarcation

Radiology: Volume 270: Number 3—March 2014 n radiology.rsna.org 663 SPECIAL REVIEW: Clinical Proton MR Spectroscopy in Central Nervous System Disorders Öz et al

Figure 3

Figure 3: 1H MR spectroscopy in glioblastomas. Contrast-enhanced T1-weighted MR images and MR spectroscopy grid (3.0 T, PRESS, 1700/30, three repetitions, section thickness = 20 mm, matrix size = 16 3 16, total acquisition time = 6 minutes 46 seconds) are shown together with representative spectra from voxels in contrast-enhancing areas. (a) Image and spectrum from patient with recurrent gliobastoma multiforme exhibits elevated tCho/tCr ratio as well as elevated lipid (Lip) and Lac levels. (b) Image and spectrum from histologically proven case of postradiation injury exhibits markedly elevated lipid (Lip) and Lac levels along with normal-appearing tCho/tCr ratio. of brain parenchyma adjacent to MR contrast MR imaging was reported to persistence of high Lac is associated imaging–delineated tumor (56). In ad- have 100% positive and negative predic- with poor outcome (77). MR spectros- dition, recent intraoperative 1H MR tive values for discriminating posttreat- copy can be used as a means to assess spectroscopy at 3.0 T helped differen- ment change, which is more accurate treatment efficacy of hypothermia, a tiate tumor from a nontumoral abnor- than both conventional MR imaging proven neuroprotective treatment for mality, as indicated by a high tCho/tCr (positive predictive value, 50%) and fluo- perinatal asphyxia (78). ratio and the presence of Lac, in 57% rine 18 deoxyglucose positron emission Although rare, inherited metabolic of suspected cases and had a positive tomography (PET) (positive predictive disorders are a significant disease entity effect on surgical success and patient value, 67%; negative predictive value, in neuropediatrics. Clinical symptoms in outcome (68). 60%) (72). However, dynamic suscep- certain inherited metabolic diseases are MR spectroscopy can help avoid the tibility contrast MR imaging showed a due to the accumulation of metabolites incorrect diagnosis of tumor progres- substantial false-positive rate, which was that are either neurotoxic or interfere sion, which can lead to inappropriate not the case with MR spectroscopy— with normal function. If the accumulating surgery, other treatment, and patient a finding that points to an incremental substance is visible at MR spectroscopy, distress in cases of posttreatment-in- value of MR spectroscopy in separating its presence or elevation in the spectrum duced changes that are ambiguous at tumor recurrence and posttreatment in- can be used for diagnosis. MR spec- conventional MR imaging. For exam- jury (72). troscopy has proved clinically useful in ple, the tCho/tNAA ratio was shown to In summary, MR spectroscopy adds neonates suspected of having metabolic reliably differentiate recurrent glioma diagnostic and prognostic benefits to disorders (79–81) owing to the unique from postradiation injury (69) (Fig 3). MR imaging and aids in treatment plan- ability to noninvasively detect the meta- Similarly, MR spectroscopy (tCho/ ning and monitoring of brain cancers. bolic defect in vivo (82–85). For example, water), either alone or in combination the presence of pyruvate (plus Lac and/ with conventional MR imaging, can fur- Pediatric Disorders: Hypoxia-Ischemia, or alanine) and succinate are early indi- ther contribute to the assessment of Inherited Metabolic Diseases, and cators of pyruvate and succinate dehydro- response to anticancer treatment (70). Traumatic Brain Injury genase complex deficiencies, respectively MR spectroscopy (tCho/tCr and tCho/ 1H MR spectroscopy was used for pe- (79,86–88). Detection of elevated gly- tNAA) and dynamic susceptibility con- diatric brain imaging as early as 1990– cine, in particular at long TEs, is clinically trast MR imaging in isolation showed 1991 (73–75), and it is part of routine diagnostic in nonketotic hyperglycinemia diagnostic accuracy of 84.6% and 86%, imaging protocols in many specialized (82), although intracerebral hemorrhage respectively; the accuracy increased to academic health centers and children’s presents a confound in the interpretation 93.3% when combined data were used hospitals. For the newborn infant, of high glycine levels (89). A grossly ele- for tumor regrowth and posttreatment quantitative assessment of cerebral vated tNAA level is a diagnostic hallmark injury (71). MR spectroscopy (tNAA/ Lac due to hypoxia-ischemia is one of of Canavan disease (90). tCho ratio and tCho concentration) in the earliest imaging signs indicative of In other inherited diseases, the combination with dynamic susceptibility clinical brain injury (37,76) (Fig 4), and reduction of metabolites owing to

664 radiology.rsna.org n Radiology: Volume 270: Number 3—March 2014 SPECIAL REVIEW: Clinical Proton MR Spectroscopy in Central Nervous System Disorders Öz et al

Figure 4

Figure 4: 1H MR spectroscopy in early assessment of perinatal hypoxia-ischemia in the newborn. MR imaging was performed (a–c) 12 hours and (d, e) 10 days after perinatal asphyxia. (a) Axial T2-weighted image shows no signal abnormalities, whereas (b) spectrum (1.5 T, PRESS, 1500/288, 192 repetitions) obtained from the right basal ganglia shows markedly increased Lac resonance with preserved tNAA, tCr, and tCho resonances. (c) Diffusion-weighted image (echo-planar imaging; 30 directions; b value, 700 sec/mm2) with axial apparent diffusion coefficient map shows no diffusion abnormalities. (d) Axial T2-weighted images show areas of high (white arrows) and low (black arrows) signal intensity in putamen and thalamus, representing clear ischemohemorrhagic lesions. (e) Axial proton density images demonstrate prominent detection of lesion’s extension. (Reprinted, with permission, from reference 76.) decreased synthesis or transport can be More minor changes in single or For effective clinical management, ob- detected with MR spectroscopy. An ab- multiple metabolites require care- jective means to evaluate long-term sent or severely reduced tCr level pre- ful quantification of the MR spectra outcome are required, especially for sents a limited differential diagnosis of and comparison with well-established comatose patients. In a cohort of chil- three underlying genetic defects (91): normal values. It is quite challenging dren with traumatic brain injury, a The lowest tCr levels are observed in to obtain these data in the pediatric regression model, incorporating age, untreated children with a Cr synthesis population owing to limitations associ- initial Glasgow coma scale, and pres- defect (guanidinoacetate methyltrans- ated with imaging healthy children, but ence of retinal hemorrhage and sup- ferase or arginine:glycine amidinotrans- they are particularly crucial because plemented with tNAA/tCr ratio and ferase deficiency), and treatment leads of developmental changes in metabo- MR spectroscopy–visible Lac within to at least partial normalization of cere- lite levels (98). This challenge can be the 1st month after incidence, was bral tCr (92,93) (Fig 5). In males with overcome by using normative data from shown to differentiate between good a Cr-transporter deficiency, brain tCr children who undergo MR imaging and and poor outcomes (102). In pediat- concentrations are reduced by four- to spectroscopy for the investigation of ric near-drowning accidents, an MR fivefold compared with that in healthy suspected neurologic conditions. This spectroscopy index based on tNAA, control subjects. These patients do not approach has proved useful in Hunter Lac, and combined Glu and Gln was benefit from Cr therapy either with or syndrome, a mucopolysaccoroidosis shown to correctly differentiate be- without additional arginine and glycine (99), and propionic acidaemia (100). tween good and poor outcomes— (94,95). The absence of tNAA owing Traumatic brain injury is a major with no false-positive results (103). to a defect in NAA synthesis (96) has cause of disability and death among These data support the clinical utility been described in a case study (97). children younger than 14 years (101). of MR spectroscopy in combination

Radiology: Volume 270: Number 3—March 2014 n radiology.rsna.org 665 SPECIAL REVIEW: Clinical Proton MR Spectroscopy in Central Nervous System Disorders Öz et al

Figure 5 an elevated tCho/tCr ratio, normal or reduced tNAA/tCr ratio (35), and elevated macromolecular signals, possibly arising from myelin breakdown products (114). The tNAA/tCr ratio in the normal-appearing white matter of patients with varying clinical presenta- tions helps differentiate patients from healthy control subjects (115,116) and inversely correlates with disability scores—especially at an early stage (117). In addition, the tCho/tCr ratio is elevated in normal-appearing white matter months before lesions become detectable at conventional MR imaging (118). These observations underscore the ability of MR spectroscopy to characterize white matter abnormality in evolving multiple sclerosis (119). In Figure 5: 1H MR spectroscopy of neurometabolic disorder. (a, b) White matter spectra (1.5 T, PRESS MR addition, increasing evidence for gray spectroscopic imaging, 3000/30, six weighted averages, nominal voxel size = 10 3 10 3 15 mm3) in girl matter involvement in multiple sclero- with guanidinoacetate methyltransferase deficiency before treatment at age 3 years 2 months (a) and after sis (120) provides motivation to study 3.5 months of treatment with oral creatine supplementation (b). Resonance from creatine-containing metab- these lesions with MR spectroscopy as olites (tCr) returned to normal in this region as well as in other investigated brain areas. well (113). Finally, MR spectroscopic imaging might play an important role in the differential diagnosis of multi- with clinical measures for predicting follow-up is an indication for treatment ple sclerosis, with acute disseminated outcome. with hematopoietic stem cell transplan- encephalomyelitis showing recovery of tation (Fig 6). Hematopoietic stem cell tNAA signal losses as a favorable prog- Demyelinating Diseases transplantation performed before sub- nostic sign (121). MR spectroscopy plays an important stantial tissue degeneration as assessed role alone (104) or in addition to other with tNAA results in clinical stabilization Focal Lesions Caused by Infectious semiquantitative MR techniques (105) (108). In patients who are newly diag- Agents in the differential diagnosis of heredi- nosed with juvenile or adult metachro- Brain infections can be life threaten- tary leukoencephalopathies. MR spec- matic leukodystrophy, a combination of ing and, hence, require an early diag- troscopy provides valuable information MR imaging and MR spectroscopy can nosis for optimal clinical management. about tissue pathophysiology for at be used to judge the state of brain tissue Definitive laboratory diagnostic tests least three different metabolic profiles: inflammation (109,110). Although mIns can be time consuming, thus delaying (a) hypomyelination, (b) white matter is typically increased even in the early therapy. MR spectroscopy is valuable in rarefaction, and (c) demyelination, stages of metachromatic leukodystro- the differential diagnosis of intracranial which were differentiated with tCho/ phy, as long as tNAA is still within the ring-enhancing lesions. When a ring- tCr and tNAA/tCr ratios in a study of normal range, hematopoietic stem cell enhancing mass lesion manifests with 70 children (104). transplantation is indicated (111,112). nonspecific clinical and conventional Hematopoietic stem cell transplan- The clinical use of MR spectros- MR imaging features, 1H MR spectros- tation is currently the only treatment copy in multiple sclerosis, an acquired copy can help confirm the definitive option for inherited demyelinating dis- demyelinating disease, remains limited diagnosis of pyogenic abscess and pro- orders such as X-linked adrenoleukodys- despite the various insights into disease vide information about the type of in- trophy, metachromatic leukodystrophy, pathology that it has offered as well as fective agent (12,122). Demonstration and globoid cell leukodystrophy (106). its ability to assess the burden of ax- of succinate, acetate, alanine, leucine, MR spectroscopy is used to monitor the onal damage (113). MR spectroscopy isoleucine, and valine are considered onset of demyelination in neurologically of chronic multiple sclerosis plaques specific for pyogenic abscess (Fig 7)— asymptomatic patients with X-linked ad- in white matter shows a consistently even in the absence of reduced diffu- renoleukodystrophy with high genotypic reduced tNAA/tCr ratio (5,35) and, sivity at MR imaging. Similarly, para- variability (14,32,107). Interval elevation sometimes, an elevated tCho/tCr ra- sitic cysts contain succinate and acetate of mIns/tNAA and tCho/tNAA ratios tio (35). Spectra from plaques un- in the absence of amino acids, which in normal-appearing white matter at dergoing active inflammation show helps differentiate them from anaerobic

666 radiology.rsna.org n Radiology: Volume 270: Number 3—March 2014 SPECIAL REVIEW: Clinical Proton MR Spectroscopy in Central Nervous System Disorders Öz et al

Figure 6 typically localized to the region(s) affected by the degenerative process (124,125). The tNAA levels reflect pathologic severity (33,126) (Fig 8) and correlate with clinical measures in cross-sectional studies (127,128). Con- sistently, tNAA/tCr tends to be lower in subjects with mild cognitive impairment who convert to dementia compared with those who remain stable (129). Therefore, the tNAA/tCr ratio or tNAA concentration may be a valuable prognostic indicator of disease progression, either alone or in combination with volumetric measurements (130). Other 1H MR spectroscopy changes associated with neurodegeneration include a decreased Glu level (128,131,132), an elevated tCho level (125), and an elevated mIns level (132,133). The elevation in mIns may be associated with glial or microglial activation, a characteristic feature of these diseases (134). An elevated mIns level appears early in dementia, pre- ceding the decrease in tNAA concentration (Fig 8), atrophy, and associated neuronal loss and cognitive impairment, as demonstrated in presymptomatic carriers for familial Alzheimer Figure 6: Single-voxel 1H MR spectroscopy shows progression of disease in boy with X-linked adreno- disease (135) and in patients with leukodystrophy. At baseline, T2-weighted signal abnormalities on conventional MR image are seen only in frontotemporal lobar degeneration posterior third of centrum semiovale, and spectrum (4.0 T, STEAM, 4500/5, 64 repetitions) is normal. One mutations (136). year later, MR image shows progression of T2 signal abnormalities in middle third of centrum semiovale. 1H MR spectroscopy may also be Spectrum in anterior third of centrum semiovale already shows increased choline (tCho) and mIns in associa- used to monitor treatment response tion with tNAA signal loss. As predicted with the spectrum at 1 year, MR image obtained 2 years later shows in neurodegenerative diseases. For ex- further progression of signal changes, with spectrum showing further mIns signal increase and tNAA loss. ample, a transient increase in tNAA Also note progressive changes in Glu/Gln ratio and accumulation of mobile lipids (Lip) plus Lac at 1- and concentration was associated with 2-year follow-up. short-term functional response during donepezil treatment in Alzheimer disease, suggesting that tNAA also re- abscesses (122,123). MR spectroscopy flects functional integrity and recovery Neurologic Diseases in Which 1H helps in the differentiation of tubercu- (137). Other studies have shown a de- MR Spectroscopy May Contribute to loma with solid caseation from other creased mIns/tCr ratio following done- Patient Management nontuberculous lesions that have a sim- pezil treatment (138) and an increased ilar appearance at conventional MR im- Glu level after galantamine treatment aging. In vivo 1H MR spectroscopy from Neurodegenerative Diseases for Alzheimer disease (139). tuberculous abscess shows only Lac and Neurodegenerative diseases such as lipid signals and is devoid of cytosolic Alzheimer disease, Parkinson disease, Epilepsy amino acids. Magnetization transfer ra- Huntington disease, amyotrophic lateral Epilepsy is a common disorder, with a tio MR imaging and amino acid signals sclerosis, and spinocerebellar ataxias prevalence of 0.5%–1.0% worldwide. in 1H MR spectroscopy help differenti- are debilitating conditions that result The specific etiology underlying the sei- ate pyogenic from tuberculous abscess in progressive neuronal degeneration zures can be variable, with 60%–70% of (12). MR spectroscopy therefore plays and death. The characteristic feature of all patients responding to medications a role in the diagnosis and clinical man- neurodegenerative diseases at 1H MR (140,141). Surgical intervention can be agement of focal brain infections. spectroscopy is a decrease in tNAA, effective in the remaining 30%–40% of

Radiology: Volume 270: Number 3—March 2014 n radiology.rsna.org 667 SPECIAL REVIEW: Clinical Proton MR Spectroscopy in Central Nervous System Disorders Öz et al

Figure 7

Figure 7: 1H MR spectroscopy of pyogenic abscess in cerebellum. (a) Axial T2-weighted image shows well-defined hyperintense lesion with hypointense wall. (b) Axial T1-weighted image shows hypointense lesion with isointense wall. (c) Diffusion-weighted image shows restricted diffusion in lesion. (d) Postcontrast T1-weighted image shows ring enhancement. (e) In vivo 1H-MR spectrum (3.0 T, PRESS, 3000/144, 128 repetitions) from center of lesion shows resonances of amino acids (AA, 0.9 ppm), lipid (Lip) and Lac (1.3 ppm), alanine (Ala, 1.5 ppm), acetate (Ac, 1.9 ppm), and succinate (Suc, 2.4 ppm). The resonances from alanine, Lac, and amino acids are inverted at the TE used owing to J evolution. patients (142,143). In the more com- means of invasive electroencephalo- acid in patients with epilepsy at ultra- mon type of focal epilepsy, surgical graphic measurements. high field strengths (150). outcomes are improved if the region Given the close physiologic relation- The most common abnormality in of seizure onset can be clearly defined ship between brain function and metab- temporal lobe epilepsy is mesial tempo- (142,143). Conventional MR imaging olism (144), MR spectroscopy has been ral sclerosis, which may often be effec- can accurately localize the seizure on- extensively used to better understand tively treated with unilateral temporal set region, for example, by identifying and localize human epilepsy (145,146). lobectomy. Multimodal evaluation, which unilateral hippocampal atrophy or mal- Abnormalities in tNAA concentration involves scalp or intracranial electro- formations of cortical development. and the tNAA/tCr ratio have been use- encephalography, conventional MR im- However, MR imaging may often be ful for detecting injured brain in the sei- aging, and/or metabolic imaging with negative or ambiguous (eg, bilateral in- zure onset focus (145–149). MR spec- PET, is commonly used to lateralize the volvement) and, in some cases, lesions troscopic imaging measures have also epileptogenic zone in mesial temporal seen at MR imaging may not match been extended to neurotransmitters, sclerosis. A meta-analysis of 1H MR spec- the focus of seizure onset identified by for example, to assess g-aminobutyric troscopy literature comprising 22 studies

668 radiology.rsna.org n Radiology: Volume 270: Number 3—March 2014 SPECIAL REVIEW: Clinical Proton MR Spectroscopy in Central Nervous System Disorders Öz et al

Figure 8

Figure 8: 1H MR spectroscopic findings at different pathologic and clinical stages of Alzheimer disease. Top panel: Antemortem 1H MR spectroscopic findings in posterior cingulate gyrus voxel (T1-weighted midsagittal image) are associated with postmortem pathologic diagnosis of Alzheimer disease (low, intermediate, and high likelihood). For each pathologic diagnosis, plot shows individual values, a box plot of the distribution, and estimated mean and 95% confidence interval for the mean. A strong association is observed with tNAA/mIns ratio (R 2 = 0.40; P , .001). (Reprinted, with permission, from reference 33.) Bottom panel: Examples of 1H MR spectra (1.5 T, PRESS, 2000/30, 128 repetitions) in patients with mild cognitive impairment (MCI) and Alzheimer disease (AD) are compared with that from a cognitively normal subject (control). mIns is elevated as an early marker of subsequent neurodegenerative changes in patient with mild cognitive impairment. tNAA is decreased and mIns is further elevated in patient with Alzheimer disease.

(19 performed with 1.5-T units) indicates with no abnormality at conventional MR may also have value for the assessment that ipsilateral MR spectroscopy abnor- imaging or in those with a bilateral epilep- of epilepsy in children, with a low tNAA mality is associated with good outcome togenic zone on electroencephalographic concentration serving as an important in- following surgery (151). Decreased recordings. However, MR spectroscopy dex of disease state (154). tNAA/tCr and/or tNAA/(tCr + tCho) is still considered a research tool in the ratios were the most common MR spec- context of surgical planning for epilepsy Acute Stroke and Brain Ischemia troscopy indexes for epileptogenic zone. (151). This picture may change when 3.0- Overall, MR imaging plays a limited role MR spectroscopy may offer potential in T (152) or higher ﬁeld MR systems (153) in decision making for clinical manage- presurgical decision making in patients are more widely used. MR spectroscopy ment of patients with acute stroke,

Radiology: Volume 270: Number 3—March 2014 n radiology.rsna.org 669 SPECIAL REVIEW: Clinical Proton MR Spectroscopy in Central Nervous System Disorders Öz et al

Table 2 Guidelines for Choosing Single- or Multivoxel MR Spectroscopy (MR Spectroscopic Imaging) Technique When to Choose Technique Advantages Disadvantages

Single-voxel spectroscopy Single focal lesion or diffuse disease; Acquisition parameters optimized for Selected volume is large and block shaped; to answer a single question (eg, tumor volume of interest result in high data region of interest must be placed vs abscess); complement to quality; fast (~2–5 min) if large voxel accurately at time of investigation; no MR spectroscopic imaging in focal size (eg, 6–8 mL) is chosen; automatic information on spatial heterogeneity of lesion (eg, short TE single-voxel water reference acquisition standard on lesion and “normal” brain regions; time spectroscopy to complement long most clinical units; data acquisition can be consuming if multiple locations are to TE MR spectroscopic imaging); aborted and limited dataset can usually still be measured in areas of interest close to skull or be used; voxel boundaries generally better difficult to obtain an acceptable shim defined than with MR spectroscopic imaging MR spectroscopic imaging Undefined, multiple, or heterogeneous Information on tissue heterogeneity; data More exacting system criteria necessary to lesions; comparison of brain regions format compatible with conventional minimize spectral loss due to insufficient in time-efficient manner; diffuse MR imaging (spectroscopic image display lipid and water suppression (shim over disease (if reliable short TE integrates with other imaging modalities); large volumes worse than in single- MR spectroscopic imaging with larger anatomic coverage; smaller voxel voxel spectroscopy); longer acquisition quantification is available) volumes (~1 mL and below) are typically times when using conventional used to assess metabolite distributions; encoding (~6–30 min depending retrospective selection of region of interest on resolution); water reference within the investigated volume for quantification adds substantial acquisition time; more experience is needed to plan MR spectroscopic imaging

usually because of a lack of immediate The concentration of tNAA in brain pa- Therefore, quantitative metabolite data availability of the imaging unit and of renchyma after ischemia (158) and in for tNAA and Lac are of value for evalu- patient-related MR imaging safety in- chronic infarction may even decrease ating the nature of ischemia and predict- formation. The decision to thrombolyse below the level of detection with in vivo ing risk for new ischemic events (161). or to apply any other form of therapeu- 1H MR spectroscopy (4). Measurement tic intervention in the hyperacute phase of tNAA levels could influence patient is based on clinical grounds and exclu- management; severely decreased tNAA Technical Considerations sively involves computed tomography appears to be related to clinical stroke to rule out either brain hemorrhage or syndrome and more extensive infarc- Data Acquisition very large ischemic lesions, which usu- tion, both indexes of poor clinical out- Any application of MR spectroscopy ally have unfavorable outcomes (155). come (36). A decrease in tNAA on fol- to a clinical question starts with the Diffusion-weighted and perfusion MR low-up MR spectroscopy data has been decision about a pulse sequence and imaging are superior imaging tech- associated with ongoing ischemia and parameters. In general, this choice niques for detecting acute ischemia and progressing infarction (159). is dictated by the disease (Table 2). highlight the penumbra, but they are Lac is another metabolite with po- When the affected brain region is well rarely used outside of specialized acute tential value for clinical evaluation in defined, single-voxel spectroscopy is the stroke clinics that have rapid access to stroke. Lac is the end product of non- preferred method and provides robust MR imaging. Similarly, 1H MR spectros- oxidative glucose consumption and is metabolite quantification in the se- copy offers great potential after the hy- commonly considered as a signature lected volume of interest, whereas MR peracute phase of stroke (beyond 4.5 of hypoxia and/or ischemia. Elevated spectroscopic imaging is the method hours) to assess several key character- Lac in the core of ischemic tissue cor- of choice in diseases where the focal istics of ischemic brain for prognostic relates with final infarct size and clini- point of pathology is unclear, if there purposes, such as severity of ischemia cal outcome (159). The presence of Lac are multiple lesions, or if the lesions and neuronal dysfunction and damage. with a concomitant reduction in tNAA are heterogeneous. For example, MR Preclinical work has shown that was observed in large infarcts with poor spectroscopic imaging is advantageous tNAA decreases in ischemic brain pa- outcome (36). Lac levels that are per- in the accurate evaluation of tissue renchyma in a linear fashion for the first sistently elevated for weeks in infarcted status in localization-related epilepsy 6 hours, followed by a slower decrease brain parenchyma have been associated (Table 1) and in the investigation of the for the subsequent 24 hours (156,157). with inflammatory macrophages (160). heterogeneity of large tumors (67). In

670 radiology.rsna.org n Radiology: Volume 270: Number 3—March 2014 SPECIAL REVIEW: Clinical Proton MR Spectroscopy in Central Nervous System Disorders Öz et al

Figure 9 potential gains in SNR and spectral res- can be used for visual inspection of olution, it is important to note that field spectra and basic quantification of strength is not the sole determinant of metabolite ratios. In addition, off-line the information content of spectra. In postprocessing tools (165) and sophis- fact, a spectrum obtained at 1.5 T with ticated quantification packages such as a protocol adhering to spectral quality LCModel (166) are widely used. These standards (Appendix E1 [online], Fig 9) packages provide quantitative error provides more reliable metabolite in- estimates for metabolite quantifica- formation than a poor-quality spectrum tion (eg, Cramér-Rao lower bounds), obtained at 3.0 T. Overall, clinical MR with which the reliability of metabo- spectroscopy can be successfully per- lite concentrations can be assessed formed at either 1.5 or 3.0 T for the (see Appendix E1 [online] for recom- majority of applications. Although the mended criteria). The availability of potential gains at magnetic fields higher error estimates is an important re- than 3.0 T for clinical MR spectroscopy quirement for clinical decision making are still being assessed, significant im- when using quantitative MR spectros- provements in spectral and spatial reso- copy measures; therefore, vendors of Figure 9: Minimum technical requirements lution at 7.0 T have been reported. For clinical imaging units are encouraged 1 to ensure that a H MR spectrum is clinically example, previously inaccessible alter- to implement more robust, U.S. Food interpretable. SNR is calculated from a nonapodized ations in low-concentration metabolites and Drug Administration–approved spectrum by using maximum height of largest may be uncovered at 7.0 T (162). For MR spectroscopy analysis packages signal (typically tNAA) divided by standard deviation MR spectroscopic imaging, the nominal that provide such quantitative error of noise. Note that these SNR limits are given only spatial resolution can be reduced to estimates. for visual assessment of spectra for ratio changes in major metabolites or for presence or absence 0.14 mL at 7.0 T (163). For clinical use, single-voxel spec- of metabolites such as Lac. Higher SNR levels are Finally, the importance of spectroscopy data can be reported numeri- necessary for reliable quantification of metabolites. tral quality generated with the chosen cally as metabolite concentrations or as FWHM = full width at half maximum. pulse sequence, parameters, and field ratios, ideally supplemented with visual- strength cannot be underestimated. For ization of volume of interest placement reliable clinical decision making based (167). On the other hand, information many abnormalities, single-voxel spec- on MR spectroscopy data, obtaining from two- or three-dimensional MR troscopy and MR spectroscopic imaging high-quality, artifact-free spectra is cru- spectroscopic imaging must be made can be used in combination; for exam- cial. The sources and forms of artifacts available to the clinician in a quick and ple, MR spectroscopic imaging to first in MR spectra have been reviewed in easy image format to incorporate into identify the lesion location and single- detail (164) and are summarized in Ap- the clinical routine. In addition, imple- voxel spectroscopy to quantify metabo- pendix E1 (online). The detection of mentation of MR spectroscopy into lites that can be reliably obtained from such artifacts and exclusion of spectra picture archiving and communication high-quality, short TE spectra in the based on predefined quality criteria re- systems is recommended to facilitate identified lesion (Table 1). lies on the human expert in most ap- easy access to MR spectroscopy data in All of the major clinical MR imaging plications of single-voxel spectroscopy, the standard work environment. vendors provide MR spectroscopy pro- whereas automated quality assessment tocols, primarily with use of the basic of MR spectroscopic imaging data is Reproducibility and Clinical Translation PRESS (1,2) and STEAM (3) sequences preferred. A practical guide to deter- Ultimately, test-retest reproducibility (Table 1). In addition, other state-of- mine whether a spectrum is adequate of measured metabolite levels deter- the-art single-voxel spectroscopy and for clinical use is provided in Figure mines the utility of MR spectroscopy MR spectroscopic imaging sequences, 9. Further considerations regarding for disease assessment. To be of clin- which offer various advantages over the the choices for clinical MR spectros- ical value, experimental and biologic basic STEAM and PRESS sequences, copy data acquisition, including pulse variability in the quantified metabo- have been implemented on some clin- sequence, parameters, field strength, lite levels must be smaller than their ical platforms (Appendix E1 [online]). and radiofrequency coils, as well as changes caused by disease. Test-retest Which field strength is optimal for recommendations for spectral quality coefficients of variance reported at 1.5 a particular clinical application of MR assessments, are detailed in Appendix and 3.0 T (168–173) show improved ac- spectroscopy is another important ques- E1 (online). curacy for several metabolites at higher tion for the practicing neuroradiologist fields and shorter TEs. Test-retest co- and clinical trialist. Although 3.0 T is Data Analysis and Reporting efficients of variance of 6% or less are becoming the preferred platform over All clinical imaging units provide MR achievable for five metabolites (tNAA, 1.5-T for MR spectroscopy owing to spectroscopy analysis software, which tCr, tCho, mIns, Glu) with single-voxel

Radiology: Volume 270: Number 3—March 2014 n radiology.rsna.org 671 SPECIAL REVIEW: Clinical Proton MR Spectroscopy in Central Nervous System Disorders Öz et al

Figure 10 that identical and optimized acquisition protocols and calibration schemes are used (Fig 10). When it is desired to use clinical MR spectroscopy data to make decisions affecting management, it is essential that an adequate cohort of subjects has been studied such that the classifier used is robust in terms of both sensitivity and specificity. Recommend- ed cohort size will depend somewhat on the nature of the data, but anything less than several hundred subjects, both healthy subjects and those with the condition of interest, would not yield a classifier that would be certified for use by a regulatory agency. A detailed discussion of these issues has appeared recently (176,177). In addition, validation of MR spectroscopy biomarkers for clinical use requires their incorporation in robust prospective multicenter clinical trials, where patient selection and treatment meets prespecified criteria and the statistical methodology is set before the trial commences. This requires careful MR spectroscopy protocol design that can be adhered to at all the participat- ing centers. In addition, effective, real- time quality control measures must be put in place to ensure that data that need to be discarded are kept to a minimum to avoid bias and ensure general- izability of the results. Appendix E1 (online) highlights further recommendations to facilitate translation of MR spectroscopy to routine use in the clinical environment, including steps that must be taken for integration with clinical imaging and for quality management in single- and mul- tisite studies (Fig 10) and a discussion on reimbursement issues, a frequently Figure 10: Comparison of MR spectral quality at multiple sites. 1H MR spectra were acquired at three cited impediment to the widespread different sites from cerebellar volume of interest (10 3 25 3 25 mm3, as shown on T1-weighted midsagittal use of clinical MR spectroscopy. image) in three healthy individuals. Spectra were obtained with 3.0-T MR unit (Tim Trio; Siemens Healthcare, Erlangen, Germany) with same acquisition protocol (fast automatic shimming technique by mapping along projections, or FASTMAP, semi-LASER [localization by adiabatic selective refocusing] [175], 5000/28, 64 Conclusions and Recommendations repetitions). (Spectrum from Hôpital de la Salpêtrière courtesy of Fanny Mochel, MD, PhD.) MGH = Mas- sachusetts General Hospital. In conclusion, MR spectroscopy is used worldwide as an adjunct to MR imaging in several common neurologic diseases, spectroscopy at 3.0 T (174), and coef- (172,173). Importantly, standard clin- including brain neoplasms, inherited ficients of variance less than 10% were ical hardware generates reproducible metabolic disorders, demyelinating reported for tNAA, tCr, tCho, and mIns MR spectroscopy data from the human disorders, and infective focal lesions. with MR spectroscopic imaging at 3.0 T brain in a multicenter setting provided The spectrum of disorders for which

672 radiology.rsna.org n Radiology: Volume 270: Number 3—March 2014 SPECIAL REVIEW: Clinical Proton MR Spectroscopy in Central Nervous System Disorders Öz et al

MR spectroscopy will be clinically used Ramón Gilberto González, MD, PhD; Stephan ical Centre Utrecht, Utrecht, the Netherlands is likely to expand; potential examples Gruber, PhD; Rolf Gruetter, PhD; Rakesh K. (D.W.J.K., P.R.L.); Philips Healthcare, Best, Gupta, MD; Arend Heerschap, PhD; Anke the Netherlands (M.J.K.); CR-UK and EPSRC include neurodegenerative diseases Henning, PhD; Hoby P. Hetherington, PhD; Cancer Imaging Centre, Institute of Cancer Re- and epilepsy. The standardization of Franklyn A. Howe, DPhil; Petra S. Hüppi, MD; search, University of London, Sutton, England MR spectroscopy data acquisition and Ralph E. Hurd, PhD; Kejal Kantarci, MD; Den- (M.O.L.); Centre for MR in Health, Centre for analysis techniques for clinical use is nis W. J. Klomp, PhD; Roland Kreis, PhD; Clinical Spectroscopy, Brigham and Women’s Marijn J. Kruiskamp, PhD; Martin O. Leach, Hospital, Harvard University Medical School, encouraged, along with the publication PhD; Alexander P. Lin, PhD; Peter R. Luijten, Boston, Mass (A.P.L., C.E.M.); Department of normative data obtained with these PhD; Małgorzata Marjańska, PhD; Andrew A. of Radiology, University of Miami, Miami, Fla techniques. Multicenter trials are en- Maudsley, PhD; Dieter J. Meyerhoff, Dr rer nat; (A.A.M.); DVA Medical Center and Department Carolyn E. Mountford, DPhil; Sarah J. Nelson, of Radiology and Biomedical Imaging, University couraged to establish the utility of MR PhD; M. Necmettin Pamir, MD; Jullie W. Pan, of California San Francisco, San Francisco, Calif spectroscopy in large enough sample MD, PhD; Andrew C. Peet, MD, PhD; Harish (D.J.M.); Centre for MR in Health, University sizes to definitively establish the value Poptani, PhD; Stefan Posse, PhD; Petra J. W. of Newcastle, Newcastle, Australia (C.E.M.); of MR spectroscopy in specific clinical Pouwels, PhD; Eva-Maria Ratai, PhD; Brian D. Department of Radiology and Biomedical Imag- Ross, MD; Tom W. J. Scheenen, PhD; Christian ing, University of California San Francisco, San applications. Where possible, these Schuster, PhD; Ian C. P. Smith, OC, PhD, DSc; Francisco, Calif (S.J.N., D.B.V.); Department of should include assessment of the im- Brian J. Soher, PhD; Ivan Tkáč, PhD; Daniel B. Neurology, University of Pittsburgh, Pittsburgh, pact on clinical outcome and economic Vigneron, PhD; and Risto A. Kauppinen, MD, Pa (J.W.P.); Department of Cancer Sciences, benefit. Clinical imaging centers spe- PhD. University of Birmingham, Birmingham, Eng- land (A.C.P.); Department of Radiology, Uni- cializing in combined use of MR imag- Author affiliations: Center for Magnetic Res- versity of Pennsylvania, Philadelphia, Pa (H.P.); ing and spectroscopy should be estab- onance Research, University of Minnesota, Department of Neurology, University of New lished in all major clinical neurologic 2021 6th St SE, Minneapolis, MN 55455 (G.O., Mexico, Albuquerque, NM (S.P.); VU Univer- P.J.B., M.M., I.T.); Department of Radiology, sity Medical Center Amsterdam, Amsterdam, centers that offer standardized MR University of California–Los Angeles, Los An- the Netherlands (P.J.W.P.); Huntington Medi- spectroscopy procedures for improved geles, Calif (J.R.A.); Department of Radiology, cal Research Center, Pasadena, Calif (B.D.R.); patient management. Manufacturers Johns Hopkins University School of Medicine, Siemens Healthcare, Erlangen, Germany (C.S.); of MR units and third-party companies Baltimore, Md (P.B.B.); Robarts Research In- Innovative Biodiagnostics, Winnipeg, Canada stitute, University of Western Ontario, London, (I.C.P.S.); Department of Radiology, Duke Uni- (eg, vendors of analysis software) are Ont, Canada (R.B.); Neuroradiology Unit, In- versity Medical Center, Durham, NC (B.J.S.); encouraged to continue to develop their stitute Clinico Humanitas IRCCS, Milan, Italy and School of Experimental Psychology and products to incorporate recent techni- (A.B.); Departments of Radiology and Clinical Clinical Research and Imaging Centre, Univer- Research, University of Bern, Bern, Switzer- sity of Bristol, Bristol, England (R.A.K.) cal advances, to obtain U.S. Food and land (C.B., R.K.); Department of Biochemistry, Drug Administration approval for clin- University of Cambridge, Cambridge, England Disclosures of Conflicts of Interest: G.O. No ical use, and to provide products with (K.M.B.); Laboratory for Functional and Meta- relevant conflicts of interest to disclose. J.R.A. manufacturer-independent standard- bolic Imaging (LIFMET), Center for Biomedical No relevant conflicts of interest to disclose. Imaging (CIBM), Ecole Polytechnique Fédérale P.B.B. Financial activities related to the present ized outputs. de Lausanne (EPFL), Lausanne, Switzerland article: is a consultant for Olea Medical. Finan- (C.C., R.G.); Acibadem University, School of cial activities not related to the present article: Acknowledgments: The initiative for the MRS Medicine, Istanbul, Turkey (A.D., M.N.P.); none to disclose. Other relationships: none to Consensus paper was proposed by Risto Kaup- School of Health Sciences, Purdue University, disclose. R.B. Financial activities related to the pinen, MD, PhD, in the MR spectroscopy ses- West Lafayette, Ind (U.D.); Oxford Centre for present article: none to disclose. Financial activ- sion of the 8th Biennial Minnesota Workshop Functional MRI of the Brain (FMRIB), John ities not related to the present article: receives on High and Ultra-high Field Imaging held in Radcliffe Hospital, Headington, Oxford, England personal fees from Bioscape Imaging Solutions. Minneapolis, Minn, in October 2011. Drs Kaup- (U.E.E.); Biomedizinische NMR Forschungs Other relationships: none to disclose. A.B. No pinen and Öz invited the expert contributors to GmbH am Max-PIank-Institut für biophysika- relevant conflicts of interest to disclose. C.B. No the paper in due course and organized telecon- lische Chemie, Göttingen, Germany (J.F.); De- relevant conflicts of interest to disclose. P.J.B. ferences and special interest group meetings partment of Radiology, Massachusetts General No relevant conflicts of interest to disclose. held in connection to the International Society Hospital, Harvard University, Boston, Mass K.M.B. No relevant conflicts of interest to dis- for Magnetic Resonance in Medicine (ISMRM) (R.G.G., E.M.R.); Department of Radiology, close. C.C. No relevant conflicts of interest to conferences to coordinate the preparation of Medical University of Vienna, Vienna, Austria disclose. A.D. No relevant conflicts of interest to the manuscript. A subcommittee formed by (S.G.); Department of Radiology, Sanjay Gan- Jens Frahm, PhD, Roland Kreis, PhD, Peter disclose. U.D. Financial activities related to the dhi Post Graduate Institute of Medical Sciences, present article: none to disclose. Financial activ- Barker, DPhil, Andrew Peet, MD, PhD, and Lucknow, India (R.K.G.); Department of Radiol- Alberto Bizzi, MD, played a major role in edit- ities not related to the present article: receives ogy, Radboud University Nijmegen Medical Cen- payment for board membership from GyroTools. ing drafts of the manuscript. The contributors ter, Nijmegen, the Netherlands (A. Heerschap, would like to thank Philips Healthcare for mak- Other relationships: none to disclose. U.E.E. No T.W.J.S.); Max Planck Institute of Biological relevant conflicts of interest to disclose. J.F. No ing the first special interest group session at Cybernetics, Tubingen, Germany and Institute ISMRM possible. relevant conflicts of interest to disclose. R.G.G. for Biomedical Engineering, University and ETH No relevant conflicts of interest to disclose. S.G. Zurich, Zurich, Switzerland (A. Henning); De- No relevant conflicts of interest to disclose. R.G. Complete list of authors (listed in alphabetic partment of Radiology, University of Pittsburgh, No relevant conflicts of interest to disclose. order except for first and last authors): Gü- Pittsburgh, Pa (H.P.H.); Clinical Sciences, St R.K.G. No relevant conflicts of interest to dis- lin Öz, PhD; Jeffry R. Alger, MPhil, PhD; Peter George’s, University of London, London, Eng- close. A. Heerschap No relevant conflicts of in- B. Barker, DPhil; Robert Bartha, PhD; Alberto land (F.A.H.); Department of Pediatrics, Uni- terest to disclose. A. Henning No relevant con- versity of Geneva, Geneva, Switzerland (P.S.H.); Bizzi, MD; Chris Boesch, MD, PhD; Patrick J. flicts of interest to disclose. H.P.H. No relevant General Electric Healthcare, Menlo Park, Ca- Bolan, PhD; Kevin M. Brindle, DPhil; Cristina conflicts of interest to disclose. F.A.H. Financial lif (R.E.H.); Department of Radiology, Mayo Cudalbu, PhD; Alp Dinçer, MD; Ulrike Dydak, activities related to the present article: none to Clinic, Rochester, Minn (K.K.); University Med- PhD; Uzay E. Emir, PhD; Jens Frahm, PhD; disclose. Financial activities not related to the

Radiology: Volume 270: Number 3—March 2014 n radiology.rsna.org 673 SPECIAL REVIEW: Clinical Proton MR Spectroscopy in Central Nervous System Disorders Öz et al

present article: has a patent pending. Other re- References 14. Öz G, Tkác I, Charnas LR, et al. Assess- lationships: none to disclose. P.S.H. No relevant ment of adrenoleukodystrophy lesions by 1. Bottomley PA. Selective volume method for conflicts of interest to disclose. R.E.H. Financial high field MRS in non-sedated pediatric performing localized NMR spectroscopy. activities related to the present article: none to patients. Neurology 2005;64(3):434–441. disclose. Financial activities not related to the U.S. patent 4,480,228. 1984. present article: is employed by GE Healthcare. 15. Oakden WK, Noseworthy MD. Propylene 2. Ordidge RJ, Bendall MR, Gordon RE, Con- Other relationships: none to disclose. K.K. Fi- glycol is essential in the LCModel basis set nelly A. Volume selection for in vivo bio- nancial activities related to the present article: for pediatric 1H-MRS. J Comput Assist To- logical spectroscopy. In: Govil G, Khetrapal institution received fees for participation in re- mogr 2005;29(1):136–139. view activities such as data monitoring boards, CL, Saran A, eds. Magnetic resonance in statistical analysis, end point committees and biology and medicine. New Delhi, India: 16. Fein G, Meyerhoff DJ. Ethanol in human the like from Takeda Global Research, Pfizer, McGraw Hill, 1985; 387–397. brain by magnetic resonance spectroscopy: and Janssen Alzheimer’s Immunotherapy. Finan- correlation with blood and breath levels, re- 3. Frahm J, Merboldt KD, Hänicke W. Local- cial activities not related to the present article: laxation, and magnetization transfer. Alcohol ized proton spectroscopy using stimulated institution receives payment for board member- Clin Exp Res 2000;24(8):1227–1235. ship from Takeda Global Research, Pfizer, and echoes. J Magn Reson 1987;72(3):502–508. Janssen Alzheimer’s Immunotherapy. Other re- 17. Gruetter R, Weisdorf SA, Rajanayagan V, 4. Bruhn H, Frahm J, Gyngell ML, Mer- lationships: none to disclose. D.W.J.K. No rel- et al. Resolution improvements in in vivo 1H boldt KD, Hänicke W, Sauter R. Cerebral evant conflicts of interest to disclose. R.K. No NMR spectra with increased magnetic field metabolism in man after acute stroke: relevant conflicts of interest to disclose. M.J.K. strength. J Magn Reson 1998;135(1):260– Financial activities related to the present article: new observations using localized proton 264. none to disclose. Financial activities not related NMR spectroscopy. Magn Reson Med to the present article: is employed by Philip 1989;9(1):126–131. 18. Otazo R, Mueller B, Ugurbil K, Wald L, Healthcare. Other relationships: none to dis- Posse S. Signal-to-noise ratio and spectral 5. Arnold DL, Matthews PM, Francis G, An- close. M.O.L. Financial activities related to the linewidth improvements between 1.5 and 7 present article: none to disclose. Financial activ- tel J. Proton magnetic resonance spectros- Tesla in proton echo-planar spectroscopic ities not related to the present article: receives copy of human brain in vivo in the eval- imaging. Magn Reson Med 2006;56(6): nonfinancial support from Siemens, Philips, and uation of multiple sclerosis: assessment 1200–1210. GE; received a grant from Philips. Other rela- of the load of disease. Magn Reson Med tionships: none to disclose. A.P.L. No relevant 1990;14(1):154–159. 19. Frahm J, Bruhn H, Gyngell ML, Merboldt conflicts of interest to disclose. P.R.L. No rel- KD, Hänicke W, Sauter R. Localized proton evant conflicts of interest to disclose. M.M. No 6. Bruhn H, Frahm J, Gyngell ML, et al. Non- NMR spectroscopy in different regions of relevant conflicts of interest to disclose. A.A.M. invasive differentiation of tumors with use the human brain in vivo: relaxation times No relevant conflicts of interest to disclose. of localized H-1 MR spectroscopy in vivo: and concentrations of cerebral metabolites. D.J.M. No relevant conflicts of interest to dis- initial experience in patients with cerebral Magn Reson Med 1989;11(1):47–63. close. C.E.M. Financial activities related to the tumors. Radiology 1989;172(2):541–548. present article: none to disclose. Financial activ- 20. Mlynárik V, Gruber S, Moser E. Proton T ities not related to the present article: none to 7. Bottomley PA. The trouble with spectroscopy (1) and T (2) relaxation times of human disclose. Other relationships: has a patent with papers. Radiology 1991;181(2):344–350. brain metabolites at 3 Tesla. NMR Biomed Innovative Biodiagnostics. S.J.N. No relevant conflicts of interest to disclose. M.N.P. No rel- 8. Lin AP, Tran TT, Ross BD. Impact of 2001;14(5):325–331. evant conflicts of interest to disclose. J.W.P. No evidence-based medicine on magnetic 21. Hofmann L, Slotboom J, Jung B, Maloca P, relevant conflicts of interest to disclose. A.C.P. resonance spectroscopy. NMR Biomed Boesch C, Kreis R. Quantitative 1H-mag- Financial activities related to the present article: 2006;19(4):476–483. netic resonance spectroscopy of human none to disclose. Financial activities not related brain: influence of composition and param- to the present article: institution has grants/ 9. Fryback DG, Thornbury JR. The efficacy grants pending with Action Medical Research. of diagnostic imaging. Med Decis Making eterization of the basis set in linear com- Other relationships: none to disclose. H.P. No 1991;11(2):88–94. bination model-fitting. Magn Reson Med relevant conflicts of interest to disclose. S.P. Fi- 2002;48(3):440–453. nancial activities related to the present article: 10. Govindaraju V, Young K, Maudsley AA. 22. Mekle R, Mlynárik V, Gambarota G, Hergt none to disclose. Financial activities not related Proton NMR chemical shifts and coupling to the present article: none to disclose. Other constants for brain metabolites. NMR M, Krueger G, Gruetter R. MR spectros- relationships: has patents. P.J.W.P. No relevant Biomed 2000;13(3):129–153. copy of the human brain with enhanced conflicts of interest to disclose. E.M.R. No rele- signal intensity at ultrashort echo times on vant conflicts of interest to disclose. B.D.R. No 11. De Graaf RA. In vivo NMR spectroscopy: a clinical platform at 3T and 7T. Magn Re- relevant conflicts of interest to disclose. T.W.J.S. principles and techniques. 2nd ed. Hobo- son Med 2009;61(6):1279–1285. Financial activities related to the present article: ken, NJ: Wiley, 2007. none to disclose. Financial activities not related 23. Tkác I, Öz G, Adriany G, Uğurbil K, Gru- 12. Gupta RK. Magnetic resonance spectros- 1 to the present article: has a grant with Siemens etter R. In vivo H NMR spectroscopy of Healthcare. Other relationships: none to dis- copy in intracranial infection. In: Gillard the human brain at high magnetic fields: close. C.S. Financial activities related to the pre- JH, Waldman AD, Barker PB, eds. Clinical metabolite quantification at 4T vs. 7T. sent article: none to disclose. Financial activities MR neuroimaging. 2nd ed. London, Eng- Magn Reson Med 2009;62(4):868–879. not related to the present article: is employed by land: Cambridge University Press, 2010; Siemens Healthcare. Other relationships: none 426–454. 24. Deelchand DK, Van de Moortele PF, Adri- to disclose. I.C.P.S. No relevant conflicts of in- any G, et al. In vivo 1H NMR spectroscopy terest to disclose. B.J.S. No relevant conflicts of 13. Howe FA, Barton SJ, Cudlip SA, et al. of the human brain at 9.4 T: initial results. interest to disclose. I.T. No relevant conflicts of Metabolic profiles of human brain tumors J Magn Reson 2010;206(1):74–80. interest to disclose. D.B.V. No relevant conflicts using quantitative in vivo 1H magnetic res- of interest to disclose. R.A.K. No relevant con- onance spectroscopy. Magn Reson Med 25. Marjańska M, Auerbach EJ, Valabrègue flicts of interest to disclose. 2003;49(2):223–232. R, Van de Moortele PF, Adriany G, Gar-

674 radiology.rsna.org n Radiology: Volume 270: Number 3—March 2014 SPECIAL REVIEW: Clinical Proton MR Spectroscopy in Central Nervous System Disorders Öz et al

wood M. Localized 1H NMR spectroscopy malities in the infarct, blood flow, and clinical 48. Julià-Sapé M, Coronel I, Majós C, et al. in different regions of human brain in vivo outcome. Stroke 1998;29(8):1618–1624. Prospective diagnostic performance evalu- at 7 T: T2 relaxation times and concentra- ation of single-voxel 1H MRS for typing and 37. Groenendaal F, Veenhoven RH, van der tions of cerebral metabolites. NMR Biomed grading of brain tumours. NMR Biomed Grond J, Jansen GH, Witkamp TD, de 2012;25(2):332–339. 2012;25(4):661–673. Vries LS. Cerebral lactate and N-acetyl- 26. Urenjak J, Williams SR, Gadian DG, Noble aspartate/choline ratios in asphyxiated full- 49. Opstad KS, Ladroue C, Bell BA, Griffiths M. Specific expression of N-acetylaspar- term neonates demonstrated in vivo using JR, Howe FA. Linear discriminant analysis tate in neurons, oligodendrocyte-type-2 proton magnetic resonance spectroscopy. of brain tumour 1H MR spectra: a compar- astrocyte progenitors, and immature ol- Pediatr Res 1994;35(2):148–151. ison of classification using whole spectra igodendrocytes in vitro. J Neurochem versus metabolite quantification. NMR 38. Lange T, Dydak U, Roberts TP, Rowley HA, 1992;59(1):55–61. Biomed 2007;20(8):763–770. Bjeljac M, Boesiger P. Pitfalls in lactate 27. Tallan HH. Studies on the distribution of N- measurements at 3T. AJNR Am J Neuro- 50. Crawford FW, Khayal IS, McGue C, et al. acetyl-L-aspartic acid in brain. J Biol Chem radiol 2006;27(4):895–901. Relationship of pre-surgery metabolic and 1957;224(1):41–45. physiological MR imaging parameters to 39. Gill SS, Thomas DG, Van Bruggen N, et al. survival for patients with untreated GBM. 28. Moffett JR, Namboodiri MA, Cangro CB, Proton MR spectroscopy of intracranial tu- J Neurooncol 2009;91(3):337–351. Neale JH. Immunohistochemical localiza- mours: in vivo and in vitro studies. J Com- tion of N-acetylaspartate in rat brain. Neu- put Assist Tomogr 1990;14(4):497–504. 51. Murphy PS, Rowland IJ, Viviers L, Brada roreport 1991;2(3):131–134. M, Leach MO, Dzik-Jurasz AS. Could as- 40. Peeling J, Sutherland G. High-resolution sessment of glioma methylene lipid res- 29. Miller BL. A review of chemical issues in 1H NMR spectroscopy studies of extracts onance by in vivo 1H-MRS be of clinical 1H NMR spectroscopy: N-acetyl-L-aspar- of human cerebral neoplasms. Magn Reson value? Br J Radiol 2003;76(907):459– tate, creatine and choline. NMR Biomed Med 1992;24(1):123–136. 463. 1991;4(2):47–52. 41. Florian CL, Preece NE, Bhakoo KK, Wil- 52. Poptani H, Gupta RK, Roy R, Pandey R, 30. Rigotti DJ, Inglese M, Gonen O. Whole- liams SR, Noble M. Characteristic meta- Jain VK, Chhabra DK. Characterization of brain N-acetylaspartate as a surrogate bolic profiles revealed by 1H NMR spectros- intracranial mass lesions with in vivo pro- marker of neuronal damage in diffuse neu- copy for three types of human brain and ton MR spectroscopy. AJNR Am J Neuro- rologic disorders. AJNR Am J Neuroradiol nervous system tumours. NMR Biomed radiol 1995;16(8):1593–1603. 2007;28(10):1843–1849. 1995;8(6):253–264. 53. Kovanlikaya A, Panigrahy A, Krieger MD, 31. Bates TE, Strangward M, Keelan J, Davey 42. Hourani R, Brant LJ, Rizk T, Weingart et al. Untreated pediatric primitive neuro- GP, Munro PM, Clark JB. Inhibition of N- JD, Barker PB, Horská A. Can proton MR ectodermal tumor in vivo: quantitation of acetylaspartate production: implications spectroscopic and perfusion imaging dif- taurine with MR spectroscopy. Radiology for 1H MRS studies in vivo. Neuroreport ferentiate between neoplastic and nonneo- 2005;236(3):1020–1025. 1996;7(8):1397–1400. plastic brain lesions in adults? AJNR Am J Neuroradiol 2008;29(2):366–372. 54. Davies NP, Wilson M, Natarajan K, et 32. Pouwels PJ, Kruse B, Korenke GC, Mao X, al. Non-invasive detection of glycine as Hanefeld FA, Frahm J. Quantitative proton 43. Law M, Yang S, Wang H, et al. Glioma grad- a biomarker of malignancy in childhood magnetic resonance spectroscopy of child- ing: sensitivity, specificity, and predictive brain tumours using in-vivo 1H MRS at 1.5 hood adrenoleukodystrophy. Neuropediat- values of perfusion MR imaging and pro- Tesla confirmed by ex-vivo high-resolution rics 1998;29(5):254–264. ton MR spectroscopic imaging compared magic-angle spinning NMR. NMR Biomed with conventional MR imaging. AJNR Am 2010;23(1):80–87. 33. Kantarci K, Knopman DS, Dickson DW, J Neuroradiol 2003;24(10):1989–1998. et al. Alzheimer disease: postmortem 55. McKnight TR, Lamborn KR, Love TD, neuropathologic correlates of antemor- 44. García-Gómez JM, Luts J, Julià-Sapé M, et al. Correlation of magnetic resonance tem 1H MR spectroscopy metabolite mea- et al. Multiproject-multicenter evaluation spectroscopic and growth characteristics surements. Radiology 2008;248(1):210– of automatic brain tumor classification by within grades II and III gliomas. J Neuro- 220. magnetic resonance spectroscopy. MAGMA surg 2007;106(4):660–666. 2009;22(1):5–18. 34. Licata SC, Renshaw PF. Neurochemistry of 56. Chang SM, Nelson S, Vandenberg S, et al. drug action: insights from proton magnetic 45. Vicente J, Fuster-Garcia E, Tortajada S, et Integration of preoperative anatomic and resonance spectroscopic imaging and their al. Accurate classification of childhood brain metabolic physiologic imaging of newly di- relevance to addiction. Ann N Y Acad Sci tumours by in vivo ¹H MRS: a multi-centre agnosed glioma. J Neurooncol 2009;92(3): 2010;1187:148–171. study. Eur J Cancer 2013;49(3):658–667. 401–415.

35. Richards TL. Proton MR spectroscopy in 46. Tate AR, Underwood J, Acosta DM, et al. 57. Chawla S, Zhang Y, Wang S, et al. Pro- multiple sclerosis: value in establishing Development of a decision support system ton magnetic resonance spectroscopy in diagnosis, monitoring progression, and for diagnosis and grading of brain tumours differentiating glioblastomas from primary evaluating therapy. AJR Am J Roentgenol using in vivo magnetic resonance single voxel cerebral lymphomas and brain metastases. 1991;157(5):1073–1078. spectra. NMR Biomed 2006;19(4):411–434. J Comput Assist Tomogr 2010;34(6):836– 841. 36. Wardlaw JM, Marshall I, Wild J, Dennis MS, 47. Choi C, Ganji SK, DeBerardinis RJ, et Cannon J, Lewis SC. Studies of acute ische- al. 2-hydroxyglutarate detection by mag- 58. Wijnen JP, Idema AJ, Stawicki M, et al. mic stroke with proton magnetic resonance netic resonance spectroscopy in IDH- Quantitative short echo time 1H MRSI of spectroscopy: relation between time from mutated patients with gliomas. Nat Med the peripheral edematous region of human onset, neurological deﬁcit, metabolite abnor- 2012;18(4):624–629. brain tumors in the differentiation between

Radiology: Volume 270: Number 3—March 2014 n radiology.rsna.org 675 SPECIAL REVIEW: Clinical Proton MR Spectroscopy in Central Nervous System Disorders Öz et al

glioblastoma, metastasis, and meningioma. tinguishing glioma recurrence from post- disorders in childhood. In: Bachelard HS, J Magn Reson Imaging 2012;36(5):1072– treatment effects. J Magn Reson Imaging ed. Magnetic resonance spectroscopy and 1082. 2012;35(1):56–63. imaging in neurochemistry. New York, NY: Plenum, 1997; 329–402. 59. Al-Okaili RN, Krejza J, Woo JH, et al. In- 70. Murphy PS, Viviers L, Abson C, et al. Moni- traaxial brain masses: MR imaging–based toring temozolomide treatment of low-grade 81. Engelke U, Moolenaar S, Hoenderop S, et diagnostic strategy—initial experience. Ra- glioma with proton magnetic resonance al. Handbook of 1H NMR spectroscopy in diology 2007;243(2):539–550. spectroscopy. Br J Cancer 2004;90(4): inborn errors of metabolism: body fluid 781–786. NMR spectroscopy and in vivo MR spec- 60. Hock A, Henning A, Boesiger P, Kollias troscopy. 2nd ed. Heilbronn, Germany: 1 SS. H-MR spectroscopy in the human 71. Matsusue E, Fink JR, Rockhill JK, Ogawa SPS Verlagsgesellschaft, 2007. spinal cord. AJNR Am J Neuroradiol T, Maravilla KR. Distinction between gli- 2013;34(9):1682–1689. oma progression and post-radiation change 82. Heindel W, Kugel H, Roth B. Noninvasive by combined physiologic MR imaging. Neu- detection of increased glycine content by 61. Steffen-Smith EA, Shih JH, Hipp SJ, Bent roradiology 2010;52(4):297–306. proton MR spectroscopy in the brains R, Warren KE. Proton magnetic resonance of two infants with nonketotic hypergly- spectroscopy predicts survival in children 72. Prat R, Galeano I, Lucas A, et al. Relative cinemia. AJNR Am J Neuroradiol 1993; with diffuse intrinsic pontine glioma. J value of magnetic resonance spectroscopy, 14(3):629–635. Neurooncol 2011;105(2):365–373. magnetic resonance perfusion, and 2-(18F) fluoro-2-deoxy-D-glucose positron emission 83. Detre JA, Wang ZY, Bogdan AR, et al. 62. Blüml S, Panigrahy A, Laskov M, et al. tomography for detection of recurrence or Regional variation in brain lactate in Elevated citrate in pediatric astrocytomas 1 grade increase in gliomas. J Clin Neurosci Leigh syndrome by localized H magnetic with malignant progression. Neurooncol 2010;17(1):50–53. resonance spectroscopy. Ann Neurol 2011;13(10):1107–1117. 1991;29(2):218–221. 73. Hüppi PS, Posse S, Lazeyras F, Burri R, 63. Wilson M, Cummins CL, Macpherson L, Bossi E, Herschkowitz N. Magnetic reso- 84. Wilichowski E, Pouwels PJ, Frahm J, et al. Magnetic resonance spectroscopy nance in preterm and term newborns: 1H- Hanefeld F. Quantitative proton magnetic metabolite profiles predict survival in spectroscopy in developing human brain. resonance spectroscopy of cerebral meta- paediatric brain tumours. Eur J Cancer Pediatr Res 1991;30(6):574–578. bolic disturbances in patients with MELAS. 2013;49(2):457–464. Neuropediatrics 1999;30(5):256–263. 74. van der Knaap MS, van der Grond J, van 64. Einstein DB, Wessels B, Bangert B, et al. 85. Bruhn H, Kruse B, Korenke GC, et al. Rijen PC, Faber JA, Valk J, Willemse K. Phase II trial of radiosurgery to magnetic Proton NMR spectroscopy of cerebral Age-dependent changes in localized proton resonance spectroscopy-defined high-risk metabolic alterations in infantile peroxi- and phosphorus MR spectroscopy of the tumor volumes in patients with glioblas- somal disorders. J Comput Assist Tomogr brain. Radiology 1990;176(2):509–515. toma multiforme. Int J Radiat Oncol Biol 1992;16(3):335–344. Phys 2012;84(3):668–674. 75. Hanefeld F, Bauer HJ, Christen HJ, Kruse 86. Zand DJ, Simon EM, Pulitzer SB, et al. B, Bruhn H, Frahm J. Multiple sclerosis in 65. Stadlbauer A, Moser E, Gruber S, et al. In vivo pyruvate detected by MR spec- childhood: report of 15 cases. Brain Dev Improved delineation of brain tumors: troscopy in neonatal pyruvate dehydroge- 1991;13(6):410–416. an automated method for segmentation nase deficiency. AJNR Am J Neuroradiol based on pathologic changes of 1H-MR- 76. Lodygensky GA, Menache CC, Hüppi PS. 2003;24(7):1471–1474. SI metabolites in gliomas. Neuroimage Magnetic resonance imaging’s role in the 87. Ghezzi D, Goffrini P, Uziel G, et al. SD- 2004;23(2):454–461. care of the infant at risk for brain injury. HAF1, encoding a LYR complex-II specific In: Perlman JM, ed. Neurology: neonatol- 66. Kallenberg K, Bock HC, Helms G, et al. assembly factor, is mutated in SDH-de- ogy questions and controversies. 2nd ed. Untreated glioblastoma multiforme: infective infantile leukoencephalopathy. Nat Amsterdam, the Netherlands: Elsevier, creased myo-inositol and glutamine levels Genet 2009;41(6):654–656. 2013. in the contralateral cerebral hemisphere 88. Ohlenbusch A, Edvardson S, Skorpen J, et at proton MR spectroscopy. Radiology 77. Hanrahan JD, Cox IJ, Edwards AD, et al. al. Leukoencephalopathy with accumulated 2009;253(3):805–812. Persistent increases in cerebral lactate succinate is indicative of SDHAF1 related concentration after birth asphyxia. Pediatr 67. Scheenen TW, Klomp DW, Wijnen JP, complex II deficiency. Orphanet J Rare Dis Res 1998;44(3):304–311. Heerschap A. Short echo time 1H-MRSI 2012;7(1):69. of the human brain at 3T with minimal 78. Azzopardi DV, Strohm B, Edwards AD, et 89. Manley BJ, Sokol J, Cheong JL. Intrace- chemical shift displacement errors using al. Moderate hypothermia to treat perina- rebral blood and MRS in neonatal non- adiabatic refocusing pulses. Magn Reson tal asphyxial encephalopathy. N Engl J Med ketotic hyperglycinemia. Pediatr Neurol Med 2008;59(1):1–6. 2009;361(14):1349–1358. 2010;42(3):219–222.

68. Pamir MN, Özduman K, Dinçer A, Yildiz 79. van der Knaap MS, Pouwels PJ. Magnetic 90. Austin SJ, Connelly A, Gadian DG, Benton E, Peker S, Özek MM. First intraoperative, resonance spectroscopy: basic principles JS, Brett EM. Localized 1H NMR spectros- shared-resource, ultrahigh-ﬁeld 3-Tesla and application in white matter disorders. copy in Canavan’s disease: a report of two magnetic resonance imaging system and its In: van der Knaap MS, Valk J, eds. Mag- cases. Magn Reson Med 1991;19(2):439– application in low-grade glioma resection. J netic resonance of myelination and my- 445. Neurosurg 2010;112(1):57–69. elin disorders. 3rd ed. Berlin, Germany: 91. Mercimek-Mahmutoglu S, Stöckler-Ipsiro- Springer, 2005; 859–880. 69. Fink JR, Carr RB, Matsusue E, et al. Com- glu S, Salomons GS. Creatine deﬁciency parison of 3 Tesla proton MR spectroscopy, 80. Frahm J, Hanefeld F. Localized proton syndromes. GeneReviews 2009. http:// MR perfusion and MR diffusion for dis- magnetic resonance spectroscopy of brain www.ncbi.nlm.nih.gov/books/NBK3794/.

676 radiology.rsna.org n Radiology: Volume 270: Number 3—March 2014 SPECIAL REVIEW: Clinical Proton MR Spectroscopy in Central Nervous System Disorders Öz et al

Updated August 18, 2011. Accessed April 103. Kreis R, Arcinue E, Ernst T, Shonk TK, by proton magnetic resonance spectros- 27, 2012. Flores R, Ross BD. Hypoxic encephalopa- copy. Lancet 1993;341(8845):630–631. thy after near-drowning studied by quan- 92. Stöckler S, Hanefeld F, Frahm J. Crea- 115. Hannoun S, Bagory M, Durand-Dubief F, titative 1H-magnetic resonance spectros- tine replacement therapy in guanidino- et al. Correlation of diffusion and meta- copy. J Clin Invest 1996;97(5):1142–1154. acetate methyltransferase deficiency, a bolic alterations in different clinical forms novel inborn error of metabolism. Lancet 104. Bizzi A, Castelli G, Bugiani M, et al. of multiple sclerosis. PLoS ONE 2012;7(3): 1996;348(9030):789–790. Classification of childhood white matter dis- e32525. orders using proton MR spectroscopic im- 93. Ndika JD, Johnston K, Barkovich JA, et al. aging. AJNR Am J Neuroradiol 2008;29(7): 116. Vrenken H, Barkhof F, Uitdehaag BM, Developmental progress and creatine res- 1270–1275. Castelijns JA, Polman CH, Pouwels PJ. MR toration upon long-term creatine supple- spectroscopic evidence for glial increase mentation of a patient with arginine:glycine 105. van der Voorn JP, Pouwels PJ, Hart AA, but not for neuro-axonal damage in MS amidinotransferase deficiency. Mol Genet et al. Childhood white matter disorders: normal-appearing white matter. Magn Re- Metab 2012;106(1):48–54. quantitative MR imaging and spectroscopy. son Med 2005;53(2):256–266. Radiology 2006;241(2):510–517. 94. van de Kamp JM, Pouwels PJ, Aarsen FK, 117. De Stefano N, Narayanan S, Francis GS, et et al. Long-term follow-up and treatment 106. Cartier N, Aubourg P. Hematopoietic stem al. Evidence of axonal damage in the early in nine boys with X-linked creatine trans- cell gene therapy in Hurler syndrome, glo- stages of multiple sclerosis and its rele- porter defect. J Inherit Metab Dis 2012; boid cell leukodystrophy, metachromatic vance to disability. Arch Neurol 2001;58(1): 35(1):141–149. leukodystrophy and X-adrenoleukodystro- 65–70. phy. Curr Opin Mol Ther 2008;10(5):471– 95. Bizzi A, Bugiani M, Salomons GS, et al. 478. 118. Tartaglia MC, Narayanan S, De Stefano N, X-linked creatine deficiency syndrome: a et al. Choline is increased in pre-lesional 107. Eichler FS, Barker PB, Cox C, et al. Proton novel mutation in creatine transporter gene normal appearing white matter in multiple MR spectroscopic imaging predicts lesion SLC6A8. Ann Neurol 2002;52(2):227–231. sclerosis. J Neurol 2002;249(10):1382– progression on MRI in X-linked adrenoleu- 1390. 96. Wiame E, Tyteca D, Pierrot N, et al. Mo- kodystrophy. Neurology 2002;58(6):901– lecular identification of aspartate N-acetyl- 907. 119. De Stefano N, Filippi M, Miller D, et al. transferase and its mutation in hypoacety- Guidelines for using proton MR spectros- 108. Wilken B, Dechent P, Brockmann K, et al. laspartia. Biochem J 2010;425(1):127–136. copy in multicenter clinical MS studies. Quantitative proton magnetic resonance Neurology 2007;69(20):1942–1952. 97. Martin E, Capone A, Schneider J, Hennig spectroscopy of children with adrenoleuko- J, Thiel T. Absence of N-acetylaspartate in dystrophy before and after hematopoietic 120. de Graaf WL, Kilsdonk ID, Lopez-Soriano the human brain: impact on neurospectros- stem cell transplantation. Neuropediatrics A, et al. Clinical application of multi-con- copy? Ann Neurol 2001;49(4):518–521. 2003;34(5):237–246. trast 7-T MR imaging in multiple sclerosis: increased lesion detection compared to 109. Kruse B, Hanefeld F, Christen HJ, et al. 98. Pouwels PJ, Brockmann K, Kruse B, et al. 3 T confined to grey matter. Eur Radiol Alterations of brain metabolites in meta- Regional age dependence of human brain 2013;23(2):528–540. metabolites from infancy to adulthood as chromatic leukodystrophy as detected by lo- detected by quantitative localized proton calized proton magnetic resonance spectros- 121. Bizzi A, Uluğ AM, Crawford TO, et al. MRS. Pediatr Res 1999;46(4):474–485. copy in vivo. J Neurol 1993;241(2):68–74. Quantitative proton MR spectroscopic imaging in acute disseminated enceph- 99. Davison JE, Hendriksz CJ, Sun Y, Davies 110. Eichler F, Grodd W, Grant E, et al. Meta- chromatic leukodystrophy: a scoring system alomyelitis. AJNR Am J Neuroradiol NP, Gissen P, Peet AC. Quantitative in 2001;22(6):1125–1130. vivo brain magnetic resonance spectro- for brain MR imaging observations. AJNR scopic monitoring of neurological involve- Am J Neuroradiol 2009;30(10):1893–1897. 122. Chang KH, Song IC, Kim SH, et al. In vivo ment in mucopolysaccharidosis type II 111. i Dali C, Hanson LG, Barton NW, Fogh J, single-voxel proton MR spectroscopy in in- (Hunter syndrome). J Inherit Metab Dis Nair N, Lund AM. Brain N-acetylaspartate tracranial cystic masses. AJNR Am J Neu- 2010;33(Suppl 3):395–399. levels correlate with motor function in meta- roradiol 1998;19(3):401–405. chromatic leukodystrophy. Neurology 2010; 100. Davison JE, Davies NP, Wilson M, et al. 123. Agarwal M, Chawla S, Husain N, Jaggi RS, 75(21):1896–1903. MR spectroscopy–based brain metabolite Husain M, Gupta RK. Higher succinate profiling in propionic acidaemia: metabolic 112. Ding XQ, Bley A, Kohlschütter A, Fiehler than acetate levels differentiate cerebral changes in the basal ganglia during acute J, Lanfermann H. Long-term neuroimag- degenerating cysticerci from anaerobic decompensation and effect of liver trans- ing follow-up on an asymptomatic juvenile abscesses on in-vivo proton MR spectros- plantation. Orphanet J Rare Dis 2011;6:19. metachromatic leukodystrophy patient copy. Neuroradiology 2004;46(3):211–215. after hematopoietic stem cell transplanta- 101. Langlois JA, Rutland-Brown W, Thomas 124. Sturrock A, Laule C, Decolongon J, et tion: evidence of myelin recovery and on- KE. The incidence of traumatic brain in- al. Magnetic resonance spectroscopy bio- going brain maturation. Am J Med Genet jury among children in the United States: markers in premanifest and early Hunting- A 2012;158A(1):257–260. differences by race. J Head Trauma Rehabil ton disease. Neurology 2010;75(19):1702– 2005;20(3):229–238. 113. Sajja BR, Wolinsky JS, Narayana PA. Pro- 1710. ton magnetic resonance spectroscopy in 102. Aaen GS, Holshouser BA, Sheridan C, 125. Kantarci K, Jack CR Jr, Xu YC, et al. Re- multiple sclerosis. Neuroimaging Clin N et al. Magnetic resonance spectros- gional metabolic patterns in mild cogni- Am 2009;19(1):45–58. copy predicts outcomes for children tive impairment and Alzheimer’s disease: with nonaccidental trauma. Pediatrics 114. Davie CA, Hawkins CP, Barker GJ, et al. a 1H MRS study. Neurology 2000;55(2): 2010;125(2):295–303. Detection of myelin breakdown products 210–217.

Radiology: Volume 270: Number 3—March 2014 n radiology.rsna.org 677 SPECIAL REVIEW: Clinical Proton MR Spectroscopy in Central Nervous System Disorders Öz et al

126. Adalsteinsson E, Sullivan EV, Kleinhans N, pharmacol Biol Psychiatry 2008;32(3): meta-analysis. Epilepsy Res 2006;71(2-3): Spielman DM, Pfefferbaum A. Longitudi- 786–793. 149–158. nal decline of the neuronal marker N-acetyl 139. Penner J, Rupsingh R, Smith M, Wells JL, 152. Fountas KN, Tsougos I, Gotsis ED, Gianna- aspartate in Alzheimer’s disease. Lancet Borrie MJ, Bartha R. Increased glutamate kodimos S, Smith JR, Kapsalaki EZ. Tem- 2000;355(9216):1696–1697. in the hippocampus after galantamine treat- poral pole proton preoperative magnetic 127. Öz G, Hutter D, Tkác I, et al. Neurochem- ment for Alzheimer disease. Prog Neuropsy- resonance spectroscopy in patients under- ical alterations in spinocerebellar ataxia chopharmacol Biol Psychiatry 2010;34(1): going surgery for mesial temporal sclero- type 1 and their correlations with clinical 104–110. sis. Neurosurg Focus 2012;32(3):E3. status. Mov Disord 2010;25(9):1253–1261. 140. Kwan P, Brodie MJ. Early identification of 153. Pan JW, Avdievich N, Hetherington HP. 128. Unschuld PG, Edden RA, Carass A, et al. refractory epilepsy. N Engl J Med 2000; J-refocused coherence transfer spectro- Brain metabolite alterations and cognitive 342(5):314–319. scopic imaging at 7 T in human brain. dysfunction in early Huntington’s disease. Magn Reson Med 2010;64(5):1237–1246. 141. Del Felice A, Beghi E, Boero G, et al. Early Mov Disord 2012;27(7):895–902. versus late remission in a cohort of pa- 154. Cross JH, Connelly A, Jackson GD, John- 129. Kantarci K, Weigand SD, Petersen RC, et al. tients with newly diagnosed epilepsy. Epi- son CL, Neville BG, Gadian DG. Proton Longitudinal 1H MRS changes in mild cog- lepsia 2010;51(1):37–42. magnetic resonance spectroscopy in chil- nitive impairment and Alzheimer’s disease. dren with temporal lobe epilepsy. Ann 142. Wiebe S, Jetté N. Epilepsy surgery utiliza- Neurobiol Aging 2007;28(9):1330–1339. Neurol 1996;39(1):107–113. tion: who, when, where, and why? Curr 130. Griffith HR, Stewart CC, den Hollander JA. Opin Neurol 2012;25(2):187–193. 155. González RG. Clinical MRI of acute ische- Proton magnetic resonance spectroscopy mic stroke. J Magn Reson Imaging 2012; in dementias and mild cognitive impair- 143. Englot DJ, Wang DD, Rolston JD, Shih TT, 36(2):259–271. ment. Int Rev Neurobiol 2009;84:105–131. Chang EF. Rates and predictors of long- term seizure freedom after frontal lobe 156. van der Toorn A, Verheul HB, Berkelbach 131. Rupsingh R, Borrie M, Smith M, Wells JL, epilepsy surgery: a systematic review and van der Sprenkel JW, Tulleken CA, Nicolay Bartha R. Reduced hippocampal glutamate meta-analysis. J Neurosurg 2012;116(5): K. Changes in metabolites and tissue water in Alzheimer disease. Neurobiol Aging 2011; 1042–1048. status after focal ischemia in cat brain as- 32(5):802–810. sessed with localized proton MR spectros- 144. Siesjö BK. Brain energy metabolism. Lon- copy. Magn Reson Med 1994;32(6):685–691. 132. Öz G, Iltis I, Hutter D, Thomas W, Bushara don, England: Wiley, 1978. KO, Gomez CM. Distinct neurochemical 157. Sager TN, Laursen H, Hansen AJ. Changes profiles of spinocerebellar ataxias 1, 2, 6, 145. Pan JW, Williamson A, Cavus I, et al. Neu- in N-acetyl-aspartate content during focal and cerebellar multiple system atrophy. rometabolism in human epilepsy. Epilepsia and global brain ischemia of the rat. J Cereb Cerebellum 2011;10(2):208–217. 2008;49(Suppl 3):31–41. Blood Flow Metab 1995;15(4):639–646.

133. Miller BL, Moats RA, Shonk T, Ernst T, 146. Capizzano AA, Vermathen P, Laxer KD, et 158. Saunders DE. MR spectroscopy in stroke. Woolley S, Ross BD. Alzheimer disease: al. Temporal lobe epilepsy: qualitative read- Br Med Bull 2000;56(2):334–345. depiction of increased cerebral myo-inosi- ing of 1H MR spectroscopic images for pre- tol with proton MR spectroscopy. Radiol- surgical evaluation. Radiology 2001;218(1): 159. Parsons MW, Li T, Barber PA, et al. Com- 1 ogy 1993;187(2):433–437. 144–151. bined H MR spectroscopy and diffusion- weighted MRI improves the prediction of 134. Ross BD, Bluml S, Cowan R, Danielsen E, 147. Connelly A, Jackson GD, Duncan JS, King stroke outcome. Neurology 2000;55(4): Farrow N, Tan J. In vivo MR spectroscopy MD, Gadian DG. Magnetic resonance 498–505. of human dementia. Neuroimaging Clin N spectroscopy in temporal lobe epilepsy. Am 1998;8(4):809–822. Neurology 1994;44(8):1411–1417. 160. Petroff OA, Graham GD, Blamire AM, et al. Spectroscopic imaging of stroke in humans: 135. Godbolt AK, Waldman AD, MacManus 148. Simister RJ, McLean MA, Barker GJ, Dun- histopathology correlates of spectral chang- DG, et al. MRS shows abnormalities before can JS. Proton MR spectroscopy of metab- es. Neurology 1992;42(7):1349–1354. symptoms in familial Alzheimer disease. olite concentrations in temporal lobe epi- Neurology 2006;66(5):718–722. lepsy and effect of temporal lobe resection. 161. Klijn CJ, Kappelle LJ, van der Grond J, Al- Epilepsy Res 2009;83(2-3):168–176. gra A, Tulleken CA, van Gijn J. Magnetic 136. Kantarci K, Boeve BF, Wszolek ZK, et al. resonance techniques for the identiﬁcation MRS in presymptomatic MAPT mutation 149. Maudsley AA, Domenig C, Ramsay RE, of patients with symptomatic carotid artery carriers: a potential biomarker for tau- Bowen BC. Application of volumetric occlusion at high risk of cerebral ischemic mediated pathology. Neurology 2010;75(9): MR spectroscopic imaging for localiza- events. Stroke 2000;31(12):3001–3007. 771–778. tion of neocortical epilepsy. Epilepsy Res 2010;88(2-3):127–138. 162. Emir UE, Tuite PJ, Öz G. Elevated pontine 137. Krishnan KR, Charles HC, Doraiswamy and putamenal GABA levels in mild-moder- PM, et al. Randomized, placebo-controlled 150. Pan JW, Duckrow RB, Spencer DD, Avdi- ate Parkinson disease detected by 7 Tesla trial of the effects of donepezil on neuronal evich NI, Hetherington HP. Selective ho- proton MRS. PLoS ONE 2012;7(1):e30918. markers and hippocampal volumes in Al- monuclear polarization transfer for spec- zheimer’s disease. Am J Psychiatry 2003; troscopic imaging of GABA at 7T. Magn 163. Scheenen TW, Heerschap A, Klomp DW. 160(11):2003–2011. Reson Med 2013;69(2):310–316. Towards 1H-MRSI of the human brain at 7T with slice-selective adiabatic refocusing 138. Bartha R, Smith M, Rupsingh R, Rylett J, 151. Willmann O, Wennberg R, May T, Woer- pulses. MAGMA 2008;21(1-2):95–101. Wells JL, Borrie MJ. High ﬁeld 1H MRS of mann FG, Pohlmann-Eden B. The role of 1H the hippocampus after donepezil treatment magnetic resonance spectroscopy in pre- 164. Kreis R. Issues of spectral quality in clinical in Alzheimer disease. Prog Neuropsycho- operative evaluation for epilepsy surgery: a 1H-magnetic resonance spectroscopy and

678 radiology.rsna.org n Radiology: Volume 270: Number 3—March 2014 SPECIAL REVIEW: Clinical Proton MR Spectroscopy in Central Nervous System Disorders Öz et al

a gallery of artifacts. NMR Biomed 2004; 169. Schirmer T, Auer DP. On the reliability of of human brain at 3T. Magn Reson Med 17(6):361–381. quantitative clinical magnetic resonance 2011;66(2):324–332. spectroscopy of the human brain. NMR 165. Soher BJ, Semanchuk P, Todd D, Steinberg Biomed 2000;13(1):28–36. 174. Terpstra M, Emir UE, Eberly LE, Öz G. J, Young K. Vespa: integrated applications Test-retest repeatability of human neuro- for RF pulse design, spectral simulation 170. Hammen T, Stadlbauer A, Tomandl B, et al. chemical profiles measured at 3 versus 7 T 1 and MRS data analysis [abstr]. In: Pro- Short TE single-voxel H-MR spectroscopy [abstr]. In: Proceedings of the Twentieth ceedings of the Nineteenth Meeting of the of hippocampal structures in healthy adults Meeting of the International Society for International Society for Magnetic Reso- at 1.5 Tesla: how reproducible are the re- Magnetic Resonance in Medicine. Berke- nance in Medicine. Berkeley, Calif: Inter- sults? NMR Biomed 2005;18(3):195–201. ley, Calif: International Society for Mag- national Society for Magnetic Resonance in 171. Li BS, Babb JS, Soher BJ, Maudsley AA, netic Resonance in Medicine, 2012; 1739. Medicine, 2011; 1410. Gonen O. Reproducibility of 3D proton 175. Öz G, Tkáč I. Short-echo, single-shot, full- 166. Provencher SW. Automatic quantitation of spectroscopy in the human brain. Magn intensity proton magnetic resonance spec- localized in vivo 1H spectra with LCModel. Reson Med 2002;47(3):439–446. troscopy for neurochemical profiling at 4 T: NMR Biomed 2001;14(4):260–264. 172. Wijnen JP, van Asten JJ, Klomp DW, et validation in the cerebellum and brainstem. 1 Magn Reson Med 2011;65(4):901–910. 167. Scheidegger O, Wingeier K, Stefan D, et al. Short echo time H MRSI of the human al. Optimized quantitative magnetic res- brain at 3T with adiabatic slice-selective 176. Smith IC, Somorjai RL. Deriving biomed- onance spectroscopy for clinical routine. refocusing pulses: reproducibility and vari- ical diagnostics from NMR spectroscopic Magn Reson Med 2013;70(1):25–32. ance in a dual center setting. J Magn Reson data. Biophys Rev 2011;3(1):47–52. Imaging 2010;31(1):61–70. 168. Gu M, Kim DH, Mayer D, Sullivan EV, Pfef- 177. Somorjai RL. Creating robust, reliable, ferbaum A, Spielman DM. Reproducibility 173. Gasparovic C, Bedrick EJ, Mayer AR, et clinically relevant classifiers from spectro- study of whole-brain 1H spectroscopic im- al. Test-retest reliability and reproducibility scopic data. Biophys Rev 2009;1(4):201– aging with automated quantification. Magn of short-echo-time spectroscopic imaging 211. Reson Med 2008;60(3):542–547.

Radiology: Volume 270: Number 3—March 2014 n radiology.rsna.org 679 TECHNICAL REPORTS

2-hydroxyglutarate detection by magnetic resonance spectroscopy in subjects with IDH-mutated gliomas

Changho Choi1,2, Sandeep K Ganji1,2, Ralph J DeBerardinis3–5, Kimmo J Hatanpaa5–7, Dinesh Rakheja6,8, Zoltan Kovacs1, Xiao-Li Yang5,7,9, Tomoyuki Mashimo5,7,9, Jack M Raisanen5–7, Isaac Marin-Valencia3, Juan M Pascual3,10,11, Christopher J Madden5,7,12, Bruce E Mickey5,7,12, Craig R Malloy1,2,9,13, Robert M Bachoo5,7,9,10 & Elizabeth A Maher5,7,9,10

Mutations in isocitrate dehydrogenases 1 and 2 (IDH1 tumors2,6. Immunohistochemistry with a commercially available anti- and IDH2) have been shown to be present in most World body to the R132H mutation of IDH1 identifies approximately 93% Health Organization grade 2 and grade 3 gliomas in adults. of the mutations, but the remaining 7% of tumors carrying a different These mutations are associated with the accumulation of IDH1 or an IDH2 mutation require direct sequencing for detection7. 2-hydroxyglutarate (2HG) in the tumor. Here we report the As 2HG is produced by all known IDH-mutant enzymes, evaluation noninvasive detection of 2HG by proton magnetic resonance of 2HG abundance is an alternative indirect method for determining spectroscopy (MRS). We developed and optimized the pulse IDH status. The finding that 2HG is present at high levels in IDH- sequence with numerical and phantom analyses for 2HG mutated gliomas has raised the possibility that this metabolite could be detection, and we estimated the concentrations of 2HG using detected noninvasively by MRS. Brain magnetic resonance imaging is spectral fitting in the tumors of 30 subjects. Detection of 2HG the primary modality for clinical evaluation of people with gliomas, so correlated with mutations in IDH1 or IDH2 and with increased the ability to detect 2HG by MRS could provide important diagnostic levels of D-2HG by mass spectrometry of the resected tumors. and prognostic information. Noninvasive detection of 2HG may prove to be a valuable Here we report the noninvasive detection of 2HG in glioma in vivo diagnostic and prognostic biomarker. by MRS at 3 tesla (T). We optimized the point-resolved spectroscopy (PRESS)8 and difference editing9 sequences with quantum-mechanical Isocitrate dehydrogenase converts isocitrate to a-ketoglutarate (aKG) in and phantom analyses for detection of 2HG in the human brain and the cytosol (IDH1) and mitochondria (IDH2). The recent identification applied them to tumor masses in 30 adults with all grades of gliomas. © 2012 Nature America, Inc. All rights reserved. America, Inc. © 2012 Nature of mutations in IDH1 and IDH2 among most humans with World Health Analysis of MRS data was blinded to IDH status. For each case in which Organization (WHO) grade 2 and 3 gliomas1,2 has directed attention 2HG was detected by MRS, we confirmed an IDH1 or IDH2 mutation to the role of abnormal metabolism in the pathogenesis and progres- in the tumor. Failure to detect 2HG by MRS was associated with the npg sion of these primary brain tumors. The mutations are confined to the detection of wild-type IDH1 and IDH2 in each case. The sensitivity and active site of the enzyme and result in a gain of function that generates specificity of the method described here and the ease with which it could 2HG (ref. 3) and induces DNA hypermethylation4,5. The abundance of be incorporated into standard magnetic resonance imaging suggests that this metabolite, normally present in vanishingly small quantities, can be 2HG detection by MRS may be an important biomarker in the clinical elevated by orders of magnitude in gliomas with IDH1 or IDH2 muta- management of these patients. tions. Intracellular concentrations on the order of several micromoles per gram of tumor have been reported3. RESULTS Although the metabolic consequences and downstream molecular Optimization of MRS methods effects of these mutations are yet to be elucidated, their potential value A 2HG molecule has five nonexchangeable scalar-coupled protons, as diagnostic and prognostic markers in gliomas has been established resonating at 4.02 p.p.m., 2.27 p.p.m., 2.22 p.p.m., 1.98 p.p.m. and 1.83 from their clear association with improved overall survival when p.p.m.10, giving rise to multiplets at approximately three locations at 3 outcomes are compared between IDH-mutated and IDH wild-type T; that is, 4.02 p.p.m. (H2), ~2.25 p.p.m. (H4 and H4ʹ) and ~1.9 p.p.m.

1Advanced Imaging Research Center, University of Texas Southwestern Medical Center, Dallas, Texas, USA. 2Department of Radiology, University of Texas Southwestern Medical Center, Dallas, Texas, USA. 3Department of Pediatrics, University of Texas Southwestern Medical Center, Dallas, Texas, USA. 4McDermott Center for Human Growth and Development, University of Texas Southwestern Medical Center, Dallas, Texas, USA. 5Harold C. Simmons Cancer Center, University of Texas Southwestern Medical Center, Dallas, Texas, USA. 6Department of Pathology, University of Texas Southwestern Medical Center, Dallas, Texas, USA. 7Annette Strauss Center for Neuro-Oncology, University of Texas Southwestern Medical Center, Dallas, Texas, USA. 8Children’s Medical Center, Dallas, Texas, USA. 9Department of Internal Medicine, University of Texas Southwestern Medical Center, Dallas, Texas, USA. 10Department of Neurology and Neurotherapeutics, University of Texas Southwestern Medical Center, Dallas, Texas, USA. 11Department of Physiology, University of Texas Southwestern Medical Center, Dallas, Texas, USA. 12Department of Neurological Surgery, University of Texas Southwestern Medical Center, Dallas, Texas, USA. 13Veterans Affairs North Texas Health Care System, Dallas, Texas, USA. Correspondence should be addressed to C.C. ([email protected]) or E.A.M. ([email protected]).

Received 26 April 2011; accepted 17 August 2011; published online 26 January 2012; corrected online 31 January 2012 (details online); doi:10.1038/nm.2682

624 VOLUME 18 | NUMBER 4 | APRIL 2012 NATURE MEDICINE TECHNICAL REPORTS

Figure 1 Theoretical and a 2.5 2.0 p.p.m. c experimental spectra of 2HG. (a) Quantum-mechanically calculated 115 Calculated Phantom spectra of the 2HG H4 resonances, 105 at 3 T, are plotted against TE and 95 1 Glycine Glycine TE2 of PRESS (subecho times of 85 the first slice– and second slice– 75 selective 180° radiofrequency PRESS (ms)

2 H4 65 (TE , TE ) H4 pulses, respectively). (b) Calculated 1 2 H4ʹ TE H4ʹ 55 (32,65) ms H3 difference-edited multiplets of the H2 H2 H3 H3ʹ H3ʹ 2HG H2 resonance are plotted against 45 subecho times TE1 and TE2 of scalar 35 difference editing. Shown for each 25 TE -TE pair are, top to bottom, Difference 1 2 20 30 40 50 60 70 80 90 100 110 editing E180-on (brown) and E180-off TE (ms) (TE , TE ) (green) subspectra, and the difference 1 1 2 (26,80) ms between the two subspectra (blue). b 4.25 3.75 p.p.m. Here, E180 denotes editing 180° E180-on (A) E180-on 85 E180-off (B) pulses tuned to 1.9 p.p.m. PRESS (A–B) / 2 (A) and edited spectra are all broadened to a singlet line width of 4 Hz. Spectra 80 in a and b are scaled equally for direct E180-off comparison. Relaxation effects were (B) 75 (ms)

not included in the calculations. 2

(c) Calculated and phantom spectra TE of 2HG for PRESS and difference 70 Difference editing. The echo times were 97 ms (A–B) / 2 and 106 ms for PRESS and editing. 65 The concentrations of 2HG and 4.0 3.5 3.0 2.5 2.0 4.0 3.5 3.0 2.5 2.0 25 35 45 55 65 75 85 95 glycine in the phantom were both Chemical shift (p.p.m.) Chemical shift (p.p.m.) TE (ms) 10 mM (pH = 7.0). Spectra are 1 scaled with respect to the glycine singlet at 3.55 p.p.m.

(H3 and H3ʹ) (Supplementary Fig. 1). The 2HG resonances are all sca- occurred in the glutamate multiplets, allowing 2HG to be measured lar coupled, and, consequently, the spectral pattern and signal strength with high selectivity against the background signals of adjacent reso- vary with changing echo time of MRS sequence. A maximum 2HG sig- nances (Supplementary Fig. 2c,d). The optimized echo time is rela- nal may be expected at ~2.25 p.p.m. where the H4 and H4ʹ spins resonate tively long, so signal loss due to transverse relaxation effects may be proximately to each other. Because of its capability of full refocusing, in considerable in vivo. However, given that 2HG does not have a well- the present study we used a PRESS sequence as a major tool for 2HG defined spectral pattern at short echo times (for example, 30 ms), the © 2012 Nature America, Inc. All rights reserved. America, Inc. © 2012 Nature measurement. 2HG signals can be better resolved at the optimized long echo time, We conducted quantum mechanical simulations to search for opti- benefiting from the suppressed complex baseline signals of mac- mal experimental parameters. The simulation indicated that the 2HG romolecules. The signal yield of 2HG in difference editing was low npg H4 resonances give rise to a maximum multiplet at total echo time of (38%) compared to that in PRESS (Fig. 1), but the editing provides 90–100 ms, for which the first subecho time, TE1, is shorter than the a useful tool for proving 2HG elevation because the edited signal at second subecho time, TE2 (Fig. 1a). Given the large spectral distance of 4.02 p.p.m. is uniquely generated via the coupling connections of the H2 resonance from its weak coupling partners (H3 spins), we also 2HG. In vivo, because the difference editing uses spectral difference measured the H2 resonance by means of difference editing. Selective induced by selective 180° rotations tuned at approximately 1.9 p.p.m., 180° rotation of the H3 spins was switched on and off within a PRESS the 4.15-p.p.m. resonance of the glutamate moiety of N-acetylaspartyl- sequence in alternate scans to induce unequal H2 multiplets in sub- glutamate11 is co-edited, but the resonance is relatively distant from spectra. Subtraction between the spectra generated an edited 2HG H2 the 2HG 4.02-p.p.m. resonance and thus does not interfere with 2HG multiplet, canceling other resonances that were not affected by the edit- editing (Supplementary Fig. 3). The lactate resonance at 4.1 p.p.m.12 ing 180° pulses. The computer simulation indicated that a large edited is not co-edited because the coupling partners at 1.31 p.p.m. are not H2 signal can be obtained using a short-echo-time set in which TE1 influenced by the editing 180° pulse. should be the shortest possible (Fig. 1b). We optimized the subecho For spectral fitting, in the present study we used model spectra that times of the PRESS and difference editing sequences to (TE1, TE2) = were calculated including the effects of the volume-localized radiofre- (32, 65) ms and (26, 80) ms, respectively. We tested these optimized MRS quency pulses used for in vivo measurements, allowing spectral fitting sequences in an aqueous solution with 2HG that was synthesized in by signal patterns identical to those obtained by experiment. Calculation house. The spectral pattern and signal intensity of 2HG were consistent of spectra at numerous echo times for MRS sequence optimization was between calculation and experiment (Fig. 1c). efficiently accomplished using the product operator-based transfor- The optimized PRESS provided a 2HG multiplet at approximately mation-matrix algorithm in the quantum-mechanical simulations13–15 2.25 p.p.m. with maximum amplitude among echo times greater than (Supplementary Methods). The spectral pattern of 2HG is pH depen- 40 ms (Supplementary Fig. 2a,b). Moreover, the optimized echo time dent10, with large shifts noted for pH < 6 (Supplementary Fig. 4). We gave rise to narrowing of the multiplet and substantial reduction of performed computer simulations and MRS sequence optimization for 2HG signals at approximately 1.9 p.p.m. Similar signal modulation 2HG detection assuming pH ~7.0 in tumors16–18.

NATURE MEDICINE VOLUME 18 | NUMBER 4 | APRIL 2012 625 TECHNICAL REPORTS

Table 1 Correlation between 2HG detection by MRS PRESS and IDH1 and IDH2 mutational status 2HG by MRS IDH1 and IDH2 mutations by IDH1 R132H by Histological diagnosis (mM (CRLB)) DNA sequencing immunohistochemistry Oligodendroglioma (WHO grade 2) 2.7 (13%) IDH2 (R172K) Negative Oligodendroglioma (WHO grade 2) 3.3 (11%) IDH1 (R132H) Positive Oligodendroglioma (WHO grade 2) 2.6 (14%) IDH1 (R132C) Negative Oligodendroglioma (WHO grade 2) 1.7 (17%) IDH1 (R132H) Positive Oligodendroglioma (WHO grade 2) 3.3 (7%) IDH1 (R132C) Negative Astrocytoma (WHO grade 2) 4.2 (10%) IDH1 (R132H) Positive Astrocytoma (WHO grade 3) 2.1 (16%) IDH1 (R132H) Positive Oligoastrocytoma (WHO grade 3) 3.9 (6%) IDH1 (R132H) Positive Oligodendroglioma (WHO grade 3) 8.9 (3%) IDH1 (R132H) Positive Oligoastrocytoma (WHO grade 3) 3.4 (8%) IDH2 (R172W) Negative Astrocytoma (WHO grade 3) 2.7 (11%) IDH2 (R172G) Negative Astrocytoma (WHO grade 3) 5.3 (6%) IDH1 (R132H) Positive Astrocytoma (WHO grade 3) 2.5 (16%) IDH1 (R132C) Negative Astrocytoma (WHO grade 3) 2.2 (15%) IDH1 (R132C) Negative Astrocytoma (WHO grade 3) ND None Negative Secondary glioblastoma (WHO grade 4) 2.1 (15%) IDH1 (R132H) Positive Glioblastoma (WHO grade 4) ND None Negative Glioblastoma (WHO grade 4) ND None Negative Glioblastoma (WHO grade 4) ND None Negative Glioblastoma (WHO grade 4) ND None Negative Glioblastoma (WHO grade 4) ND None Negative Glioblastoma (WHO grade 4) ND None Negative Glioblastoma (WHO grade 4) ND None Negative Glioblastoma (WHO grade 4) ND None Negative Glioblastoma (WHO grade 4) ND None Negative Glioblastoma (WHO grade 4) ND None Negative Glioblastoma (WHO grade 4) ND None Negative Glioblastoma (WHO grade 4) ND None Negative Glioblastoma (WHO grade 4) ND None Negative Glioblastoma (WHO grade 4) ND None Negative The MRS measures labeled ‘ND’ (not detected) were 2HG estimates ≤0.08 mM with CRLB ≥85%. The MRS estimates of 2HG concentrations were significantly different between mutated and wild-type IDH (P = 6 × 10-8, unpaired t-test). © 2012 Nature America, Inc. All rights reserved. America, Inc. © 2012 Nature

2HG detected in MRS spectra from subjects with gliomas g-Aminobutyric acid (GABA), glutamate and glutamine have reso- npg We included 30 subjects with gliomas in the 2HG MRS analysis (Table 1). nances of 2.1–2.4 p.p.m. Thus, the signals are partially overlapped with The optimized PRESS was applied to the tumor mass. A representative the 2HG 2.25-p.p.m. signal in PRESS spectra and can interfere with normal brain spectrum (Fig. 2a) showed the expected pattern of cho- 2HG estimation depending on their signal strengths. Spectral fitting line, creatine and N-acetylaspartate (NAA) without evidence of 2HG. with the calculated spectra enabled resolution of the metabolites with In contrast, the classic pattern of elevated choline with decreased cre- CRLB < 20% for concentrations above ~2 mM (Fig. 2). We validated atine and NAA was present in all glioma grades (Fig. 2b–f). A signal the PRESS detection of 2HG using two methods. First, we compared attributed to 2HG was discernible at 2.25 p.p.m. in the WHO grade 2 spectral fitting outputs from a basis set with or without a 2HG signal. and 3 tumors (Fig. 2c–f), but not in the glioblastoma (Fig. 2b). For spectra without measurable 2HG signals (Fig. 3a), the residuals were We identified an IDH1 or IDH2 mutation in each of these cases. essentially identical between the two fitting methods. However, spectra We analyzed the single-voxel–localized PRESS data with lin- with a noticeable signal at 2.25 p.p.m., when fitted using a basis set with- ear combination of model (LCModel) software19, using spec- out 2HG, resulted in large residuals at 2.25 p.p.m. (Fig. 3a). For spectra tra of 20 metabolites as basis sets, calculated incorporating the with intermediate 2HG concentrations, the residuals were progressively volume-localized pulses. We estimated the concentration of larger with increasing 2HG estimates. This result shows that the signal at 2HG using the brain water signal from the voxel as reference and 2.25 p.p.m. is primarily attributable to 2HG without substantial interfer- adjusted the relaxation effects on the observed metabolite sig- ence from the neighboring resonances. Second, we used difference edit- nals using published relaxation times of brain metabolites for ing to confirm the PRESS measurements of 2HG in seven subjects. When 3 T20–22. With a 2-min scan on 2 × 2 × 2 cm3 areas of brain tissue, 2HG a signal at 2.25 p.p.m. was discernible in PRESS spectra, we detected an was measurable for concentrations >1.5 mM, with Cramér-Rao lower edited H2 signal at 4.02 p.p.m. (Fig. 3b). When 2HG was not mea- bound (CRLB) < 18% (Table 1). With the use of precisely calculated surable in PRESS spectra, there was no observable edited peak at 4.02 model spectra for spectral fitting, the LCModel fits reproduced the in p.p.m. (Fig. 3b). This co-detection of the PRESS 2.25 p.p.m. peak and vivo spectra closely, resulting in residuals at the noise levels that did the edited 4.02 p.p.m. signal supports the idea that the signals are both not show considerable chemical-shift dependences. attributable to 2HG. The similarity between the 2HG concentrations

626 VOLUME 18 | NUMBER 4 | APRIL 2012 NATURE MEDICINE TECHNICAL REPORTS

a b c the presence of a 2HG peak in Normal brain Grade 4 glioblastoma Grade 2 oligodendroglioma MRS is 100% correlated with the (wild-type IDH1 and IDH2) (IDH1 mutated) presence of a mutation in IDH1 or IDH2 and elevated concen- NAA Cho trations of d-2HG in the tumor. Cr Cho Cr Cho Cr NAA Lip Cr Moreover, the absence of a 2HG NAA Lac Lac peak when MRS is performed In vivo Fit in a tumor mass is 100% corre- Residuals 2HG (0) 2HG (0) 2HG (3.3 ± 0.2) lated with wild-type IDH1 and GABA (1.1 ± 0.2) GABA (0.5 ± 0.2) GABA (0.4 ± 0.2) IDH2 and lack of accumulation Glu (9.0 ± 0.2) Glu (4.6 ± 0.2) Glu (1.4 ± 0.2) Gln (2.4 ± 0.2) Gln (5.3 ± 0.2) Gln (2.9 ± 0.2) of d-2HG in the tumor tissue. 4 3 2 1 p.p.m. 4 3 2 1 p.p.m. 4 3 2 1 p.p.m. Thus, the ability to detect 2HG d e f by MRS in a tumor mass is both Grade 2 oligodendroglioma Grade 3 oligodendroglioma Grade 3 astrocytoma highly sensitive and specific. (IDH1 mutated) (IDH1 mutated) (IDH1 mutated) Cho Development of multivoxel Cho Lip imaging of 2HG We extended the optimized Cr Cho Cr Gly PRESS echo time method to NAA Lac NAA Lac Cr NAA multivoxel imaging of 2HG. The subject with grade 3 oligo- 2HG (9.1 ± 0.3) 2HG (8.9 ± 0.3) 2HG (4.6 ± 0.3) dendroglioma, whose single- GABA (0.2 ± 0.2) GABA (0) GABA (0.1 ± 0.1) Glu (0.5 ± 0.2) Glu (2.1 ± 0.2) Glu (1.4 ± 0.2) voxel MRS data are shown in Gln (2.0 ± 0.2) Gln (2.3 ± 0.2) Gln (2.6 ± 0.2) Figure 2e, was scanned with 4 3 2 1 p.p.m. 4 3 2 1 p.p.m. 4 3 2 1 p.p.m. 1 × 1 cm2 resolution on a slice 1.5 cm thick that included the Figure 2 In vivo 1H spectra and analysis. (a–f) In vivo single-voxel–localized PRESS spectra from normal brain (a) and tumors (b–f), at 3 T, are shown together with spectral fits (LCModel) and the components of 2HG, GABA, glutamate tumor mass (Fig. 4a). The and glutamine, as well as voxel positioning (2 × 2 × 2 cm3). Spectra are scaled with respect to the water signal patterns of the single voxel– from the voxel. Vertical lines are drawn at 2.25 p.p.m. to indicate the H4 multiplet of 2HG. Shown in brackets is acquired spectra were repro- the estimated metabolite concentration (mM) ± s.d. Cho, choline; Cr, creatine; Glu, glutamate; Gln, glutamine; Gly, duced in spectra obtained with glycine; Lac, lactate; Lip, lipids. Scale bars, 1 cm. the multi-voxel MRS method. The 2HG signal at 2.25 p.p.m. estimated by PRESS and editing provides evidence that the PRESS mea- was clearly discernible in spectra from the tumor regions (Fig. 4b). surement of 2HG is valid, as the edited 2HG signal at 4.02 p.p.m. was Spectra from contralateral normal brain showed no 2HG signals at 2.25 generated without substantial interference from the scalar coupling con- p.p.m. (Fig. 4c). A map of 2HG concentrations (Fig. 4d) showed that nection between the 4.02 p.p.m. and ~1.9 p.p.m. resonances, which is a 2HG was concentrated at the center of the T2w-FLAIR hyperintensity © 2012 Nature America, Inc. All rights reserved. America, Inc. © 2012 Nature unique feature of 2HG among known brain metabolites12. region. We estimated the 2HG concentrations using the normal brain NAA concentration of 12 mM12 as reference, giving a 2HG concentra- Validation of MRS measures of 2HG by tissue analysis tion approximately 9 mM at the center of the tumor mass, in agreement npg We analyzed each tumor for IDH gene status by immunohistochemistry with the 2HG estimates by single-voxel MRS. The spatial distribution for the IDH1 R132H mutation and by gene sequencing of IDH1 and pattern of choline was similar to that of 2HG in the region of T2w- IDH2 (Table 1). Of the 30 subjects studied, 15 had measurable 2HG by FLAIR hyperintensity but, as expected, was found throughout the brain, MRS, and in each case we confirmed an IDH1 (12 of 15 subjects) or an whereas 2HG showed rapid drop off in normal brain. The NAA con- IDH2 (3 of 15 subjects) mutation. The remaining 15 subjects did not centrations were low in the tumor mass, and the choline/NAA ratio in have detectable 2HG by MRS (<0.08 mM, CRLB > 85%), and analysis tumors was high relative to that of normal brain tissue. Because of their of IDH1 and IDH2 revealed no mutations. The MRS estimates of 2HG ability to detect 2HG in small volumes, the metabolic measures by the concentrations were significantly different between subjects with IDH multivoxel MRS method may contain reduced partial-volume effects mutations and wild-type IDH genes (unpaired t-test; P = 6 × 10-8). compared to the single-voxel MRS. As 2HG is unique to tumor cells, We validated these results further by measuring d-2HG and l-2HG the specificity of detection is a key advance in clinical MRS for IDH- concentrations in tumor samples by liquid chromatography–tandem mutated gliomas. mass spectrometry (LC-MS/MS) for the 13 subjects for whom sufficient frozen material from the initial tumor resection was available DISCUSSION (Supplementary Table 1). Samples from the brain tissue adjacent to We have detected 2HG noninvasively by optimized MRS methods tumors were available for analysis from three subjects; thus, this tissue in subjects with gliomas and have shown concordance of 2HG levels served as relative normal controls (Supplementary Fig. 5). l-2HG and with mutations in IDH1 and IDH2, as well as accumulation of 2HG d-2HG were clearly differentiated in the spectra. Five wild-type IDH1 in tumor tissue. To our knowledge, this is currently the only direct and IDH2 glioblastoma tumors had similar concentrations of l-2HG metabolic consequence of a genetic mutation in a cancer cell that can and d-2HG. In all tumor samples, l-2HG was <1.0 nmol per mg protein. be identified through noninvasive imaging. The signal overlaps of 2HG In marked contrast, d-2HG levels in IDH1- and IDH2-mutated tumors with GABA, glutamate and glutamine, which occur in short-echo-time were 20-fold to 2,000-fold higher than those in wild-type IDH glioblas- standard data acquisitions, were overcome with multiplet narrowing tomas (Supplementary Fig. 6). Taken together, these data show that by MRS sequence optimization and spectral fitting using precisely

NATURE MEDICINE VOLUME 18 | NUMBER 4 | APRIL 2012 627 TECHNICAL REPORTS

a b external reference signal such as 24 Fitting with 2HG Fitting without 2HG Residuals PRESS Difference editing from a phantom , referencing (threefold magnified) (fourfold magnified) with respect to brain water sig- NAA NAA Cho Cr Cr With 2HG Without 2HG nals may be a realistic means of Cho Cho 2HG 2HG 2HG NAA estimating metabolite concentrations in tumors25. Cr Two studies have previously reported in vivo detection of 2HG 2HG (9.1 mM) (8.7 mM) 2HG in the brains of people with 2-hydroxyglutaric acid- uria26,27. Large singlet-like

2HG 2HG signals at 2.5–2.6 p.p.m. were (5.7 mM) (5.4 mM) assigned to 2HG, although this chemical shift assignment is not consistent with in vitro high- 2HG 2HG resolution magnetic resonance spectra of 2HG at neutral pH10. A 2HG signal at approximately 2.5 p.p.m., which is actually a 2HG 2HG multiplet, can occur only at low

432143212.5 2.0 2.5 2.0 43214321 pH (~2.5), as reported previ- p.p.m. p.p.m. p.p.m. p.p.m. p.p.m. p.p.m. ously10 and confirmed in this work (Supplementary Fig. 4). Figure 3 Validation of 2HG PRESS measurements. (a) LCModel fitting results (fits and residuals) of PRESS spectra obtained with basis set with or without 2HG. Data are displayed in order of increasing 2HG estimates, above to below. The pH measured noninvasively (b) PRESS and difference-edited spectra from four subjects are shown in pairs, together with LCModel fits and in a wide range of tumors ranges 2HG signal components. Vertical lines are drawn at 2.25 p.p.m. and 4.02 p.p.m. in the PRESS and edited spectra, between 6.8 and 7.2. Even using respectively. microelectrode studies, the lowest pH measured was ~6.0. An calculated basis spectra of metabolites. The methods presented here intracellular pH of ~7.0 has been reported in cancer cells16–18. The estimated metabolite concentrations using the brain water signal as chemical shifts and coupling constants, used for MRS data analysis reference in tumors, assuming an equal contribution of gray and white in the present study, were measured at pH 7.0 (ref. 10). As the pro- matter. The metabolite estimation may be valid only when the water ton NMR spectrum of 2HG is close to constant between pH 6.5–7.5 concentration is similar among regions of the brain and between nor- (Supplementary Fig. 4), the efficiency of detecting 2HG in gliomas mal brain and tumor tissues. The water concentration in tumors could should not be sensitive to tumor pH. be increased as a result of the effects of high cellularity or brain edema, A PRESS sequence used for 2HG measurement in the present study which would result in an underestimate of metabolite concentrations is commonly available in clinical magnetic resonance systems. The © 2012 Nature America, Inc. All rights reserved. America, Inc. © 2012 Nature in the present study. Given a maximum possible water concentration field strength used here, 3 T, is becoming more commonly used in the of 55.6 M (bulk water), the true metabolite concentrations could be up academic and clinical magnetic resonance community, and the data- to 30% higher than our estimates, which were obtained using a water acquisition method could be implemented on standard hardware already npg concentration of 42.3 M, calculated from the published values for the in place in many magnetic resonance imaging centers. Without the need water concentrations in gray and white matter23. Although uncertain- for more specialized instrumentation or the production of expensive ties in metabolite estimates can theoretically be minimized by using an exogenous probes, the detection of 2HG by MRS is a method that could

Cho a b d [2HG] map mM Figure 4 Spectroscopic imaging 9 of 2HG. (a) Multivoxel imaging 6 spectra from a subject with a WHO Cr grade 3 oligodendroglioma are Cr Gly NAA 3 2HG overlaid on the T2w-FLAIR image. The grid size is 1 × 1 cm, with slice In vivo 0 Fit thickness 1.5 cm. The spectra are [CHO] map displayed between 4.1 p.p.m. and Residuals 4 1.8 p.p.m. (left to right). (b,c) Two 4.0 3.5 3.0 2.5 2.0 p.p.m. representative spectra (one from 3 the tumor and another from the c 2 contralateral normal brain) are shown NAA 1 together with LCModel fits and 0 Cho residuals. Mins, myoinositol. (d) The Cr [NAA] map estimated concentrations of 2HG, Cr 12 Mins choline and NAA in individual voxels Glu were color coded for comparison. The In vivo 8 Fit NAA level in gray matter in normal Residuals 4 brain was assumed to be 12 mM. Scale bars, 1 cm. 4.0 3.5 3.0 2.5 2.0 p.p.m. 0

628 VOLUME 18 | NUMBER 4 | APRIL 2012 NATURE MEDICINE TECHNICAL REPORTS

be quickly adopted for clinical use. As the presence of an IDH1 or IDH2 COMPETING FINANCIAL INTERESTS mutation makes the diagnosis of glioma when evaluating a brain mass, The authors declare no competing financial interests. the ability to detect 2HG by MRS will be a valuable diagnostic tool. Published online at http://www.nature.com/naturemedicine/. Although not obviating the need for a surgical procedure to determine Reprints and permissions information is available online at http://www.nature.com/ tumor grade, the presence of 2HG on MRS would differentiate a tumor reprints/index.html. from a non-neoplastic process such as demyelinating disease. Moreover, 1. Balss, J. et al. Analysis of the IDH1 codon 132 mutation in brain tumors. Acta the association of IDH1 and IDH2 mutations with improved survival Neuropathol. 116, 597–602 (2008). among gliomas makes the detection of 2HG an important prognostic 2. Yan, H. et al. IDH1 and IDH2 mutations in gliomas. N. Engl. J. Med. 360, 765–773 marker as well. (2009). 3. Dang, L. et al. Cancer-associated IDH1 mutations produce 2-hydroxyglutarate. Nature The additional clinical value of this biomarker may be in its dynamic 462, 739–744 (2009). measurement over the time course of treatment and follow up. If 2HG 4. Figueroa, M.E. et al. Leukemic IDH1 and IDH2 mutations result in a hypermethylation concentrations reflect changes in tumor cellularity, then proliferation phenotype, disrupt TET2 function, and impair hematopoietic differentiation. Cancer Cell 18, 553–567 (2010). would lead to increased 2HG concentrations, and tumor cells killed by 5. Christensen, B.C. et al. DNA methylation, isocitrate dehydrogenase mutation, and survival radiation or chemotherapy would lead to decreased 2HG concentra- in glioma. J. Natl. Cancer Inst. 103, 143–153 (2011). 6. Parsons, D.W. et al. An integrated genomic analysis of human glioblastoma multiforme. tions. Stable disease would be expected to have stable 2HG concen- Science 321, 1807–1812 (2008). trations. Although these correlations are still speculative, potential 7. von Deimling, A., Korshunov, A. & Hartmann, C. The next generation of glioma biomarkers: clinical applications include the follow up of patients with WHO grade 2 MGMT methylation, BRAF fusions and IDH1 mutations. Brain Pathol. 21, 74–87 (2011). 8. Bottomley, P.A. Selective volume method for performing localized NMR spectroscopy. gliomas who typically have a long period of minimal progression that US Patent 4,480,228 (1984). is followed rapidly by aggressive growth and transformation to high 9. Mescher, M., Merkle, H., Kirsch, J., Garwood, M. & Gruetter, R. Simultaneous in vivo grade. Similarly, the availability of an imaging biomarker to follow for spectral editing and water suppression. NMR Biomed. 11, 266–272 (1998). 10. Bal, D. & Gryff-Keller, A. 1H and 13C NMR study of 2-hydroxyglutaric acid and lactone. the detection of recurrent disease or response to chemotherapy would Magn. Reson. Chem. 40, 533–536 (2002). be a major advance in the clinical management of patients with gliomas. 11. Krawczyk, H. & Gradowska, W. Characterisation of the 1H and 13C NMR spectra of N-acetylaspartylglutamate and its detection in urine from patients with Canavan disease. Additional studies of the dynamic properties of 2HG measurement will J. Pharm. Biomed. Anal. 31, 455–463 (2003). address these potential uses. 12. Govindaraju, V., Young, K. & Maudsley, A.A. Proton NMR chemical shifts and coupling constants for brain metabolites. NMR Biomed. 13, 129–153 (2000). METHODS 13. Ernst, R.R., Bodenhausen, G. & Wokaun, A. Principles of nuclear magnetic resonance in one and two dimensions Ch. 2 (Clarendon Press, Oxford, UK, 1987). Methods and any associated references are available in the online version 14. Thompson, R.B. & Allen, P.S. Sources of variability in the response of coupled spins to of the paper at http://www.nature.com/nm/. the PRESS sequence and their potential impact on metabolite quantification. Magn. Reson. Med. 41, 1162–1169 (1999). 15. Choi, C. et al. Improvement of resolution for brain coupled metabolites by optimized 1H Note: Supplementary information is available on the Nature Medicine website. MRS at 7T. NMR Biomed. 23, 1044–1052 (2010). 16. Gillies, R.J., Raghunand, N., Garcia-Martin, M.L. & Gatenby, R.A. pH imaging. A review ACKNOWLEDGMENTS of pH measurement methods and applications in cancers. IEEE Eng. Med. Biol. Mag. This work was supported by US National Institutes of Health grants 23, 57–64 (2004). RC1NS0760675, R21CA159128, and RR02584 and by the Cancer Prevention 17. Griffiths, J.R. Are cancer cells acidic? Br. J. Cancer 64, 425–427 (1991). Research Institute of Texas grant RP101243-P04. We thank C. Sheppard for expert 18. McLean, L.A., Roscoe, J., Jorgensen, N.K., Gorin, F.A. & Cala, P.M. Malignant gliomas management of the patient database and for coordinating research scans and tissue display altered pH regulation by NHE1 compared with nontransformed astrocytes. Am. samples; S. McNeil for expert human subject care during scanning; C. Foong for J. Physiol. Cell Physiol. 278, C676–C688 (2000). 19. Provencher, S.W. Estimation of metabolite concentrations from localized in vivo proton expert assistance with pathological analysis of tumor; and R.L. Boriack for expert

© 2012 Nature America, Inc. All rights reserved. America, Inc. © 2012 Nature NMR spectra. Magn. Reson. Med. 30, 672–679 (1993). assistance with 2HG measurements by mass spectrometry. 20. Mlynárik, V., Gruber, S. & Moser, E. Proton T (1) and T (2) relaxation times of human brain metabolites at 3 tesla. NMR Biomed. 14, 325–331 (2001). AUTHOR CONTRIBUTIONS 21. Träber, F., Block, W., Lamerichs, R., Gieseke, J. & Schild, H.H. 1H metabolite relaxation C.C. developed the MRS methodology for 2HG detection, designed and performed times at 3.0 tesla: measurements of T1 and T2 values in normal brain and determination npg the magnetic resonance experiments and data analysis, supervised the MRS study, of regional differences in transverse relaxation. J. Magn. Reson. Imaging 19, 537–545 prepared figures and wrote the manuscript. E.A.M. led all aspects of the human (2004). study, contributed to data analysis, preparation of the figures and writing of the 22. Ganji, S.K. et al. T2 measurement of J-coupled metabolites in the human brain at 3T. published online, doi:10.1002/nbm.1767 (15 August 2011). manuscript. S.K.G. carried out magnetic resonance data acquisition and contributed NMR Biomed. 23. Norton, W.T., Poduslo, S.E. & Suzuki, K. Subacute sclerosing leukoencephalitis. II. to data analysis. D.R. performed mass spectrometry analysis on resected tumors Chemical studies including abnormal myelin and an abnormal ganglioside pattern. and contributed to manuscript preparation. Z.K. synthesized 2HG and prepared J. Neuropathol. Exp. Neurol. 25, 582–597 (1966). a figure. R.J.D. and C.R.M. contributed to the conceptual approach, review of the 24. Keevil, S.F. et al. Absolute metabolite quantification by in vivo NMR spectroscopy: II. data and manuscript preparation. K.J.H. and J.M.R. contributed to tumor sample A multicentre trial of protocols for in vivo localised proton studies of human brain. Magn. collection and validation, neuropathological evaluation and diagnosis, evaluation Reson. Imaging 16, 1093–1106 (1998). of immunohistochemical stains and manuscript preparation. X.-L.Y. and T.M. 25. Tong, Z., Yamaki, T., Harada, K. & Houkin, K. In vivo quantification of the metabolites performed the tissue evaluation of IDH mutations. I.M.-V. and J.M.P. contributed in normal brain and brain tumors by proton MR spectroscopy using water as an internal standard. 22, 1017–1024 (2004). to conceptual analysis. B.E.M. recruited subjects and contributed to clinical data Magn. Reson. Imaging 26. Sener, R.N. L-2 hydroxyglutaric aciduria: proton magnetic resonance spectroscopy and analysis and manuscript preparation. C.J.M. recruited subjects and contributed diffusion magnetic resonance imaging findings. J. Comput. Assist. Tomogr. 27, 38–43 to manuscript preparation. R.M.B. contributed to the conceptual approach, led (2003). the tumor analysis workup, and contributed to data analysis and manuscript 27. Goffette, S.M. et al. L-2-Hydroxyglutaric aciduria: clinical, genetic, and brain MRI char- preparation. acteristics in two adult sisters. Eur. J. Neurol. 13, 499–504 (2006).

NATURE MEDICINE VOLUME 18 | NUMBER 4 | APRIL 2012 629 ONLINE METHODS Magnetic resonance spectroscopy data analysis. Data were analyzed as Subject inclusion. We selected subjects from two University of Texas described28. Following a 1-Hz apodization, spectra were fitted with LCModel Southwestern Medical Center (UTSW) Institutional Review Board–approved software19, using calculated spectra of 20 metabolites as basis functions. The basis brain tumor clinical protocols that have magnetic resonance imaging and set included spectra of 2HG, NAA, GABA, glutamate, glycine, creatine, myoino- include MRS as part of the study procedures. We obtained informed consent sitol, glutamine, lactate, alanine, acetate, aspartate, ethanolamine, glutathione, for each subject. Scans from 53 subjects were screened and were included for phosphorylethanolamine, scyllo-inositol, taurine, N-acetylaspartylglutamate, analysis of 2HG if (i) there was a visible tumor mass by standard magnetic glucose and choline. The metabolite concentrations were estimated with respect resonance sequences (gadolinium enhancement or T2w-FLAIR signal abnor- to the short-echo-time water signal. Assuming an equal composition of gray and mality), (ii) MRS had been performed in at least 1 voxel in the tumor that was white matter in tumors, we used a water concentration value of 42.3 M, calcu- of acceptable spectral quality (singlet line width <6 Hz), and (iii) there was lated from the literature values23 for the water concentrations in gray and white adequate tissue available for IDH gene sequencing. Scans from 30 subjects met matter. Relaxation effects on metabolite signals were corrected using published these criteria, and 29 subjects had been imaged before initial surgery or after metabolite T2 and T1: T2 = 150 ms, 230 ms and 280 ms for Cr, Cho and NAA, a limited surgical procedure (biopsy or subtotal resection) and had not been and 180 ms for other metabolites, respectively; T1 = 1.2 ms for 2HG, glutamate, treated with chemotherapy or radiation. One subject with secondary GBM glutamine and myo-inositol, and 1.5 ms for other metabolites20–22. was imaged at the time of recurrence, 3 years after treatment with radiation. A search was done in each case for the availability of frozen tissue. In 13 of Immunohistochemistry for detection of the R132H mutation in IDH1. Paraffin 30 cases, a frozen tumor sample was identified, with three cases having both sections were cut at 4-µm thickness and prepared according to standard clinical tumor and a sample of adjacent, non–tumor-bearing brain available for analysis. methods. The primary antibody for IDH1 R132H (Dianova) was diluted 1 in 20.

Magnetic resonance spectroscopy data acquisition. Experiments were car- IDH1 and IDH2 DNA sequencing. DNA was prepared from frozen tissue by ried out on a 3-T whole-body scanner (Philips Medical Systems). A body coil standard methods or from formalin-fixed paraffin embedded samples accord- was used for radiofrequency transmission and an eight-channel head coil was ing to a published method29. Primers for IDH1 and IDH2 sequencing were used used for reception. Data were acquired according to our published methods28. according to published methods1,3. PRESS8 and scalar difference editing9 were used for measuring 2HG in brain tumors. For editing, two 20-ms Gaussian 180° pulses, tuned to 1.9 p.p.m., Measurement of 2HG enantiomers by mass spectrometry. d-2HG and l-2HG were switched on and off in alternate scans to generate an edited H2 signal at were extracted from tumor tissue and normal brain, and LC-MS/MS analyses 30,31 4.02 p.p.m. in difference spectra. The echo times of PRESS and editing were were performed as described . 97 ms and 106 ms, respectively. The quantum-mechanical simulations were Statistical analyses. Cramér-Rao lower bounds of metabolite estimates, which carried out by means of the product operator–based transformation matrix represent the lower bounds of the precision, were obtained with the built-in algorithm (Supplementary Methods). For in vivo magnetic resonance scans, algorithm of the LCModel software19. following the survey imaging, T2w-FLAIR images were acquired to identify 3 tumor regions. For single-voxel–localized data acquisition, a 2 × 2 × 2 cm voxel Additional methods. Detailed methodology for 2HG synthesis and transfor- was positioned in the tumor mass. PRESS acquisition parameters included a mation matrix–incorporated density-matrix simulations is described in the sweep width of 2500 Hz, 2,048 sampling points, a repetition time of 2 s, and Supplementary Methods. 64 averages (scan time 2.1 min). Editing data were acquired with 384 averages (scan time 13 min). An unsuppressed water signal was acquired with an echo 28. Choi, C. et al. Measurement of glycine in the human brain in vivo by 1H-MRS at 3 T: time of 14 ms and a repetition time of 20 s for use as a reference in metabolite application in brain tumors. Magn. Reson. Med. 66, 609–618 (2011). 29. Maher, E.A. et al. Marked genomic differences characterize primary and secondary quantification. Spectroscopic imaging data were acquired, using the optimized glioblastoma subtypes and identify two distinct molecular and clinical secondary glio- PRESS echo time, from a slice 1.5 cm thick with a resolution of 1 × 1 cm2. blastoma entities. Cancer Res. 66, 11502–11513 (2006). 30. Rakheja, D., Mitui, M., Boriack, R.L. & DeBerardinis, R.J. Isocitrate dehydrogenase 1/2

© 2012 Nature America, Inc. All rights reserved. America, Inc. © 2012 Nature We carried out undersampling of k-space data by 20%, the scan time being mutational analyses and 2-hydroxyglutarate measurements in Wilms tumors. Pediatr. approximately10 min (two averages, repetition time = 1.3 s). Data were zero Blood Cancer 56, 379–383 (2011). filled for the unacquired k-space points and filtered with a cosine function 31. Rakheja, D. et al. Papillary thyroid carcinoma shows elevated levels of 2-hydroxyglutarate. before Fourier transformation. Tumour Biol. 32, 325–333 (2011). npg

NATURE MEDICINE doi:10.1038/nm.2682 correction notice

Nat. Med.; doi:10.1038/nm.2682 2-hydroxyglutarate detection by magnetic resonance spectroscopy in subjects with IDH-mutated gliomas

Changho Choi, Sandeep K Ganji, Ralph J DeBerardinis, Kimmo J Hatanpaa, Dinesh Rakheja, Zoltan Kovacs, Xiao-Li Yang, Tomoyuki Mashimo, Jack M Raisanen, Isaac Marin-Valencia, Juan M Pascual, Christopher J Madden, Bruce E Mickey, Craig M Malloy, Robert M Bachoo & Elizabeth A Maher

In the version of this supplementary file originally posted online, the equations [4], [5], [6] and [8] were incorrect. The error has been corrected in this file as of 31 January 2012. 1

SUPPLEMENTARY FIGURES

Supplementary Figure 1. NMR characteristics of 2HG. (a) Molecular structure of 2HG. The

non-exchangeable protons of C2, C3 and C4 are detectable in 1H MRS. (b) Scalar coupling

connections between the non-exchangeable protons. (c) A 1H MR spectrum of 2HG calculated

for single-pulse acquisition at 3T is shown together with the spectral locations of the H2, H3 and

Nature Medicine doi:10.1038/nm.2682 2

H4 spin resonances. The calculated time-domain signal was multiplied by a 1-Hz exponential

function before Fourier transformation.

Nature Medicine doi:10.1038/nm.2682 3

Supplementary Figure 2. (a,b) Comparison of PRESS 2HG spectra calculated with full-

Hamiltonian model (shaped slice-selective radio-frequency pulses used for in vivo) and with

simplified-Hamiltonian model (1-μs non-selective radio-frequency pulses) for various PRESS

Nature Medicine doi:10.1038/nm.2682 4

subecho time pairs (TE1, TE2). Spectra, calculated for [2HG]/[Gly] = 1, are scaled with respect

to the Gly singlet at 3.55 p.p.m. in each spectrum. The 2HG spectral pattern and signal

intensities differ between the models, largely due to the finite bandwidth of the slice-selective

180° RF pulses and the coherence proliferation during the radio-frequency pulses. For short

echo times (echo time = TE1+TE2 < 70 ms), the H4 and H3 signals are similar, and the H2

multiplet is different between the models. As echo time increases, the discrepancy between the

H3 multiplets from the two models becomes substantial. The difference in the H4 multiplet is

noticeable at long echo times. This result indicates that the full Hamiltonian model should be

used for optimizing the MRS sequences and creating basis spectra for spectral fitting. (c,d)

Similarly, calculated spectra of glutamate (c) and N-acetylaspartylglutamate (NAAG) (d) show

discrepancy between the simplified- and full-Hamiltonian models. NAAG exhibits substantial

differences between the models due to the relatively large spectral distances between the

coupling partners (i.e., ~2.6 and 4.6 p.p.m. of aspartyl moiety, and 1.9 and 4.15 p.p.m. of

glutamate moiety). Spectra in a, c, and d were calculated for an identical concentration, ignoring

T1 and T2 relaxation effects. Calculated and phantom spectra were all broadened to singlet

linewidth of 4 Hz.

Nature Medicine doi:10.1038/nm.2682 5

Supplementary Figure 3. (a) The refocusing profile of the 20-ms Gaussian 180° radio-

frequency pulse (truncated at 12%; bandwidth = 56 Hz) used for difference editing is shown

together with the 1H resonances of 2HG and metabolites that have resonances in the proximity of

Nature Medicine doi:10.1038/nm.2682 6

the 2HG H2 resonance at 4.02 p.p.m., and those of metabolites which are co-edited. Green

(dashed) lines indicate scalar coupling connections of metabolite resonances. The editing 180°

pulse (E180) is tuned to 1.9 p.p.m. to selectively rotate the 2HG H3 spins through 180° in a

subscan (E180-on). An edited 2HG H2-spin signal at 4.02 p.p.m. is obtained via subtraction

between this subspectrum and an E180-off subspectrum. The 4.15 p.p.m. resonance of the

glutamate moiety of NAAG, that has coupling partners at ~1.9 p.p.m., is coedited. However, the

coedited NAAG signal may not interfere with 2HG detection substantially since the coedited

resonance is fairly separated (by 0.13 p.p.m.) from the 2HG H2 resonance (4.02 p.p.m.). In

practice, a noticeable coedited signal at 4.15 p.p.m. was not detected in-vivo due to the relatively

low concentrations of NAAG in brain. The 3.92 p.p.m. singlet of creatine is identical in

subspectra, so canceled via subtraction. Lactate has a resonance at 4.1 p.p.m. which is coupled

to the 1.31 p.p.m. resonance. This resonance is not affected by the E180, thus the lactate 4.1

p.p.m. resonance is not coedited. There are several additional coupled resonances in the

proximity of the 2HG H2 resonance, but these resonances are canceled via subtraction since their

coupling partners are not influenced by the editing 180° pulse. Glutamate, glutamine, GABA

and NAA have resonances at ~1.9 p.p.m. and the resulting non-zero signals at 1.8 - 2.3 p.p.m.

overlap with the 2HG H4 and H3 multiplets in the difference spectra. However, since the edited

2HG H2 signal uniquely appears at 4.02 p.p.m., the signal overlaps at 1.8 - 2.3 p.p.m. do not

interfere with 2HG estimation considerably. (b) Calculated edited spectra of 2HG and NAA for

equal concentrations (left) are illustrated together with sum spectra at 2HG-to-NAA

concentration ratios of 0.25, 0.5, 1, 2, and 4 (right).

Nature Medicine doi:10.1038/nm.2682 7

1 Supplementary Figure 4. The pH dependence of the H NMR spectrum of 2HG in D2O at 298

K and 9.4 T. Each spectrum was recorded after adjusting the pH of sodium 2-HG with DCl to

the desired pH (7.5 - 2.0). The free induction decay was acquired following a standard single

pulse excitation. Each spectrum was recorded with an internal standard of tert-butanol (1.24

p.p.m.) whose resonance does not depend on pH. A vertical dotted line is drawn at 2.25 p.p.m..

Nature Medicine doi:10.1038/nm.2682 8

Supplementary Figure 5. Measurement of 2HG in tissue samples. (a) Tracings from liquid

chromatography/tandem mass spectroscopy of 2HG in 3 patients (red peaks). Upper panel: blue

peaks represent labeled internal standards for L-2HG and D-2HG. GBM with wild type IDH

show similar levels of L- and D-2HG. Grade 2 oligodendroglioma (middle panel) and secondary

GBM (lower panel), samples diluted 10 fold, demonstrate markedly elevated levels of D-2HG.

(b) D-2HG levels in 3 tissue samples of brain adjacent to tumor and 13 tumor samples. Red

bars: patients who had measurable 2HG by MRS. Blue bars: patients without detectable 2HG

by MRS.

Nature Medicine doi:10.1038/nm.2682 9

Supplementary Figure 6. Linear regression of 1H MRS estimates of 2HG concentration vs.

mass spectrometry measures of 2HG. The coefficient of determination (R2) was 0.63. The MRS

and mass spectrometry data obtained at different time points (specified with asterisks in

Supplementary Table 1) are not included.

Nature Medicine doi:10.1038/nm.2682 10

SUPPLEMENTARY TABLE

2HG (MRS) D-2HG L-2HG Histological Diagnosis IDH1 IDH2 mM (CRLB) nmol/mg prot nmol/mg prot Oligodendroglioma (WHO Grade 2) 2.7 (13%) 48.68 0.12 WT mutant Astrocytoma (WHO Grade 3) 2.1 (16%) 20.61 0.47 mutant WT Astrocytoma (WHO Grade 3) 2.7 (11%) 113.39 0.17 WT mutant Oligoastrocytoma (WHO Grade 3) 3.9 (6%) 28.41 0.21 mutant WT Oligodendroglioma (WHO Grade 3) 8.9 (3%) 401.79 0.79 mutant WT Oligoastrocytoma (WHO Grade 3) 3.4 (8%)* 1673.50 0.05 WT mutant Sec Glioblastoma (WHO Grade 4) 2.1 (15%)* 2118.10 0.23 mutant WT Glioblastoma (WHO Grade 4) Not Detected 0.14 0.07 WT WT Glioblastoma (WHO Grade 4) Not Detected 0.20 0.17 WT WT Glioblastoma (WHO Grade 4) Not Detected 0.38 0.36 WT WT Glioblastoma (WHO Grade 4) Not Detected 0.44 0.29 WT WT Glioblastoma (WHO Grade 4) Not Detected 0.61 0.19 WT WT Glioblastoma (WHO Grade 4) Not Detected 0.84 0.77 WT WT

Supplementary Table 1: Correlation between 2HG concentration measured by MRS (PRESS) and tissue levels

of D- and L-2HG by liquid chromatography-tandem mass spectrometry. The symbol * refers to patients whose

first scans to evaluate 2HG were done at the time of tumor recurrence, >2 years after the time of the initial tumor

resection. The remainder of the patients had the initial scan to evaluate 2HG at the time of the initial surgical

resection. Abbreviation: WT – wild type.

Nature Medicine doi:10.1038/nm.2682 11

Supplementary Methods

Synthesis of disodium salt of 2-hydroxyglutarate:

Disodium salt of 2-hydroxyglutarate was in-house synthesized and used for making a 2HG

phantom (pH = 7.0) for validating the MRS methods. (S)-(+)-5-Oxo-2-tetrahydrofurancarboxylic

acid (500 mg, Aldrich) was dissolved in water (2 mL) and two equivalents of sodium hydroxide

solution (2M, 3.84 mL) were added. The mixture was stirred at room temperature overnight and

then lyophilized to produce a quantitative yield of the disodium salt of 2-hydroxyglutarate.

2NaOH NaOOC COONa O COOH O H2O OH

Quantum-mechanical simulations by the product-operator-based transformation matrix

algorithm:

Quantum-mechanical simulations were carried out to optimize the echo times of the PRESS and

difference editing for 2HG measurement and to create the model spectra of metabolites for

LCModel spectral fitting. The time evolution of the density operator was calculated numerically

incorporating the shaped 90° and 180° radio-frequency and gradient pulses. The product-

operator-based transformation matrix method 1,2 was employed to calculate the spectra at

numerous echo times. An echo time that gives maximum 2HG signal was then selected for each

sequence and applied in the patient study. The density matrix simulations were programmed

with Matlab (The MathWorks Inc.). Published chemical shift and coupling constants were used

in the simulation 3-5.

The time evolution of the density operator ρ is described by the Liouville-von Neumann

equation 1

∂ρ/∂t = –i [H, ρ], [1]

Nature Medicine doi:10.1038/nm.2682 12

which has a solution

ρ = exp(–iHt) ρ0 exp(iHt), [2]

for a time-independent Hamiltonian H. The Hamiltonian H may include the chemical shift (CS)

and scalar coupling (J) terms and the radio-frequency (RF) and gradient (G) pulse terms,

H = HCS + HJ + HRF + HG, [3]

in the rotating frame.

For a spin system with N coupled protons (spin = 1/2), 4N product operators (PO) can

constitute a complete set in Liouville space 6. The density matrix ρ may be written as a linear

sum of the PO terms α,

4 N = ρ  c αii , [4] i =1

where ρ and α are 2N×2N square matrices with complex entries, and the coefficient c is real. The

density operator can be expressed as a column vector σ which is composed of the coefficients c,

c   1  c  σ = 2 . [5]       c N  4 

The density operator evolution during an radio-frequency pulse can be put in terms of a single

matrix multiplication 1,

′  TTT N c   2,11,1 4,1  c   1     1  ′ c   2,21,2  TTT N  c  ′ =σ 2 =  4,2  2 T σ= , [6]             ′      c N N N  TTT NN c N  4   4,42,41,4   4 

where the transformation matrix T is a 4N×4N square matrix with real entries. The T-matrix was

constructed for each spatially/spectrally-selective shaped radio-frequency pulse and used for

Nature Medicine doi:10.1038/nm.2682 13

calculating the time evolution of the density operator during the MRS sequences for each

metabolite.

For a time-dependent radio-frequency pulse whose envelope consists of n numbers as a

function of time, HRF and consequently H may be constant during each time period Δt. The

density operator following the RF pulse was calculated using a (total) time evolution operator

Vtotal,

−1 ρ = Vtotal ρ0 Vtotal, [7]

where

= total 21  VVVVV ni . [8]

The time evolution operator for the i-th period of the radio-frequency pulse, Vi, was obtained

using

diag Δ−= −1 i ( i ) UtHiexpUV , [9]

diag –1 where Hi (= U HiU) and U are the diagonalized matrix and the unitary matrix of the

Hamiltonian of the i-th period, Hi, respectively.

When a gradient pulse was applied during an radio-frequency pulse for slice selection,

since HG and consequently H are position dependent, the space was divided into small segments

and the calculation of Eq. [7] was undertaken for individual segments, assuming uniform HG

within each segment. The simulation for slice selection was conducted on a 20 mm thick slice at

the center of a 30 mm sample. The sample space was divided into 150 segments, the spatial

resolution being 1% with respect to the slice thickness (i.e., 0.01 = 30/150/20). The 90° and

180° slice-selective radio-frequency pulse envelopes consisted of 500 and 200 digits for radio-

frequency amplitude/phase variations, respectively. The bandwidths of the slice-selective 90°

and 180° pulses were 4220 and 1260 Hz, respectively. With a radio-frequency carrier at 3

Nature Medicine doi:10.1038/nm.2682 14

p.p.m., the slices of resonances between 1 – 5 p.p.m. were all included within the sample

dimension for both 90° and 180° radio-frequency pulses. For the 180° pulse, two density

matrices were calculated with two orthogonal radio-frequency phases (i.e., 0 and π/2), and the

slice-localized density matrix was obtained via subtraction between the matrices,

ρslice = (ρφ=0 − ρφ=π/2)/2. [10]

The square matrix ρslice was then converted to a column vector σ, whose i-th element was

calculated from

ci = trace(αi ρslice), [11]

N where αi is the normalized i-th PO term of the spin system and i = 1, 2, …, 4 . A single-column

matrix was calculated from each PO term as an initial density operator prior to the radio-

frequency pulse, and placed in the corresponding column of the T-matrix. For 2HG with 5

coupled spins, a 45×45 transformation matrix was constructed from the 45 column vectors, each

from each PO term. The calculation of the T-matrix of 2HG for the slice selective 180° pulse

was completed in ~7 hours in a PC. The T-matrix calculation time for the PRESS 90° RF pulse

was relatively minimal (~1 min) because the calculation was to be done only for a single PO

term (i.e., Iz). For calculating a spectrum following a PRESS sequence

90 – TD1 − 180 – TD2 − 180 – TD3 − Acquisition, [12]

| ← TE1 → | ← TE2 → |

the simulation began with the calculated density matrix of the slice-selective 90° pulse. The time

evolution during the inter-radio-frequency pulse delay (TD1) was calculated using

–1 ρ =V ρ0 V, [13]

diag –1 diag where V = U exp(–i H TD1) U , and H and U were formed from H = HCS + HJ. After this,

the square density ρ matrix was converted to a column matrix σ using Eq.[11] and multiplied by

Nature Medicine doi:10.1038/nm.2682 15

the 180° pulse T-matrix (Eq. [6]), giving a column matrix at the end of the ss180. This column

matrix was then converted to a square matrix ρ for calculating the density operator evolution

during the subsequent time delay (TD2). The calculation of the density operator evolution was

continued to obtain the density operator ρ at the end of the sequence. The expectation values of

single-quantum coherences were then extracted from the ρ, using trace(I–ρ), to construct a time-

domain signal, which was Fourier transformed to obtain a spectrum in the frequency domain.

The spoiling gradients symmetric about the PRESS 180° pulses were omitted in the simulation

because the 2-step phase cycling in the T-matrix calculation eliminated the outer-band

magnetization completely. With this transformation matrix method, a 3D-localized PRESS

spectrum of 2HG was calculated in < 0.5 s in a PC.

For calculating a difference edited spectrum from a sequence scheme 7,

90 – 180 – E180 − 180 – E180 – Acquisition [14]

| ← TE1 → | ← TE2 → |

the density operator evolutions during the slice-selective 90° and 180° RF pulses and the inter-

RF pulse delays were calculated similarly to the PRESS simulation. Two spectra were

calculated; one with editing 180° pulses (E180) turned on (subscan-A) and another with E180

turned off (subscan-B). The spoiling gradient pulses were applied similarly to the published

scheme 7. Since the E180 pulses, tuned to 1.9 p.p.m., nullify the transverse components of the

2HG H3 spins, the T-matrix of the E180 was constructed from averaging over two density

matrices obtained with RF phases 0 and π/2,

ρE180 = (ρφ=0 + ρφ=π/2)/2. [15]

In the E180-on subscan, the coupling of the H3 spins to the H2 spin was decoupled, thereby

leading to an inphase H2 multiplet. The E180-off subscan was essentially the same as PRESS.

Nature Medicine doi:10.1038/nm.2682 16

A difference spectrum was obtained from the two subspectra, (A–B)/2. The Gaussian envelope

of the E180 consisted of 500 numbers. The T-matrix calculation time for the 2HG 5-spin system

was ~20 min. Using this T-matrix algorithm, a 3D-localized difference spectrum of 2HG was

calculated in ~1 s.

Supplementary References

1. Ernst, R.R., Bodenhausen, G. & Wokaun, A. Principles of nuclear magnetic resonance in one and two dimensions, (Clarendon Press, Oxford, 1987).

2. Thompson, R.B. & Allen, P.S. Sources of variability in the response of coupled spins to the PRESS sequence and their potential impact on metabolite quantification. Magn Reson Med 41, 1162-1169 (1999).

3. Fan, T.W.-M. Metabolite profiling by one- and two-dimensional NMR analysis of complex mixtures. J Prog Nuc Magn Reson Spec 28, 161-219 (1996).

4. Govindaraju, V., Young, K. & Maudsley, A.A. Proton NMR chemical shifts and coupling constants for brain metabolites. NMR Biomed 13, 129-153 (2000).

5. Bal, D. & Gryff-Keller, A. 1H and 13C NMR study of 2-hydroxyglutaric acid and its lactone. Magn Reson Chem 40, 533-536 (2002).

6. Sorensen, O.W., Eich, G.W., Levitt, M.H., Bodenhausen, G. & Ernst, R.R. Product operator formalism for the description of NMR pulse experiments. J Prog Nuc Magn Reson Spec 16, 163-192 (1983).

7. Mescher, M., Merkle, H., Kirsch, J., Garwood, M. & Gruetter, R. Simultaneous in vivo spectral editing and water suppression. NMR Biomed 11, 266-272 (1998).

Nature Medicine doi:10.1038/nm.2682 NeuroImage 86 (2014) 43–52

Contents lists available at ScienceDirect

NeuroImage

journal homepage: www.elsevier.com/locate/ynimg

Review Current practice in the use of MEGA-PRESS spectroscopy for the detection of GABA Paul G. Mullins a,⁎, David J. McGonigle b,c, Ruth L. O'Gorman d,e, Nicolaas A.J. Puts f,g, Rishma Vidyasagar h, C. John Evans b, Cardiff Symposium on MRS of GABA and Richard A.E. Edden f,g a Bangor Imaging Unit, School of Psychology, Bangor University, Bangor, LL57 2AS, UK b CUBRIC, School of Psychology, Cardiff University, Cardiff, CF10 3AT, UK c School of Biosciences, Cardiff University, Cardiff, CF10 3AT, UK d University Children's Hospital, Steinwiesstrasse 75, 8032 Zürich, Switzerland e Center for Integrative Human Physiology (ZIHP), University of Zürich, Switzerland f Russell H Morgan Department of Radiology and Radiological Science, The Johns Hopkins University School of Medicine, 600 North Wolfe Street, Baltimore, MD 21287, USA g F.M. Kirby Center for Functional Brain Imaging, Kennedy Krieger Institute, 707 North Broadway Street, Room G-25, Baltimore, MD 21205, USA h Biomedical Imaging Institute, School of Cancer and Enabling Sciences, Manchester University, Stopford Building, Oxford Road, Manchester, M13 9PL,UK article info abstract

Available online 13 December 2012 There is increasing interest in the use of edited proton magnetic resonance spectroscopy for the detection of GABA in the human brain. At a recent meeting held at Cardiff University, a number of spectroscopy groups Keywords: met to discuss the acquisition, analysis and interpretation of GABA-edited MR spectra. This paper aims to GABA set out the issues discussed at this meeting, reporting areas of consensus around parameters and procedures Edited MRS in the field and highlighting those areas where differences remain. It is hoped that this paper can fulfill two MEGA-PRESS needs, providing a summary of the current ‘state-of-the-art’ in the field of GABA-edited MRS at 3 T using MRS analysis MEGA-PRESS and a basic guide to help researchers new to the field to avoid some of the pitfalls inherent in the acquisition and processing of edited MRS for GABA. © 2012 Elsevier Inc. All rights reserved.

Contents

Introduction ...... 44 MEGA-PRESS ...... 44 Acquisition ...... 44 Basic parameters ...... 45 PRESS timing ...... 45 PRESS pulses ...... 45 Editing pulse timing ...... 45 Editing pulse bandwidth ...... 46 Phase cycling ...... 46 Voxel size ...... 46 Unsuppressed water ...... 46 Pre-processing of data before fitting ...... 46 Signal quantification and fitting ...... 47 Concentration estimation and the use of Internal Standards ...... 48 Contamination of spectra by co-edited macromolecular signal ...... 49 Future developments ...... 49 Acknowledgments ...... 50 Appendix 1. Processing pipelines ...... 50 A1.1. AMARES/jMRUI ...... 50 A1.2. Gannet ...... 5051

⁎ Corresponding author. Fax: +44 1248 382599. E-mail address: [email protected] (P.G. Mullins).

A1.3. LCModel ...... 5051 A1.4. Tarquin ...... 5051 References ...... 5051

Introduction through the bonding electron network, which alters the appearance of the spectrum and the time-evolution of spins during an experiment. In Magnetic resonance spectroscopy (MRS) provides a non invasive the context of MEGA-PRESS, applying an RF pulse to one coupled spin technique to measure neurometabolites in vivo. In particular a large can modify the time-evolution of a coupling partner and therefore the number of studies into methods to accurately and reliably measure the appearance of the corresponding peak in the spectrum. We refer the in- neurotransmitters glutamate and GABA (γ-aminobutyric acid) have terested reader to de Graaf (2007) and Keeler (2011) for further explana- taken place, a few of which are referenced here (Edden and Barker, tion of scalar coupling that is beyond the scope of this article. 2007; Evans et al., 2010; Hancu, 2009; Henry et al., 2010; Hurd et al., A difference-edited technique, MEGA-PRESS involves the collection 2004; Jang et al., 2005; Jensen et al., 2005a; Mullins et al., 2008; of two interleaved datasets which differ in their treatment of the GABA Rothmanetal.,1993;Waddelletal.,2007). GABA in particular has spin system. In one dataset, an editing pulse is applied to GABA spins at proven to be difficult to reliably measure in-vivo with standard sin- 1.9 ppm in order to selectively refocus the evolution of J-coupling to gle voxel techniques, in large part due to the spectral overlap of the the GABA spins at 3 ppm (often referred to as ‘ON’). In the other, the in- main GABA peaks with peaks of other neurotransmitters which are version pulse is applied elsewhere so that the J-coupling evolves freely present in much greater concentrations, in particular the creatine (Cr) throughout the echo time (often referred to as ‘OFF’). The majority of peak at 3.0 ppm. While high field (3 T–7 T) short echo sequences have peaks in the spectrum are unaffected by the editing pulses, so subtraction shown some promise in allowing detection of GABA (Hu et al., 2007; of the refocused ON spectrum from the non-refocused OFF spectrum Mekle et al., 2009; Napolitano et al., 2012; Stagg et al., 2011), recent tech- removes all these peaks from the spectrum and retains only those nical advances and increased availability of spectral editing sequences peaks that are affected by the editing pulses. This process is demonstrat- have resulted in a rapid growth in the use of edited proton MRS to detect ed in Fig. 1. Thus, in-vivo, the edited spectrum contains signals close to the inhibitory neurotransmitter GABA in both the healthy and diseased 1.9 ppm (those directly affected by the pulses), the GABA signal at brain (Puts and Edden, 2012). In response to this development, re- 3 ppm (coupled to GABA spins at 1.9 ppm), the combined glutamate/ searchers from several spectroscopy groups met in August 2011 to glutamine/glutathione (Glx) peaks at 3.75 ppm (coupled to the Glx res- discuss current practice for the use of the MEGA-PRESS sequence onances at approximately 2.1 ppm), and J-coupled macromolecular for GABA-edited MRS (Edden and Barker, 2007; Mescher et al., 1998; (MM) peaks. The ON and OFF spectra are generally collected in an inter- Rothman et al., 1993; Terpstra et al., 2002). The purpose of the meeting leaved fashion to limit the impact of subject and hardware instabilities, was to present recent findings (Boy et al., 2010, 2011; Foerster et al., and subtraction is performed in post-processing. For more specific infor- 2012a, 2012b; Michels et al., 2012; O'Gorman et al., 2011b; Petrou et mation regarding the MEGA-PRESS technique and its acquisition and al., 2012; Puts et al., 2011) and to discuss issues of acquisition, processing post-processing, readers are directed towards the following articles and quantification encountered in performing these studies. (Edden and Barker, 2007; Evans et al., 2010; Henry et al., 2010; This paper focuses on MEGA-PRESS (Mescher et al., 1996, 1998) Mescher et al., 1998; Near et al., 2011; O'Gorman et al., 2011b; Terpstra editing for GABA at 3 T, as this is currently the most widely used MRS et al., 2002). technique for quantifying GABA and therefore the most promising MEGA-PRESS has become the most widely used technique for MRS ground on which to build consensus. This paper is not intended to be a measurements of GABA, largely due to ease of implementation within prescriptive rule book for MEGA-PRESS or GABA acquisition and analy- pre-existing PRESS sequences and the distribution of a research se- sis; rather, it is intended as a useful guide to current “minimal-best” prac- quence by at least one major vendor. Implementing MEGA-PRESS within tice as reached from consensus amongst researchers in the MRS field a stock PRESS sequence at the simplest level involves adding two RF present at the meeting. Although it is likely that some of the discussion pulses and altering the timing of the gradient pulses. MEGA-PRESS was presented here may generalize to different editing strategies and field published contemporaneously with the BASING method (Star-Lack et strengths, this paper does not cover the entire range of alternate acquisi- al., 1998) which similarly adds two editing pulses to the PRESS sequence. tion schemes for GABA measurement which includes 2D J-resolved MRS The distinction between the two lies mainly in the choice of gradient (Jensen et al., 2005b; Ryner et al., 1995), alterations in TE or sequence scheme used for coherence transfer pathway selection; both allow for si- timing parameters (Hu et al., 2007; Mullins et al., 2008; Thompson and multaneous water suppression, a feature that is often not used due to the Allen, 2001), unedited spectra at 7 T (Mekle et al., 2009), CT-PRESS development of excellent pre-saturation methods. Many GABA experi- (Mayer et al., 2006) or other editing sequences (Choi et al., 2005)(fora ments being applied currently could be loosely described either as broader review see Puts and Edden (2012)). However we believe that MEGA- or BASING-edited. They are possibly almost uniformly referred applied research into GABAergic function in both the healthy and dis- to as MEGA-PRESS because that method was originally applied to editing eased brain will benefit greatly from standardization of acquisition and GABA, whereas BASING was originally applied to lactate. analysis practices, allowing the quantitative comparison of data from different brain regions, studies, and scanner platforms. Acquisition

MEGA-PRESS Almost inevitably, the implementation and application of the MEGA- PRESS method differs between makes and models of scanners, field MEGA-PRESS (MEshcher–GArwood Point RESolved Spectroscopy), strength, operating systems and research groups. However, the acquired named after the authors who first proposed the MEGA suppression spectrashareanumberofstereotypical characteristics, as shown in scheme, is quickly becoming the standard technique used in MRS mea- Fig. 2. The differences in implementation originate largely from differ- surements of GABA. It allows GABA signals to be separated from the ences in the timing, slice profile and bandwidth of slice-selective pulses stronger overlying signals of other metabolites by taking advantage of used in the base PRESS experiment and differences in the timing and known couplings within the GABA molecule. Scalar coupling is an inter- bandwidth of the editing pulses. Despite these differences in sequence action between different hydrogen nuclei within a molecule, transmitted implementation, spectra from all three major MR system vendors are P.G. Mullins et al. / NeuroImage 86 (2014) 43–52 45

Fig. 1. Schematic diagram of MEGA-PRESS editing for GABA. (a) Editing pulses applied at 1.9 ppm modulate the shape of the GABA signals at 3 ppm (b). Subtracting scans acquired without these pulses (labeled OFF) from scans acquired with the editing pulses (ON) removes overlying creatine signals from the edited spectrum, revealing the GABA signal in the difference spectrum (labeled DIFF). (b) shows the effect of editing pulses on signals at 3 ppm only.

sufficiently similar to be amenable to a common series of processing and PRESS pulses analysis steps (Fig. 2). The following discussion of the intricacies of MEGA-PRESS implemen- It has been widely reported that losses in editing efficiency occur due tation assumes that the MEGA-PRESS sequence (an example is shown in to finite bandwidth slice-selective refocusing in edited PRESS (Edden and Fig. 3b) consists of two editing pulses added into a base PRESS sequence Barker, 2007; Near et al., 2011). The bandwidth of slice-selective (an example is shown in Fig. 3a), usually the vendor-default PRESS refocusing pulses (which is generally inherited from the base PRESS implementation. sequence) is determined by the maximum B1 available and the RF waveform used. The slice-profile of these pulses (i.e. how rectangular the slice selection is) will therefore also impact the appearance of the acquired Basic parameters doublet. This is one of the reasons for slightly different phantom spectral results between different vendor systems (Fig. 2a). This figure highlights Echo times (TE) are largely standardized to the 68 ms duration the fact that even under optimum conditions and when using a phan- suggested in the original GABA editing paper (Rothman et al., 1993). tom, the edited spectra acquired deviate from the theoretical ideal, and Repetition times vary between groups, resulting in variable degrees of small changes in sequence parameters can have unintended conse- T1-weighting of the signals; this should be considered in the light of quences in the final acquired spectrum. This highlights the need to be the significant MM contribution to the edited signal, which is discussed exact when stipulating the RF pulses used in numerical simulations. further below. Editing pulse timing

PRESS timing In order to fully refocus the evolution of coupling during TE, and therefore maximize editing efficiency, the two editing pulses should At medium and long echo times, the first spin echo duration (TE1) be separated by TE/2 (as shown in Fig. 3b). However this optimal of the PRESS sequence is typically kept as short as possible to mini- timing constrains the maximum duration of the editing pulses and mize the degree of excitation of multiple-quantum coherence path- may not always be adopted. The absolute timing of the first editing ways. The value of TE1 varies between vendors due to differences in pulse is often set to be halfway between the slice-selective excitation maximum B1 strength, maximum gradient strength and slew rate, pulse (specifically the zero-phase time point during the pulse) and and gradient areas chosen to achieve slice-selection. Changing the rel- the second slice-selective refocusing pulse. The second editing pulse is ative values of TE1 and TE2 (while maintaining TE) will modulate the often set halfway between that second excitation pulse and the end of appearance of the edited GABA signal detected (see Fig. 2 for in vitro ex- the echo time. Combined with the differences in resulting edited spectra amples) as it does for other J-coupled metabolite peaks (Thompson and seen in Fig. 3, it would therefore seem prudent for the researcher to be Allen, 2001). fully aware of the timing schemes used in their implementation of 46 P.G. Mullins et al. / NeuroImage 86 (2014) 43–52

Fig. 3. Representative PRESS (a) and MEGA-PRESS (b) pulse sequences, showing the addition of the editing pulses symmetrically around the second 180° pulse.

whether phase cycling for localization (which can greatly improve water suppression, for example) is a higher priority than rapid (e.g. contiguous) interleaving of ON and OFF spectra for MEGA-PRESS subtraction. For ease of implementation, phase cycling of between 2 and 16 steps is often prioritized over interleaving, but the phase cycling and edit ON/OFF interleaving scheme should be selected carefully to minimize subtraction artefacts resulting from potential drift or motion between collection of the ON and OFF lines. The degree of phase cycling also impacts the time resolution of data available for pre-processing frequency- and phase-correction (see “Pre-processing of data before ﬁtting”) in some data export formats.

Voxel size

Typically, the voxel size used in MEGA-PRESS is large when compared Fig. 2. Comparison of GABA-edited spectra across three vendors. The Siemens data is ac- to that used for other neuroimaging modalities (for example, fMRI). quired with a vendor-distributed sequence, whereas the GE and Philips sequences are cus- Primarily, the large voxel size is necessary to offset the inherent low tomer implementations by and available from RAEE. (a) Phantom data (acquired in a signal to noise ratio (SNR) for GABA (reported to be between 0.7 and 10 mM GABA solution in phosphate-buffered saline) show similar edited signal in each 3 ‘ ’ 1.4 mM/cm in concentration (Govindaraju et al., 2000; Petroff, 2002; case. Note that the commonly anticipated pseudo-doublet is not observed in any imple- 3 mentation, and that implementations differ signiﬁcantly in the extent to which the ‘center Rothman et al., 1993)), and so voxels on the order of 3×3×3 cm are peak’ is edited. (b) In-vivo edited spectra. Typical parameters are: TE 68 ms; TR 2 s; commonly used as a compromise between localization and signal qual- 3×3×3 cm3 voxel; acquisition time 10 min; editing pulses applied at 1.9 ppm (ON) and ity. As with all spectroscopic techniques when SNR may be limited, re- 7.46 ppm (OFF). ductions in voxel size may require increases in acquisition times and vice versa to ensure data reliability. MEGA-PRESS, and how they may differ when comparing results to other vendors or versions of the sequence. Unsuppressed water

Editing pulse bandwidth To allow concentration reference to tissue water, unsuppressed water spectra are acquired after the metabolite spectra as a separate The bandwidth of editing pulses, a key parameter as it determines scan (or as part of the standard PRESS acquisition). Typically 16 spec- the degree of co-editing of macromolecular signal (see Section 7 tral averages are collected for the water reference scan, allowing for a “Contamination of spectra by co-edited macromolecular signal”), is full phase cycle, but as long as spectral quality is good, any number determined by both the RF waveform used for editing and the dura- should suffice. tion of the editing pulses. The editing pulse duration is in turn determined by the remaining period of the echo time not already occupied Pre-processing of data before fitting by PRESS localization pulses and gradients. Where greater maximum B1 is available, slice selective pulses can be shorter and editing pulses lon- Processing of edited single voxel MRS data follows the same work- ger, resulting in narrower bandwidth editing pulses with less co-editing flow as that for unedited single voxel measurements: if a phased array of macromolecular signals. receive coil is used, data from the individual coils are phase-corrected and combined either using the default vendor approach or off-line; ex- Phase cycling ponential line broadening (3–4 Hz is typical) is usually applied (unless LCModel is to be used for analysis); Fourier transformation is applied; Although pulsed field gradients are used in the PRESS sequence for and frequency- and phase-correction are applied to time-resolved fre- localization and coherence transfer pathway selection, localization is quency domain data prior to temporal averaging and subsequent fitting improved with additional phase cycling. It is an unresolved question (Zhu et al., 1992). As the data are usually collected as a series of ON and P.G. Mullins et al. / NeuroImage 86 (2014) 43–52 47

OFF editing pairs they are already in a format that facilitates within-scan spectra. Fitting methods that use predefined models of spectral peaks correction of each pair for any potential frequency drift. Indeed, fre- corresponding to the spectrum of expected neurometabolites, called basis quency correction of this type is particularly beneficial as it improves sets, such as LCModel (Provencher, 1993, 2001), TARQUIN (Wilson et al., line width in the difference spectrum (Waddell et al., 2007)andcanre- 2010), and QUEST in jMRUI (Naressi et al., 2001; Stefan et al., 2009), orig- duce or remove subtraction artefacts associated with frequency and inally designed for unedited spectra, can be applied to edited spectra with phase instability. appropriate modifications to basis sets and control parameters. Some The details of frequency correction vary extensively between groups. tools, which are custom-written for GABA-MRS analysis, such as Gannet Typically a high-signal-to-noise peak is fitted to determine the frequency (available through gabamrs.blogspot.com), can perform both preprocess- and phase corrections to be applied to each spectral pair across the entire ing and fitting. Table 1 summarizes the pre-processing and fitting steps acquisition. While the residual water peak is often chosen for frequency commonly used in the various MRS quantification tools. and phase corrections due to its higher SNR and spectral separation Where used, basis sets should be acquired experimentally, or simulat- from GABA signals, the frequency and phase of the residual water signal ed using full 3-dimensional voxel simulation (rather than hard-pulse or may respond in an unpredictable manner to any frequency instability, on-resonance approximations), with the exact pulse widths and timings and the efficacy of this approach varies according to the degree of of the individual experiment used as approximations may fail to capture water suppression applied (which differs between vendors). One method the significant spatial variation in intensity and multiplet structure of the for dealing with this frequency instability involves the addition of an in- edited signal that occur throughout the voxel (see e.g. (Edden and Barker, terleaved water navigator to the MEGA-PRESS sequence (Bhattacharyya 2007)). Software such as VESPA (scion.duhs.duke.edu,n.d.) is freely avail- et al., 2007). Creatine (Cr) can also be used as a frequency and phase ref- able and has been used by several groups to simulate basis spectra. How- erence, and may be preferable to the water peak due to the common lo- ever, validation of simulations against phantom data is advised as there is cation of origin of the GABA and Cr signals. However, Cr has lower SNR on-going debate as to the values of the couplings in GABA and the extent than water, and the form of the Cr signal is expected to differ between to which two-bond proton–proton couplings need to be included OFF and ON scans due to changes in the underlying GABA signal at the (Govindaraju et al., 2000; Kreis and Bolliger, 2012). A database of same chemical shift. For this reason, pair-wise corrections (applying simulated spectra for different sequence implementations would the same correction parameters to both spectra in each ON–OFF pair) be useful for future researchers and would promote common anal- should be applied if Cr is to be used as a frequency and phase reference ysis pathways. There has been mention of such a database on both (Evans et al., 2012), except in cases where large frequency drifts are the TARQUIN and VESPA users mailing lists, and the VESPA project present, since frequency correction in the presence of large frequency has a repository for user contributions http://scion.duhs.duke.edu/ drifts may result in incomplete subtraction of choline and creatine sig- vespa/contrib, although as yet no metabolite results for any se- nals. The N-acetyl-D-aspartate (NAA) signal does not appear in ON spec- quence have been contributed. tra, so could only be used with a pair-wise correction based on the OFF Using the AMARES fitting routine in jMRUI is another option where spectra (at the cost of temporal resolution). An additional benefit of fre- prior knowledge and a degree of manual user intervention can be used quency correction is the possibility of using the magnitude of frequency to fit both the GABA and other metabolite peaks as a mixture of single shifts as an exclusion criterion for data confounded by excessive subject Gaussian or Lorentzian peaks. motion (Bhattacharyya et al., 2007). While the simple triplet model of the 3-ppm GABA signal supports Once the frequency- and phase-corrections have been applied, time the use of a pseudo-doublet model for GABA fitting, both simulations averaging is performed and the edited difference spectrum calculated. and phantom data show significant contribution from the center peak In addition, the sum of the editing OFF spectra alone can be calculated, (as seen in Figs. 1 and 2), and the appearance of in-vivo spectra from essentially a standard press sequence, so that other metabolites such different implementations vary in the extent to which a splitting is as NAA, Cr and Cho can be quantified either as the internal quantifica- seen in-vivo, with estimates varying from 20% to 60% of spectra showing tion reference (see below) or for additional metabolic information. Al- doublet character. Note that small-to-moderate Cr subtraction artefacts ternatively, reference Cr levels can be calculated from the sum of edit can easily be misinterpreted as ‘true’ doublet splittings, and the use of ON and OFF spectra, with the advantage of improved SNR (relative to doublet character as a benchmark for spectral quality cannot be univer- the edit OFF spectra alone). sally applied. Trials of model-type within one pipeline (GANNET) suggest that the integral that results from fitting by either a Gaussian singlet or Signal quantification and fitting doublet model agree with an R2 of 0.9872 for a dataset of 144 spectra from occipital, sensorimotor and DLPFC (as shown in Fig. 4). This is per- As with unedited spectroscopy, there are a number of tools avail- haps not surprising given the general line broadening seen in in vivo able for fitting and quantification of GABA concentration from edited data, and the presence of underlying macromolecule peaks (discussed

Table 1 Summary description of analysis methods.

Fitting method AMARES (jMRUI) Gannet LC-MODEL Tarquin

DATA format GE, Philips, Siemens GE, Philips, Siemens GE, Philips, Siemens GE, Philips, Siemens Frequency correction Optional automatic processing Automatic Optional pre-processing step. Not Optional automatic part of the LC Model package processing Water subtraction Optional semi-automatic processing Optional automatic processing Not performed Optional automatic processing Calculation of edited spectrum Manual Automatic Pre-processing required Automatic GABA model User choice — singlet or doublet, prior Singlet, Gaussian Basis set uses simulated or Two Gaussian singlets knowledge for Gaussian or Lorentzian phantom spectra. Concentration estimates User choice of reference to water, Automatic reference to water and Cr Automatic reference to water or Automatic reference to Cr NAA user choice of NAA, Cr water or user choice of Cr, NAA Quality measures SD of residual SD of residual, rejection of data with CRLB SD of residuals poor quality, or excessive motion Analysis of non-edited spectrum Single, or double peak model Single peak model Metabolite basis set analysis Metabolite basis set analysis 48 P.G. Mullins et al. / NeuroImage 86 (2014) 43–52

Table 2 Within-session reliability of different analysis methods.

Analysis method GABA test–retest reproducibility (% CV from consecutive scans)

AMARES 9% Gannet 8% LCModel 7% TARQUIN 8%

variation of the GABA concentrations derived for each participant, averaged across the subject group. Coefficients of variation ranged from 7% to 9% for all fitting methods (Table 2). Further details with regard to the processing methods are given in Table 1. While MEGA-PRESS has been shown to be reliable in vivo, researchers are recommended to validate the technique on their systems against phantoms of known concentration, and to perform this step on a regular basis as part of a normal quality assurance protocol. Validation of in vivo measures is of course difficult as apart from other MRS techniques, other non-invasive methods to measure neuronal GABA are Fig. 4. The effect of fit model on the estimation of GABA: Analysis was performed on non-existent. This may be considered one potential downfall of any 144 in-vivo datasets using both single Gaussian and double Gaussian functions to fit MRS methods, in that ground truth for concentration is not known, the GABA peak. The high correlation between the resulting integral values demonstrates that, for in-vivo data, fitting a single Gaussian is equivalent to using a more and all measures by their nature are relative to some proposed reference, complex model. whose concentration may itself be based on assumptions and estima- tions. However, the use of appropriate assumptions, and literature values

for key constants (e.g. T1,T2, and water concentrations) as discussed in in greater detail later), and highlights that even with edited spectra, dis- the next section is widely accepted, and should provide measures that crimination of individual spectral peaks is difficult in vivo. are comparable between research groups. If LCModel is to be used to fit edited spectra, the default fit parameters (including an unconstrained baseline and the default macromo- Concentration estimation and the use of Internal Standards lecular fitting) are not recommended as these may lead to variable results, often with minimal detection of GABA due to LCModel appor- The calculation of concentration values from fitted edited spectra tioning the majority of the peak at 3.0 ppm to the MM component. follows the same principles (and suffers from the same limitations) The use of LCModel for fitting edited spectra is limited by the fact as for standard single-voxel MRS. In general, concentrations are calcu- that its default settings tend to assume a spectrum of sharp peaks lated relative to an experimentally acquired internal reference of (an superimposed on a broad positive baseline. In in vivo edited spectra, assumed) known concentration. In published studies to date, three the GABA signal is not particularly narrow (compared to the MM sig- references have been used — Cr, Water and NAA. Each has its merits: nal) and the majority of the MM baseline is removed upon editing. Cr and NAA have the advantage that they are acquired during the These problems can be addressed either by constraining the baseline MEGA-PRESS scan, rather than as a separate scan at the beginning (by setting VITRO=T or NOBASE=T in LCModel) or by explicitly or end of the GABA acquisition, so potential effects of subject motion modeling the co-edited macromolecules in the basis set (see below) are minimized. Using Cr as a reference is familiar amongst the clinical and are not limited to LCModel. However, LCModel is viewed in some community and has been shown to perform well (Bogner et al., 2010). 1 circles as the default tool for the analysis of H-MRS spectra and many In addition, the main Cr resonance is very close to the edited GABA sig- literature examples of variable or inappropriate fitting arise from mis- nal; therefore chemical shift displacement issues will be negligible. The use of LCModel (e.g. Fig. 5 in Taki et al., 2009). water signal has higher SNR and is more easily modeled, although TARQUIN is an MRS analysis package that uses a linear combina- chemical shift effects (Howe et al., 1993; Weinreb et al., 1985)mean tion of basis functions to fit spectra, similar to LCModel. The fitting that the acquired water signal (in some implementations) may be is performed in the time-domain and simulated basis sets based on from a different location than the acquired voxel. Within the Gannet quantum calculations can be automatically generated to match common pipeline, referencing to both Cr and water is used (when water data acquisition protocols. However, for edited MEGA-PRESS data TARQUIN are provided), which is useful in establishing that it is the GABA numer- uses a simple predefined basis set which models the GABA peak as two ator, rather than the denominator, that drives observed effects as single Gaussian peaks. Full simulation of the MEGA-PRESS sequence is changes in NAA, Cr and water signals are all well documented in under development by the developers of TARQUIN. disease. The reliability of these various fitting/analysis methods has been Water-scaled GABA concentrations can be estimated in institu- assessed in a recent study investigating the (within-session) test–retest tional units from metabolite peak amplitudes according to the follow- reproducibility of GABA concentrations in a group of healthy adult vol- ing equation: unteers (O'Gorman et al., 2011b). Results from this study indicate that the reproducibility of GABA quantification is similar across methods fi − −TR −TE (see Table 2). The reliability of GABA quanti cation was assessed from 1 exp T exp T 3 ½¼SGABA ½ 1H2O 2H2 O MMcor ð Þ MEGA-PRESS data acquired from a 25×40×30 mm voxel in the dorso- GABA H2O VISH O 1 S 2 − −TR −TE eff – H2O 1 exp exp lateral prefrontal cortex in sixteen healthy adults (age 25 38 years; T1GABA T2GABA with signed informed consent and local ethics board approval) (O'Gorman et al., 2011b). Four consecutive resting MEGA-PRESS spectra were acquired from each participant and water-scaled GABA concentra- where SGABA and SH2O are the averaged raw GABA and water signals, re- tions were derived with LCModel, jMRUI, TARQUIN and Gannet. For spectively, [H2O] is the brain water concentration (e.g. 55,550 mmol/l), each fitting method, the reliability was quantified as the coefficient of VISH2O is the water visibility (e.g. 0.65 in white matter), eff is the editing P.G. Mullins et al. / NeuroImage 86 (2014) 43–52 49

efficiency (0.5), TR is the repetition time, T1H2O is the T1 of water (1.1 s) studies (Behar et al., 1994; Hofmann et al., 2001; McLean et al., 2004). (Wansapura et al., 1999), T2H2O is the T2 of water (0.095 s) (Wansapura However using this technique to account for the MM contributions et al., 1999), T1GABA is the T1 of GABA (0.8 s), T2GABA is the T2 of GABA has several disadvantages; (i) it increases experiment time, (ii) in the (0.13 s) (Träber et al., 2004), and MMcor is a macromolecular correc- MEGA-PRESS implementation it reduces the amplitude of the peak at tion factor given by the fraction of GABA thought to occupy the 3.0 ppm, reducing fit reliability, (iii) it increases the noise level in the GABA+peak (0.45). resulting MM-corrected GABA spectrum by √ 2 (assuming the residual The use of water as a concentration reference for spectroscopy has noise of the metabolite-nulled acquisition is the same as that of the been discussed extensively elsewhere, and while it has many advan- GABA acquisition), and, (iv) it requires the subtraction of two non- tages, users are recommended to acquaint themselves fully with current interleaved measurements, resulting in a greater sensitivity to partici- practices and cautions (Gasparovic et al., 2006). The most important pant motion. Pilot data (O'Gorman et al., 2011a) show that metabolite- issue with regard to quantification using water as an internal reference nulling significantly reduces the reliability of GABA measures (with is the accurate correction for partial volume effects within the voxel — CRLBs increasing from 20 to 44%) and reduces the sensitivity of this is especially important when performing group comparisons or cor- MEGA-PRESS GABA levels. These factors would suggest that metabo- relations with other parametric measures. It is most important to correct lite nulling is not a useful technique for control of MM contributions the measured GABA signal for the CSF-fraction of the voxel, such that the to the MEGA-PRESS experiment. quoted GABA pseudo-concentration is per unit brain tissue, excluding CSF Symmetric suppression of MM signals is a promising approach, and (Kreis et al., 1993). Correction of the water reference signal for apparent has been shown to be somewhat effective. Caution is advised when ap- water density and relaxivity in each tissue type is recommended plied at 3 T however as the editing pulses commonly used are insuffi- (Gasparovic et al., 2006). Most current image analysis packages for ciently selective to suppress MM without also significantly impacting structural/anatomical neuroimaging data have the ability to perform on the GABA signals via co-editing. Recent work has suggested that tissue segmentation, the reliability of which will depend on several the use of a slightly increased echo time over the typical 68 ms in factors, not the least of which is the quality of the anatomical images MEGA-PRESS allows for editing pulses that are sufficiently selective to collected. While ensuring correct registration of the voxel of interest avoid this co-editing (Edden et al., 2012). Validation of this technique, with the anatomical brain structures is seen as the biggest hurdle, or others with more selective editing pulses is one potentially reward- the importance of performing tissue segmentation and correction ing area of future development. was universally acknowledged in all discussions, especially if water Finally, explicitly modeling the MM contributions in the basis set is was used as a concentration reference. Readers are directed to Alger advantageous in that this method does not require additional scans (2010) for a detailed discussion of the issues inherent in quantitative (Murdoch and Dydak, 2011). However, fitting of edited spectra for measures in MRS. both GABA and MM is heavily dependent on data quality and fitting con- Correction of GABA measurements for voxel gray matter (GM) and straints, and reproducibility data from one group suggests that modeling white matter (WM) fraction is more controversial. While some groups the co-edited macromolecules may introduce additional variability into have found that the concentration of GABA is two times higher in gray the estimated GABA levels. While simple to implement, other fitting matter than white matter (Jensen et al., 2005a; Petroff et al., 1988, methods like AMARES, and Gannet do not yet allow for simulation of 1989), it is probably more appropriate to utilize GM:WM ratios explic- MM or accurate modeling of the baseline beyond a linear fit. In this itly as covariates in any statistical analysis rather than to attempt to cor- case fitting of the edited peak at 3.0 ppm as either a singlet or a doublet rect measures based on these reported differences in concentration will yield an estimate for GABA that should be reported as having a sig- between tissue types — indeed it is not obvious what such corrected nificant contribution from MM, often referred to as GABA+ in the quantities would correspond to, as they no longer represent concentra- literature. tions. Corrections for voxel GM:WM fraction invariably also ignore the While unfortunate, the contribution of MM to the GABA signal must heterogeneous excitation of GABA signal within the voxel (Edden and be acknowledged, and the likelihood of MM driving observed results Barker, 2007) and are as likely to inject tissue-fraction-driven effects must be assessed on a case-by-case basis. For example, studies that hy- into the data as they are to correct for these effects. pothesize strong relationships between GABA-MRS measures and behavioral function (e.g. Edden et al., 2009; Puts et al., 2011; Yoon et al., Contamination of spectra by co-edited macromolecular signal 2010) from the outset, or examine experimentally-induced changes within session due to a task (Floyer-Lea et al., 2006; Michels et al., As with any spectra acquired with a short-to-medium TE, macro- 2012) are perhaps less likely to be confounded by MM. However, molecular (MM) contamination represents a major area of concern. group studies comparing patients with neurotypical controls may be af- Unfortunately, spectral editing does not separate GABA signal from fected by MM changes between groups, particularly in light of recent at least one MM component, arising from spins at 3 ppm coupled to data suggesting that aging may influence the macromolecular contribu- spins at 1.7 ppm which are affected by the editing pulses. As such re- tion (Aufhaus et al., 2012). As such work on methods to account for MM duction of the MM signal component is perhaps one of the most im- contributions is an important area of development in GABA measures. portant areas for development in the measurement of GABA. While MM contributions are an acknowledged problem, there is as Three main approaches have been proposed to separate GABA from yet no real consensus on the best way to deal with them, and those co-edited MM signals: direct measurement of the MM baseline through methods that are available are not yet widely applied. This widespread metabolite nulling (Behar et al., 1994; Hofmann et al., 2001; McLean et failure to account for MM, while perhaps understandable given the lim- al., 2004); symmetric editing-based suppression of MM (Henry et al., itations of the methods available, stands as the greatest single limitation 2001; Mescher et al., 1998; Rothman et al., 1993)andfitting of data of the area to date and should be acknowledged as such. While prob- with GABA and MM basis functions (Murdoch and Dydak, 2011). Cur- lematic, accounting for MM contribution also represents an area of con- rently each method has some detrimental effect on data quality or siderable on-going work, both in the MEGA-PRESS community and time of acquisition, thus the most common approach to date has been within MRS in general. All those interested in using MRS to measure to accept MM contamination as a limitation of the MEGA-PRESS method GABA should follow these future advances closely. at 3 T. A discussion of the three main techniques applied currently, and their caveats follows. Future developments Direct measurement of the MM baseline through the use of an inversion pulse prior to the PRESS sequence to cause metabolite nulling has Methodological developments continue, both in terms of acquisition been proposed for normal short echo MRS as well as MEGA-PRESS and processing, and this review is only intended to present a snapshot 50 P.G. Mullins et al. / NeuroImage 86 (2014) 43–52 of a rapidly developing field. Quantification of GABA at fields higher The Welsh Institute of Cognitive Neuroscience than 3 T, although not widely available, will benefitfromincreased NIH grants P41 EB015909 and R21 NS077300. SNR, increased spectral dispersion and increased selectivity of editing pulses, advancing our understanding of GABAs role in the brain even Cardiff Symposium on MRS of GABA, 22–23 August 2011: further. There is also considerable work being done on non-edited Dr. Matthew J. Brookes. Sir Peter Mansfi eld Magnetic Resonance MRS techniques to detect GABA reliably (Hu et al., 2007; Napolitano Centre, Nottingham University, UK et al., 2012) at 3 T, however, MEGA-PRESS is still the most commonly Adrian Garcia — University of Birmingham, Birmingham, UK applied MRS technique for GABA measurement, warranting this review Bradley R. Foerster, MD — Department of Radiology, University of and discussion of appropriate implementation and current problems Michigan, Ann Arbor, MI & Ann Arbor VA Healthcare System, Ann with this technique. Arbor, MI Absolute quantification in meaningful concentration units is one use- Myria Petrou, MA MBChB MS, Department of Radiology, Universi- ful goal, requiring the measurement of GABA relaxation times (Edden et ty of Michigan, Ann Arbor, MI al., 2011), editing efficiency, improved handling of MM contamination, Darren Price — Sir Peter Mansfield Magnetic Resonance Centre, and extensive validation of both acquisition and processing steps. How- Nottingham University, UK ever, much like other MRS measures at present, until a standard practice Bhavana S. Solanky — NMR Research Unit, Department of is settled upon by the wider community most GABA measures should be Neuroinflammation, University College London, Institute of Neurology, thought of as provisional or institutional estimates, and not true absolute UK concentrations when making comparisons between studies at different Inês R. Violante — Institute of Biomedical Research in Light and sites. Provided the same techniques are applied however, studies from Image, Faculty of Medicine, University of Coimbra, Coimbra, Portugal the same sites, should be directly comparable. Steve Williams — Biomedical Imaging Institute, School of Cancer Determining a standardized approach can be problematic in any and Enabling Sciences, Manchester University, Manchester, UK academic field where several differing view points vie for prominence Martin Wilson — Cancer Sciences, University of Birmingham, however a summary of the consensus reached at the MEGA-PRESS Birmingham, UK meeting regarding minimal best practice follows. Data should be collected at the standard TE of 68 ms although recent data suggests that Appendix 1. Processing pipelines the use of 80 ms allows for better MM removal through the use of symmetric-editing (Edden et al., 2012). Phase cycling of at least 2 A number of MRS research groups were represented at the Cardiff steps should be performed, and allows for repeated interleaving of meeting, all at differing stages of development of a processing pipeline. ON and OFF edited spectra. The interleaving of spectra allows for fre- Illustrative examples of the processing steps for two of the three main quency correction and monitoring across a spectral acquisition and is clinical MRI system manufacturers are presented below to outline cur- fi recommended to both improve line width of the nal spectra, and to rent procedures. Areas of similarity in each will then be discussed and catch subject motion which may lead to subtraction artefacts. Researchers considered in terms of appropriateness and utility. It should be pointed should also be aware of the differences in timing and pulse shapes be- out that the MEGA-PRESS sequences are not currently standard “product” tween vendors and sequence versions and the effects this may have on sequences for these manufacturers and are distributed as research the data acquired. Voxel size, or acquisition time can be tailored to the patches or sequences specific to platform.1 question at hand, but should also be tested to ensure they provide ade- The example pipeline includes steps which can be thought of as fi quate signal to noise for reliable tting. There are several processing and reflecting best-practice beliefs amongst the participants and formed fi tting implementations available, if basis sets are used, models from the basis for round table discussion on day two of the meeting. We have phantoms or from full simulation with appropriately modeled RF pulses set out below discussion of those aspects of these processing steps that are recommended over simpler approximations, however use of either are most important to consider, highlighting both areas of consensus a single or double Gaussian peak for GABA has been shown to produce and those where further refinement and consideration of the technical equally reliable results. Concentration referencing to an unsuppressed issues is recommended. Note that for the AMARES, LCModel, and water scan is recommended, although the Cr and NAA peaks from the un- TARQUIN pipelines, automated drift correction is not applied but can be edited spectrum can also be used as long as whichever is used should be implemented as an additional pre-processing step using custom software. prominently mentioned when reporting the results. If water is used, appropriate correction for tissue content and relaxation differences is re- A1.1. AMARES/jMRUI quired. Finally, MM contributions need to be addressed, either through acknowledgement of their contribution to the GABA signal reported, or GE pre-processing through the use of some technique to control for these. While several In SAGE, data are coil combined and phase corrected, using the first techniques exist, researchers are advised to be aware of the caveats and point of the FID of the unsuppressed water signal for each coil. The potential pitfalls of each as discussed previously. coil-combined unsuppressed water lines are then averaged and saved The measurement of GABA concentration using edited MRS at 3 T is to a sage data file for subsequent processing, and the coil-combined mean extremely powerful approach for investigating the role of GABAergic tabolite lines are subtracted and saved to a separate sage data file. (If Cr inhibition in healthy brain function and the pathophysiology of neuro- or other metabolite information is desired then the coil-combined un- logical and psychiatric disease. Development of common acquisition edited lines are saved to a different data file.) Automated drift correc- and analysis strategies across sites and vendors will enhance the inter- tion is not applied. The binary sage data files are converted to text pretability of the literature and widen uptake of the methodology to files for subsequent processing in jMRUI. sites without pre-existing MRS expertise. It is our hope that this report can form the foundation of such a common acquisition, processing and analysis framework. Philips pre-processing Data are exported from the scanner as *.SDAT and *.SPAR files and Acknowledgments opened in jMRUI as a time series of editing pairs (ON and OFF spectra). The ON and OFF spectra are 180° out of phase, so to produce the edited This work is funded in part by: spectrum a simple addition across all the spectra is required. Phase The University Research Priority Program “Integrative Human Physiology” at the University of Zurich 1 These sequences are freely available from Dr Edden ([email protected]). P.G. Mullins et al. / NeuroImage 86 (2014) 43–52 51 correction may be required. If individual editing pairs are summed first, Quantification then the inverted NAA peak can be used to check for frequency drifts Water-scaled GABA concentrations are calculated according to a before summation to produce the final edited spectrum. HSVD re- modified version of Eq. (1). The editing efficiency and macromolecu- moval of any residual water can then be performed and 4 Hz Gausian lar correction factors are not applied automatically. apodization applied. A1.4. Tarquin Processing The residual water signal is removed by Hankel SVD. GABA levels are Pre-processing evaluated with the AMARES algorithm, after manually defining the center None required. frequency and width of the GABA peak. GABA is modeled as a Gaussian singlet (phase 0°) and NAA is modeled as a single inverted Lorentzian Processing peak (phase 180°). GABA levels are quantified either relative to NAA, Coil-combination and phasing are performed automatically (see unsuppressed water or creatine (modeled as a single Lorentzian in the Table 1 for details). Frequency correction can be applied, with the op- unedited lines). Alternatively, GABA/Cr can be calculated by multiplying tion to use both the Cr and Cho peaks, or the Cr peak alone. GABA is the GABA/NAA ratio from the edited spectrum by the NAA/Cr concentra- modeled as two Gaussian singlets, and both the water-scaled GABA tion calculated by LCModel from the unedited lines (Donahue et al., concentration and the GABA/Cr ratio are calculated. A simple single 2010). Gausian model is also included for the MM peak.

Quantification fi Water-scaled GABA concentrations must be calculated manually, Quanti cation but can be estimated in institutional units using Eq. (1). Water-scaled GABA concentrations are calculated according to a modified version of Eq. (1), excluding the relaxation correction since Tarquin partially accounts for relaxation effects by using a scaling factor A1.2. Gannet in its estimate. Readers are advised that this scaling factor is set by default, and may not take into account the true relaxation effects for Pre-processing MEGA-PRESS sequences and so should either explicitly set this scaling None required. factor to one, or correct for its effect post hoc. The editing efficiency and macromolecular correction factors are not applied automatically. Processing Coil-combination, phasing, apodization, and frequency correction are performed automatically (see Table 1 for details). GABA is References modeled as a single Gaussian superimposed on a linear baseline, Alger, J.R., 2010. Quantitative proton magnetic resonance spectroscopy and spectro- and both the water-scaled GABA concentration and the GABA/Cr scopic imaging of the brain. Top. Magn. Reson. Imaging 21 (2), 115–128. ratio are calculated. Aufhaus, E., Weber-Fahr, W., Sack, M., Tunc-Skarka, N., Oberthuer, G., Hoerst, M., Meyer- Lindenberg, A., Boettcher, U., Ende, G., 2012. Absence of changes in GABA concentrations with age and gender in the human anterior cingulate cortex: a MEGA-PRESS Quantification study with symmetric editing pulse frequencies for macromolecule suppression. Water-scaled GABA concentrations are calculated according to Eq. (1), Magn. Reson. Med. http://dx.doi.org/10.1002/mrm.24257. fi Behar, K.L., Rothman, D.L., Spencer, D.D., Petroff, O.A., 1994. Analysis of macromolecule taking into account the editing ef ciency and approximate macromolec- resonances in 1H NMR spectra of human brain. Magn. Reson. Med. 32 (3), 294–302. ular contributions to the GABA+ peak. Bhattacharyya, P.K., Lowe, M.J., Phillips, M.D., 2007. Spectral quality control in motion- corrupted single-voxel J-difference editing scans: an interleaved navigator approach. Magn.Reson.Med.58(4),808–812. A1.3. LCModel Bogner, W., Gruber, S., Doelken, M., Stadlbauer, A., Ganslandt, O., Boettcher, U., Trattnig, S., et al., 2010. In vivo quantification of intracerebral GABA by single-voxel (1)H-MRS — – GE pre-processing how reproducible are the results? Eur. J. Radiol. 73 (3), 526 531. http://dx.doi.org/ 10.1016/j.ejrad.2009.01.014. Data are coil combined and phase corrected in SAGE, as described Boy, F., Evans, C.J., Edden, R.A.E., Singh, K.D., Husain, M., Sumner, P., 2010. Individual above for AMARES/jMRUI. The coil-combined unsuppressed water differences in subconscious motor control predicted by GABA concentration in – lines are then averaged and saved to a sage data file for subsequent SMA. Curr. Biol. 20 (19), 1779 1785. Boy, F., Evans, C.J., Edden, R.A.E., Lawrence, A.D., Singh, K.D., Husain, M., Sumner, P., processing, and the coil-combined metabolite lines are subtracted 2011. Dorsolateral prefrontal gamma-aminobutyric acid in men predicts individual and saved to a separate sage data file. No apodization is performed. differences in rash impulsivity. BPS 70 (9), 866–872. Automated drift correction is not applied. The binary sage data files Choi, I.-Y., Lee, S.-P., Shen, J., 2005. In vivo single-shot three-dimensionally localized multiple quantum spectroscopy of GABA in the human brain with improved spec- are converted to LCModel .RAW format. tral selectivity. J. Magn. Reson. 172 (1), 9–16. de Graaf, R.A., 2007. In Vivo NMR Spectroscopy: Principles and Techniques, 2nd ed. Philips pre-processing Wiley-Interscience. Donahue, M.J., Near, J., Blicher, J.U., Jezzard, P., 2010. Baseline GABA concentration and SDAT files are converted to LCModel .RAW format and averaged. fMRI response. NeuroImage 53 (2), 392–398. http://dx.doi.org/10.1016/j.neuroimage. (Since ON and OFF spectra are acquired 180° out of phase, the 2010.07.017. MEGA-PRESS difference spectrum can be derived from a sum of all spec- Edden, R.A.E., Barker, P.B., 2007. Spatial effects in the detection of gamma-aminobutyric acid: improved sensitivity at high fields using inner volume saturation. Magn. tra, so prior subtraction is not necessary). If frequency correction is to be Reson. Med. 58 (6), 1276–1282. applied it should be done before the averaging step. No apodization is Edden, R.A.E., Muthukumaraswamy, S.D., Freeman, T.C.A., Singh, K.D., 2009. Orientation performed. discrimination performance is predicted by GABA concentration and gamma oscillation frequency in human primary visual cortex. J. Neurosci. 29 (50), 15721–15726. Edden, R.A., Intrapiromkul, J., Zhu, H., Cheng, Y., Barker, P.B., 2011. Measuring T(2) in vivo Processing with J-difference editing: application to GABA at 3 Tesla. J. Magn. Reson. Imaging 35 – The (subtracted) metabolite and water files are processed in (1), 229 234. Edden, R., Puts, N., Barker, P.B., 2012. Macromolecule-suppressed GABA-edited mag- LCModel using either an experimental or simulated basis set includ- netic resonance spectroscopy at 3T. Magn. Reson. Med. 68 (3), 657–661. ing basis spectra for GABA, glutamate, glutamine, Glx, NAA and gluta- Evans, C.J., Puts, N.A.J., Robson, S.E., Boy, F., McGonigle, D.J., Sumner, P., Singh, K.D., et thione, either with the control parameters VITRO=T, NOBASE=T, or al., 2010. Diurnal stability of gamma-aminobutyric acid concentration in visual and sensorimotor cortex. J. Magn. Reson. Imaging 31 (1), 204–209. with the coedited macromolecular signal included in the basis set Evans, C.J., Puts, N.A., Robson, S.E., Boy, F., McGonigle, D.J., Sumner, P., Singh, K.D., (Murdoch and Dydak, 2011). Edden, R.A., 2012. Subtraction artifacts and frequency (Mis-)alignment in J- 52 P.G. Mullins et al. / NeuroImage 86 (2014) 43–52

difference GABA editing. J. Magn. Reson Imaging. http://dx.doi.org/10.1002/ Near, J., Simpson, R., Cowen, P., Jezzard, P., 2011. Efficient gamma-aminobutyric acid jmri.23923 (Epub ahead of print). editing at 3T without macromolecule contamination: MEGA-SPECIAL. NMR Biomed. Floyer-Lea, A., Wylezinska, M., Kincses, T., Matthews, P.M., 2006. Rapid modulation of GABA 24 (10), 1277–1285. concentration in human sensorimotor cortex during motor learning. J. Neurophysiol. 95 O'Gorman, R.L., Edden, R.A., Michels, L., Murdoch, J.B., Martin, E., 2011a. Precision and (3), 1639–1644. repeatability of in vivo GABA and glutamate quantification. Presented at the Intl. Foerster, B.R., Callaghan, B.C., Petrou, M., Edden, R.A., Chenevert, T.L., Feldman, E.L., Soc. Mag. Reson. Med., Montreal, Canada, vol. 19, p. 3434. 2012a. Decreased motor cortex gamma-aminobutyric acid in amyotrophic lateral O'Gorman, R.L., Michels, L., Edden, R.A., Murdoch, J.B., Martin, E., 2011b. In vivo detec- sclerosis. Neurology 78 (20), 1596–1600. tion of GABA and glutamate with MEGA-PRESS: reproducibility and gender effects. Foerster, B.R., Petrou, M., Edden, R.A.E., Sundgren, P.C., Schmidt-Wilcke, T., Lowe, S.E., J. Magn. Reson. Imaging 33 (5), 1262–1267. Harte, S.E., et al., 2012b. Reduced insular γ-aminobutyric acid in fibromyalgia. Arthri- Petroff, O.A.C., 2002. Book review: GABA and glutamate in the human brain. Neuroscien- tis Rheum. 64 (2), 579–583. tist 8 (6), 562–573. Gasparovic, C., Song, T., Devier, D., Bockholt, H.J., Caprihan, A., Mullins, P.G., Posse, S., et al., Petroff, O.A., Ogino, T., Alger, J.R., 1988. High-resolution proton magnetic resonance 2006. Use of tissue water as a concentration reference for proton spectroscopic imag- spectroscopy of rabbit brain: regional metabolite levels and postmortem changes. ing. Magn. Reson. Med. 55 (6), 1219–1226. http://dx.doi.org/10.1002/mrm.20901. J. Neurochem. 51 (1), 163–171. Govindaraju, V., Young, K., Maudsley, A.A., 2000. Proton NMR chemical shifts and cou- Petroff, O.A., Spencer, D.D., Alger, J.R., Prichard, J.W., 1989. High-field proton magnetic pling constants for brain metabolites. NMR Biomed. 13 (3), 129–153. resonance spectroscopy of human cerebrum obtained during surgery for epilepsy. Hancu, I., 2009. Optimized glutamate detection at 3T. J. Magn. Reson. Imaging 30 (5), Neurology 39 (9), 1197–1202. 1155–1162. Petrou, M., Pop-Busui, R., Foerster, B.R., Edden, R.A., Callaghan, B.C., Harte, S.E., Harris, R.E., Henry, P.-G., Dautry, C., Hantraye, P., Bloch, G., 2001. Brain GABA editing without mac- et al., 2012. Altered excitation–inhibition balance in the brain of patients with diabetic romolecule contamination. Magn. Reson. Med. 45 (3), 517–520. neuropathy. Acad. Radiol. 19 (5), 607–612. Henry, M.E., Lauriat, T.L., Shanahan, M., Renshaw, P.F., Jensen, J.E., 2010. Accuracy and Provencher, S.W., 1993. Estimation of metabolite concentrations from localized in vivo stability of measuring GABA, glutamate, and glutamine by proton magnetic reso- proton NMR spectra. Magn. Reson. Med. 30 (6), 672–679. nance spectroscopy: a phantom study at 4 Tesla. J. Magn. Reson. 1–9. Provencher, S.W., 2001. Automatic quantitation of localized in vivo 1H spectra with Hofmann, L., Slotboom, J., Boesch, C., Kreis, R., 2001. Characterization of the macromole- LCModel. NMR Biomed. 14 (4), 260–264. cule baseline in localized (1)H-MR spectra of human brain. Magn. Reson. Med. 46 Puts, N.A.J., Edden, R.A.E., 2012. In vivo magnetic resonance spectroscopy of GABA: a (5), 855–863. methodological review. Prog. Nucl. Magn. Reson. Spectrosc. 60 (C), 29–41. Howe, F.A., Stubbs, M., Rodrigues, L.M., Griffiths, J.R., 1993. An assessment of artefacts Puts, N.A., Edden, R.A., Evans, C.J., McGlone, F., McGonigle, D.J., 2011. Regionally spe- in localized and non-localized 31P MRS studies of phosphate metabolites and pH cifi c human GABA concentration correlates with tactile discrimination thresh- in rat tumours. NMR Biomed. 6 (1), 43–52. olds. J. Neurosci. 31 (46), 16556–16560. Hu, J., Yang, S., Xuan, Y., Jiang, Q., Yang, Y., Haacke, E.M., 2007. Simultaneous detection Rothman, D.L., Petroff, O.A., Behar, K.L., Mattson, R.H., 1993. Localized 1H NMR mea- of resolved glutamate, glutamine, and γ-aminobutyric acid at 4T. J. Magn. Reson. surements of gamma-aminobutyric acid in human brain in vivo. Proc. Natl. Acad. 185 (2), 204–213. Sci. U. S. A. 90 (12), 5662–5666. Hurd, R., Sailasuta, N., Srinivasan, R., Vigneron, D.B., Pelletier, D., Nelson, S.J., 2004. Mea- Ryner, L.N., Sorenson, J.A., Thomas, M.A., 1995. Localized 2D J-resolved 1H MR spectroscopy: surement of brain glutamate using TE-averaged PRESS at 3T. Magn. Reson. Med. 51 strong coupling effects in vitro and in vivo. Magn. Reson. Imaging 13 (6), 853–869. (3), 435–440. scion.duhs.duke.edu, d. scion.duhs.duke.edu. Retrieved June 9, 2012, from http://scion. Jang, D.-P., Lee, J.-M., Lee, E., Park, S., Kim, J.-J., Namkoong, K., Yoon, K.-J., et al., 2005. duhs.duke.edu. Interindividual reproducibility of glutamate quantification using 1.5-T proton Stagg, C.J., Bestmann, S., Constantinescu, A.O., Moreno, L.M., Allman, C., Mekle, R., magnetic resonance spectroscopy. Magn. Reson. Med. 53 (3), 708–712. Woolrich, M., et al., 2011. Relationship between physiological measures of excitabil- Jensen, J.E., deB Frederick, B., Renshaw, P.F., 2005a. Grey and white matter GABA level ity and levels of glutamate and GABA in the human motor cortex. J. Physiol. 589 differences in the human brain using two-dimensional, J-resolved spectroscopic (Pt 23), 5845–5855. imaging. NMR Biomed. 18 (8), 570–576. Star-Lack, J., Spielman, D., Adalsteinsson, E., Kurhanewicz, J., Terris, D.J., Vigneron, D.B., Jensen, J.E., Frederick, B.D., Wang, L., Brown, J., Renshaw, P.F., 2005b. Two-dimensional, 1998. In vivo lactate editing with simultaneous detection of choline, creatine, NAA, J-resolved spectroscopic imaging of GABA at 4 Tesla in the human brain. Magn. and lipid singlets at 1.5 T using PRESS excitation with applications to the study of Reson. Med. 54 (4), 783–788. brain and head and neck tumors. J. Magn. Reson. 133 (2), 243–254. Keeler, J., 2011. Understanding NMR Spectroscopy, 2nd ed. Wiley. Stefan, D., Cesare, F.D., Andrasescu, A., Popa, E., Lazariev, A., Vescovo, E., Strbak, O., et al., Kreis, R., Bolliger, C.S., 2012. The need for updates of spin system parameters, illustrated 2009. Quantitation of magnetic resonance spectroscopy signals: the jMRUI soft- for the case of γ-aminobutyric acid. NMR Biomed. 25, 1401–1403. http://dx.doi.org/ ware package. Meas. Sci. Technol. 20 (10), 104035. 10.1002/nbm.2810. Taki, M.M., Harada, M., Mori, K., Kubo, H., Nose, A., Matsuda, T., Nishitani, H., 2009. High Kreis, R., Ernst, T., Ross, B.D., 1993. Development of the human brain: in vivo quantifi- gamma-aminobutyric acid level in cortical tubers in epileptic infants with tuberous cation of metabolite and water content with proton magnetic resonance spectros- sclerosis complex measured with the MEGA-editing J-difference method and a copy. Magn. Reson. Med. 30 (4), 424–437. three-Tesla clinical MRI Instrument. Neuroimage 47 (4), 1207–1214. Mayer, D., Kim, D.-H., Adalsteinsson, E., Spielman, D.M., 2006. Fast CT-PRESS-based spi- Terpstra, M., Ugurbil, K., Gruetter, R., 2002. Direct in vivo measurement of human cere- ral chemical shift imaging at 3 Tesla. Magn. Reson. Med. 55 (5), 974–978. bral GABA concentration using MEGA-editing at 7 Tesla. Magn. Reson. Med. 47 (5), McLean, M.A., Simister, R.J., Barker, G.J., Duncan, J.S., 2004. Discrimination between 1009–1012. neurochemical and macromolecular signals in human frontal lobes using short Thompson, R.B., Allen, P.S., 2001. Response of metabolites with coupled spins to the echo time proton magnetic resonance spectroscopy. Faraday Discuss. 126, STEAM sequence. Magn. Reson. Med. 45 (6), 955–965. 93–102 (discussion 169–83). Träber, F., Block, W., Lamerichs, R., Gieseke, J., Schild, H.H., 2004. 1H metabolite relaxation Mekle, R., Mlynárik, V., Gambarota, G., Hergt, M., Krueger, G., Gruetter, R., 2009. MR spec- times at 3.0 Tesla: measurements of T1 and T2 values in normal brain and determi- troscopy of the human brain with enhanced signal intensity at ultrashort echo times nation of regional differences in transverse relaxation. J. Magn. Reson. Imaging 19 on a clinical platform at 3T and 7T. Magn. Reson. Med. 61 (6), 1279–1285. (5), 537–545. Mescher, M., Tannus, A., Johnson, M.O., Garwood, M., 1996. Solvent suppression using Waddell, K.W., Avison, M.J., Joers, J.M., Gore, J.C., 2007. A practical guide to robust detection selective echo dephasing. J. Magn. Reson. Ser. A 123 (2), 226–229. of GABA in human brain by J-difference spectroscopy at 3 T using a standard volume Mescher, M., Merkle, H., Kirsch, J., Garwood, M., Gruetter, R., 1998. Simultaneous in coil. Magn. Reson. Imaging 25 (7), 1032–1038. vivo spectral editing and water suppression. NMR Biomed. 11 (6), 266–272. Wansapura, J.P., Holland, S.K., Dunn, R.S., Ball, W.S., 1999. NMR relaxation times in the Michels, L., Martin, E., Klaver, P., Edden, R., Zelaya, F., Lythgoe, D.J., Lüchinger, R., et al., human brain at 3.0 tesla. J. Magn. Reson. Imaging 9 (4), 531–538. 2012. Frontal GABA levels change during working memory. PLoS One 7 (4), e31933. Weinreb, J.C., Brateman, L., Babcock, E.E., Maravilla, K.R., Cohen, J.M., Horner, S.D., Mullins, P.G., Chen, H., Xu, J., Caprihan, A., Gasparovic, C., 2008. Comparative reliability 1985. Chemical shift artifact in clinical magnetic resonance images at 0.35 T. of proton spectroscopy techniques designed to improve detection of J-coupled me- AJR Am. J. Roentgenol. 145 (1), 183–185. tabolites. Magn. Reson. Med. 60 (4), 964–969. Wilson, M., Reynolds, G., Kauppinen, R.A., Arvanitis, T.N., Peet, A.C., 2010. A constrained Murdoch, J.B., Dydak, U., 2011. Modeling MEGA-PRESS macromolecules for a better grasp least-squares approach to the automated quantitation of in vivo 1H magnetic res- of GABA. Presented at the Intl. Soc. Mag. Reson. Med., Montreal, vol. 19, p. 1394. onance spectroscopy data. Magn. Reson. Med. 65 (1), 1–12. Napolitano, A., Kockenberger, W., Auer, D.P., 2012. Reliable gamma aminobutyric acid Yoon, J.H., Maddock, R.J., Rokem, A., Silver, M.A., Minzenberg, M.J., Ragland, J.D., Carter, measurement using optimized PRESS at 3 T. Magn. Reson. Med. http://dx.doi.org/ C.S., 2010. GABA concentration is reduced in visual cortex in schizophrenia and 10.1002/mrm.24397. correlates with orientation-specific surround suppression. J. Neurosci. 30 (10), Naressi, A., Couturier, C., Devos, J.M., Janssen, M., Mangeat, C., de Beer, R., Graveron- 3777–3781. Demilly, D., 2001. Java-based graphical user interface for the MRUI quantitation pack- Zhu, G., Gheorghiu, D., Allen, P.S., 1992. Motional degradation of metabolite signal age. Magn. Reson. Mater. Phys. 12 (2–3), 141–152. strengths when using STEAM: a correction method. NMR Biomed. 5 (4), 209–211. Neurochem Res (2014) 39:1–36 DOI 10.1007/s11064-013-1199-5

OVERVIEW

A Guide to the Metabolic Pathways and Function of Metabolites Observed in Human Brain 1H Magnetic Resonance Spectra

Caroline D. Rae

Received: 28 September 2013 / Revised: 8 November 2013 / Accepted: 11 November 2013 / Published online: 21 November 2013 Ó Springer Science+Business Media New York 2013

Abstract The current knowledge of the normal bio- fMRS Functional magnetic resonance chemistry of compounds that give rise to resonances in spectroscopy human brain proton magnetic resonance spectra measure- GABA c-Aminobutyric acid able at readily available field strengths (i.e. B3T)is GABA(A)R GABA-A receptor reviewed. Molecules covered include myo- and scyllo- GABA(B)R GABA B receptor inositol, glycerophospho- and phospho-choline and cho- GAMT Guanidinoacetate methyl transferase line, creatine and phosphocreatine, N-acetylaspartate, GSH Glutathione N-acetylaspartylglutamate, glutamate, glutamine, c-ami- GSSG Oxidized glutathione nobutyrate, glucose, glutathione and lactate. The factors IQ Intelligence quotient which influence changes in the levels of these compounds MEGA-PRESS Mescher–Garwood point resolved are discussed. As most proton resonances in the brain at spectroscopy low field are derived from a combination of moieties whose NA N-Acetyl containing resonance biochemistry is complex and interrelated, an understanding NAA N-Acetylaspartate of the mechanisms underlying why these species change is NAAG N-Acetylaspartylglutamate crucial to meaningful interpretation of human brain NAT8L N-Acetyltransferase-8 like enzyme spectra. NMDAR N-Methyl-D-aspartate receptor MRS Magnetic resonance spectroscopy Keywords N-acetylaspartate Choline Creatine PCr Phosphocreatine Lactate myo-inositol c-Aminobutyric acid PRESS Point-resolved spectroscopy

Abbreviations ASPA Aspartoacylase Cho Choline-containing resonance COSY Correlation spectroscopy Introduction Cre Creatine-containing resonance CRT Creatine transporter Advances in magnetic resonance spectroscopy have made it possible to obtain, with continual improvements in resolution, non-invasive information about the biochemistry of the human brain. The resulting magnetic resonance C. D. Rae (&) Neuroscience Research Australia, Barker St, Randwick, spectrum is derived from a mixture of compounds, domi- NSW 2031, Australia nated by those with high proton concentrations ([0.1 mM) e-mail: [email protected] not dispersed by spin coupling or, depending upon choice of acquisition parameters, those with resonances that C. D. Rae Brain Sciences, The University of New South Wales, resolve and/or refocus to give coherent signals. Interpre- Kensington, NSW 2052, Australia tation of these spectra is complicated by several factors. 123 2 Neurochem Res (2014) 39:1–36

1. Resonance frequency overcrowding, especially at low which influence changes in the levels of these compounds field strengths, means that few resonances are derived are discussed and possible interpretations of changes are from a single compound. For example, the main offered (Table 1). For more information on the assignment methyl singlet resonances which appear in most brain of resonances and structures of metabolites, the reader is spectra (NA, Cho, Cre, Fig. 1) are themselves con- referred to the comprehensive article of Govindaraju et al. tributed to by at least two compounds, and may also [1] although see comments by Kreis and Bolliger [2]; for have other resonances underlying them. information on interpretation of spectra in neoplastic brain 2. Most compounds giving rise to resonances are influ- tissue, to Tosi et al. [3]; for information on common enced by more than one biochemical pathway, and can spectral artefacts to the review and comprehensive artefact give rise to, or result from, more than one biochemical gallery of Kreis [4]; and for information on spectral event. This means that there is rarely one interpretation acquisition, analysis and for a more historical context, to which can be applied universally and that spectra Mountford et al. [5]. should not be considered out of context. Factors like disease etiology, the age and gender of the subject as well as known biochemical information about the N-Acetylaspartyl Compounds disease itself must be considered. 3. There is still a paucity of information about the The N-acetyl resonance (NA) at 2.008 ppm is derived biochemical role of compounds contributing to the mainly from N-acetylaspartate (NAA), with a small con- resonances. Research in this area is aided and stimu- tribution from N-acetylaspartylglutamate (NAAG) [6, 7]. lated by data derived from spectra obtained in vivo, The relative contribution from each of these compounds especially from disease states where biochemical varies across brain regions [8]. NAA levels decrease with regulation may be abnormal. However, most informa- age, albeit slowly (reported values range from 0.5–3 %/ tion comes from experiments conducted in vitro (in decade, depending on region) [9, 10]. cultured cells, or tissue slices) and/or via extrapolation from animal models. Synthesis and Catabolism This review focuses on the resonances in human brain proton magnetic resonance spectra (Fig. 1) which are more N-acetylaspartate (NAA) is one of the highest concentrated frequently reported in the literature and measureable at of all free amino acids in the brain, second only to gluta- readily available field strengths (i.e. B3 T). The factors mate. It can be present in some neurons at concentrations

Fig. 1 Example 1H magnetic resonance spectra acquired from frontal (TR = 2 s, 1,024 data points, 2 Hz exponential multiplication). The grey matter using the PRESS sequence at 3T with a range of echo large baseline distortion present in the 32 ms TE spectrum can be times. Spectra represent the sum of 32 transients and were acquired attributed to resonances from macromolecules [456, 457] which can from an 8 cm3 volume of interest in the left anterior cingulate cortex be problematic to estimate and adequately account for [458]

123 Neurochem Res (2014) 39:1–36 3

Table 1 Possible roles for resonance markers Resonance Change due to Marker for

N-acetyl compounds (NA) Neuronal viability Neuronal loss Neuronal function IQ, BMI, diabetic status, brain metabolic activity, mitochondrial viability Neuronal density Neuronal loss Cell packing Inborn error Canavan’s disease Error in synthesis pathway Sickle cell disease [455] Myelin degradation/synthesis Demyelination Temperature change Temperature Choline-containing compounds (Cho) Membrane/phospholipid turnover Neurodegeneration/inﬂammation Altered cell density Total membrane content of VOI White/grey matter differences White/grey contributions to VOI Creatine-containing compounds (Cre) Bioenergetics Hypoxia (preconditioning effects) Metabolic activity of area Also related to creatine kinase activity in area Inborn error Transporter error Synthetic enzyme error c-amino-butyrate (GABA) Metabolic/functional activity May increase or decrease depending on the cell types responsible: decreased by mainstream inhibitory activity Increased by some benzodiazepines Marker for gamma synchrony Marker for tonic inhibition Glucose Glucose use Blood glucose levels Glucose transport Transporter expression Supply and demand Blood glucose concentration and metabolic activity Glutamate Metabolic activity Increases with increased metabolic activity May be marker for addiction Glutamine Metabolic activity Correlates strongly with glutamate Hyperammonaemia Glutathione (GSH) Increased oxidative stress Preconditioning, stress levels Age Decline in ability to maintain reductive capacity myo-Inositol Organic osmolyte Osmotic stress/oedema Second messenger Mood/depression Inborn error SMIT transporter expression Beta-amyloid binding Neurodegeneration Scyllo-inositol Increased blood concentration Diet Beta-amyloid binding Neurodegeneration

Lactate Local brain activity Transient lag in O2 supply on activation: decreased pyruvate clearance Oxygen supply Ischaemia/hypoxia Inborn error Mitochondrial encephalopathies Reduced clearance Altered BOLD response/decreased vascularisation of area Post-ischemia Macrophage inﬁltration

up to 20 mM. It is synthesized from acetyl coenzyme-A responsible for synthesis of NAA [12]. NAT8L is localized and aspartate by an NAA synthase, recently identiﬁed as N- in neurons (Fig. 2) where it is present in mitochondria [13] acetyltransferase-8 like protein (NAT8L) [11] and and cytoplasm [14] as well as possibly in oligodendrocytes

123 4 Neurochem Res (2014) 39:1–36

[15]. Despite neurons possessing the synthetic machinery ASPA [19]. The acetate produced is used to synthesize for NAA, the compound does not enter into any further myelin lipids [20–22] and may also play a role in the metabolic pathways in the neuron, apart from further acetylation of enzymes and histone proteins in the nucleus metabolism of a small amount to NAAG [16]. NAAG is [23, 24]. Enzymes regulating NAA levels can be induced limited to glutamatergic pathways and reactive microglia by activation of dopaminergic receptors (NAT8L) [11] and [17]. glutamatergic receptors (ASPA) [25], suggesting a role for N-acetylaspartate (NAA) and NAAG are catabolised in NAA in storing acetyl-Co units which are then provided glial cells (Fig. 2). NAA is de-acetylated mainly in oligo- when cell demand increases [26, 27]. dendrocytes, which contain large amounts of the catabolic N-acetylaspartate (NAA) turns over slowly [16], with a enzyme aspartoacylase (ASPA: EC 3.5.1.15) [18] although half-life of *7 h, measured in human brain [28]. Levels some neurons and microglia also possess small amounts of can be lowered by inhibition of mitochondrial respiration

Fig. 2 Metabolism and compartmentation of N-acetylaspartyl com- [40]. NAA has been shown to leave the brain, being released into pounds. N-Acetylaspartate (NAA) is synthesized in neurons in a blood plasma and thence into urine such that there is a net efﬂux from reaction catalyzed by N-acetyltransferase-8-like protein (Nat8 l). brain *1 % per day [459, 460]. NAA can enter oligodendrocytes NAA is also the precursor in neurons for the peptide N-acetylaspar- from neurons [20]. The transporters responsible are speculated to tylglutamate (NAAG), a probable neurotransmitter. NAA can be involve gap-junctional mechanisms or NaC3 [461]. In oligodendro- catabolised in neurons by aspartoacylase (ASPA) where the acetate cytes, NAA is catabolised to aspartate and acetate via ASPA. The units can be used in the nucleus to acetylate proteins such as histones. acetate is then used in the synthesis of fatty acids, such as myelin. NAA can leave neurons (transporter as yet unknown but it may be the NAAG is catabolized by glutamate carboxypeptidase II (GCPII) and sodium dependent dicarboxylate transporter NaCD2/NaC2 [35], since NAAG peptidase and the resulting glutamate taken up by excitatory the analogue of this transporter NaC3 is capable of transporting NAA amino acid transporters (EAAT) in the astrocyte in astrocytes [34]). In practice, very little NAA is found in astrocytes

123 Neurochem Res (2014) 39:1–36 5

[29]. Turnover rates in rat vary from 0.2 to 0.72 lmol/g [45] and it is clear that the utility of NAA as a neuronal wet weight/h depending on the brain region measured. marker may not be universal, depending on the underlying Rates of synthesis of NAA are comparable between rats pathology. In practice, it is likely that both situations may and humans (0.55 lmol/g wet weight/h), despite the two- hold to some degree. times disparity in overall metabolic rate, but repeat data NAA is a known osmolyte, providing *7 % of neuronal from human is limited. Rates of synthesis also appear osmolarity [48]. It is released into the extracellular space unrelated to depth of anaesthesia [30]. In a study where after K? stimulation [36], and hypotonic shock [49]. This ethanol was infused intravenously, NAA levels were shown release likely represents the neuronal response to the to fall by 8 % in a study where ethanol was infused applied osmotic stress [42, 50]. Recently, there has been intravenously for 1 h. This was suggested to be due to some argument about whether NAA levels can change in reduced mitochondrial activity subsequent to decreased functional situations [51] with the expressed hypothesis metabolic activity [31]. that NAA is acting as a water pump under these circum- NAAG is similarly catabolised by glial cells (Fig. 2) stances. In this role, extrusion of NAA represents a using mostly glutamate carboxypeptidase II, present on the mechanism whereby neurons can extrude the water pro- cell surface of virtually all astrocytes [32] but absent in duced by metabolism [52]. It is now demonstrated that the cultured oligodendrocytes and microglia [33] and to a change in NAA seen on visual stimulation is an artifact of much lesser extent glutamate carboxypeptidase III [32]. magnetic susceptibility [53]. NAA is likely transported into glia (and possibly into oli- A recent interesting addition to the discussion about godendrocytes) via the sodium-dependent dicarboxylate NAA and its physicochemical role maintains that devel- transporter NaC3/NaCD3 [34] which also transports Krebs opmental replacement of taurine by NAA increases diffu- cycle intermediates succinate and citrate and is inhibited by sivity of high-energy phosphate compounds and is key in lithium ion [35]. brain maturation and the move towards higher energy use In displaying an intercellular metabolic cycle, NAA and [54]. Brain maturation of mice lacking the NAA-degrading NAAG are exhibiting similarities to well-established neu- enzyme ASPA is delayed [24] and coincides with patho- rotransmitters such as glutamate and GABA. However, logically increased taurine levels at the adult stage [55]. NAA appears not to act as a neurotransmitter [36], High correlation has been reported between NAA levels although NAAG does exhibit neurotransmitter activity [17] and water apparent diffusion coefficient (ADC) [56]. and levels of NAAG/Cr correlate with cognitive ability in Recovery of NAA after insult has also been reported. De human MS patients [37]. Inhibition of glutamate car- Stefano et al. reported recovery of NAA signal following boxypeptidase activity increases brain NAAG levels and lesion resolution in MS patients and in those with mito- improves cognitive ability in mice with experimental chondrial disorders showing that NAA loss was not irre- autoimmune encephalomyelitis [37] and in rat models of versible [57], although recovery of signal is plainly traumatic brain injury [38]. Turnover of NAAG is faster mechanism dependent. Obstructive sleep apnea patients than that of NAA [16, 30]. show no NAA recovery following continuous positive airway pressure (CPAP) therapy [58] although there may A Role for NAA be partial reversal of cognitive symptoms [59], suggesting neuronal loss underlies the decrease seen here in NAA NAA has long been considered a neuronal marker, based rather than functional deficits. on the fact that brain tumours of glial origin did not contain NA is decreased in nonspecific mental retardation [60], NAA while those of neuronal origin did [39]. Using cul- and correlates positively in some areas of the brain with tured cells, NAA was localised to neurons and immature mental function [44, 61–63] although the basis of this oligodendrocytes [40]. Urenjak and coworkers subse- correlation is not well understood [64]. It may be because quently determined that mature oligodendrocytes, the cell of increased neuronal density in those persons with better type involved in formation of the myelin sheath, do not mental function, or because of increased mitochondrial contain NAA [41] although this was later challenged [15] efficiency [65] or because of increased density of mito- as the synthetic enzyme NAT8L is also found in oligo- chondria. The correlation is sufficiently robust to recom- dendrocytes [14]. mend the matching of mental function (e.g. IQ) in Some have suggested that NAA may be a poor marker experimental subjects and controls as closely as possible of neuronal density and may relate more to neuronal dys- where NAA is to be used as an experimental marker, function [42–44] although this view is questioned by others particularly in frontal and parietal lobe areas [64]. [45–47] from studies of neurodegenerative conditions. A subject has been identified with no detectable NAA Little data is available showing both NAA measurements signal in the brain [66]. This subject displayed develop- and histopathological correlations in non-diseased brain mental retardation and moderately delayed myelination. 123 6 Neurochem Res (2014) 39:1–36

The authors suggested caution in blanket application of the phosphatidylcholine (lecithin) or lysophosphatidylcholine ‘‘NAA as neuronal marker’’ dogma and it is clear that [77, 78]. Free diffusion of choline across the blood brain mechanisms exist for compensating for NAA in its barrier does occur, but at quantitatively insignificant rates absence. More recently the cause of NAA absence was at normal physiological blood choline concentrations [79]. identified as a mutation in NAT8L [12] the NAA synthe- The efficacy of uptake from dietary consumption sizing enzyme and it was suggested that the reason for only depends upon the source of the choline (whether free a moderate decrease in myelination is because NAT8L choline or lecithin (phosphatidylcholine)). Free choline can deficit does not prevent the availability of acetyl-CoA for be catabolised by gut bacteria, reducing total uptake; lec- myelin synthesis, although it may disrupt acetyl-CoA ithin is the supplement of choice. Initial reports that oral reserves [14]. However, a follow up study on that patient supplementation with choline in humans resulted in a dose- showed that the loss of NAA was not well tolerated as dependent increase in the Cho resonance [80] were not neurological function deteriorated severely after 5 years of replicated [81, 82] although this was suggested in a sub- age (i.e. epilepsy, microcephaly, motor and cognitive sequent replication by the original group to be due to lack impairments). Surprisingly peripheral (urine) levels of of statistical power [83]. In animal (rat) brain oral choline NAA and NAAG were normal [67]. Recently a NAT8L supplementation results in a significant increase in the knock-out mouse has been described, also with minimal concentration of nicotinic cholinergic receptors [84] and an neurological effects and mildly altered behavior, including increase in acetylcholine concentrations [85]. Further, there decreased social interaction, but normal memory and is good correlation between the Cho resonance and brain learning [68]. In keeping with most metabolites there are acetylcholine levels [86]. Recently, a correlation between redundancies in the system and examples at the extremes of total dietary choline and cognition has been reported in steady state are to be extrapolated from with caution. healthy humans [87], illustrating the importance of maintaining brain choline supply. Within the brain, free choline is transported into cells via Choline two main transport mechanisms [88]. There is a low affinity

(KM *20–200 lM) system widely distributed in neuronal The total choline (Cho) signal at d = 3.2 arises from the tissues [89] which operates as a means of choline uptake for nine protons in the N-methyl moiety of choline-containing phospholipid synthesis. Uptake by this system is altered by compounds. These are mostly glycerophosphocholine pH and membrane polarization. A high-affinity (KM *0.5– (GPC) and phosphocholine (PC) with a possible small 3.0 lM) sodium-dependent system has a regional distribu- contribution from free choline [69]. The presence of free tion in brain parallel to that of choline acetyltransferase [90] choline has been confirmed in animal brain [70], but and is classically described as being specific for cholinergic remains undetermined in normal human brain due to the nerve terminals [91]. Since the ratio of cholinergic to non- difficulty of obtaining sufficiently fresh tissue. In addition, cholinergic terminals in the CNS is low, quantitatively, the the ethanolamines, namely phosphoethanolamine and gly- low-affinity uptake system is vastly more prevalent [92]. The cerophosphoethanolamine, coresonate with the 1H Cho molecular details of choline transport are still emerging and resonance [71]. much is left to be identified [93]. Active transport is also responsible for keeping the con- Choline, Supply and Transport centration of choline in the CSF and interstitial fluid low by continuous removal, occurring mainly in the choroid plex- Because the brain cannot synthesise choline de novo, uses [94]. Brain cells therefore function with less choline in uptake of choline from extracellular fluids is essential [72]. their microenvironment than cells outside the brain [91]. Plasma choline concentrations are fairly constant at Choline is at first formed as the choline moiety of *10 lM[73] although they fluctuate as high as 20 lM phosphatidylcholine (Fig. 3) and only afterwards can it be following a choline-rich meal, and may decline by up to released as a free compound. The various phospholipases 50 % in dietary deficiency [74]. Choline is readily trans- convert phosphatidylcholine into lysophosphatidylcholine ported through the blood brain barrier by a carrier-medi- (phospholipase A), phosphocholine (phospholipase C) or ated process which is unsaturated at physiological levels of choline (phospholipase D). Free choline is also released plasma choline and is hence sensitive to alterations in from glycerophosphocholine by the enzyme glycerophos- plasma choline [75]. The cerebral compartment of free phocholine diesterase (E.C. 3.1.4.2) which is present in choline is thus open to exchange with the circulation. many tissues, including the brain [95]. Indeed, arteriovenous difference studies have shown that The organ responsible for the production of most new brain produces more free choline than it consumes [76]. choline in mammals is the liver. Symptoms of choline- Much choline is supplied to the brain via the blood as deficiency are not known in humans who are (possibly) 123 Neurochem Res (2014) 39:1–36 7

Fig. 3 Reactions affecting the Cho signal. Figure shows reactions signal, particularly changes in the activity of the various phospho- involving the choline moiety and underlines how the steady-state lipases. PLC phospholipase C, PLD phospholipase D equilibria between the different choline forms will influence the Cho capable as adults of producing all the choline required by choline-containing phospholipids or acetylcholine do not de novo synthesis [91]. Requirement for choline is espe- [99]. Situations that impair energy supply in the brain cially high under conditions of high growth, such as in (hypoxia/ischaemia, hypoglycaemia) result in high brain children. Choline supply affects brain development and is levels of free choline with the concomitant risk of loss of required for normal development of memory [96]. this choline to the circulation. This in turn has the capacity Synthesis of phosphatidyl choline from labeled etha- further to impair synthesis of phospholipids and acetyl- nolamine by phosphatidylethanolamine methyltransferases choline [100]. A strong relationship exists between the has been demonstrated both in neurons and glia, although activity of phospholipase A2 (see Fig. 3) and the size of the the rate of synthesis is considerably higher in glia [97]. brain choline resonance [101, 102] but to date the influence However, the brain cannot synthesise choline de novo at a of other phospholipases has not been examined. significant rate [98]. Reactions using choline require Taken together, these data outline the high degree of energy, while those that produce it by degradation of interconnectedness of all the choline moieties and how the

123 8 Neurochem Res (2014) 39:1–36 various steady-state equilibria between reactions in choline are in fast, near-equilibrium exchange (example reported -1 -1 anabolism and catabolism are key in determining the values for kf; 0.24 s in rat brain [108], 0.42 s (resting) eventual size of the Cho resonance. and 0.76 s-1 (activated) visual cortex in human [109]) on the NMR time scale, are interconverted by the enzyme The Cho Resonance as a Membrane Marker creatine kinase in reactions catalysed in both the mitochondria and the cytosol by appropriate isoforms of the Given the above conclusion about the equilibria contrib- enzyme (Eq. 1): uting to the Cho resonance then it follows that the Cho Creatine þ ATP $ PCr þ ADP þ Hþ; ð1Þ resonance is also in equilibrium with membrane phospholipids. Free choline concentrations correlate best of any Binding of creatine to creatine kinase has been shown in choline containing metabolite with the integral of the Cho creatine kinase knockout mice to play only a minor role, resonance, although a mix of free choline and glycero- suggesting that the binding phenomenon has little phosphocholine gives a better fit than either variable alone consequence for the thermodynamic availability of creatine [69]. Tissue cellularity also correlates strongly with the in the reaction [110]. Phosphocreatine has a significantly Cho resonance [69], indicating that the Cho resonance also shorter T2 than creatine [111] and therefore each component reflects cell density (i.e. the total contribution of cell will contribute differently to the Cre resonance, depending on membrane to the resonance in the volume of interest), as the echo-time chosen for spectral acquisition. Dynamic does the level of immortalisation and transformation of the alterations in the ratio of phosphocreatine to creatine can tissue sampled [103]. therefore alter the intensity of the Cre peak in 1H MRS spectra Phospholipids constitute around 40 % of myelin. The on a seconds time scale [111, 112] particularly at long echo Cho signal, while possibly having utility as a marker of times. Shorter echo times may therefore better reflect total active demyelination is of limited utility as a marker of creatine levels and be less influenced by the creatine/ myelin status [104]. Studies which validate any links phosphocreatine steady-state equilibrium. between myelin status and Cho (or other metabolites) in Magnetisation-transfer experiments have shown a humans are few and are sorely needed. decrease in the brain creatine resonance, [113, 114] sug- Correlations have been reported between the Cho reso- gesting that a portion of brain creatine is bound transiently nance and brain function. In recently abstinent alcoholics to the immobile proton pool [115] and therefore ‘‘NMR the resultant increase in the Cho resonance correlated invisible’’. This pool has been suggested to be as large as positively with mental performance [105]. This was most 15 % of the total pool [116]. likely due to previous dietary deficiencies. Cho levels correlate with mood in normal subjects [106]. In healthy Source brain, a negative correlation has been reported between the Cho resonance and IQ [62] and a relationship between Creatine in the body is derived both from the diet and from choline levels and oestrogen has been reported in women, synthesis. Average daily requirements are *2 g/day as with acute effects of sex hormones on the size of the res- *1.7 % of total body creatine is lost through spontaneous onance [107]. cyclisation of creatine and phosphocreatine to the break- Changes in the Cho resonance therefore reflect non-steady- down produce creatinine. Creatine, derived from the Greek state alterations in membrane turnover (either increased kreas, meaning flesh, is found exclusively in meat. Tissue membrane synthesis, or breakdown), or changes in cell den- levels are generally only fractionally higher in omnivores sity. The former interpretation is more likely in pathological than vegetarians, indicating that the synthetic route is states such as Alzheimer disease, tumours, inflammations, capable of meeting the majority of requirement for crea- infections and schizophrenia. In non-diseased brain, and tine. Oral supplementation of creatine in omnivores results possibly in developmental disorders, differences in cell den- in increased brain creatine levels (9 %), although there is sity are more likely to account for differences between spectra considerable intersubject variability in this [117] presum- although alterations in the steady-state equilibria of mem- ably due to variation in meat intake and/or transporter branes must also be considered. expression. Brain is capable of synthesising creatine [118–120]. The two main enzymes for creatine synthesis are present in both Creatine neurons and glia [118] (Fig. 4) although few cells express both enzymes simultaneously [121]. It is likely that the The ‘‘Cre’’ resonance at 3.02 ppm arises mainly from the synthetic route is important in maintaining creatine levels N-methyl moiety of two compounds; creatine and its under normal levels of creatine consumption, as uptake of phosphorylation product, phosphocreatine. The two, which creatine from the blood under these circumstances may be 123 Neurochem Res (2014) 39:1–36 9 limited [122]. Those with guanidinoacetate methyltrans- membrane via diffusion, has shown efficacy in mouse ferase deficiency have very low (0.3 mM) levels of brain models [130]. creatine, which can be increased over a period of months The apparent contradiction whereby creatine has been by creatine supplementation [123]. Metabolic defects such shown to be absent in the brain despite the (assumed) as ornithine D-transaminase deficiency reduce brain crea- presence of normal creatine synthesis machinery may be tine levels [124]. Creatine transporter deficiency has been explained by an intercellular creatine cycle (Fig. 4). The described where creatine was also absent in the brain existence of cycling has been proposed by: despite the assumed presence of normal synthetic 1. Dringen et al. [120] on the basis that astrocytes have machinery [125]. The clinical symptoms of all these defi- been shown to release guanidinoacetate (the immediate ciencies were very similar, suggesting that they arose from precursor of creatine); absence of creatine/phosphocreatine rather than excess of 2. Mo¨ller and Hamprecht [131] on the grounds that precursors. Creatine transporter deficiency cases have creatine transport (uptake) has been shown to be proved unresponsive to creatine supplementation, and have minimal in cultured neuronal cell lines; not responded well to supplementation with creatine pre- 3. Braissant et al. [118] who showed absent expression of cursors. One study has reported neuropsychological the creatine transporter CRT-1 in rat astrocytes [132] improvement in creatine transporter deficiency following and that SLC6A8 takes up both guanidine acetate and arginine supplementation [126] while others have reported creatine [133]. no effect [127–129]. Treatment with cyclocreatine, which can substitute for creatine but which may cross the cell

Fig. 4 Current understanding of metabolism, compartmentation and A third subpopulation (3) possess none of the synthetic enzymes and synthesis of creatine and its analogues in the brain. The synthesis and are absolutely reliant on SLC6A8 to take up creatine produced by uptake of creatine is differentially active in subpopulations of cells in other cells (Type 1 or 2). There is also a fourth subpopulation that the brain [462]. Some cells, which include neuronal, astrocytic and does not possess any synthetic enzymes or transporters and does not oligodendrocyte subpopulations contain both enzymes of the creatine use creatine (4). This means that both GAA and creatine are released synthesis pathway (designated 1 in the ﬁgure) AGAT (Arginine– from sub-populations of cells. Both of these compounds have been glycine aminidino transferase) and GAMT (Guanidinoacetate methyl shown to act as neurotransmitters or neuromodulators; GAA at transferase). Others, (designated 2) possess only the ﬁrst synthetic GABA(C) receptors [178] and creatine at glutamatergic NMDA enzyme AGAT, and are therefore net producers of guanidinoacetate receptors [179, 181]. There is currently no information the creatine (GAA), or only the second enzyme (GAMT) and therefore need the synthesizing status of cells possessing these receptors. SaM creatine transporter SLC6A8 in order to take up the precursor GAA. S-adenosylmethionine, SaHC S-adenosylhomocysteine

123 10 Neurochem Res (2014) 39:1–36

Transport Role of Creatine

Creatine enters the brain via SLC6A8, a sodium-dependent Creatine plays a pivotal role in brain energy homeostasis transporter [122], one of two creatine transporters expres- (Fig. 5). It acts in concert with multiple ATP-producing sed in humans [134]. Creatine, and its precursor guanidi- and requiring reactions as a buffer for high energy phos- noacetate, also enter cells in the brain via the sodium- phate bonds. The high energy phosphate bond in phos- dependent creatine transporter [131]. The specific ability of phocreatine has a higher free energy of hydrolysis than that individual cell types in the brain to transport creatine is not in ATP (DG°0 kJ/mol =-45.0 cf. -31.8, respectively; fully resolved. One study has shown a saturable Na? [122]). Energy can thus be stored as phosphocreatine, dependent uptake system in cultured astroglial cells, but buffering ATP levels. Phosphocreatine is a smaller, less little uptake in neuron-rich primary cultures [131], while negatively charged molecule than ATP, making it more others have shown high dendritic neuronal expression in energetically favourable for the cell to accumulate, and cultured hippocampal neurons [135]. In rat brain, neurons creatine kinase can quickly interconvert the two. ATP and oligodendrocytes, but not astrocytes, express the cre- produced via both glycolysis and oxidative phosphoryla- atine transporter CRT1 [118]; with expression particularly tion is buffered this way by cytosolic and mitochondrial common in highly oxidative areas of the brain [136]. pools of creatine, respectively. Conversely, phosphocrea- Braissant et al. reported that cells lacking guanidinoacetate tine is also an excellent source of ATP when it is demanded methyltransferase take up guanidinoacetate via SLC6A8 by ATP-hydrolysing reactions. Given ample PCr, the cre- and convert it to creatine, providing an explanation as to atine kinase reaction regenerates ATP at a rate 40 times why persons with SLC6A8 deficiency may not respond to faster than oxidative phosphorylation and 10 times faster arginine supplementation [133]. Deficiency of brain crea- than glycolysis [148]. tine has also been reported in those with a CRT mutation in The ATP buffering role of phosphocreatine is of con- the presence of normal muscle creatine levels, suggesting siderable importance in the brain. Upon functional activa- that muscle may have alternative routes for creatine uptake tion, phosphocreatine levels acutely decrease [149], mainly which are not available to brain [137]. GAMT-deficient via activation of the Na?/K?ATP-ase, which is directly mice show a time delay in creatine supplementation functionally coupled to creatine kinase. An increase in between muscle and brain changes, with brain levels much cerebral pH with functional activation has also been slower to increase than muscle [138] suggesting that observed, which likely arises due to the sequestering of H? muscle creatine levels are not good indicators of brain by ADP in the resynthesis of ATP [150], indicating that levels; brain levels ought to be measured directly. In gen- ATP turnover is actually occurring, despite the reported eral, evidence seems to suggest that one must supplement lack of change in the signal from ATP in 31P fMRS [151]. humans with 5 g creatine per day for more than 5 days in Indeed phosphocreatine performs such an efficient role as order to increase brain creatine levels [117, 139]. Note that an ATP buffer that ATP levels are not seen to decrease this is the dietary equivalent of eating *2 kg/day of raw until quite severe levels of O2 depletion are reached [152] meat! and ATP has not been observed to decrease measurably The concentration of creatine varies across regions of under normal heavy mental workloads. Saturation transfer the brain, being higher in cortical grey matter than white experiments show that the rate of the forward reaction of matter [140], and higher still in the cerebellum [10, 141]. creatine kinase is proportional to the degree of workload The regional distribution of creatine and of the creatine imposed, and a better indicator of the level of work than the transporter follows the expression and activity of creatine rate of decline of phosphocreatine [153]. However, whether kinase, which is highest in areas of highest synaptic the role of phosphocreatine as a buffer for ATP is essential activity [142]. Creatine levels across the brain correlate and/or irreplaceable is still open to conjecture. Creatine significantly with the expression of the creatine trans- kinase knockout mice have apparently normal brain func- porter [143]. Neurons in culture contain high levels of tion, with slightly higher ATP levels than normal [154] but phosphocreatine [144], although others have reported these mice are impaired under conditions of high energy primary astrocytes to exhibit higher PCr/ATP than pri- demand [155]. Changes to the glycolytic network and mary neurons when grown under similar external crea- mitochondrial structure in the muscle of these mice may tine concentrations [145]. Cellular variation in creatine- partly compensate for the defect [156] and similar changes kinase isoforms has also been reported in human brain may well exist in the brain. The buffering role of creatine [146] replicating earlier findings in the mouse [147] with kinase is also supported by cerebral vascular reactivity; in neurons containing high levels of mitochondrial creatine conditions where vascular reactivity may be impaired such kinase and astrocytes relatively high levels of the cyto- as during the transient hypoxia associated with obstructive solic form. sleep apnea, the cytosolic creatine kinase system plays 123 Neurochem Res (2014) 39:1–36 11

Fig. 5 Creatine and cellular bioenergetics. The ﬁgure shows cytosolic and mitochondrial compartments and their respective creatine kinase isoforms. Cytosolic creatine kinase interacts with ATP produced via substrate-level phosphorylation (e.g. via glycolysis) and also acts as a rapid supplier of ATP for hydrolytic cytosolic reactions, such as those involving the Na?K?-ATPase. Mitochondrial creatine kinase (MiCK) forms a metabolon with a porin (a voltage-dependent anion channel), which when combined with mitochondrial creatine kinase prefers a cationic form favoring creatine inﬂux. These proteins also combine with the adenine nucleotide translocase, ATP-synthaseF0F1 and an inorganic phosphate carrier. This metabolon is visible on electron micrographs [463]

little role in ATP buffering [157] with the workload largely favours transport of creatine over phosphocreatine. The borne by the mitochondria. Indeed, brain tissue heteroge- porin also allows efflux of high energy phosphate bonds neity may play a significant role in the variability in results from the mitochondrion in the form of phosphocreatine, obtained using different approaches to perturb the system although ATP and ADP can also pass through porin in its [158]. A better understanding of creatine and creatine anionic form. Creatine itself influences mitochondrial res- kinase compartmentation will aid our interpretation of piration rates, by altering the apparent KM of the mito- brain bioenergetics. chondrial voltage-dependent anion channel for ADP [162]. Mitochondrial creatine kinase is involved in regulating Phosphocreatine, through its smaller charge and size, oxidative phosphorylation [159, 160] most likely through also has a higher diffusibility than both ATP and ADP controlling the availability of ADP [161] by maintaining a [163]. Phosphocreatine therefore provides a vehicle by complex between inner and outer mitochondrial mem- which high-energy phosphate bonds can be distributed branes. Octomeric mitochondrial creatine kinase assembles throughout the cell. The relative contribution of phospho- in a complex which includes adenine nucleotide translo- creatine as an energy shuttle is dependent on the activity case and a porin, which, when entered into the complex, status of the cell as well as the local environment including 123 12 Neurochem Res (2014) 39:1–36 the distributions of creatine kinase and adenylate kinase amount of activity occurring in a particular region, or to the [164]. degree of vascularization of the area. Creatine acts as an organic osmolyte in the brain [49, 165] serving a more dominant role as the brain matures [166]. Efflux of creatine occurs through a pharmacologi- c-Aminobutyric Acid (GABA) cally distinct pathway to that of taurine [167]. Supple- mentation with creatine in rats is associated with an GABA, the major inhibitory neurotransmitter in the brain, increase in the brain myo-inositol signal [168]. Creatine contributes three coupled resonances to the proton spec- supplementation leads to increased water retention [169] trum, each arising from the three CH2 moieties in an which may in turn result in increased myo-inositol. AA0MM0XX0 spin system. There may also be significant In keeping with its pivotal role in cell energy homeostasis, contribution from the GABA moiety in homocarnosine, a higher creatine levels are associated with increased neuro- dipeptide of histidine and GABA [183, 184]. Editing protection in subsequent hypoxia. Creatine supplementation sequences are available (e.g. [185–187]) which resolve the is of neuroprotective benefit in a range of disorders including triplet resonance of –CH2(NH3) from the overlapping cre- neurological and atherosclerotic diseases [170] and is ben- atine singlet with good reliability [188, 189] although eficial to brain function. It reduces mental fatigue and oxygen concerns have been raised about the effect of underlying requirement when doing a repetitive task [171] and improves macromolecule resonances [190]. An assymetric PRESS performance at cognitive tasks in vegetarians [172] and approach, not widely tested, has shown comparable coef- omnivores [173]. Creatine has been suggested to be active at ficient of variation to MEGA–PRESS [191] but faster the GABA receptor [174, 175] although authors performing acquisition time [192]. Resolution of GABA is possible direct experiments have reported activity only for creatinine using two-dimensional techniques such L-COSY and (the cyclic breakdown product) not creatine [176, 177]. The J-PRESS [193, 194]), or single shot 1D methods using precursor, guanidinoacetate, is active at GABA receptors double quantum filtering [195]. being an agonist at GABA(A)R [177] and antagonist at GABA levels are higher in gray than in white matter

GABA(A)rhoR[178], in the latter case at least at physio- [196] and do not show measurable diurnal variation [197] logically relevant concentrations. Creatine may also be and decline with age [198]. active at the receptor responding to the major excitatory neurotransmitter glutamate, the NMDA receptor (NMDAR). Synthesis and Catabolism Royes et al. [179] have implicated creatine in population spike amplitude modulation and the NMDAR polyamine GABA is synthesized from glutamate by glutamate binding site may be a possible site of activity [180] with decarboxylase and catabolised via GABA-transaminase to creatine being released in response to neural activity [174]. succinic semi-aldehyde, and thence to succinate. This latter Creatine mediates a direct inhibitory action on the NMDAR- pathway for GABA catabolism is known as the GABA mediated calcium response [181]. Creatine supplementation shunt, and is irreversible. In practice, the carbon backbone decreases the degree of the BOLD response to photic stim- for much GABA synthesis is derived originally from glu- ulation [173], suggesting that creatine may increase basal tamine via glutamate, as blockage of glutamine transport metabolic rate. This has serious implications for fMRI, (i.e. depletes GABA pools [199–202] showing that astrocytic the size of the BOLD response can be varied considerably by glutamine is a major source for GABA synthesis (Fig. 6). creatine) especially considering that creatine/phosphocrea- Additional pathways for production of GABA have also tine can change on a very short time scale [111]. Since cre- been described although the evidence for these is so far less atine and guanidinoacetate cycle between cells, the creatine well supported. GABA may be produced from the dipep- cycle may modulate neuronal activity via the above reported tide homocarnosine, through the action of carnosinase activity at receptors. This explanation provides a better [203]. GABA can be synthesized from amines such as kinetic framework for the many reported cognitive and ornithine and putrescine in astrocytes [204] which may be neuroprotective effects of creatine than a purely energetic part of an inhibitory control mechanism linking glutamate argument [170, 182], although the role of creatine in directly uptake to astrocytic release of GABA [205]. altering mitochondrial respiration cannot be ignored either Only a fraction of brain GABA is neurotransmitter [162]. GABA at any one time. The rest constitutes a metabolic Creatine is by no means a constant compound in the pool of GABA, which is most probably located in cell brain and ought never to be assumed to be such. The bodies [206–208]. Similar to the situation with glutamate, practice of using creatine as a reference compound or as a it is likely that any pool of GABA may be drawn on for denominator in metabolic ratios is of questionable reli- neurotransmitter GABA and that the turnover of GABA is ability. Creatine levels are most likely related to the considerable [209]. 123 Neurochem Res (2014) 39:1–36 13

Fig. 6 Metabolism of GABA in the GABAergic synapse. GABA is synthesized in the GABA- ergic neuron via glutamate decarboxylase (GAD). It is packaged into vesicles with the assistance of the vesicular GABA transporter (VGAT) and released into the synapse. From there it can bind to GABA receptors and be taken up via one of the GABA transporters (GAT). In glial cells, it enters the GABA shunt via GABA- transaminase and then enters the Krebs cycle. The carbon backbone for GABA synthesis is supplied to neurons via glutamine. The glutamine is deaminated to glutamate, most probably via a phosphate activated glutaminase

GABA has two receptor types; the ionotropic GABA(A) on neurons and astrocytes. Studies from knockout mice and receptor and the metabotropic GABA(B) receptor. The epilepsy models have shown that GABA transporters play a receptors are found widely throughout the brain and are significant role in regulating brain functions including located both pre- and post- as well as extra-synaptically, cognition, excitability, pain and analgesia. providing extensive opportunity to alter neural activity. Four transporters have been cloned (GAT 1–4), their Modulation of GABA receptors can cause either an increase cellular distribution partly worked out and their pharma- or a decrease in the amount of GABA, depending on the cological properties studied extensively [215]. An esti- relative amount of inhibitory activity induced [210, 211]. mated *20 % of GABA is taken up into astrocytes, The GABAergic system is involved in maintenance and indicating that astrocytic GABA uptake has a modulatory modulation of many important physiological processes, effect on GABA-ergic synapses. Indeed, pharmacological including sleep, pain, motor control, mechanisms of inhibition of astrocytic GABA uptake has been shown to anaesthesia, and mechanisms of anxiety [212]. Further, have anti-convulsant activity [216] while inhibition of dysfunction in the system has been implicated in the neuronal uptake has been shown to be pro-convulsant. It is pathological processes underlying epilepsy, anxiety, this modulation of activity that is most interesting about schizophrenia, as well as developmental and neurodegen- GABA transporters. Rather than simply controlling the erative disorders. duration of excitatory post-synaptic potentials there is evidence that transporters are involved in limiting overspill Transport and cross talk to surrounding neurons; transporters have been found in extrasynaptic membranes where they would GABA as a neuroactive substance must be removed rapidly not be involved in termination of synaptic transmission and thoroughly from the synaptic cleft in order to maintain [217] as well as on non-GABAergic neurons [218]. Fur- signal to noise in the synapse. The concentration gradient is ther, there is evidence for regulation of transporter activity large; ambient brain GABA levels are at the low end of by GABA [219]. Extrasynaptically located transporters detection for MRS, being *1mM[213] intracellularly and may operate in reverse depending on the membrane *2 lM extracellularly [214]. GABA neurotransmission is potential [220–222] allowing release of GABA which may terminated by high-affinity transport mediated by carriers then bind to high affinity extrasynaptic GABA receptors.

123 14 Neurochem Res (2014) 39:1–36

Interpretation GABAergic activity and temporal coordination of network oscillations [238, 239]. Overall GABA levels are therefore The recent more widespread use of GABA editing sequences better markers of GABAergic tone (i.e. levels of tonic is resulting in many discoveries about the relationship of inhibition) than of inhibitory activity per se [240, 241]. GABA concentration to brain function [223]. Levels of GABA are altered by motor learning [224] and relate to motor decision speed, with higher GABA associated with Glucose faster decisions; the authors interpreted this as meaning that higher GABA was related to less distractibility [225]. Sub- A highly-coupled spin system is one where the difference conscious motor control is also predicted by GABA levels in between the coupling constants and the resonance fre- the supplementary motor area [226]. Lower thalamic levels quencies for each spin system is relatively small, while a of GABA are associated with neuropathic pain; the authors second order spin system is one where the coupling con- have interpreted this as relating to lost inhibition of the stants cannot be resolved by simple inspection but need to perception of pain [227]. In the visual cortex, orientation be calculated. Glucose, as a polyhydroxylated sugar, has a discrimination is predicted by GABA levels and by gamma highly-coupled, second order spin system further compli- oscillation frequency [228]. In frontal lobe there is a genetic cated by the presence of glucose anomers, a and b. These relationship, with glutamate decarboxylase (GAD1) and are in chemical exchange, but have separate spin systems. catechol-o-methyltransferase (COMT) polymorphisms A resonance arising from strongly coupled protons of demonstrating significant interactions with GABA levels glucose is resolvable in human brain spectra B3T at [229] in the anterior cingulate cortex. The degree of negative 3.43 ppm [242] and other resonances of glucose can be BOLD response in the default network (anterior cingulate observed using editing techniques [243–245]. cortex) is also related to GABA levels [230]; this relationship Glucose is the mandatory, major exogenous fuel for the is also seen in the visual cortex [231]. brain. It serves as the substrate for oxidative metabolism, as Studies of drugs which alter the activity of the GAB- well as providing carbon backbone for other synthetic Aergic system have shown a range of outcomes, with some processes. Levels of glucose are related to brain function; authors suggesting that uptake of GABA, for example, is decrease of blood glucose below *4 mM in humans energy neutral and unlikely to affect fMRI or PET signals results in a graded decrease in brain cellular function and [232]. Some argue that GABAergic activity decreases fMRI integrity, culminating in coma and death [246]. The abso- signal while others argue that increased inhibition is lute requirement of the brain for glucose is most likely due accompanied by increased membrane permeability, which to the requirement for cytosolic (glycolytic) ATP by the requires increased pumping of ions and thus is energy-con- Na?/K?ATPase [247, 248]. suming [233–236]. Indeed, on theoretical grounds, GABA- ergic input, when considered in the context of brain metab- Glucose Transport olism may, depending on current local circumstances, either increase or decrease regional energy consumption [237]. Glucose crosses the blood–brain barrier mostly via a non- Activity at GABA(B) receptors can indeed produce increa- insulin dependent glucose transporter (GLUT1), whose ses or decreases in metabolic activity, with agonists and kinetic properties make it sensitive to physiological chan- antagonists capable of producing both increased or ges in blood glucose concentration [249]. GLUT1 is the decreased Krebs cycle activity [211]. This is because the end primary transporter expressed in astrocytes [250] and is outcome of GABAergic activity is a sum of inhibitory and also expressed in neurons. Local glucose use is tightly excitatory activity across all the cells affected by the GAB- coupled to GLUT1 density both anatomically [251] and Aergic modulation. It is this summed activity which one developmentally [252]. Glucose may also leave the brain would be measuring with 1H MRS. Levels of GABA itself via GLUT1, particularly under hyperglycaemic conditions have little to do with excitatory activity and are mostly [253]. modified by GABAergic activity, with increases in levels of GLUT3 is a glucose transporter which is specific to GABA correlating with increased engagement of the GAB- neurons [249, 254]. GLUT1 and GLUT3 are subject to Aergic compartment (i.e. that fraction of cells which are trans-acceleration (whereby transport into a cell is more GABAergic) [208]. Researchers measuring GABA levels probable when the transporter can simultaneously bind and should therefore be aware that increased GABA levels do not transport a molecule out of the cell) and are asymmetric, so necessarily mean that there is increased inhibition. that the kinetics of transport are greatly different depending The level of metabolic GABA relates to levels of upon the direction of transport. Expression of GLUT3 by ambient extracellular GABA via reversal of local GABA neurons provides an advantage over GLUT1; the rate of transporters [222]. This contributes to local tonic uptake (zero-trans) by GLUT3 is sevenfold faster than 123 Neurochem Res (2014) 39:1–36 15

GLUT1 [255], so GLUT3 performs well when local protons are particularly affected, each being coupled to extracellular glucose is low (e.g. under high demand after four other protons in the molecule. The cCH2 resonances functional activation). are the cluster used in MRS in vivo although it should be The distribution of glucose in the brain was measured as understood that these resonance patterns are also heavily approximately even in the extra- and intracellular space second order. Further, at 3T, they are not resolved from the - [256–258] indicating that the signal seen in vivo is largely CH2 resonances of GABA (CH2–COO ) or, in particular, intracellular, since this compartment has been measured in the cCH2 of glutamine. Judicious choice of echo time [275, a range of animals to represent *80 % of the volume 276], or use of an asymmetric PRESS sequence [277] can [259]. Concentrations measured in white matter are con- help to weight the spectrum in favour of glutamate, and sistently higher by 15–20 % than in grey matter [260]. The acquisition of 2D spectra can also be of assistance [278, resonance of glucose is decreased by functional activation 279]. When single echo single voxel spectra are used, [261] suggesting that the rate of use of glucose is faster glutamate can be estimated using fitting approaches such as than the rate of transport/diffusion under functional con- LC Model [280] or jMRUI [281] although careful attention ditions [262]. Repeated studies have shown that glucose should be paid to the reliability and accuracy of the esti- transport rates are high enough to sustain large increases in mates [282]. glucose use; the possibility remains of a short (s) lag before Glutamate is the most abundant free amino acid in the the full activation of reactive hyperaemia and fuel delivery CNS and the major excitatory neurotransmitter [283, 284]. [263]. It is synthesized in a number of reactions; including by Glial cells can store glucose polymerised as glycogen glutamate dehydrogenase in an NAD? requiring reaction [264]. Astrocytic glycogen can be generated endogenously from 2-oxoglutarate (a-ketoglutarate), a Krebs cycle from exogenous glucose, or from gluconeogenic substrates intermediate, and by relatively fast [285] transamination [265, 266]. Such stores are small compared to the brain’s reactions where a-ketoglutarate and glutamate react with glucose requirements and can only supply brain energy other amino-acid and ketoacid pairs, such as alanine and requirements for a few minutes. Turnover of brain glyco- pyruvate, aspartate and oxaloacetate, leucine and ketoi- gen stores is dynamic and related to both blood glucose socaproate [286]. Glutamate levels are therefore sensitive concentration and brain activity [267, 268] suggesting that to ammonia levels (via glutamate dehydrogenase), and the these energy stores are involved in maintaining fuel supply low concentration of 2-oxoglutarate also makes this mol- during times of functional activation. Glial glycogen has ecule a point of metabolic control (Fig. 7). been shown to be essential for maintaining K? homeostasis Glutamate exists in several metabolic pools in the brain in astrocytes [269]. Glial cells do not possess glucose with differing turnover rates [287, 288]. It is not useful to 6-phosphatase [270], which means that any catabolized classify glutamate as ‘‘neurotransmitter’’ or ‘‘metabolic’’ glycogen must be exported as pyruvate, lactate or a similar glutamate. The relative amount that is neurotransmitter metabolite (Krebs cycle intermediate or amino acid), and glutamate varies depending on the cell type and location in not as glucose. Glial glycogen is therefore held to be the brain with estimates ranging from[45 % in some areas incapable of serving as a glucose supply for neurons. to almost insignificant in others (e.g. substantia nigra). It Neurons can also make glycogen under circumstances appears that almost any pool of glutamate is available as a when the normal machinery for suppressing synthesis is source of glutamate for neurotransmission, and that syn- dysfunctional [271]. thesis of neurotransmitter is not rate limiting in any case Levels of glucose in the brain are therefore most likely [284, 289]. related to blood glucose levels and secondarily to the rate Glutamate has four major subtypes of neurotransmitter of brain glucose use. receptor; three ionotropic receptors named after the compounds originally used selectively to activate them, (N- methyl-D-aspartate (NMDA), AMPA and Kainate) and Glutamate metabotropic receptors (mGluR) [290, 291]. The concentration of glutamate is relatively high within cells (milli- Glutamate, (2-aminopentanedioic acid, Glu), is an a-amino molar), and comparatively low extracellularly (low acid which is present in brain in relatively high (\12 mM) micromolar). This is to maintain the signal to noise gra- concentrations [272]. It contains two CH2 moieties in dient for neurotransmission. Glutamate removal from the addition to the aCH [273]. Each CH2 is pro-chiral due to extracellular space is mediated by high affinity active the presence of the stereogenic (chiral) centre in the amino transport systems which are located on both neurons and acid backbone [274]. Therefore, each proton is chemically glia [292] as well as by a cystine/glutamate antiporter non-equivalent; the result of which is that the resonance [293]. In alcoholics, downregulation of high affinity spectrum of glutamate is highly second order. The bCH2 glutamate transporters can lead to increased uptake of 123 16 Neurochem Res (2014) 39:1–36

being related to increased metabolic activity, although the reasons for this may be manifold.

Glutamine

Glutamine (2-amino-4-carbamoylbutanoic acid) is an amino acid present in relatively high concentrations in the brain and cerebrospinal fluid. Values in the order of 4– 6 mM have been reported in freshly biopsied human brain tissue [302] with 1 to 4 mM reported using magnetic resonance spectroscopy [303–305] The lower levels reported using MRS may be due to a short T2 compartment of glutamine [305]. Like glutamate, glutamine exhibits a Fig. 7 Metabolic pathways involving glutamate and glutamine and strongly coupled second order spin system, with significant their role in nitrogen metabolism. Glutamate is a substrate for a range overlap with the signals from glutamate as well as those of fast, near-equipibrium transaminase reactions. Shown here are the reactions catalysed by alanine aminotransferase (Ala-AT) and aspar- from the aspartyl moiety of NAA (Fig. 1). Brain glutamine tate aminotransferase (Asp-AT) but similar reactions involving is, consequently, a significant challenge to measure accu- branched chain amino acids such as leucine also exist in the brain. rately in vivo [305], especially with single-shot acquisition This latter reaction has been estimated to provide up to 30 % of methods although averaging methods offer some glutamate nitrogen [286]. Glutamate is also involved in nitrogen regulation via glutamate dehydrogenase (Glu deH) where glutamate is improvement [306]. The proportion of the c-CH2 glutamate converted to 2-oxoglutarate (a-ketoglutarate), an intermediate in the and glutamine resonances can be shifted using an asym- Krebs cycle and the substrate of a highly regulated enzyme a- metric PRESS sequence [277] giving smaller uncertainties ketoglutarate-dehydrogenase. This proximity to the Krebs cycle in the measurement of glutamine [307]. makes glutamate levels a good marker for Krebs cycle activity. Glutamine is produced from glutamate in astrocytes via glutamine Glutamine is synthesized from glutamate in a reaction synthetase [308] and converted to glutamate in neurons via a requiring ATP and catalyzed by glutamine synthetase (see phosphate activated glutaminase (PAG) Fig. 7), an enzyme localized exclusively in glia [308] although it may also be expressed in neurons under certain glutamate by the cystine/glutamate antiporter with con- pathological conditions [309]. Glutamine synthetase is the comitant reduction in cystine uptake. This can then result major user of free ammonia in the brain; brain levels of in decreased intracellular levels of the redox tripeptide glutamine are thus sensitive to blood ammonia levels [310]. glutathione (L-c-glutamyl-L-cysteinylglycine; GSH) as well The main source of carbon backbone for synthesis of as increased excitatory activity resulting from higher than GABA is glutamine [202, 311]. It plays a major role in the usual glutamate concentrations in the synapse [294]. The brain as a by-product of glutamate neurotransmission. increased glutamate observed in the brain of drinkers likely Glutamate released by neurons is taken up by astrocytes, results from increased metabolic activity rather than being converted to glutamine and released where it can be taken increased extracellular glutamate. Glutathione (see below) up by neurons and deaminated to form glutamate once may serve as a ‘‘sink’’ for brain glutamate as the two pools more. Estimates of neuronal glutamate derived from glu- are in intimate exchange [295]. tamine are in the order of *60–70 % [312, 313]. The Glutamate can be neurotoxic, mostly through its actions majority of glutamine in the brain is less metabolically at N-methyl-D-aspartate (NMDA) receptors where it can active, a ‘‘large, slow-turnover pool’’ which is probably cause mitochondrial apoptosis through the release of Ca2? located in neurons [201], with another smaller pool in glia [296]. which can be supplied by acetate [314]. Glutamine enters Glutamate has also been shown to act as a substrate and leaves cells via a range of amino acid exchangers [315, under a range of circumstances [297] although levels of it 316]. increase under functional situations such as visual stimu- The difficulties associated with quantifying glutamine lation [298–300]. Changes in glutamate levels are most by MRS at clinical field strengths mean there are little data closely linked to metabolic activity (Krebs cycle flux) with available on how this metabolite relates to brain function. increased excitatory activity relating to increased glutamate From experiments in vitro we know that total cortical levels in a roughly linear fashion [241, 301]. There is large glutamate and glutamine are highly correlated in healthy capacity in the system with dynamic concentration changes tissue. Glutamine is decreased in anhedonic major of several fold being possible. In normal brain it is there- depression with normal glutamate and GABA levels [317]. fore acceptable to interpret increased glutamate levels as The authors interpreted this as due to altered glutamatergic 123 Neurochem Res (2014) 39:1–36 17 metabolism. Others have shown increases in glutamine extracts at high field ([14 T). Editing schemes have been in vivo following administration of ketamine, also inter- proposed including those targeting the cysteinyl moiety preted as increased glutamatergic activity [318]; it is dif- [324] or that from glycine [325]. Some have demonstrated ficult to know exactly what is meant by this. Increases have a reasonably robust measurement from non-edited short been reported following administration of topiramate, an echo spectra using fitting routines such as LCModel [326]. anti-epileptic [319] where the increase has been attributed Glutathione exists in the reduced form and as an oxi- to activity at the GABA(A) receptor, and also following dized disulphide, GSSG. In healthy brain the ratio of GSH/ administration of a range of drugs active at the benzodi- GSSG is over 100 [327] and GSSG is below NMR detec- azepine site of the GABA(A) receptor [320]. More work tion levels [328] but this changes in pathology [329]. GSH/ directly investigating glutamine is required for better GSSG is a useful indicator of cellular redox status [330]. understanding. Astrocytes generally have higher GSH concentrations than neurons [331], although levels in some neurons are Relative Concentrations and Cell Compartments comparable with glia [332–334]. Levels of GSH in white matter are also generally lower than in grey [335], and Observations have been made that glutamate levels are levels may be higher in females than in males [336]. generally higher in neurons while glutamine is higher in astrocytes [41, 321] raising the suggestion that the relative Synthesis and Transport concentrations of these metabolites could be used to draw histological assumptions about the underlying tissue. Glutathione is synthesized in brain as it is in other major Extreme care should be taken with this approach. A range organs via a two-step pathway (Fig. 8). Glutamate is of ratios for Glu/Gln has been identified, for example in the combined with cysteine via c-glutamylcysteine synthetase cerebellum, ranging from 2 for parallel fibres to 0.3 for to form c-glutamylcysteine, and this dipeptide combines glial rich areas but this work was done at the spatial res- with glycine in a reaction catalyzed by glutathione syn- olution of the electron and/or the light microscope [322]. thetase [337]. Both reactions require ATP. GSH is ‘‘recy- Measures made at much lower resolution by MRS neces- cled’’ via the c-glutamyl cycle [338] which also serves sarily reflect a wide range of many different cell popula- further to regulate levels of cysteine [339]. tions making statements about cell types based on MRS Synthesis of GSH depends on the availability of pre- overly simplistic as well as being a significant inverse cursors, most particularly cysteine. Cysteine is neurotoxic problem. Oligodendrocytes, often disregarded, also have at high concentrations [340] so the balance between sup- relatively high levels of both glutamate and glutamine [41, plying adequate cysteine for GSH synthesis while keeping 322]. There is significant, dynamic exchange of glutamate cysteine levels low is highly regulated. There are a number and glutamine between cell types meaning that changes in of possible sources of the cysteinyl thiol moiety (Fig. 8) neuronal glutamate due to increased astrocytic synthesis including cysteine itself, cystine, and methionine via [323] could easily, for example, be erroneously ascribed to transsulfuration [341, 342]. This latter pathway links underlying cell population changes. In many cases the homocysteine levels with those of GSH. The major source relative differences in glial and neuronal glutamate or of cysteine for GSH synthesis in astrocytes is from cystine, glutamine levels are small [321] and are subject to dynamic which enters astrocytes via the cystine/glutamate alterations potentially greater than the relative differences exchanger. and also may be subject to measurement uncertainties Astrocytes release GSH [343]; some of this serves as a which are in some cases bigger than the differences substrate for the exogenous astrocytic enzyme c-glutam- between cell types. Before it can be recommended, the yltranspeptidase, which releases c-glutamyl-aminoacid utility of this approach needs concurrent MRS and histo- conjugates and cys–gly [344]. Neurons also possess an logical verification, which is currently unavailable. exogenous amino peptidase which cleaves the cys–gly dipeptide to its constituent amino acids; these can then be taken up into neurons and used to synthesise GSH (Fig. 8). Glutathione Neuronal GSH levels are crucially supported by astrocytic supply of these precursors [345]. Glutathione is a tripeptide (L-c-glutamyl-L-cysteinylglycine; GSH) found ubiquitously throughout the body and in Role relatively high concentrations in the brain (2–3 mM) as well as in the extracellular and cerebrospinal fluid. It is GSH is generally the most prevalent intracellular thiol and problematic to measure in vivo due to significant resonance plays a major role in the reductive process essential to overlap with other metabolites; this remains true even in maintain life. In an organ as highly oxidative as the brain, 123 18 Neurochem Res (2014) 39:1–36

Fig. 8 Brain synthesis, compartmentation and metabolism of gluta- three of which are important in brain [464]. Extracellularly, GSH is thione. Glutathione (GSH) is synthesized from glutamate, cysteine cleaved by c-glutamyltranspeptidase, releasing cys-gly and c-glut- and glycine in a two-step pathway with both steps requiring ATP. amyl-aminoacid conjugates. These dipeptides are also cleaved by GSH is a negative feedback regulator of its own synthetic pathway. neuronal extracellular aminopeptidases and the free amino acids can GSH is converted to GSSG via redox reactions and restored to GSH be taken up by their respective neuronal amino acid exchangers. via GSSG reductase, which requires the cofactor NADPH. NADPH is Cysteine is taken up by the excitatory amino acid carrier EAAC1; generated by a range of dehydrogenase enzymes using NADP?/ cells deﬁcient in this carrier are also deﬁcient in GSH [465]. Only NADPH as cofactor, which includes enzymes of the pentose- *20 % of cysteine uptake is via the ASC transporter or system L phosphate pathway. GSSG can also leave the cell via a multidrug [466]. In astrocytes, cysteine may also be derived via transsulfuration resistance pump either as GSH, GSSG or as a mixed conjugated with the sulfur group provided by methionine, via S-adenosylmethi- disulphide. There are nine known multidrug resistance pumps at least onine, S-adenosylhomocysteine, homocysteine and cystathione

GSH is of prime importance. GSH is essential for the depletion results in mitochondrial dysfunction and synthesis and degradation of proteins and formation of the decreased NAA levels [347]. Glutathione levels decline deoxyribonucleotide precursors of DNA. It protects the cell with age but are generally elevated by mild stress, possibly against reactive oxygen compounds and conjugates with a preconditioning response which better positions the body foreign compounds such as drugs. It is also a cofactor in to manage further increases in stress levels [327]. several reactions, such as the glyoxalase pathway which is In summary, GSH levels are connected to mitochondrial essential for removal of highly reactive ketoaldehydes that function, levels of glutamate and cysteine availability. leak from the glycolytic pathway [346]. Glutathione They are linked to oxidative stress and also to age.

123 Neurochem Res (2014) 39:1–36 19 myo-Inositol 2. Major component of intracellular second messenger system. myo-Inositol is one of nine possible isomers of hexahydr- 3. An organic osmolyte, involved in maintenance of cell oxycyclohexane; it is by far the most prevalent isomer [348], volume. comprising 95 % of inositols in the human body. It is myo-inositol is an essential component of a post- acquired both by synthesis (mainly in kidney) and by receptor, second messenger signalling system found in ingestion. In the brain it is derived from three sources: (1) many cells. It has been critically linked to receptors in the receptor stimulation (a salvage pathway), (2) de novo syn- central nervous system, including the glutamatergic thesis from glucose, and (3), uptake of dietary myo-inositol (metabotropic), muscarinic, serotonergic, adrenergic and through plasma membrane myo-inositol transporters. histaminergic systems. Upon stimulation of these receptors, The main, resolved resonance of interest derived from guanosine triphosphate binding protein activates the myo-inositol is that at d = *3.56 ppm ([242, 349], for enzyme phospholipase C, acting upon membrane inositol more information see [1]). This resonance, at whole body phospholipids and releasing second messenger inositol MR field strengths, also contains contributions from the 1,4,5 triphosphate (IP ; which releases intracellular cal- amino acids glycine, valine and threonine [1]. It is not 3 cium) and diacylglycerol (DAG; which activates protein likely that significant contributions are made to the signal kinase C). under normal conditions from inositol 1-phosphate, A number of different inositol phospholipids have been although levels of this have been shown to increase sig- identified that use inositol monophosphate as a precursor nificantly after lithium administration. Use of non-standard [360]. myo-inositol can be phosphorylated in a number of sequences can improve myo-inositol detection at 3T [350]. positions, creating the second messenger inositol 1,4,5 In human brain, the concentration of myo-inositol varies triphosphate, and a variety of other possible second mes- from region to region [351, 352], ranging from 2 to senger candidates. Some of these have been isolated in 15 mmol/kg wet weight [353], averaging around 8 mmol/ various tissues in vivo but their roles have yet to be elu- kg wet weight [354], and does not correlate with age. cidated. It is possible that in disease states, one or more myo-Inositol has been proposed as a glial marker [145] inositol phospholipids is increased or decreased, causing but this is disputed. In cultured cells and immortal cell lines imbalance in the second messenger signaling system [360]. the level of myo-inositol in astrocytes is higher than in For all pathways in the inositol lipid cycle, the formation neurons. The astrocyte’s ability to accumulate myo-inositol of myo-inositol depends upon the hydrolysis of inositol is greater [355, 356] but some neuronal cell lines contain monophosphates; this reaction is catalysed by inositol high levels of myo-inositol [357]. Unlike N-acetylaspar- monophosphatase [361]. This enzyme is inhibited uncom- tate, where the synthetic enzyme is generally present only petitively by lithium ion, meaning that inhibition of inositol in one cell type, there is no evidence that inositol-associ- monophosphatase can occur selectively in abnormally ated anabolic or catabolic activity is confined solely to one stimulated brain cells [360]. Inositol monophosphatase cell type; indeed synthetic activity is mainly confined to the activity varies considerably between individuals (up to a vasculature [358]. Uptake of myo-inositol could be stim- full order of magnitude [354]) and also increases with age. ulated by differentiation [357]. Further, since inositol myo-inositol may have modulating effects on certain uptake and phosphoinositide turnover are highly variable enzymes, receptors or transcription factors in addition to depending upon the region of the brain under investigation, second-messenger-linked effects. For example, phosphati- it follows that the regional distribution of myo-inositol is dyl inositol has potent effects upon tyrosine hydroxylase, similarly highly variable [357]. Indeed the concentration of an important controlling enzyme in the biosynthesis of myo-inositol is higher in grey (e.g. 4.9 ± 0.5) than white catecholamines, dopamine and noradrenaline and is (3.1 ± 0.4) matter [140] (but see also [141]) and shows a involved in anchoring acetylcholinesterase [362]. Inositol regional (and possibly gender) variation [141]. There is phosphate and DAG metabolism are linked to Na?/ also little correlation between inositol concentrations and K?ATPase activity [363] and activity of the insulin levels of glial-fibrillary acidic protein, a glial cell marker receptor [364]. [359]. Concentrations measured by MRS are highest in myo-inositol administration alters behaviour/mood (and cerebellum [141] and pons [351] but others have reported even makes goldfish more active! [365]), although the high concentrations in the frontal and temporal lobes [353]. precise mechanism by which this occurs remains unclear. myo-Inositol has a number of known roles; some of Dietary supplementation with inositol has been used to which are interlinked. treat a range of psychiatric and behavioral conditions [360, 1. Constituent of phosphoglycerides (phosphatidylinosi- 366] with increased inositol serving to enhance motor tol) and hence lipid component of biomembranes. activity, reduce depression and the frequency and severity

123 20 Neurochem Res (2014) 39:1–36 of panic attacks, but enhance hyperactivity in children with myo-inositol is therefore involved with the functions of attention deficit disorder. brain cell signaling, and/or volume regulation. Each func- myo-inositol acts as an organic osmolyte in the brain. tion draws upon separate pools of responsive myo-inositol Whilst it is well accepted that exogenous myo-inositol [389]. The blanket assumption that myo-inositol is a glial plays a role in the response to osmotic shock [367], and marker is an over-generalisation. that levels of myo-inositol in the brain respond to chronic application of osmotic stress [368, 369], endogenous myo- inositol is probably not extruded in response to acute Scyllo-Inositol hypotonic shock [49]. Elevation of myo-inositol has been reported in conditions where altered intracranial pressure Present in the brain at 5–12 % of the concentration of myo- may ensue, such as stroke [370] and head injury [371]. It inositol, scyllo-inositol can also be detected by MRS, with also changes dynamically (min) in response to direct cur- a resonance at 3.35 ppm [390]. scyllo-Inositol is formed rent stimulation; this was attributed either to osmotic from myo-inositol via an epimerase [391] although the alteration or altered membrane turnover [372]. amount of brain scyllo-inositol actually derived from this myo-inositol crosses the blood brain barrier both by source remains unclear; some report strong correlations simple diffusion and via a sterospecific, saturable transport between levels of myo- and scyllo-inositol, while others system [373] although large doses are generally required report no or little relationship between the two [392, 393]. for effective uptake [374]. Penetration of exogenous ino- There is a strong relationship between blood and CSF sitol varies across brain regions; it may not penetrate the levels of scyllo-inositol and brain levels. Brain levels also cerebellum at all [375]. myo-inositol is transported into respond well to dietary supplementation [394]. scyllo- cells against a concentration gradient by the Na?/myo- inositol is a substrate for the sodium/myo-inositol co- inositol cotransporters (SMIT1 (SLC5A3) and SMIT2 transporters (SMITs, see above). Brain levels have been (SLC5A13)). Expression of SMIT1 is required for normal shown to be highly elevated in alcoholic encephalopathy brain development [376]. In fetal brain the expression of along with elevated CSF and plasma levels. Glucose-6- SMIT1 is high. It is gradually down-regulated as the brain phosphate is converted to myo-inositol-1-phosphate, then matures [377], corresponding to the developmental change myo-inositol, which is then converted to scyllo-inositol. in intracellular myo-inositol concentrations. Levels of Therefore, the authors suggested that decreased glycolytic myo-inositol are higher in fetuses [378] and in children flux would allow more glucose-6-phosphate to be con- [242] than in adults. This may be due to the relative verted to scyllo-inositol, although they did not report a immaturity of the blood brain barrier and the consequent corresponding increase in myo-inositol [392]. myo-Inositol requirement for tighter volume regulation. Little is known administration does not cause a corresponding increase in about SMIT2 although it is expressed in the brain [379] and scyllo-inositol levels [395]. is regulated by osmolarity [380]. A H?-myo-inositol co- scyllo-Inositol and myo-inositol have both been shown transporter (HMIT) has been described which is highly to have anticonvulsant properties [396]. scyllo-Inositol is expressed in the brain [381]. It may regulate neuronal of particular interest due to its ability to decrease accu- phosphoinositide synthesis in times of high activity [382] mulation of amyloid-beta protein [388] making it a possi- and may account for the heterogeneity of brain myo-ino- ble therapeutic in Alzheimer’s disease. sitol levels. Dietary supplementation with myo-inositol results in increased brain levels of myo-inositol within 4 days, but Lactate levels return to baseline after 8 days of supplementation [383]. This is most likely due to downregulation of the Lactic acid (2-hydroxypropanoic acid), a three carbon number of SMIT, although with the caveat that transporter hydroxyacid, contains a methyl group which gives rise to a 3 expression in response to increased myo-inositol levels doublet resonance ( JHH = 6.93 Hz) at *1.3 ppm. Lactic varies depending upon the brain region examined [384]. acid exists mainly in anionic form (lactate) in vivo; the pKa SMIT1 may play a role in the elevation of myo-inositol is 3.86. seen in Down syndrome [385] as the gene for SMIT1 is located on chromosome 21 [386]. myo-Inositol may also be Synthesis and Catabolism of Lactate involved in the development of pathology seen in older Down syndrome subjects and Alzheimer disease, as well as Lactate is synthesized from pyruvate by the enzyme lactate in head injury [387], as evidence emerges of the role of dehydrogenase. The equilibrium of this reaction lies in myo-inositol in stabilising beta-amyloid protein complexes favor of lactate, by *20:1. In vivo, the mass action ratio of [388]. the reaction participants shifts the steady-state equilibrium 123 Neurochem Res (2014) 39:1–36 21 such that in some cell types, especially those with strongly rather than as an entity with its own deterministic strategy. aerobic requirements, the mass action ratio of the reaction Since all of the main reactions involving lactate (lactate lies mildly in favor of pyruvate (*12:8). There are a dehydrogenase, lactate transport) are reasonably fast in number of different isoforms of lactate dehydrogenase relative terms, metabolic fluxes can be argued to support [397], whose different mass action ratios reflect the lactate- the hypothesis. Cerebral activation by induction of a producing requirements of the cells in which they are mental work load, results in a transient oxygen and glucose contained. The lactate dehydrogenase reaction is a rela- [416] debt which may produce a transient ‘‘uncoupling’’ of tively fast one in addition to being near-equilibrium [398]. glycolysis and the Krebs cycle [417–420] with inhibition of This means that lactate can be both quickly produced and the malate-aspartate shuttle [421]. This creates an oppor- quickly catabolised, but that the reaction can play only a tunity for lactate levels to increase [151, 418]. They are very limited role in flux regulation, and may only do this in further increased by breakdown of glial glycogen as glia any significant way during fast transitions in energy are stimulated by increased extracellular glutamate levels metabolism [399]. and lagging glucose delivery. When lactate levels exceed falling local extracellular glucose concentrations and rise Transport of Lactate sufficiently high to pass in volume through the relatively low affinity monocarboxylate transporter, lactate may then Lactate enters and leaves cells via non-ionic diffusion (e.g. become the more used fuel. However, calculations have *5 % of lactate movement in red blood cells is via this suggested that lactate is responsible for replacing at most route) as well as via proton symport by the mono-carbox- 25 % of brain energy demands normally fulfilled by glu- ylate transporter (MCT) [400], and sodium-dependent cose [422]. Lactate diffuses rapidly out of cells and away lactate transporters (SLC5A8; SMCT1[401] and SLC5A12, from the area of metabolic activation, possibly through SMCT2 [402]). There are a number of different isoforms of gap-junction mediated means [423, 424]. A not inconsid- the MCT, with different affinities for lactate [403, 404] and erable amount of it may also leave the brain, depending on their expression in different cell types can somewhat the circumstances under which it is produced [425], via account for differing speeds and affinities of lactate uptake leakage into sinuses and perivascular spaces [426]. Others by neurons and glia [403, 405, 406]. Pyruvate is trans- have also called into question whether the situation in ported in and out of the mitochondrion, via the mito- cultured cells applies in vivo [427]. Lactate, and the related chondrial pyruvate carrier [407, 408] which shows little monocarboxylate, pyruvate, increase the oxygen con- affinity for lactate [409, 410]. The sodium-dependent lac- sumption rate of mitochondria through an unknown tate transporter SMCT1 has widespread neuronal expres- mechanism [428] and, in high enough concentrations, can sion and broad substrate affinity, including ketone bodies cause convulsions [429]. This increased metabolic rate also [404]. SMCT2 has been reported to be expressed in ought to be taken into consideration when comparing lac- astrocytes [411] but there is little information yet available tate and glucose use in the brain. about the physiological role of either of these transporters. The metabolic pool of lactate in normal brain is thus Lactate crosses the blood brain barrier via the MCT. best considered as an entity with a fast, dynamic turnover Uptake of lactate into the brain is considerably more (Fig. 9), whose flux levels relate to pyruvate clearance important in the fetal or newborn brain [412], where the rates, cellular redox potential and lactate efflux rates. transporter is more highly expressed in the blood brain Although it has been reported that substrate lactate can barrier, and where intrinsic blood lactate levels are higher, support neural activity [430], and is obligatory after pro- than in adults [413]. Lactate decreases in importance as an longed oxygen deprivation [431], this result was not found exogenous fuel as synaptic activity (and cerebral glucose by others [432]. Indeed, it has been suggested to be an use) increases and blood-brain barrier lactate transport artefact of the tissue preparation method [433] or of the fact decreases [414]. that most inhibitors used to block plasma membrane lactate There is still considerable argument about the role of uptake also act to block pyruvate uptake into the mito- lactate in the brain. Cell culture studies have shown lactate chondrion [410, 434]. being produced in reasonable quantities by astrocytes, which have been suggested [415] to ‘‘predigest’’ glucose The Lactate Resonance (or glycogen stores) and to pass it to neurons for consumption following neuronal release of glutamate; the The resonance of lactate is not generally observed in ‘‘astrocyte–neuron lactate shuttle hypothesis’’. This conventional MR spectra of normal brain, but has been hypothesis has proven very seductive but, as with all noted under conditions of functional activation [151], or examples of post-modernist biochemistry, lactate should conditions where the flow of substrates to and from the properly be viewed as part of the thermodynamic system brain is altered, such as hyperventilation [435], where 123 22 Neurochem Res (2014) 39:1–36

Fig. 9 Lactate metabolism and transport in the brain. Lactate is brain via the circulatory system, or the perivascular space. It may also produced by lactate dehydrogenase from pyruvate in a rapid, near- rapidly be trafficked between cells via gap junctions. In summary, equilibrium reaction. Lactate may then be converted back to pyruvate these numerous processes are rapid and act to disperse lactate. The or leave the cell via an array of passive (exchange equilibrium) one possible exemption is that lactate may be accumulated via the transporters (monocarboxylate transporters, MCTs). The relative sodium-dependent lactate transporter. It is not currently known what affinities of these transporters for the different monocarboxylates relative effect this transporter may have. A collation of literature KM are shown next to the respective transporter. It may then leave the for the various transporters may be found in [467] increased brain pH accelerates glycolysis, and decreased which is common to small molecules with carboxylic acids cerebral blood flow putatively decreases lactate clearance groups [444, 445] and, in chronic hypoxia, by production [436]. Lactate has been detected in brain following vigor- of lactate by infiltrating leukocytes [446] and at long echo ous exercise [437] although fast metabolism may limit the times by changes in the relaxation properties due to altered amount accumulated [438], and following lactate infusion lactate compartmentation [447]. [439]. Extracellular lactate may also be a useful sleep The resonance of lactate is also of interest in functional indicator [440]. Lactate can reliably be observed in normal MRS, where production of lactate has been observed in brain using an editing sequence such as MEGA–PRESS disorders such as migraine with aura [448] and dyslexia [191, 441], although the signal to noise and resolution of [449]. In spreading depression both free glucose and oxy- this approach is less than optimal considering the time- gen supplies are transiently decreased [416] which would scale upon which lactate levels may alter. Lactate is probe expected to result in increased lactate. Under these duced rapidly in response to hypoxia/ischaemia on a sec- conditions, brain lactate leaves the brain and enter the onds timescale. It is not a marker of severe ischaemia due blood [425]. to the proximity of lactate dehydrogenase to equilibrium. The mechanism for lactate production in dyslexic brain Alanine or glycerol-3-phosphate are probably better linear is less transparent but potentially instructive. Reading markers for the degree of hypoxia [442, 443]. In addition, remediation results in undetectable lactate under the same observation and quantification of the lactate resonance is stimulus [450]. fMRS-detectable production of lactate complicated by the phenomenon of ‘‘NMR invisibility’’ tends to occur mainly under conditions that impose a

123 Neurochem Res (2014) 39:1–36 23 relatively high workload. It may be that, in dyslexia, neu- 4. Kreis R (2004) Issues of spectral quality in clinical H-1-mag- ronal organization and blood supply in affected areas are netic resonance spectroscopy and a gallery of artifacts. NMR Biomed 17:361–381 such that metabolic conditions are suboptimal. Remedia- 5. Mountford CE, Stanwell P, Lin A, Ramadan S, Ross B (2010) tion, which would act to strengthen and improve the effi- Neurospectroscopy: the past, present and future. Chem Rev cacy of synaptic connections and improve vascularization 110:3060–3086 [451], would produce conditions less favorable to lactate 6. Choi CH, Ghose S, Uh J, Pate A, Dimitrov IE, Lu HZ, Douglas D, Ganjil S (2010) Measurement of N-acetylaspartylglutamate production and more favorable to lactate clearance. in the human frontal brain by (1)H-MRS at 7 T. Magn Reson The bottom line for the presence of lactate in the brain Med 64:1247–1251 remains the pyruvate clearance rate. Lactate will increase if 7. Frahm J, Michaelis T, Merboldt K-D, Ha¨nicke W, Gyngell ML, pyruvate clearance is slowed. Whether these increased Bruhn H (1991) On the N-acetyl methyl resonance in localised 1 h NMR spectra of human brain in vivo. NMR Biomed 4:201– lactate levels are detectable by MRS is then dependent on 204 factors such as the local vascularisation/perfusion of the 8. Pouwels PJW, Frahm J (1997) Differential distribution of NAA tissue and the lactate compartmentation. and NAAG in human brain as determined by quantitative localised proton MRS. NMR Biomed 10:73–79 9. Brooks JCW, Roberts N, Kemp GJ, Gosney MA, Lye M, Interpretation of Spectra Whitehouse GH (2001) A proton magnetic resonance spectroscopy study of age-related changes in frontal lobe metabolite What should have become apparent from the information concentrations. Cereb Cortex 11:598–605 provided in this review is the complexity of the biochemical 10. Maudsley AA, Domenig C, Govind V, Darkazanli A, Studholme C, Bloomer C (2009) Mapping of brain metabolite distributions interrelationships that underlie changes in brain metabolites. by volumetric proton MR spectroscopic imaging (MRSI). Magn Spectra need to be interpreted using information about the Reson Med 61:548–559 status of the subject as there are multiple possible causes for 11. Niwa M, Nitta A, Mizoguchi H, Ito Y, Noda Y, Nagai T, changes in metabolites (Table 1). MRS by itself is suffi- Nabeshima T (2007) A novel molecule ‘‘Shati’’ is involved in methamphetamine-induced hyperlocomotion, sensitization, and ciently diagnostic in only a handful of cases, such as Cana- conditioned place preference. J Neurosci 27:7604–7615 van’s disease, creatine synthesis or transporter deficiencies. 12. Wiame E, Tyteca D, Pierrot N, Collard F, Amyere M, Noel G, It can be of use in other disorders when considered within Desmedt J, Nassogne MC, Vikkula M, Octave JN, Vincent MF, context or as part of a wider ranging examination. Improved Courtoy PJ, Boltshauser E, Van Schaftingen E (2010) Molecular identification of aspartate N-acetyltransferase and its mutation in analyses, which take into account all the changes in a spec- hypoacetylaspartia. Biochem J 425:127–136 trum, such as the multivariate methods used in metabolomics 13. Patel TB, Clark JB (1979) Synthesis of N-acetyl-L-aspartate by that consider overall metabolic profiles, rather than indi- rat brain mitochondria and its involvement in mitochondrial vidual metabolites in isolation [452], will prove useful in the cytosolic carbon transport. Biochem J 184:539–546 14. Ariyannur PS, Moffett JR, Manickam P, Pattabiraman N, Arun P, future as will fast quantitative methods that can be used Nitta A, Nabeshima T, Madhavarao CN, Namboodiri AMA routinely in a clinical setting [453, 454]. A thorough (2010) Methamphetamine-induced neuronal protein NAT8L is understanding of biochemical processes underlying meta- the NAA biosynthetic enzyme: implications for specialized acetyl bolic change will remain essential. coenzyme A metabolism in the CNS. Brain Res 1335:1–13 15. Bhakoo KK, Pearce D (2000) In vitro expression of N-acetylaspartate by oligodendrocytes: implications for proton magnetic Acknowledgments This work was supported by the National resonance spectroscopy signal in vivo. J Neurochem 74:254–262 Health and Medical Research Council of Australia (Fellowship to 16. Tyson RL, Sutherland GR (1998) Labelling of N-acetylaspartate CR). The author is grateful to Prof. Stephen R. Williams (The Uni- and N-acetylaspartylglutamate in rat neocortex, hippocampus versity of Manchester), to Assoc. Prof. Matthias Klugmann (The and cerebellum from [1-13C]glucose. Neurosci Lett 251:181– University of NSW), to Prof. Stefan Broër (The Australian National 184 University) and to Prof. John Griffiths (Cancer Research, UK) for 17. Neale JH, Bzdega T, Wroblewska B (2000) N-acetylaspartylg- critical appraisal of the manuscript. lutamate: the most abundant peptide neurotransmitter in the mammalian central nervous system. J Neurochem 75:443–452 18. Klugmann M, Symes CW, Klaussner BK, Leichtlein CB, Serikawa T, Young D, During MJ (2003) Identification and References distribution of aspartoacylase in the postnatal rat brain. Neuro Report 14:1837–1840 1. Govindaraju V, Young K, Maudsley AA, Maudsley AA (2000) 19. Moffett JR, Arun P, Ariyannur PS, Garbern JY, Jacobowitz DM, Proton NMR chemical shifts and coupling constant for brain Namboodiri AMA (2011) Extensive aspartoacylase expression metabolites. NMR Biomed 13:129–153 in the rat central nervous system. Glia 59:1414–1434 2. Kreis R, Bolliger CS (2012) The need for updates of spin system 20. Chakraborty G, Mekala P, Yahya D, Wu G, Ledeen RW (2001) parameters, illustrated for the case of c-aminobutyric acid. NMR Intraneuronal N-acetylaspartate supplies acetyl groups for Biomed 25:1401–1403 myelin lipid synthesis: evidence for myelin-associated aspar- 3. Tosi MR, Fini G, Tinti A, Reggiani A, Tugnoli V (2002) toacylase. J Neurochem 78:736–745 Molecular characterisation of human healthy and neoplastic 21. Dadamo AF, Yatsu FM (1966) Acetate metabolism in nervous cerebral and renal tissues by in vitro H-1NMR spectroscopy. Int system. N-acetyl-L-aspartic acid and biosynthesis of brain lipids. J Mol Med 9:299–310 J Neurochem 13:961–1000

123 24 Neurochem Res (2014) 39:1–36

22. Madhavarao CN, Arun P, Moffett JR, Szucs S, Surendran S, 38. Gurkoff GG, Feng JF, Van KC, Izadi A, Ghiasvand R, Shahlaie Matalon R, Garbern J, Hristova D, Johnson A, Jiang W, Nam- K, Song M, Lowe DA, Zhou J, Lyeth BG (2013) NAAG pep- boodiri MAA (2005) Defective N-acetylaspartate catabolism tidase inhibitor improves motor function and reduces cognitive reduces brain acetate levels and myelin lipid synthesis in Can- dysfunction in a model of TBI with secondary hypoxia. Brain avan’s disease. Proc Natl Acad Sci USA 102:5221–5226 Res 1515:98–107 23. Hershfield JR, Madhavarao CN, Moffett JR, Benjamins JA, 39. Nadler JV, Cooper JR (1972) N-Acetyl-L-aspartic acid content Garbern JY, Namboodiri A (2006) Aspartoacylase is a regulated of human neural tumours and bovine peripheral nervous tissues. nuclear-cytoplasmic enzyme. FASEB J 20:2139–2141 J Neurochem 19:313–319 24. Kumar S, Biancotti JC, Matalon R, de Vellis J (2009) Lack of 40. Urenjak J, Williams SR, Gadian DG, Noble M (1992) Specific aspartoacylase activity disrupts survival and differentiation of expression of N-acetylaspartate in neurons, oligodendrocyte neural progenitors and oligodendrocytes in a mouse model of type-2 progenitors and immature oligodendrocytes in vitro. J canavan disease. J Neurosci Res 87:3415–3427 Neurochem 59:55–61 25. Francis JS, Strande L, Pu A, Leone P (2011) Endogenous as- 41. Urenjak J, Williams SR, Gadian DG, Noble M (1993) Proton partoacylase expression is responsive to glutamatergic activity nuclear magnetic resonance spectroscopy unambiguously iden- in vitro and in vivo. Glia 59:1435–1446 tifies different neural types. J Neurosci 13:981–989 26. Ariyannur PS, Moffett JR, Madhavarao CN, Arun P, Vishnu N, 42. Demougeot C, Garnier P, Mossiat C, Bertrand N, Giroud M, Jacobowitz DM, Hallows WC, Denu JM, Namboodiri AMA Beley A, Marie C (2001) N-acetylaspartate, a marker of both (2010) Nuclear-cytoplasmic localization of acetyl coenzyme a cellular dysfunction and neuronal loss: its relevance to studies of synthetase-1 in the rat brain. J Comp Neurol 518:2952–2977 acute brain injury. J Neurochem 77:408–415 27. Francis JS, Strande L, Markov V, Leone P (2012) Aspartoacy- 43. Jenkins BG, Klivenyi P, Kustermann E, Andreassen OA, Ferr- lase supports oxidative energy metabolism during myelination. J ante RJ, Rosen BR, Beal MF (2000) Nonlinear decrease over Cereb Blood Flow Metab 32:1725–1736 time in N-acetyl aspartate levels in the absence of neuronal loss 28. Moreno A, Ross BD, Blu¨ml S (2001) Direct determination of the and increases in glutamine and glucose in transgenic Hunting- N-acetylaspartate synthesis rate in the human brain by 13C MRS ton’s disease mice. J Neurochem 74:2108–2119 and [1-13C]glucose infusion. J Neurochem 77:347–350 44. Rae C, Karmiloff-Smith A, Lee MA, Dixon RM, Grant J, Bla- 29. Bates TE, Strangwald M, Keelan J, Davey GP, Munro PM, Clarke mire AM, Thompson CH, Styles P, Radda GK (1998) Brain JB (1996) Inhibition of N-acetylaspartate production: implica- biochemistry in Williams syndrome: evidence for a role of the tions for 1H MRS studies in vivo. Neuro Rep 7:1397–1400 cerebellum in cognition? [see comments]. Neurology 51:33–40 30. Xu S, Yang J, Shen J (2008) Measuring N-acetylaspartate syn- 45. Cheng LL, Newell K, Mallory AE, Hyman BT, Gonzalez RG thesis in vivo using proton magnetic resonance spectroscopy. J (2002) Quantification of neurons in Alzheimer and control Neurosci Meth 172:8–12 brains with ex vivo high resolution magic angle spinning proton 31. Gomez R, Behar KL, Watzl J, Weinzimer SA, Gulanski B, San- magnetic resonance spectroscopy and stereology. Magn Reson acora G, Koretski J, Guidone E, Jiang L, Petrakis IL, Pittman B, Imag 20:527–533 Krystal JH, Mason GF (2012) Intravenous ethanol infusion 46. Cheng LL, Ma MJ, Becerra L, Ptak T, Tracey I, Lackney A, decreases human cortical c-aminobutyric acid and N-acetylas- Gonzalez RG (1997) Quantitative neuropathology by high res- partate as measured with proton magnetic resonance spectroscopy olution magic angle spinning proton magnetic resonance spec- at 4 Tesla. Biol Psychiatr 71:239–246 troscopy. Proc Nat Acad Sci USA 94:6408–6413 32. Sacha P, Zamecnik J, Barinka C, Hlouchova K, Vicha A, 47. Sager TN, Topp S, Torup L, Hanson LG, Egestad B, Moller A Mlcochova P, Hilgert I, Eckschlager T, Konvalinka J (2007) (2001) Evaluation of CA1 damage using single-voxel H-1-MRS Expression of glutamate carboxypeptidase II in human brain. and un-biased stereology: can non-invasive measures of N- Neuroscience 144:1361–1372 acetyl-aspartate following global ischemia be used as a reliable 33. Berger UV, Luthi-Carter R, Passani LA, Elkabes S, Black I, measure of neuronal damage? Brain Res 892:166–175 Konradi C, Coyle JT (1999) Glutamate carboxypeptidase II is 48. Baslow MH (2000) Functions of N-acetyl-L-aspartate and N- expressed by astrocytes in the adult rat nervous system. J Comp acetyl-L-aspartylglutamate in the vertebrate brain: role in glial Neurosci 415:52–64 cell-specific signalling. J Neurochem 75:453–459 34. Fujita T, Katsukawa H, Yodoya E, Wada M, Shimada A, Okada 49. Bothwell JH, Rae C, Dixon RM, Styles P, Bhakoo KK (2001) N, Yamamoto A, Ganapathy V (2005) Transport characteristics Hypo-osmotic swelling activated release of organic osmolytes in of N-acetyl-L-aspartate in rat astrocytes: involvement of sodium- brain slices—implications for brain oedema in vivo. J Neuro- coupled high-affinity carboxylate transporter NaC3/NaDC3- chem 77:1632–1640 mediated transport system. J Neurochem 93:706–714 50. Taylor DL, Davies SEC, Obrenovitch TP, Doheny MH, Patsalos 35. Yodoya E, Wada M, Shimada A, Katsukawa H, Okada N, Ya- PN, Clark JB, Symon L (1995) Investigation in the the role of N- mamoto A, Ganapathy V, Fujita T (2006) Functional and acetylaspartate in cerebral osmoregulation. J Neurochem molecular identification of sodium-coupled dicarboxylate 65:275–281 transporters in rat primary cultured cerebrocortical astrocytes 51. Baslow MH, Hrabe J, Guilfoyle DN (2007) Dynamic relation- and neurons. J Neurochem 97:162–173 ship between neurostimulation and N-acetylaspartate metabo- 36. Taylor DL, Davies SE, Obrenovitch TP, Urenjak J, Richards lism in the human visual cortex. J Mol Neurosci 32:235–245 DA, Clark JB, Symon L (1994) Extracellular N-acetylaspartate 52. Baslow MH (2010) Evidence that the tri-cellular metabolism of in the rat brain: in vivo determination of basal levels and N-acetylaspartate functions as the brain’s ‘‘operating system’’: changes evoked by high K?. J Neurochem 62:2349–2355 how NAA metabolism supports meaningful intercellular fre- 37. Rahn KA, Watkins CC, Alt J, Rais R, Stathis M, Grishkan I, quency-encoded communications. Amino Acid 39:1139–1145 Crainiceau CM, Pomper MG, Rojas C, Pletnikov MV, Calabresi 53. Mangia S, Tkac I (2008) Letter to the editor. J Mol Neurosci PA, Brandt J, Barker PB, Slusher BS, Kaplin AI (2012) Inhi- 35:245–246 bition of glutamate carboxypeptidase II (GCPII) activity as a 54. Nakada T (2010) Conversion of brain cytosol profile from fetal treatment for cognitive impairment in multiple sclerosis. Proc to adult type during the perinatal period: taurine-NAA exchange. Natl Acad Sci USA 109:20101–20106 Proc Jpn Acad Ser B-Phys Biol Sci 86:630–642

123 Neurochem Res (2014) 39:1–36 25

55. Mersmann N, Tkachev D, Jelinek R, Roth PT, Mobius W, Ru- 72. Wurtman RJ (1992) Choline metabolism as a basis for the hwedel T, Ruhle S, Weber-Fahr W, Sartorius A, Klugmann M selective vulnerability of cholinergic neurons. Trends Neurosci (2011) Aspartoacylase-LacZ Knockin mice: an engineered 15:117–122 model of canavan disease. PLoS One 6 73. Lo¨ffelholz K, Klein J, Ko¨ppen A (1993) Choline, a precursor of 56. Irwan R, Sijens PE, Potze JH, Oudkerk M (2005) Correlation of acetylcholine and phospholipids in the brain. Prog Brain Res proton MR spectroscopy and diffusion tensor imaging. Magn 98:197–200 Reson Imag 23:851–858 74. Zeisel SH (1991) Choline, an essential nutrient for humans. 57. Destefano N, Matthews PM, Arnold DL (1995) Reversible FASEB J 5:2093–2098 decreases in N-acetylaspartate after acute brain injury. Magn 75. Cornford EM, Braun LD, Oldendorf WH (1978) Carrier medi- Reson Med 34:721–727 ated blood-brain barrier transport of choline and certain choline 58. Tonon C, Vetrugno R, Lodi R, Gallassi R, Provini F, Iotti S, analogues. J Neurochem 30:299–308 Plazzi G, Montagna P, Lugaresi E, Barbiroli B (2007) Proton 76. Aquilonius S-M, Ceder G, Lying-Tunell U, Malmlund HO, magnetic resonance spectroscopy study of brain metabolism in Schuberth J (1975) The arteriovenous difference of choline obstructive sleep apnoea syndrome before and after continuous across the brain of man. Brain Res 99:430–433 positive airway pressure treatment. Sleep 30:305–311 77. Illingworth RD, Portman OW (1972) Uptake and metabolism of 59. Ferini-Strambi L, Baietto C, Di Gioia MR, Castaldi P, Castro- plasma lysophosphatidylcholine in vivo by brain of squirrel novo C, Zucconi M, Cappa SR (2003) Cognitive dysfunction in monkeys. Biochem J 130:557–1000 patients with obstructive sleep apnea (OSA): partial reversibility 78. Ansell GB, Spanner S (1971) Studies on origin of choline in after continuous positive airway pressure (CPAP). Brain Res brain of rat. Biochem J 122:741–1000 Bull 61:87–92 79. Klein J, Koppen A, Loffelholz K, Schmitthenner J (1992) 60. Hashimoto T, Tayama M, Miyazaki M, Yoneda Y, Yoshimoto Uptake and metabolism of choline by rat-brain after acute T, Harada M, Miyoshi H, Tanouchi M, Kuroda Y (1995) choline administration. J Neurochem 58:870–876 Reduced N-acetylaspartate in the brain observed on in vivo 80. Stoll AL, Renshaw PF, de Micheli E, Wurtman RJ, Pillay SS, proton magnetic resonance spectroscopy in patients with mental Cohen BM (1995) Choline ingestion increases the resonance of retardation. Ped Neurol 13:205–208 choline-containing compounds in human brain: an in vivo pro- 61. Gadian DG, Isaacs EB, Cross JH, Connelly A, Jackson GD, ton magnetic resonance study. Biol Psychiatr 37:170–174 King MD, Neville BGR, Varghakhadem F (1996) Lateralisation 81. Dechent P, Pouwels PJW, Frahm J (1999) Neither short-term of brain function in childhood revealed by magnetic resonance nor long-term adminstration of oral choline alters metabolite spectroscopy. Neurology 46:974–977 concentrations in human brain. Biol Psychiatr 46:406–411 62. Jung RE, Brooks WM, Yeo RA, Chiulli SJ, Weers C, Sibbitt 82. Tan J, Bluml S, Hoang T, Dubowitz D, Mevenkamp G, Ross BD WL (1999) Biochemical markers of intelligence: a proton MR (1998) Lack of effect of oral choline supplementation on the spectroscopy study of normal human brain. Proc R Soc Lond B concentrations of choline metabolites in human brain. Magn 266:1375–1379 Reson Med 39:1005–1010 63. Jung RE, Haier RJ, Yeo RA, Rowland LM, Petropoulos H, 83. Babb SM, Ke Y, Lange N, Kaufman MJ, Renshaw PF, Cohen Levine AS, Sibbitt WL, Brooks WM (2005) Sex differences in BM (2004) Oral choline increases choline metabolites in human N-acetylaspartate correlates of general intelligence: an H-1 MRS brain. Psychiatr Res Neuroimag 130:1–9 study of normal human brain. Neuroimage 26:965–972 84. Morley BJ, Fleck DL (1987) A time course and dose-response 64. Jung RE, Haier RJ (2007) The parieto-frontal integration theory study of the regulation of brain nicotinic receptors by dietary (P-FIT) of intelligence: converging neuroimaging evidence. choline. Brain Res 421:21–29 Behav Brain Sci 30:135–187 85. Hirsch MJ, Wurtman RJ (1978) Lecithin consumption increases 65. Rae C, Scott RB, Lee MA, Hines N, Paul C, Simpson JM, acetylcholine concentrations in rat brain and adrenal gland. Karmiloff-Smith A, Anderson M, Styles P, Radda GK (2003) Science 202:223–225 Brain bioenergetics and cognitive ability. Dev Neurosci 25:324– 86. Wang XC, Du XX, Tian Q, Wang JZ (2008) Correlation 331 between choline signal intensity and acetylcholine level in dif- 66. Martin E, Capone A, Schneider J, Hennig J, Thiel T (2001) ferent brain regions of rat. Neurochem Res 33:814–819 Absence of N-acetylaspartate in the human brain: impact on 87. Poly C, Massaro JM, Seshadri S, Wolf PA, Cho EY, Krall E, neurospectroscopy? Ann Neurol 49:518–521 Jacques PF, Au R (2011) The relation of dietary choline to 67. Boltshauser E, Schmitt B, Wevers RA, Engelke U, Burlina AB, cognitive performance and white-matter hyper intensity in the Burlina AP (2004) Follow-up of a child with hypoacetylaspartia. Framingham offspring cohort. Am J Clin Nutr 94:1584–1591 Neuropediatr 35:255–258 88. Yamamura HI, Snyder SH (1972) Choline: high-affinity uptake 68. Furukawa-Hibi Y, Nitta A, Fukumitsu H, Somiya H, Toriumi K, by rat brain synaptosomes. Science 178:626–628 Furukawa S, Nabeshima T, Yamada K (2012) Absence of 89. Diamond I, Kennedy EP (1969) Carrier-mediated transport of SHATI/Nat8 l reduces social interaction in mice. Neurosci Lett choline into synaptic nerve endings. J Biol Chem 244:3258– 526:79–84 3263 69. Miller BL, Chang L, Booth R, Ernst T, Cornford M, Nikas D, 90. Kuharm MJ, Murrin LC (1978) Sodium-dependent, high affinity McBride D, Jenden DJ (1996) In vivo 1H MRS: correlation with choline uptake. J Neurochem 30:15–21 in vitro chemistry/histochemistry. Life Science 58:1929–1935 91. Tucek S (1984) Problems in the organization and control of 70. Barker PB, Breiter SN, Soher BJ, Chatham JC, Forder JR, acetylcholine synthesis in brain neurons. Prog Biophys Mol Biol Samphilipo MA, Magee CA, Anderson JH (1994) Quantitative 44:1–46 proton spectroscopy of canine brain: in vivo and in vitro cor- 92. Scremin OU, Jenden DJ (1993) Acetylcholine turnover and relations. Magn Reson Med 32:157–163 release: the influence of energy metabolism and systemic cho- 71. Blu¨ml S, Seymour KJ, Ross BD (1999) Developmental changes line availability. Prog Brain Res 98:191–195 in choline- and ethanolamine-containing compounds measured 93. Michel V, Yuan ZF, Ramsubir S, Bakovic M (2006) Choline with proton-decoupled 31P MRS in in vivo human brain. Magn transport for phospholipid synthesis. Exp Biol Med 231:490– Reson Med 42:643–654 504

123 26 Neurochem Res (2014) 39:1–36

94. Aquilonius S-M, Windbladh B (1972) Cerebrospinal fluid 113. Dreher W, Norris DG, Leibfritz D (1994) Magnetization transfer clearance of choline and some other amines. Acta Physiol Scand affects the proton creatine/phosphocreatine signal intensity: 85:78–90 in vivo demonstration in the rat brain. Magn Reson Med 31:81– 95. Mann SP, Hebb C (1977) Free choline in the brain of the rat. J 84 Neurochem 28:241–244 114. Helms G, Frahm J (1999) Magnetization transfer attenuation of 96. Zeisel SH (2000) Choline: needed for normal development of creatine resonances in localised proton MRS of human brain memory. J Am Coll Nutr 19:528S–531S in vivo. NMR Biomed 12:490–494 97. Dainous F, Freysz L, Mozzi R, Dreyfus H, Louis JC, Porcellati 115. Kruiskamp MJ, de Graaf RA, van der Grond J, Lamerichs R, G, Massarelli R (1982) Synthesis of choline phospholipids in Nicolay K (2001) Magnetic coupling between water and crea- neuronal and glial cell cultures by the methylation pathway. tine protons in human brain and skeletal muscle, as measured FEBS Lett 146:221–223 using inversion transfer H-1-MRS. NMR Biomed 14:1–4 98. Lakher M, Wurtman RJ (1987) In vivo synthesis of phosphati- 116. Opstad KS, Bell BA, Griffiths JR, Howe FA (2008) An dylcholine in rat brain via the phospholipid methylation path- assessment of the effects of sample ischaemia and spinning time way. Brain Res 419:131–140 on the metabolic profile of brain tumour biopsy specimens as 99. Jenden DJ (1991) The metabolism of choline. Bull Clin Neu- determined by high-resolution magic angle spinning H-1 NMR. rosci 55:99–106 NMR Biomed 21:1138–1147 100. Gibson GE, Blass JP (1976) Impaired synthesis of acetylcholine 117. Dechent P, Pouwels PJW, Wilken B, Hanefeld F, Frahm J in brain accompanying mild hypoxia and hypoglycaemia. J (1999) Increase of total creatine in human brain after oral sup- Neurochem 27:37–42 plementation of creatine-monohydrate. Am J Physiol 277:R698– 101. Sartorius A, Schloss P, Vollmayr B, Ende G, Neumann-Haefelin R704 C, Hoehn M, Henn FA (2006) Correlation between MR-spec- 118. Braissant O, Henry HML, Eilers B, Bachmann C (2001) troscopic rat hippocampal choline levels and phospholipase A2. Endogenous synthesis and transport of creatine in the rat brain: World J Biol Psychiatr 7:246–250 an in situ hybridization study. Mol Brain Res 86:193–201 102. Boulanger Y, Labelle M, Khiat A (2000) Role of phospholipase 119. Defalco AJ, Davies RK (1961) The synthesis of creatine by the A(2) on the variations of the choline signal intensity observed by brain of the intact rat. J Neurochem 7:308–312 H-1 magnetic resonance spectroscopy in brain diseases. Brain 120. Dringen R, Verleysdonk S, Hamprecht B, Wilker W, Leibfritz Res Rev 33:380–389 D, Brand A (1998) Metabolism of glycine in primary astroglial 103. Bhakoo KK, Williams SR, Florian CL, Land H, Noble M (1996) cells: synthesis of creatine, serine and glutathione. J Neurochem Immortalisation and transformation are associated with specific 70:835–840 alterations in choline metabolism. Cancer Res 56:4630–4635 121. Braissant O, Henry H (2008) AGAT, GAMT and SLC6A8 104. Laule C, Vavasour IM, Kolind SH, Li DKB, Traboulsee TL, distribution in the central nervous system, in relation to creatine Moore GRW, MacKay AL (2007) Magnetic resonance imaging deficiency syndromes: a review. J Inherit Metab Dis 31:230–239 of myelin. Neurotherapeutics 4:460–484 122. Wyss M, Kaddurah-Daouk R (2000) Creatine and creatinine 105. Martin PR, Gibbs SJ, Nimmerrichter AA, Riddle WR, Welch metabolism. Physiol Rev 80:1107–1212 LW, Willcott MR (1995) Brain proton magnetic resonance 123. Sto¨ckler S, Hanefeld F, Frahm J (1996) Creatine replacement spectroscopy studies in recently abstinent alcoholics. Alcohol therapy in guanidinoacetate methyltransferase deficiency, a Clin Exp Res 19:1078–1082 novel inborn error of metabolism. Lancet 348:789–790 106. Jung RE, Yeo RA, Love TM, Petropoulos H, Sibbitt WL, 124. Valayannopoulos V, Boddaert N, Mention K, Touati G, Barbier Brooks WM (2002) Biochemical markers of mood: a proton V, Chabli A, Sedel F, Kaplan J, Dufier JL, Seidenwurm D, magnetic resonance study of normal human brain. Biol Psychiatr Rabier D, Saudubray JM, de Lonlay P (2009) Secondary crea- 51:224–229 tine deficiency in ornithine delta-aminotransferase deficiency. 107. Craig MC, Daly EM, O’Gorman R, Rymer J, Lythgoe D, Ng G, Mol Genet Metab 97:109–113 Simmons A, Maki PM, Murphy DGM (2007) Effects of acute 125. Salomons GS, van Dooren SJM, Verhoeven NM, Cecil KM, ovarian hormone suppression on the human brain: an in vivo Ball WS, Degrauw TJ, Jakobs C (2001) X-linked creatine- H-1 MRS study. Psychoneuroendocrinology 32:1128–1132 transporter gene (SLC6A8) defect: a new creatine-deficiency 108. Xiong QA, Du F, Zhu XH, Zhang PY, Suntharalingam P, Ip- syndrome. Am J Hum Genet 68:1497–1500 polito J, Kamdar FD, Chen W, Zhang JY (2011) ATP production 126. Chilosi A, Leuzzi V, Battini R, Tosetti M, Ferretti G, Comparini rate via creatine kinase or ATP synthase in vivo a novel A, Casarano M, Moretti E, Alessandri MG, Bianchi MC, Cioni superfast magnetization saturation transfer method. Circ Res G (2008) Treatment with L-arginine improves neuropsycholog- 108:U265–U653 ical disorders in a child with Creatine transporter defect. Neu- 109. Chen W, Zhu X-H, Adriany G, Ugurbil K (1997) Increase of rocase 14:151–161 creatine kinase activity in the visual cortex of human brain 127. de Kamp JMV, Pouwels PJW, Aarsen FK, ten Hoopen LW, during visual stimulation: a 31P NMR magnetization transfer Knol DL, de Klerk JB, de Coo IF, Huijmans JGM, Jakobs C, van study. Magn Reson Med 38:551–557 der Knaap MS, Salomons GS, Mancini GMS (2012) Long-term 110. Kruiskamp MJ, van Vliet G, Nicolay K (2000) 1H and 31P follow-up and treatment in nine boys with X-linked creatine magnetization transfer studies of hindleg muscle in wild-type transporter defect. J Inherit Metab Dis 35:141–149 and creatine kinase-deficient mice. Magn Reson Med 43:657– 128. Fons C, Sempere A, Arias A, Lopez-Sala A, Poo P, Pineda M, 664 Mas A, Vilaseca MA, Salomons GS, Ribes A, Artuch R, 111. Ke Y, Cohen BM, Lowen S, Hirashima F, Nassar L, Renshaw Campistol J (2008) Arginine supplementation in four patients PF (2002) Biexponential transverse relaxation (T2) of the proton with X-linked creatine transporter defect. J Inherit Metab Dis MRS creatine resonance in human brain. Magn Reson Med 31:724–728 47:232–238 129. Valayannopoulos V, Boddaert N, Chabli A, Barbier V, Des- 112. Simister RJ, McLean MA, Salmenpera TM, Barker GJ, Duncan guerre I, Philippe A, Afenjar A, Mazzuca M, Cheillan D, JS (2008) The effect of epileptic seizures on proton MRS visible Munnich A, de Keyzer Y, Jakobs C, Salomons GS, de Lonlay P neurochemical concentrations. Epilepsy Res 81:36–43 (2012) Treatment by oral creatine, L-arginine and L-glycine in

123 Neurochem Res (2014) 39:1–36 27

six severely affected patients with creatine transporter defect. J 147. Tachikawa M, Fukaya M, Terasaki T, Ohtsuki S, Watanabe M Inherit Metab Dis 35:151–157 (2004) Distinct cellular expressions of creatine synthetic enzyme 130. Kurosawa Y, DeGrauw TJ, Lindquist DM, Blanco VM, Pyne- GAMT and creatine kinases uCK-Mi and CK-B suggest a novel Geithman GJ, Daikoku T, Chambers JB, Benoit SC, Clark JF neuron-glial relationship for brain energy homeostasis. Eur J (2012) Cyclocreatine treatment improves cognition in mice with Neurosci 20:144–160 creatine transporter deficiency. J Clin Invest 122:2837–2846 148. Walliman T, Wyss M, Brdiczka D, Nicolay K, Eppenberger HM 131. Mo¨ller A, Hamprecht B (1989) Creatine transport in cultured (1992) Intracellular compartmentation, structure and function of cells of rat and mouse brain. J Neurochem 52:544–550 creatine kinase isoenzyme in tissues with high and fluctuating 132. Braissant O, Henry H (2007) AGAT, GAMT and SLC6A8 energy demands: the phosphocreatine circuit for cellular energy distribution in the central nervous system, in relation to creatine homeostasis. Biochem J 281:21–40 deficiency syndromes: a review. In annual symposium of the 149. Rango M, Castelli A, Scarlato G (1997) Energetics of 3.5 s society-for-the-study-of-inborn-errors-of-metabolism. Hamburg, neural activation in humans: a 31P spectroscopy study. Magn Germany: Springer Reson Med 38:878–883 133. Braissant O, Beard E, Torrent C, Henry H (2010) Dissociation 150. Kemp GJ (2000) Non-invasive methods for studying brain of AGAT, GAMT and SLC6A8 in CNS: relevance to creatine energy metabolism: what they show and what it means. Dev deficiency syndromes. Neurobiol Dis 37:423–433 Neurosci 22:418–428 134. Dai WX, Vinnakota S, Qian XJ, Kunze DL, Sarkar HK (1999) 151. Sappey-Marinier D, Calabrese G, Fein G, Hugg JW, Biggins C, Molecular characterization of the human CRT-1 creatine trans- Weiner MW (1992) Effect of photic stimulation on human porter expressed in xenopus oocytes. Arch Biochem Biophys visual cortex lactate and phosphates using 1H and 31P magnetic 36:75–84 resonance spectroscopy. J Cereb Blood Flow Metab 12:584–592 135. Dodd JR, Birch NP, Waldvogel HJ, Christie DL (2010) Func- 152. Cox DWG, Morris PG, Feeney F, Bachelard HS (1983) 31P tional and immunocytochemical characterization of the creatine MRS studies on cerebral energy metabolism under conditions of transporter in rat hippocampal neurons. J Neurochem 115:684– hypoglycaemia and hypoxia in vitro. Biochem J 212:365–370 693 153. Sauter A, Rudin M (1993) Determination of creatine kinase 136. Mak CSW, Waldvogel HJ, Dodd JR, Gilbert RT, Lowe MTJ, kinetic parameters in rat brain by magnetization transfer. Cor- Birch NP, Faull RLM, Christie DL (2009) Immunohistochemi- relation with brain function. J Biol Chem 268:13166–13171 cal localisation of the creatine transporter in the rat brain. 154. t’Zandt HJAI, Jost C, Oerlemans F, Klomp DWJ, Wieringa B, Neuroscience 163:571–585 Heerschap A (2000) Brains of creatine kinase deficient mice 137. Pyne-Geithman GJ, de Grauw TJ, Cecil KM, Chuck G, Lyons lack phosphocreatine and exhibit an increased NAA level. Proc MA, Ishida Y, Clark JF (2004) Presence of normal creatine in Int Soc Magn Reson Med 8:174 the muscle of a patient with a mutation in the creatine trans- 155. Kekelidze T, Khait I, Togliatti A, Benzecry JM, Wieringa B, porter: a case study. Mol Cell Biochem 262:35–39 Holtzman D (1998) Altered brain phosphocreatine and ATP 138. Nabuurs C, Romeijn M, Veltien A, Kan H, Isbrandt D, Heers- regulation when mitochondrial creatine kinase is absent. Wiley- chap A (2009) Creatine deficiency, uptake and breakdown Liss, Los Angeles studied in brain and muscle of Arginine:glycine Amidinotrans- 156. de Groof AJC, Oerlemans FTJJ, Jost CR, Wieringa B (2001) ferase deficient mice. Proc Int Soc Magn Reson Med 19:1026 Changes in glycolytic network and mitochondrial design in 139. Wilkinson ID, Mitchel N, Breivik S, Greenwood P, Griffiths PD, creatine kinase-deficient muscles. Muscle Nerve 24:1188–1196 Winter EM, Van Beek EJR (2006) Effects of creatine supple- 157. Rae C, Bartlett D, Yang Q, Walton D, Denotti A, Sachinwalla T, mentation on cerebral white matter in competitive sportsmen. Grunstein RR (2009) Dynamic changes in brain bioenergetics Clin J Sport Med 16:63–67 during obstructive sleep apneoa. J Cereb Blood Flow Metab 140. McLean MA, Woermann FG, Barker GJ, Duncan JS (2000) 29:1421–1428 Quantitative analysis of short echo time 1HMRSI of cerebral 158. Shoubridge EA, Briggs RW, Radda GK (1982) P-31 NMR gray and white matter. Magn Reson Med 44:401–411 saturation transfer measurements of the steady-state rates of 141. Pouwels PJW, Frahm J (1998) Regional metabolite concentra- creatine kinase and ATP synthetase in the rat brain. FEBS Lett tions in human brain as determined by quantitative localised 140:288–292 proton MRS. Magn Reson Med 39:53–60 159. Holtzman D, Brown M, Ogorman E, Allred E, Wallimann T 142. Kaldis P, Hemmer W, Zanolla E, Holtzman D, Walliman T (1998) T Brain ATP metabolism in hypoxia resistant mice fed (1996) ‘Hot spots’ of creatine kinase localisation in brain: cer- guanidinopropionic acid. Dev Neurosci 20:469–477 ebellum, hippocampus and choroid plexus. Dev Neurosci 160. Holtzman D, Meyers R, Ogorman E, Khait I, Walliman T, 18:542–554 Allred E, Jensen F (1997) In vivo brain phosphocreatine and 143. Sartorius A, Lugenbiel P, Mahlstedt MM, Ende G, Schloss P, ATP regulation in mice fed a creatine analog. Am J Physiol Vollmayr B (2008) Proton magnetic resonance spectroscopic 41:C1567–C1577 creatine correlates with creatine transporter protein density in rat 161. Saks VA, Kongas O, Vendelin M, Kay L (2000) Role of the brain. J Neurosci Meth 172:215–219 creatine/phosphocreatine system in the regulation of mitochon- 144. Hertz L, Drejer J, Schousboe A (1988) Energy metabolism in drial respiration. Acta Physiol Scand 168:635–641 glutamatergic neurons, GABAergic neurons and astrocytes in 162. Monge C, Beraud N, Kuznetsov AV, Rostovtseva T, Sackett D, primary culture. Neurochem Res 13:605–610 Schlattner U, Vendelin M, Saks VA (2008) Regulation of res- 145. Brand A, Richter-Lansberg C, Richter-Lansberg C, Leibfritz D piration in brain mitochondria and synaptosomes: restrictions of (1993) Multinuclear NMR studies on the energy metabolism of ADP diffusion in situ, roles of tubulin, and mitochondrial cre- glial and neuronal cells. Dev Neurosci 15:289–298 atine kinase. Mol Cell Biochem 318:147–165 146. Lowe MTJ, Kim EH, Faull RLM, Christie DL, Waldvogel HJ 163. Yoshizaki K, Watari H, Radda GK (1990) Role of phospho- (2013) Dissociated expression of mitochondrial and cytosolic creatine in energy-transport in skeletal muscle of bullfrog creatine kinases in the human brain: a new perspective on the studied by P-31 NMR. Biochem Biophys Acta 1051:144–150 role of creatine in brain energy metabolism. J Cereb Blood Flow 164. Kemp GJ, Manners DN, Clark JF, Bastin ME, Radda GK (1998) Metab 33:1295–1306 Theoretical modelling of some spatial and temporal aspects of

123 28 Neurochem Res (2014) 39:1–36

the mitochondrion creatine kinase myofibril system in muscle. 183. Jackson MC, Lenney JF (1996) The distribution of carnosine Mol Cell Biochem 184:249–289 and related dipeptides in rat and human tissues. Inflamm Res 165. Heilig CW, Stromski ME, Blumenfield JD, Lee JP, Gullans SR 45:132–135 (1989) Characterization of the major brain osmolytes that 184. Petroff OAC, Hyder F, Rothman DL, Mattson RH (2000) accumulate in salt-loaded rats. Am J Physiol 257:F1108–F1116 Effects of gabapentin on brain GABA, homocarnosine and 166. Miller TJ, Hanson RD, Yancey PH (2000) Developmental pyrrolidinone in epilepsy patients. Epilepsia 41:675–680 changes in organic osmolytes in prenatal and postnatal rat tis- 185. Keltner JR, Wald LL, Christensen JD, Maas LC, Moore CM, sues. Comp Biochem Physiol A 125:45–56 Cohen BM, Renshaw PF (1996) A technique for detecting 167. Bothwell JH, Styles P, Bhakoo KK (2002) Swelling-activated GABA in the human brain with PRESS localization and opti- taurine and creatine effluxes from rat cortical astrocytes are mized refocusing spectral editing radiofrequency pulses. Magn pharmacologically distinct. J Membr Biol 185:157–164 Reson Med 36:458–461 168. Michaelis T, Wick M, Fujimori H, Matsumara A, Frahm J 186. Rothman DL, Behar KL, Hetherington HP, Shulman RG (1984) (1999) Proton MRS of oral creatine supplementation in rats. Homonuclear H1 double resonance difference spectroscopy of Cerebral metabolite concentration and ischemic challenge. the rat brain in vivo. Proc Nat Acad Sci United States Am-Biol NMR Biomed 12:309–314 Sci 81:6330–6334 169. Juhn MS, Tarnopolsky M (1998) Potential side effects of oral 187. Rothman DL, Petroff OAC, Behar KL, Mattson RH (1993) creatine supplementation: a critical review. Clin J Sport Med Localised 1H NMR measurements of GABA levels in human 8:298–304 brain in vivo. Proc Nat Acad Sci USA 90:5662–5666 170. Wyss M, Schulze A (2002) Health implications of creatine: can 188. Bogner W, Gruber S, Doelken M, Stadlbauer A, Ganslandt O, oral creatine supplementation protect against neurological and Boettcher U, Trattnig S, Doerfler A, Stefan H, Hammen T atherosclerotic disease? Neuroscience 112:243–260 (2010) In vivo quantification of intracerebral GABA by single- 171. Watanabe A, Kato N, Kato T (2002) Effects of creatine on voxel (1)H-MRS-How reproducible are the results? Eur J Radiol mental fatigue and cerebral hemoglobin oxygenation. Neurosci 73:526–531 Res 42:279–285 189. O’Gorman RL, Michels L, Edden RA, Murdoch JB, Martin E 172. Rae C, Digney AL, McEwan SR, Bates TC (2003) Oral creatine (2011) In vivo detection of GABA and glutamate with MEGA– monohydrate supplementation improves cognitive performance; PRESS: reproducibility and gender effects. J Magn Reson Imag a placebo-controlled, double blind, cross-over trial. Proc R Soc 33:1262–1267 Lond B 279:2147–2150 190. McLean MA, Busza AL, Wald LL, Simister RJ, Barker GJ, 173. Hammett ST, Wall MB, Edwards TC, Smith AT (2010) Dietary Williams SR (2002) In vivo GABA? measurement at 1.5 T supplementation of creatine monohydrate reduces the human using a PRESS-localized double quantum filter. Magn Reson fMRI BOLD signal. Neurosci Lett 479:201–205 Med 48:233–241 174. Almeida LS, Salomons GS, Hogenboom R, Jakobs C, Schof- 191. Mescher M, Merkle H, Kirsch J, Garwood M, Gruetter R (1998) felmeer ANM (2006) Exocytotic release of creatine in rat brain. Simultaneous in vivo spectral editing and water suppression. Synapse 60:118–123 NMR Biomed 11:266–272 175. Koga Y, Takahashi H, Oikawa D, Tachibana T, Denbow DM, 192. Napolitano A, Kockenberger W, Auer DP (2013) Reliable Furuse M (2005) Brain creatine functions to attenuate acute gamma aminobutyric acid measurement using optimised PRESS stress responses through gabaergic system in chicks. Neurosci- at 3T. Magn Reson Med 69:1523–1528 ence 132:65–71 193. Thomas MA, Hattori N, Umeda M, Sawada T, Naruse S (2003) 176. de Deyn PP, Macdonald RL (1990) Guanidino compounds that Evaluation of two-dimensional L-COSY and PRESS using a 3T are increased in cerebrospinal fluid and brain of uremic patients MRI scanner: from phantoms to human brain in vivo. NMR inhibit GABA and glycine responses on mouse neurons in cell Biomed 16:245–251 culture. Ann Neurol 28:627–633 194. Thomas MA, Yue K, Binesh N, Davanzo P, Kumar A, Siegel B, 177. Neu A, Neuhoff H, Trube G, Fehr S, Ullrich K, Roeper J, Is- Frye M, Curran J, Lufkin R, Martin P, Guze B (2001) Localised brandt D (2002) Activation of GABAA receptors by guanidi- two-dimensional shift correlated MR spectroscopy of human noacetate: a novel pathophysiological mechanism. Neurobiol brain. Magn Reson Med 46:58–67 Dis 11:298–307 195. Choi CH, Coupland NJ, Hanstock CC, Ogilvie CJ, Higgins 178. Chebib M, Gavande N, Wong KY, Park A, Premoli I, Mewett ACM, Gheorghiu D, Allen PS (2005) Brain gamma-aminobu- KN, Allan RD, Duke RK, Johnston GAR, Hanrahan JR (2009) tyric acid measurement by proton double-quantum filtering with Guanidino acids act as rho 1 GABA(C) receptor antagonists. selective J rewinding. Magn Reson Med 54:272–279 Neurochem Res 34:1704–1711 196. Choi IY, Lee SP, Merkle H, Shen J (2006) In vivo detection of 179. Royes LF, Fighera MR, Furian AF, Oliveira MS, Fiorenza NG, gray and white matter differences in GABA concentration in the Ferreira J, da Silva AC, Priel MR, Ueda ES, Calixto JB, Cav- human brain. Neuroimage 33:85–93 alheiro EA, Mello CF (2008) Neuromodulatory effect of crea- 197. Evans CJ, McGonigle DJ, Edden RAE (2010) Diurnal stability tine on extracellular action potentials in rat hippocampus: role of of gamma-aminobutyric acid concentration in visual and sen- NMDA receptors. Neurochem Int 53:33–37 sorimotor cortex. J Magn Reson Imag 31:204–209 180. Oliveira MS, Furian AF, Fighera MR, Fiorenza NG, Ferreira J, 198. Gao F, Edden RA, Li M, Puts NA, Wang G, Liu C, Zhao B, Rubin MA, Mello CF, Royes LFF (2008) The involvement of Wang H, Bai X, Zhao C, Wang X, Barker PB (2013) Edited the polyamines binding sites at the NMDA receptor in creatine- magnetic resonance spectroscopy detects an age-related decline induced spatial learning enhancement. Behav Brain Res in brain GABA levels. Neuroimage 78:75–82 187:200–204 199. Patel AB, Rothman DL, Cline GW, Behar KL (2001) Glutamine 181. Genius J, Geiger J, Bender A, Moller H-J, Klopstock T, Rujescu is the major precursor for GABA synthesis in rat neocortex D (2012) Creatine protects against excitotoxicity in an in vitro in vivo following acute GABA-transaminase inhibition. Brain model of neurodegeneration. PLoS ONE 7:e30554 Res 919:207–220 182. Andres RH, Ducraya AD, Schlattner U, Wallimann T, Widmer 200. Paulsen RE, Odden E, Fonnum F (1988) Importance of gluta- HR (2008) Functions and effects of creatine in the central ner- mine for gamma-aminobutyric acid synthesis in rat neostriatum vous system. Brain Res Bull 76:329–343 in vivo. J Neurochem 13:637–641

123 Neurochem Res (2014) 39:1–36 29

201. Rae C, Hare N, Bubb WA, McEwan SR, Broër A, McQuillan 220. Jackson MF, Esplin B, Capek R (2000) Reversal of the activity- JA, Balcar VJ, Conigrave AD, Broër S (2003) Inhibition of dependent suppression of GABA-mediated inhibition in hippo- glutamine transport depletes glutamate and GABA neurotrans- campal slices from gamma-vinyl GABA (vigabatrin)-pretreated mitter pools: further evidence for metabolic compartmentation. J rats. Neuropharmacol 39:65–74 Neurochem 85:503–514 221. Wu Y, Wang W, Richerson GB (2003) Vigabatrin induces tonic 202. Tapia R, Gonza´lez RM (1978) Glutamine and glutamate as inhibition via GABA transporter reversal without increasing precursors of the releasable pool of GABA in brain cortex slices. vesicular GABA release. J Neurophysiol 89:2021–2034 Neurosci Lett 10:165–169 222. Wu YM, Wang WG, Diez-Sampedro A, Richerson GB (2007) 203. Jackson MC, Scollard DM, Mack RJ, Lenney JF (1994) Nonvesicular inhibitory neurotransmission via reversal of the Localization of a novel pathway for the liberation of GABA in GABA transporter GAT-1. Neuron 56:851–865 the human CNS. Brain Res Bull 33:379–385 223. Puts NAJ, Edden RAE (2012) In vivo magnetic resonance 204. Seiler N (2004) Catabolism of polyamines. Amino Acid 26:217– spectroscopy of GABA: a methodological review. Prog NMR 233 Spect 60:29–41 205. Heja L, Nyitrai G, Kekesi O, Dobolyi A, Szabo P, Fiath R, 224. Floyer-Lea A, Wylezinska M, Kincses T, Matthews PM (2005) Ulbert I, Pal-Szenthe B, Palkovits M,Kardos J (2012) Astrocytes Rapid modulation of GABA concentration in human sensorimotor convert network excitation to tonic inhibition of neurons. Bmc cortex during motor learning. J Neurophysiol 95:1639–1644 Biol 10:26 225. Sumner P, Edden RAE, Bompas A, Evans CJ, Singh KD (2010) 206. Dericioglu N, Garganta CL, Petroff OA, Mendelsohn D, Wil- More GABA, less distraction: a neurochemical predictor of liamson A (2008) Blockade of GABA synthesis only affects motor decision speed. Nat Neurosci 13:825–827 neural excitability under activated conditions in rat hippocampal 226. Boy F, Evans CJ, Edden RAE, Singh KD, Husain M, Sumner P slices. Neurochem Int 53:22–32 (2010) Individual differences in subconscious motor control 207. Hanstock CC, Coupland NJ, Allen PS (2002) GABA X-2 mul- predicted by GABA concentration in SMA. Curr Biol 20:1779– tiplet measured pre- and post administration of vigabatrin in 1785 human brain. Magn Reson Med 48:617–623 227. Henderson LA, Peck CC, Petersen ET, Rae CD, Youssef AM, 208. Rae C, Nasrallah FA, Griffin JL, Balcar VJ (2009) Now I know Reeves JM, Wilcox SL, Akhter R, Murray GM, Gustin SM my ABC. A systems neurochemistry and functional metabolo- (2013) Chronic pain: lost inhibition? J Neurosci 33:7574–7582 mic approach to understanding the GABAergic system. J Neu- 228. Edden RAE, Muthukumaraswamy SD, Freeman TCA, Singh rochem 109(Suppl 1):109–116 KD (2009) Orientation discrimination performance is predicted 209. Bak LK, Schousboe A, Waagepetersen HS (2006) The gluta- by GABA concentration and gamma oscillation frequency in mate/GABA-glutamine cycle: aspects of transport, neurotrans- human primary visual cortex. J Neurosci 29:15721–15726 mitter homeostasis and ammonia transfer. J Neurochem 98:641– 229. Marenco S, Savostyanova AA, van der Veen JW, Geramita M, 653 Stern A, Barnett AS, Kolachana B, Radulescu E, Zhang FY, 210. Nasrallah F, Griffin JL, Balcar VJ, Rae C (2009) Understanding Callicott JH, Straub RE, Shen J, Weinberger DR (2010) Genetic your inhibitions. Effects of GABA and GABAA receptors on Modulation of GABA levels in the anterior cingulate cortex by brain cortical metabolism. J Neurochem 108:57–71 GAD1 and COMT. Neuropsychpharmacol 35:1708–1717 211. Nasrallah F, Griffin JL, Balcar VJ, Rae C (2007) Understanding 230. Northoff G, Walter M, Schulte RF, Beck J, Dydak U, Henning your inhibitions. Modulation of brain cortical metabolism by A, Boeker H, Grimm S, Boesiger P (2007) GABA concentra- GABA-B receptors. J Cereb Blood Flow Metab 27:1510–1520 tions in the human anterior cingulate cortex predict negative 212. Mo¨hler H (2006) GABAA receptor diversity and pharmacology. BOLD responses in fMRI. Nat Neurosci 10:1515–1517 Cell Tissue Res 326:505–516 231. Donahue MJ, Near J, Blicher JU, Jezzard P (2010) Baseline 213. Petroff OAC, Behar KL, Mattson RH, Rothman DL (1996) GABA concentration and fMRI response. Neuroimage 53:392– Human brain gamma-aminobutyric acid levels and seizure 398 control following initiation of vigabatrin therapy. J Neurochem 232. Chatton JY, Pellerin L, Magistretti PJ (2003) GABA uptake into 67:2399–2404 astrocytes is not associated with significant metabolic cost: 214. Cavelier P, Hamann V, Rossi D, Mobbs P, Attwell D (2005) implications for brain imaging of inhibitory transmission. Proc Tonic excitation and inhibition of neurons: ambient transmitter Natl Acad Sci USA 100:12456–12461 sources and computational consequences. Prog Biophys Mol 233. Ackermann RF, Finch DM, Babb TL, Engel J (1984) Increased Biol 87:3–16 glucose-metabolism during long-duration recurrent inhibition of 215. Schousboe A (2000) Pharmacological and functional charac- hippocampal pyramidal cells. J Neurosci 4:251–264 terisation of astrocytic GABA transport: a short review. Neu- 234. Palacios JM, Kuhar MJ, Rapoport SI, London ED (1982) Effects rochem Res 25:1241–1244 of c-aminobutyric acid agonist and antagonist drugs on local 216. Bolvig T, Larsson OM, Pickering DS, Nelson N, Falch E, Kro- cerebral glucose utilization. J Neurosci 2:853–860 gsgaard-Larsen P, Schousboe A (1999) Action of bicyclic isox- 235. Peyron R, Le Bars D, Cinotti L, Garcia-Larrea L, Galy G, azole GABA analogues on GABA transporters and its relation to Landais P, Millet P, Lavenne F, Froment JC, Krogsgaard-Larsen anticonvulsant activity. Eur J Pharmacol 375:367–374 P (1994) Effects of GABAA receptors activation on brain glu- 217. Richerson GB, Wu Y (2003) Dynamic equilibrium of neuro- cose metabolism in normal subjects and temporal lobe epilepsy transmitter transporters: not just for reuptake anymore. J Neu- (TLE) patients. A positron emission tomography (PET) study. rophysiol 90:1363–1374 Part 1: brain glucose metabolism is increased after GABAA 218. Yasumi M, Sato K, Shimada S, Nishimura M, Tohyama M receptors activation. Epilepsy Res 19:45–54 (1997) Regional distribution of GABA transporter 1 (GAT1) 236. Roland PE, Friberg L (1988) The effect of the GABA-A agonist mRNA in the rat brain: comparison with glutamic acid decar- THIP on regional cortical blood flow in humans. A new test of boxylase(67)(GAD(67))mRNA localization. Mol Brain Res hemispheric dominance. J Cereb Blood Flow Metab 8:314–323 44:205–218 237. Tagamets MA, Horwitz B (2001) Interpreting PET and fMRI 219. Bernstein EM, Quick MW (1999) Regulation of c-aminobutyric measures of functional neural activity: the effects of synaptic acid (GABA) transporters by extracellular GABA. J Biol Chem inhibition on cortical activation in human imaging studies. Brain 274:889–895 Res Bull 54:267–273

123 30 Neurochem Res (2014) 39:1–36

238. Gaetz W, Edgar JC, Wang DJ, Roberts TPL (2011) Relating 256. Lund-Andersen H (1979) Transport of glucose from blood to MEG measured motor cortical oscillations to resting gamma- brain. Physiol Rev 59:305–352 aminobutyric acid (GABA) concentration. Neuroimage 55:616– 257. Pfeuffer J, Tkać I, Gruetter R (2000) Extracellular-intracellular 621 distribution of glucose and lactate in the rat brain assessed 239. Traub RD, Cunningham MO, Gloveli T, LeBeau FEN, Bibbig A, noninvasively by diffusion-weighted 1H nuclear magnetic res- Buhl EH, Whittington MA (2003) GABA-enhanced collective onance spectroscopy in vivo. J Cereb Blood Flow Metab behavior in neuronal axons underlies persistent gamma-frequency 20:736–746 oscillations. Proc Natl Acad Sci USA 100:11047–11052 258. Silver IA, Erecinska M (1998) Glucose-induced intracellular ion 240. Nasrallah FA, Balcar VJ, Rae CD (2011) Activity dependent changes in sugar-sensitive hypothalamic neurons. J Neuro- GABA release controls brain cortical tissue slice metabolism. J physiol 79:1733–1745 Neurosci Res 89:1935–1945 259. Nicholson C, Sykova´ E (1998) Extracellular space structure 241. Stagg CJ, Bestmann S, Constantinescu AO, Moreno LM, All- revealed by diffusion analysis. Trends Neurosci 21:207–215 man C, Mekle R, Woolrich M, Near J, Johansen-Berg H, 260. de Graaf RA, Pan JW, Telang F, Lee JH, Brown P, Novotny EJ, Rothwell JC (2011) Relationship between physiological mea- Hetherington HP, Rothman D (2001) Differentiation of glucose sures of excitability and levels of glutamate and GABA in the transport in human brain gray and white matter. J Cereb Blood human motor cortex. J Physiol-Lond 589:5845–5855 Flow Metab 21:483–492 242. Michaelis T, Merboldt K-D, Ha¨nicke W, Gyngell ML, Bruhn H, 261. Merboldt K-D, Bruhn H, Ha¨nicke W, Michaelis T, Frahm J Frahm J (1991) On the identification of cerebral metabolites in (1992) Decrease of glucose in the human visual cortex during localised 1H NMR spectra of human brain in vivo. NMR Bio- photic stimulation. Magn Reson Med 25:187–194 med 4:90–98 262. Erecinska M, Silver IA (1994) Ions and energy in mammalian 243. Marjanska M, Henry PG, Bolan PJ, Vaughan B, Seaquist ER, brain. Prog Neurobiol 43:37–71 Gruetter R, Ugurbil K, Garwood M (2005) Uncovering hidden 263. Dienel GA (2013) Fuelling and imaging brain activation. ASN in vivo resonances using editing based on localized TOCSY. Neuro 4:267–321 Magn Reson Med 53:783–789 264. Cataldo AM, Broadwell RD (1986) Cytochemical identification 244. Keltner JR, Wald LL, Ledden PJ, Chen YCI, Matthews RT, of cerebral glycogen and glucose-6-phosphatase activity under Kuestermann E, Baker JR, Rosen BR, Jenkins BG (1998) normal and experimental conditions. II. Choroid plexus and Localized double-quantum filter for the in vivo detection of ependymal epithelia, endothelia and pericytes. J Neurocytol brain glucose. Magn Reson Med 39:651–656 15:511–524 245. de Graaf RA, Dijkhuizen RM, Biessels GJ, Braun KPJ, Nicolay 265. Bhattacharya SB, Datta AG (1993) Is brain a gluconeogenic K (2000) In vivo glucose detection by homonuclear spectral organ? Mol Cell Biochem 125:51–57 editing. Magn Reson Med 43:621–626 266. Dringen R, Schmoll D, Cesar M, Hamprecht B (1993) Incor- 246. Mitrakou A, Ryan C, Veneman T, Mokan M, Jenssen T, Kiss I, poration of radioactivity from C-14 lactate into the glycogen of Durrant J, Cryer P, Gerich J (1991) Hierarchy of glycemic cultured mouse astroglial cells—evidence for gluconeogenesis thresholds for counterregulatory hormone secretion, symptoms, in brain cells. Biol Chem Hoppe-Seyler 374:343–347 and cerebral dysfunction. Am J Physiol 260:E67–E74 267. Griffin JL, Rae C, Radda GK, Matthews PM (1999) Delayed 247. Bachelard HS, Cox DWG, Drower J (1984) Sensitivity of gui- labelling of brain glutamate after an intra-arterial [13C]glucose nea-pig hippocampal granule cell field potentials to hexoses bolus: evidence for aerobic metabolism of guinea pig brain in vitro: an effect on cell excitability? J Physiol 352:91–102 glycogen store. Biochim Biophys Acta 1450:297–307 248. Mata M, Fink DJ, Gainer H, Smith CB, Davidsen L, Savaki H, 268. Swanson R (1992) Physiological coupling of glial glycogen Schwartz WJ, Sokoloff L (1980) Activity-dependent energy metabolism to neuronal activity in brain. Can J Physiol Phar- metabolism in rat posterior pituitary primarily reflects sodium macol 70:S138–S144 pump activity. J Neurochem 34:213–215 269. Xu J, Song D, Xue Z, Gu L, Hertz L, Peng L (2013) Require- 249. Vannucci SJ, Maher F, Simpson IA (1997) Glucose tranporter ment of glycogenolysis for uptake of increased extracellular K? proteins in brain: delivery of glucose to neurons and glia. Glia in astrocytes: potential implications for K? homeostasis and 21:2–21 glycogen usage in brain. Neurochem Res 38:472–485 250. Morgello S, Uson RR, Schwartz EJ, Haber RS (1995) The 270. Nelson T, Lucignani G, Atlas S, Crane A, Dienel G, Sokoloff human blood-brain barrier glucose transporter (GLUT 1) is a L (1985) Reexamination of glucose-6-phosphatase activity in glucose transporter of gray matter astrocytes. Glia 14:43–54 the brain in vivo: no evidence for a futile cycle. Science 251. Zeller K, Rahner-Welsch S, Kuschinsky W (1997) Distribution 229:60–62 of GLUT1 glucose tranporters in different brain structures 271. Vilchez D, Ros S, Cifuentes D, Pujadas L, Valles J, Garcia- compared to glucose utilization and capillary density of adult rat Fojeda B, Criado-Garcia O, Fernandez-Sanchez E, Medrano- brains. J Cereb Blood Flow Metab 17:204–209 Fernandez I, Dominguez J, Garcia-Rocha M, Soriano E, De 252. Vannucci SJ, Clark RR, Koehler-Stec E, Li K, Smith CB, Cordoba SR, Guinovart JJ (2007) Mechanism suppressing gly- Davies P, Maher F, Simpson IA (1998) Glucose tranporter cogen synthesis in neurons and its demise in progressive expression in brain: relationship to cerebral glucose utilisation. myoclonus epilepsy. Nat Neurosci 10:1407–1413 Dev Neurosci 20:369–379 272. Choi CH, Coupland NJ, Bhardwaj PP, Kalra S, Casault CA, 253. Bachelard HS, Daniel PM, Love ER, Pratt OE (1973) The Reid K, Allen PS (2006) T-2 measurement and quantification of transport of glucose into the brain of the rat in vivo. Proc R Soc glutamate in human brain in vivo. Magn Reson Med 56:971–977 Lond B 183:71–82 273. Jardetzky O, Jardetzky CD (1958) Proton magnetic resonance 254. Nagamatsu S, Sawa H, Kamada K, Nakamichi Y, Yoshimoto K, spectra of amino acids. J Biol Chem 233:383–387 Hoshino T (1993) Neuron-specific glucose transporter (NSGT): 274. Field LD, Sternhell S, Kalman JR (2008) Organic structures CNS distribution of GLUT3 rat glucose transporter (RGT3) in from spectra, 4th edn. Wiley, Chichester rat central neurons. FEBS Lett 334:289–295 275. Schubert F, Gallinat J, Seifert F, Rinneberg H (2004) Glutamate 255. Maher F, Davies-Hill TM, Simpson IA (1996) Substrate speci- concentrations in human brain using single voxel proton mag- ficity and kinetic parameters of GLUT3 in rat cerebellar granule netic resonance spectroscopy at 3 Tesla. Neuroimage 21:1762– neurons. Biochem J 315:827–831 1771

123 Neurochem Res (2014) 39:1–36 31

276. Wijtenburg SA, Knight-Scott J (2011) Very short echo time 298. Lin Y, Stephenson MC, Xin L, Napolitano A, Morris PG (2012) improves the precision of glutamate detection at 3T in (1)H Investigating the metabolic changes due to visual stimulation magnetic resonance spectroscopy. J Magn Reson Imag 34:645– using functional proton magnetic resonance spectroscopy at 7 T. 652 J Cereb Blood Flow Metab 32:1484–1495 277. Snyder J, Wilman A (2010) Field strength dependence of 299. Mangia S, Giove F, DiNuzzo M (2012) Metabolic pathways and PRESS timings for simultaneous detection of glutamate and activity-dependent modulation of glutamate concentration in the glutamine from 1.5 to 7 T. J Magn Reson 203:66–72 human brain. Neurochem Res 37:2554–2561 278. Hancu I (2009) Optimized glutamate detection at 3T. J Magn 300. Schaller B, Mekle R, Xin LJ, Kunz N, Gruetter R (2013) Net Reson Imag 30:1155–1162 increase of lactate and glutamate concentration in activated 279. Mullins PG, Chen H, Xu J, Caprihan A, Gasparovic C (2008) human visual cortex detected with magnetic resonance spec- Comparative reliability of proton spectroscopy techniques troscopy at 7 Tesla. J Neurosci Res 91:1076–1083 designed to improve detection of J-coupled metabolites. Magn 301. Sibson NR, Dhankhar A, Mason GF, Rothman DL, Behar KL, Reson Med 60:964–969 Shulman RG (1998) Stoichiometric coupling of brain glucose 280. Provencher SW (1993) Estimation of metabolite concentrations metabolism and glutamatergic neuronal activity. Proc Natl Acad from localised in vivo proton NMR spectra. Magn Reson Med Sci USA 95:316–321 30:672–679 302. Perry TL, Hansen S, Berry K, Mok C, Lesk D (1971) Free amino 281. Naressi A, Couturier C, Devos JM, Janssen M, Mangeat C, de acids and related compounds in biopsies of human brain. J Beer R, Graveron-Demilly D (2001) Java-based graphical user Neurochem 18:521–528 interface for the MRUI quantitation package. Magma 12:141– 303. Oz G, Tkac I (2011) Short-echo, single-shot, full-intensity 152 proton magnetic resonance spectroscopy for neurochemical 282. Tkac I, Oz G, Adriany G, Ugurbil K, Gruetter R (2009) In vivo profiling at 4 T: validation in the cerebellum and brainstem. H-1 NMR spectroscopy of the human brain at high magnetic Magn Reson Med 65:901–910 fields: metabolite quantification at 4T vs. 7T. Magn Reson Med 304. Kassem MNE, Bartha R (2003) Quantitative proton short-echo- 62:868–879 time LASER spectroscopy of normal human white matter and 283. Bennett MR, Balcar VJ (1999) Forty years of amino acid hippocampus at 4 Tesla incorporating macromolecule subtrac- transmission in the brain. Neurochem Int 35:269–280 tion. Magn Reson Med 49:918–927 284. Fonnum F (1984) Glutamate—a neurotransmitter in mammalian 305. Hancu I, Port J (2011) The case of the missing glutamine. NMR brain. J Neurochem 42:1–11 Biomed 24:529–535 285. Yang JH, Xu S, Shen J (2009) Fast isotopic exchange between 306. Prescot A, Richards T, Dager SR, Choi C, Renshaw PF Phase- mitochondria and cytosol in brain revealed by relayed C-13 adjusted echo time (PATE)-averaging 1H MRS: application for magnetization transfer spectroscopy. J Cereb Blood Flow Metab improved glutamine quantification at 2.89T. NMR Biomed 29:661–669 25:1245–1252 286. Yudkoff M (1997) Brain metabolism of branched-chain amino 307. Rae C, Geng G, Williams SR (2012) Going for glutamine: acids. Glia 21:92–98 evaluation of asymmetric PRESS approaches. Proc Int Soc 287. Berl S, Lajtha A, Waelsch H (1961) Amino acid and protein Magn Reson Med 20:1753 metabolism-VI Cerebral compartments of glutamic acid 308. Martinez-Hernandez A, Bell KP, Norenberg MD (1977) Gluta- metabolism. J Neurochem 7:186–197 mine synthetase—glial localization in brain. Science 195:1356– 288. Shank RP, Leo GC, Zielke HR (1993) Cerebral metabolic com- 1358 partmentation as revealed by nuclear magnetic resonance analysis 309. Fernandes SP, Dringen R, Lawen A, Robinson SR (2010) of D-[1-13C]glucose metabolism. J Neurochem 61:315–323 Neurones express glutamine synthetase when deprived of glu- 289. Kreft M, Bak LK, Waagepetersen HS, Schousboe A (2012) tamine or interaction with astrocytes. J Neurochem 114:1527– Aspects of astrocyte energy metabolism, amino acid neuro- 1536 transmitter homoeostasis and metabolic compartmentation. Asn 310. Kanamori K, Ross BD, Chung JC, Kuo EL (1996) Severity of Neuro 4(3):e00086 hyperammonemic encephalopathy correlates with brain ammo- 290. Bordi F, Ugolini A (1999) Group I metabotropic glutamate nia level and saturation of glutamine synthetase in vivo. J receptors: implications for brain diseases. Prog Neurobiol Neurochem 67:1584–1594 59:55–79 311. Peng L, Hertz L, Huang R, Sonnewald U, Petersen SB, West- 291. Kew JNC, Kemp JA (2005) Ionotropic and metabotropic glu- ergaard N, Larsson OM, Schousboe A (1993) Utilization of tamate receptor structure and pharmacology. Psychopharmacol glutamine and of TCA cycle constituents as precursors for 179:4–29 transmitter glutamate and GABA. Dev Neurosci 15:367–377 292. Danbolt NC (2001) Glutamate uptake. Prog Neurobiol 35:1–105 312. Lieth E, LaNoue KF, Berkich DA, Xu B, Ratz M, Taylor C, 293. Albrecht P, Lewerenz J, Dittmer S, Noack R, Maher P, Methner Hutson SM (2001) Nitrogen shuttling between neurons and glial A (2010) Mechanisms of oxidative glutamate toxicity: the glu- cells during glutamate synthesis. J Neurochem 76:1712–1723 tamate/cystine antiporter system x(c)(-) as a neuroprotective 313. Sibson NR, Mason GF, Shen J, Cline GW, Herskovits AZ, Wall drug target. CNS Neurol Disord-Drug Targets 9:373–382 JEM, Behar KL, Rothman DL, Shulman RG (2001) In vivo 13C 294. Kalivas PW (2009) The glutamate homeostasis hypothesis of NMR measurement of neurotransmitter glutamate cycling, addiction. Nat Rev Neurosci 10:561–572 anaplerosis and TCA cycle flux in rat brain during [2-13C]glu- 295. Koga M, Serritella AV, Messmer MM, Hayashi-Takagi A, cose infusion. J Neurochem 76:975–989 Hester LD, Snyder SH, Sawa A, Sedlak TW (2011) Glutathione 314. Badar-Goffer R, Bachelard H, Morris P (1990) Cerebral is a physiologic reservoir of neuronal glutamate. Biochem metabolism of acetate and glucose studied by 13C NMR spec- Biophys Res Commun 409:596–602 troscopy. Biochem J 266:133–139 296. Choi DW (1987) Ionic dependence of glutamate neurotoxicity. J 315. Broër S, Brookes N (2001) Transfer of glutamine between Neurosci 7:369–379 astrocytes and neurons. J Neurochem 77:705–719 297. Dienel GA (2013) Astrocytic energetics during excitatory neu- 316. Deitmer JW, Broër A, Broër S (2003) Glutamine efflux from rotransmission: what are contributions of glutamate oxidation astrocytes is mediated by multiple pathways. J Neurochem and glycolysis? Neurochem Int 63:244–258 87:127–135

123 32 Neurochem Res (2014) 39:1–36

317. Walter M, Henning A, Grimm S, Schulte RF, Beck J, Dydak U, 334. Makar TK, Nedergaard M, Preuss A, Gelbard AS, Perumal AS, Schnepf B, Boeker H, Boesiger P, Northoff G (2009) The Cooper AJL (1994) Vitamin-E, ascorbate, glutathione, gluta- relationship between aberrant neuronal activation in the prege- thione disulfide, and enzymes of glutathione metabolism in nual anterior cingulate, altered glutamatergic metabolism, and cultures of chick astrocytes and neurons—evidence that astro- anhedonia in major depression. Arch Gen Psychiatr 66:478–486 cytes play an important role in antioxidative processes in the 318. Rowland LM, Bustillo JR, Mullins PG, Jung RE, Lenroot R, brain. J Neurochem 62:45–53 Landgraf E, Barrow R, Yeo R, Lauriello J, Brooks WM (2005) 335. Srinivasan R, Ratiney H, Hammond-Rosenbluth KE, Pelletier D, Effects of ketamine on anterior cingulate glutamate metabolism Nelson SJ (2010) MR spectroscopic imaging of glutathione in in healthy humans: a 4-T proton MRS study. Am J Psychiatr the white and gray matter at 7 T with an application to multiple 162:394–396 sclerosis. Magn Reson Imag 28:163–170 319. Moore CM, Wardrop M, Frederick BD, Renshaw PF (2006) 336. Harish G, Venkateshappa C, Mahadevan A, Pruthi N, Bharath Topiramate raises anterior cingulate cortex glutamine levels in MMS, Shankar SK (2011) Glutathione metabolism is modulated healthy men; a 4.0 T magnetic resonance spectroscopy study. by postmortem interval, gender difference and agonal state in Psychopharmacol 188:236–243 postmortem human brains. Neurochem Int 59:1029–1042 320. Henry ME, Jensen JE, Licata SC, Ravichandran C, Butman ML, 337. Meister A, Anderson ME (1983) Glutathione. Ann Rev Biochem Shanahan M, Lauriat TL, Renshaw PF (2010) The acute and late 52:711–760 CNS glutamine response to benzodiazepine challenge: a pilot 338. Meister A (1974) Glutathione—metabolism and function via pharmacokinetic study using proton magnetic resonance spec- gamma-glutamyl cycle. Life Science 15:177–190 troscopy. Psychiatr Res Neuroimag 184:171–176 339. Anderson ME (1998) Glutathione: an overview of biosynthesis 321. Petroff OAC, Pleban LA, Spencer DD (1995) Symbiosis and modulation. Chem Biol Interact 112:1–14 between in vivo and in vitro NMR spectroscopy—the creatine, 340. Janaky R, Varga V, Hermann A, Saransaari P, Oja SS (2000) N-acetylaspartate, glutamate and GABA content of the epileptic Mechanisms of L-cysteine neurotoxicity. Neurochem Res human brain. Magn Reson Imag 13:1197–1211 25:1397–1405 322. Zhang NY, Laake J, Nagelhus E, Storm-Mathisen J, Ottersen OP 341. Vitvitsky V, Thomas M, Ghorpade A, Gendelman HE, Banerjee (1991) Distribution of glutamine-like immunoreactivity in the R (2006) A functional transsulfuration pathway in the brain links cerebellum of rat and baboon (Papio anubis) with reference to to glutathione homeostasis. J Biol Chem 281:35785–35793 the issue of metabolic compartmentation. Anat Embryol 342. Kranich O, Dringen R, Sandberg M, Hamprecht B (1998) Uti- 184:213–223 lization of cysteine and cysteine precursors for the synthesis of 323. Hertz L, Dringen R, Schousboe A, Robinson SR (1999) Astro- glutathione in astroglial cultures: preference for cystine. Glia cytes: glutamate producers for neurons. J Neurosci Res 57:417– 22:11–18 428 343. Wang XF, Cynader MS (2000) Astrocytes provide cysteine to 324. Terpstra M, Henry PG, Gruetter R (2003) Measurement of neurons by releasing glutathione. J Neurochem 74:1434–1442 reduced glutathione (GSH) in human brain using LCmodel ana- 344. Dringen R, Pfeiffer B, Hamprecht B (1999) Synthesis of the lysis of difference-edited spectra. Magn Reson Med 50:19–23 antioxidant glutathione in neurons: supply by astrocytes of Cys- 325. Kaiser LG, Marjanska M, Matson GB, Iltis I, Bush SD, Soher Gly as precursor for neuronal glutathione. J Neurosci 19:562–569 BJ, Mueller S, Young K (2010) (1)H MRS detection of glycine 345. Sagara J, Miura K, Bannai S (1993) Maintenance of neuronal residue of reduced glutathione in vivo. J Magn Reson 202:259– glutathione by glial-cells. J Neurochem 61:1672–1676 266 346. O’Brien PJ, Siraki AG, Shangari N (2005) Aldehyde sources, 326. Terpstra M, Vaughan TJ, Ugurbil K, Lim KO, Schulz SC, metabolism, molecular toxicity mechanisms, and possible Gruetter R (2005) Validation of glutathione quantitation from effects on human health. Crit Rev Toxicol 35:609–662 STEAM spectra against edited H-1 NMR spectroscopy at 4T: 347. Heales SJR, Davies SEC, Bates TE, Clark JB (1995) Depletion application to schizophrenia. Magn Reson Mater Phys, Biol Med of brain glutathione is accompanied by impaired mitochondrial- 18:276–282 function and decreased N-acetyl aspartate concentration. Neu- 327. Maher P (2005) The effects of stress and aging on glutathione rochem Res 20:31–38 metabolism. Age Res Rev 4:288–314 348. Holub BJ (1986) Metabolism and function of myo-inositol and 328. Satoh T, Yoshioka Y (2006) Contribution of reduced and oxi- inositol phospholipids. Annu Rev Nutr 6:563–597 dized glutathione to signals detected by magnetic resonance 349. Cerdan S, Parrilla R, Santoro J, Rico M (1985) 1H NMR spectroscopy as indicators of local brain redox state. Neurosci detection of cerebral myo-inositol. FEBS Lett 187:167–172 Res 55:34–39 350. Hancu I, Gillen R, Cowan J, Zimmerman EA (2011) Improved 329. Ballatori N, Krance SM, Notenboom S, Shi SJ, Tieu K, Ham- myo-inositol detection through Carr-Purcell PRESS: a tool for mond CL (2009) Glutathione dysregulation and the etiology and more sensitive mild cognitive impairment diagnosis. Magn Re- progression of human diseases. Biol Chem 390:191–214 son Med 65:1515–1521 330. Schafer FQ, Buettner GR (2001) Redox environment of the cell 351. Minati L, Aquino D, Bruzzone MG, Erbetta A (2010) Quanti- as viewed through the redox state of the glutathione disulfide/ tation of normal metabolite concentrations in six brain regions glutathione couple. Free Radical Biol Med 30:1191–1212 by in vivo 1H-MR spectroscopy. J Med Phys 35:154–163 331. Rice ME, RussoMenna I (1998) Differential compartmentali- 352. Haris M, Cai KJ, Singh A, Hariharan H, Reddy R (2011) In vivo zation of brain ascorbate and glutathione between neurons and mapping of brain myo-inositol. Neuroimage 54:2079–2085 glia. Neuroscience 82:1213–1223 353. Fisher SK, Novak JE, Agranoff BW (2002) Inositol and higher 332. Keelan J, Allen NJ, Antcliffe D, Pal S, Duchen MR (2001) inositol phosphates in neural tissues: homeostasis, metabolism Quantitative imaging of glutathione in hippocampal neurons and and functional significance. J Neurochem 82:736–754 glia in culture using monochlorobimane. J Neurosci Res 354. Belmaker RH, Agam G, van Calker D, Richards MH, Kofman O 66:873–884 (1998) Behavioural reversal of lithium effects by four inositol 333. Langeveld CH, Schepens E, Jongenelen CAM, Stoof JC, Hjelle isomers correlates perfectly with biochemical effects on the PI OP, Ottersen OP, Drukarch B (1996) Presence of glutathione cycle. Neuropsychpharmacol 19:220–232 immunoreactivity in cultured neurones and astrocytes. Neuro- 355. Glanville NT, Byers DM, Cook HW, Spence MW, Palmer Rep 7:1833–1836 FBSC (1989) Differences in the metabolism of inositol and

123 Neurochem Res (2014) 39:1–36 33

phosphoinositides by cultured cells of neuronal and glial origin. 376. Berry GT, Wu S, Buccafusca R, Ren J, Gonzales LW, Ballard Biochim Biophys Acta 1004:169–179 PL, Golden JA, Stevens MJ, Greer JJ (2003) Loss of murine 356. Wiesinger H (1991) Myo-inositol transport in mouse astroglia- Na?/myo-inositol cotransporter leads to brain myo-inositol rich primary cultures. J Neurochem 56:1698–1704 depletion and central apnea. J Biol Chem 278:18297–18302 357. Novak JE, Turner RS, Agranoff BW, Fisher SK (1999) Differ- 377. Guo W, Shimada S, Tajiri H, Yamauchi A, Yamashita T, Okada entiated human NT2-N neurons possess a high intracellular S, Tohyama M (1997) Developmental regulation of Na?/myo- content of myo-inositol. J Neurochem 72:1431–1440 inositol cotransporter gene expression. Mol Brain Res 51:91–96 358. Wong Y-HH, Kalmbach SJ, Hartmann BK, Sherman WR (1987) 378. Battaglia FC, Meschia JN, Blechnew JN, Barron DH (1961) The Immunohistochemical staining and enzyme activity measure- free myo-inositol concentration of adult and fetal tissues of ments show myo-inositol-1-phosphate synthase to be localised several species. Q J Exp Med 46:188–193 in the vasculature of the brain. J Neurochem 48:1434–1442 379. Coady MJ, Wallendorff B, Gagnon DG, Lapointe JY (2002) 359. Kim JP, Lentz MR, Westmoreland SV, Greco JB, Ratai EM, Identification of a novel Na?/myo-inositol cotransporter. J Biol Halpern E, Lackner AA, Masliah E, Gonzalez RG (2005) Chem 277:35219–35224 Relationships between astrogliosis and H-1 MR spectroscopic 380. Bissonnette P, Coady MJ, Lapointe JY (2004) Expression of the measures of brain choline/creatine and myo-inositol/creatine in sodium-myo-inositol cotransporter SMIT2 at the apical mem- a primate model. Am J Neuroradiol 26:752–759 brane of Madin-Darby canine kidney cells. J Physiol-Lond 360. Vadnal R, Parthasarathy L, Parthasarathy R (1997) Role of 558:759–768 inositol in the treatment of psychiatric disorders. CNS Drugs 381. Uldry M, Ibberson M, Horisberger JD, Chatton JY, Riederer 7:6–16 BM, Thorens B (2001) Identification of a mammalian H?-myo- 361. Gani D, Downes CP, Batty I, Bramham J (1993) Lithium and inositol symporter expressed predominantly in the brain. EMBO myo-inositol homeostasis. Biochim Biophys Acta 1177:253–269 J 20:4467–4477 362. Futerman AH, Low MG, Ackerman KE, Sherman WR, Silman I 382. Uldry M, Steiner P, Zurich MG, Beguin P, Hirling H, Dolci W, (1985) Identification of covalently bound inositol in the hydro- Thorens B (2004) Regulated exocytosis of an H?/myo-inositol phobic membrane-anchoring domain of Torpedo acetylcholin- symporter at synapses and growth cones. EMBO J 23:531–540 esterase. Biochem Biophys Res Commun 129:312–317 383. Moore CM, Breeze JL, Kukes TL, Rose SL, Dager SR, Cohen 363. Elmallakh RS, Li R (1993) Is the Na?–K?-ATPase the link BM, Renshaw PF (1999) Effects of myo-inositol ingestion on between phosphoinositide metabolism and bipolar disorder. J human brain myo-inositol levels: a proton magnetic resonance Neuropsychiatr Clin Neurosci 5:361–368 spectroscopic imaging study. Biol Psychiatr 45:1197–1202 364. Jones DR, Varela-Nieto I (1999) Diabetes and the role of ino- 384. Lubrich B, Spliess O, Gebicke-Haerter P-J, van Calker D (2000) sitol-containing lipids in insulin signaling. Mol Med 5:505–514 Differential expression, activity and regulation of the sodium/ 365. Levine J, Chengappa KNR, Reddy R (1999) Acute myoinositol myo-inositol cotransporter in astrocyte cultures from different enhances swimming activity in goldfish. J Neural Transm regions of the rat brain. Neuropharmacol 39:680–690 106:433–441 385. Shonk T, Ross BD (1995) Role of increased cerebral myo-ino- 366. Fux M, Levine J, Aviv A, Belmaker RH (1996) Inositol treat- sitol in the dementia of down syndrome. Magn Reson Med ment of obsessive-compulsive disorder. Am J Psychiatr 33:858–861 153:1219–1221 386. Hattori M, Fujiyama A, Taylor TD, Watanabe H, Yada T, Park 367. Strange K, Morrison R, Shrode L, Putnam R (1993) Mechanism HS, Toyoda A, Ishii K, Totoki Y, Choi DK, Soeda E, Ohki M, and regulation of swelling-activated inositol efflux in brain glial Takagi T, Sakaki Y, Taudien S, Blechschmidt K, Polley A, cells. Am J Physiol 265:C244–C256 Menzel U, Delabar J, Kumpf K, Lehmann R, Patterson D, 368. Haüssinger D, Laubenberger J, Vom Dahl S, Ernst T, Bayer S, Reichwald K, Rump A, Schillhabel M (2000) The DNA Langer M, Gerok W, Hennig J (1994) Proton magnetic reso- sequence of human chromosome 21. Nature 405:311–319 nance spectroscopy studies on human brain myo-inositol in 387. Garnett MR, Blamire AM, Corkill RG, Cadoux-Hudson TAD, hypo-osmolarity and hepatic encephalopathy. Gastroenterol Rajagopalan B, Styles P (2000) Early proton magnetic reso- 107:1475–1480 nance spectroscopy in normal-appearing brain correlates with 369. Trachtman H, Futterweit S, Hammer E, Siegel TW, Oates P outcome in patients following traumatic brain injury. Brain (1991) The role of polyols in cerebral cell volume regulation in 123:2046–2054 hypernatremic and hyponatremic states. Life Science 49:677–688 388. McLaurin J, Golomb R, Jurewicz A, Antel JP, Fraser PE (2000) 370. Rumpel H, Lim WEH, Chang HM, Chan LL, Ho GL, Wong Inositol stereoisomers stabilize an oligomeric aggregate of MC, Tan KP (2003) Is myo-inositol a measure of glial swelling Alzheimer amyloid beta peptide and inhibit A beta-induced after stroke? A magnetic resonance study. J Magn Reson Imag toxicity. J Biol Chem 275:18495–18502 17:11–19 389. Bersudsky Y, Kaplan Z, Shapiro Y, Agam G, Kofman O, Bel- 371. Ashwal S, Holshouser B, Tong K, Serna T, Osterdock R, Gross maker RH (1994) Behavioural evidence for the existence of two M, Kido D (2004) Proton spectroscopy detected myoinositol in pools of cellular inositol. Eur Neuropsychopharmacol 4:463– children with traumatic brain injury. Pediatr Res 56:630–638 467 372. Rango M, Cogiamanian F, Marceglia S, Barberis B, Arighi A, 390. Michaelis T, Helms G, Merboldt K-D, Ha¨nicke W, Bruhn H, Biondetti P, Priori A (2008) Myoinositol content in the human brain Frahm J (1993) Identification of scyllo-inositol in proton NMR is modified by transcranial direct current stimulation in a matter of spectra of human brain in vivo. NMR Biomed 6:105–109 minutes: a 1H-MRS study. Magn Reson Med 60:782–789 391. Hipps PP, Holland WH, Sherman WR (1977) Interconversion of 373. Spector R (1988) Myo-inositol transport through the blood brain myoinositol and scyllo-inositol with simultaneous formation of barrier. Neurochemical Res 13:785–787 neo-inositol by an NADP? dependent epimerase from bovine 374. Levine J, Rapaport A, Lev L, Bersudsky Y, Kofman O, Bel- brain. Biochem Biophys Res Commun 77:340–346 maker RH, Shapiro J, Agam G (1993) Inositol treatment raises 392. Viola A, Nicoli F, Denis B, Confort-Gouny S, Le Fur Y, CSF inositol levels. Brain Res 627:168–170 Ranjeva JP, Viout P, Cozzone PJ (2004) High cerebral scyllo- 375. Patishi Y, Lubrich B, Berger M, van Calker D, Kofman O, inositol: a new marker of brain metabolism disturbances induced Belmaker RH (1996) Differential uptake of myo-inositol in vivo by chronic alcoholism. Magn Reson Mater Phys, Biol Med into rat brain areas. Eur Neuropsychopharmacol 6:73–75 17:47–61

123 34 Neurochem Res (2014) 39:1–36

393. Seaquist ER, Gruetter R (1998) Identification of a high con- 410. Halestrap AP, Denton RM (1974) Specific inhibition of pyruvate centration of scyllo-inositol in the brain of a healthy human transport in rat liver mitochondria and human erythrocytes by subject using H-1 and C-13 NMR. Magn Reson Med 39:313– a-cyano-4-hydroxycinnamate. Biochem J 138:313–316 316 411. Martin PM, Dun Y, Mysona B, Ananth S, Roon P, Smith SB, 394. Choi JK, Carreras I, Dedeoglu A, Jenkins BG (2010) Detection Ganapathy V (2007) Expression of the sodium-coupled mono- of increased scyllo-inositol in brain with magnetic resonance carboxylate transporters SMCT1 (SLC5A8) and SMCT2 spectroscopy after dietary supplementation in Alzheimer’s dis- (SLC5A12) in retina. Invest Ophthalmol Vis Sci 48:3356–3363 ease mouse models. Neuropharmacol 59:353–357 412. Cornford EM, Cornford ME (1986) Nutrient transport and the 395. Fenili D, Brown M, Rappaport R, McLaurin J (2007) Properties blood brain barrier in developing animals. Fed Proc 45:2065– of scyllo-inositol as a therapeutic treatment of AD-like pathol- 2072 ogy. J Mol Med-Jmm 85:603–611 413. Medina JM, Tabernero A, Tovar A, Martin-Barrientos J (1996) 396. Nozadze M, Mikautadze E, Lepsveridze E, Mikeladze E, Ku- Metabolic fuel utilisation and pyruvate oxidation during the chiashvili N, Kiguradze T, Kikvidze M, Solomonia R (2011) postnatal period. J Inher Metab Disord 19:432–442 Anticonvulsant activities of myo-inositol and scyllo-inositol on 414. Cremer JE, Braun LD, Oldendorf WH (1976) Changes during pentylenetetrazol induced seizures. Seizure-Eur J Epilepsy development in transport processes of the blood-brain barrier. 20:173–176 Biochim Biophys Acta 448:633–637 397. Bittar PG, Charnay Y, Pellerin L, Bouras C, Magistretti PJ 415. Magistretti PJ, Pellerin L (1996) Cellular bases of brain energy (1996) Selective distribution of lactate dehydrogenase isoen- metabolism and their relevance to functional brain imaging: zymes in neurons and astrocytes of human brain. J Cereb Blood evidence for a prominent role of astrocytes. Cereb Cortex 6:50– Flow Metab 16:1079–1089 61 398. Kashiwaya Y, Sato K, Tsuchiya N, Thomas S, Fell DA, Veech 416. Silver IA, Erecinska M (1994) Extracellular glucose concen- RA, Passonneau JV (1994) Control of glucose utilization in tration in mammalian brain: continuous monitoring of changes working perfused rat heart. J Biol Chem 269:25502–25514 during increased neuronal activity and upon limitation in oxygen 399. Quistorff B, Grunnet N (2011) The isoenzyme pattern of LDH supply in normo-, hypo- and hyperglycaemic animals. J Neu- does not play a physiological role; except perhaps during fast rosci 14:5068–5076 transitions in energy metabolism. Aging-US 3:457–460 417. Fox PT, Raichle ME (1986) Focal physiological uncoupling of 400. Deuticke B, Rickert I, Beyer E (1978) Stereoselective, SH- cerebral blood flow and oxidative metabolism during somato- dependent transfer of lactate in mammalian erythrocytes. Bio- sensory stimulation in human subjects. Proc Natl Acad Sci chim Biophys Acta 507:137–155 (USA) 83:1140–1144 401. Coady MJ, Chang MH, Charron FA, Plata C, Wallendorff B, 418. Fox PT, Raichle ME, Mintun MA, Dence C (1988) Non-oxi- Sah JF, Markowitz SD, Romero ME, Lapointe JY (2004) The dative glucose consumption during focal physiological neuronal human tumour suppressor gene SLC5A8 expresses a Na?- activity. Science 241:462–464 monocarboxylate cotransporter. J Physiol-Lond 557:719–731 419. Lowry JP, Fillenz M (1997) Evidence for uncoupling of oxygen 402. Gopal E, Umapathy NS, Martin PM, Ananth S, Gnana-Prakasam and glucose utilization during neuronal activation in rat stria- JP, Becker H, Wagner CA, Ganapathy V, Prasad PD (2007) tum. J Physiol 498(2):497–501 Cloning and functional characterization of human SMCT2 420. Madsen PL, Cruz NF, Sokoloff L, Dienel GA (1999) Cerebral (SLC5A12) and expression pattern of the transporter in kidney. oxygen/glucose ratio is low during sensory stimulation and rises Biochimica Et Biophysica Acta-Biomembranes 1768:2690– above normal during recovery: excess glucose consumption 2697 during stimulation is not accounted for by lactate efflux from or 403. Broër S, Rahman B, Pellegri G, Pellerin L, Martin JL, Ver- accumulation in brain tissue. J Cereb Blood Flow Metab leysdonk S, Hamprecht B, Magistretti PJ (1997) Comparison of 19:393–400 lactate transport in astroglial cells and transporter 1 (MCT 1) 421. Contreras L, Satrustegui J (2009) Calcium signaling in brain expressing Xenopus laevis oocytes. Expression of two different mitochondria. Interplay of malate aspartate NADH shuttle and monocarboxylate transporters in astroglial cells and neurons. J calcium uniporter/mitochondrial dehydrogenase pathways. J Biol Chem 272:30096–30102 Biol Chem 284:7091–7099 404. Martin PM, Gopal E, Ananth S, Zhuang L, Itagaki S, Prasad 422. Dienel GA, Hertz L (2001) Glucose and lactate metabolism BM, Smith SB, Prasad PD, Ganapathy V (2006) Identify of during brain activation. J Neurosci Res 66:824–838 SMCT1(SLC5A8) as a neuron-specific Na? -coupled trans- 423. Cruz NF, Ball KK, Dienel GA (2007) Functional imaging of porter for active uptake of L-lactate and ketone bodies in the focal brain activation in conscious rats: impact of [C-14]glucose brain. J Neurochem 98:279–288 metabolite spreading and release. J Neurosci Res 85:3254–3266 405. Dringen R, Wiesinger H, Hamprecht B (1993) Uptake of L- 424. Dienel G, Ball K, Popp D, Cruz NF (2001) A role for gap lactate by cultured rat brain neurons. Neurosci Lett 163:5–7 junctions in metabolite spreading? J Neurochem 78(suppl 1):86 406. Walz W, Mukerji S (1988) Lactate release from cultured 425. Cruz NF, Adachi K, Dienel GA (1999) Rapid efflux of lactate astrocytes and neurons: a comparison. Glia 1:366–370 from cerebral cortex during K?-induced spreading cortical 407. Herzig S, Raemy E, Montessuit S, Veuthey JL, Zamboni N, depression. J Cereb Blood Flow Metab 19:380–392 Westermann B, Kunji ERS, Martinou JC (2012) Identification 426. Ball KK, Cruz NF, Mrak RE, Dienel GA (2010) Trafficking of and functional expression of the mitochondrial pyruvate carrier. glucose, lactate, and amyloid-beta from the inferior colliculus Science 337:93–96 through perivascular routes. J Cereb Blood Flow Metab 30:162–176 408. Bricker DK, Taylor EB, Schell JC, Orsak T, Boutron A, Chen 427. Mangia S, Simpson IA, Vannucci SJ, Carruthers A (2009) The YC, Cox JE, Cardon CM, Van Vranken JG, Dephoure N, Redin in vivo neuron-to-astrocyte lactate shuttle in human brain: evi- C, Boudina S, Gygi SP, Brivet M, Thummel CS, Rutter JA dence from modeling of measured lactate levels during visual (2012) Mitochondrial pyruvate carrier required for pyruvate stimulation. J Neurochem 109:55–62 uptake in yeast, drosophila, and humans. Science 337:96–100 428. Levasseur JE, Alessandri B, Reinert M, Clausen T, Zhou ZW, 409. Halestrap AP (1975) The mitochondrial pyruvate carrier. Altemeni N, Bullock MR (2006) Lactate not glucose, up-regu- Kinetics and specificity for substrates and inhibitors. Biochem J lates mitochondrial oxygen consumption both in sham and lat- 148:85–96 eral fluid percussed rat brains. Neurosurgery 59:1122–1130

123 Neurochem Res (2014) 39:1–36 35

429. Gonzalez SV, Nguyen NHT, Rise F, Hassel B (2005) Brain 1 NMR spectroscopy in the human visual cortex. J Cereb Blood metabolism of exogenous pyruvate. J Neurochem 95:284–293 Flow Metab 27:1055–1063 430. Schurr A, West CA, Rigor BM (1988) Lactate-supported syn- 448. Sarchielli P, Tarducci R, Presciutti O, Gobbi G, Pelliccioli GP, aptic function in the rat hippocampal slices preparation. Science Stipa G, Alberti A, Capocchi G (2005) Functional H-1-MRS 240:1326–1328 findings in migraine patients with and without aura assessed 431. Schurr A, Payne RS, Miller JJ, Rigor BM (1997) Brain lactate is interictally. Neuroimage 24:1025–1031 an obligatory aerobic substrate for functional recovery after 449. Richards TL, Dager SR, Corina D, Serafini S, Heide AC, hypoxia: further in vitro validation. J Neurochem 69:423–426 Steury K, Strauss W, Hayes CE, Abbott RD, Craft S, Shaw D, 432. Cox DWG, Bachelard HS (1988) Partial attenuation of dentate Posse S, Berninger VW (1999) Dyslexic children have granule cell evoked activity by the alternative substrates, lactate abnormal brain lactate response to reading-related language and pyruvate: evidence for a postsynaptic action. Exp Brain Res tasks. Am J Neuroradiol 20:1393–1398 69:368–372 450. Richards TL, Corina D, Serafini S, Steury K, Echelard DR, 433. Yamane K, Yokono K, Okada Y (2000) Anaerobic glycolysis is Dager SR, Marro K, Abbott RD, Maravilla KR, Berninger VW crucial for the maintenance of neural activity in guinea pig (2000) Effects of a phonologically driven treatment for dyslexia hippocampal slices. J Neurosci Meth 103:163–171 on lactate levels measured by proton MR spectroscopy imaging. 434. McKenna MC, Hopkins IB, Carey A (2001) Alpha-cyano-4- Am J Neuroradiol 21:916–922 hydroxycinnamate decreases both glucose and lactate metabo- 451. Gjedde A, Poulsen PH, Østergaard L (1999) On the oxygenation lism in neurons and astrocytes: implications for lactate as an of hemoglobin in the human brain. Adv Exp Med Biol 471:81– energy substrate for neurons. J Neurosci Res 66:747–754 97 435. Petroff OA, Prichard JW, Behar KL, Rothman DL, Alger JR, 452. Maher AD, Solo V, Rae CD (2013) Magnetic resonance-based Shulman RG (1985) Cerebral metabolism in hyper- and hypo- metabolomics for understanding neurological disorders: current carbia: 31P and 1H nuclear magnetic resonance studies. Neu- status and considerations. Curr Metabolomics 1:2–14 rology 35:1681–1688 453. Tisell A, Dahlqvist LO, Warntjes J, Lundberg P (2013) Proce- 436. Rosenberg AA (1988) Response of the cerebral circulation to dure for quantitative 1H MRS and tissue characterization of profound hypocarbia in neonatal lambs. Stroke 19:1365–1370 human brain tissue based on the use of quantitative MRI. Magn 437. Maddock RJ, Casazza GA, Buonocore MH, Tanase C (2011) Reson Med 70:905–915 Vigorous exercise increases brain lactate and Glx (glutamate plus 454. Scheidigger O, Wingeier K, Stefan D, Graveron-Demilly D, van glutamine): a dynamic 1H-MRS study. Neuroimage 57:1324– Ormondt D, Wiest R, Slotboom J (2013) Optimized quantitative 1330 magnetic resonance spectroscopy for clinical routine. Magn Re- 438. Dalsgaard MK, Quistorff B, Danielsen ER, Selmer C, Vogel- son Med 70:25–32 sang T, Secher NH (2004) A reduced cerebral metabolic ratio in 455. Steen RG, Ogg RJ (2005) Abnormally high levels of brain N- exercise reflects metabolism and not accumulation of lactate acetylaspartate in children with sickle cell disease. Am J Neu- within the human brain. J Physiol-Lond 554:571–578 roradiol 26:463–468 439. Dager SR, Marro KI, Richards TL, Metzger GD (1992) Local- 456. Hwang JH, Graham GD, Behar KL, Alger JR, Prichard JW, ised magnetic resonance spectroscopy measurement of brain Rothman DL (1996) Short echo time proton magnetic resonance lactate during intravenous lactate infusion in healthy volunteers. spectroscopic imaging of macromolecule and metabolite signal Life Science 51:973–985 intensities in the human brain. Magn Reson Med 35:633–639 440. Naylor E, Aillon DV, Barrett BS, Wilson GS, Johnson DA, 457. Mader I, Seeger U, Karitzky J, Erb M, Schick F, Klose U (2002) Johnson DA, Harmon HP, Gabbert S, Petillo PA (2012) Lactate Proton magnetic resonance spectroscopy with metabolite nulling as a biomarker for sleep. Sleep 35:1209–1222 reveals regional differences of macromolecules in normal 441. Edden RAE, Harris AD, Murphy K, Evans CJ, Saxena N, Hall human brain. J Magn Reson Imag 16:538–546 JE, Bailey DM, Wise RG (2010) Edited MRS is sensitive to 458. Seeger U, Klose U, Mader I, Grodd W, Nagele T (2003) changes in lactate concentration during inspiratory hypoxia. J Parameterized evaluation of macromolecules and lipids in Magn Reson Imag 32:320–325 proton MR spectroscopy of brain diseases. Magn Reson Med 442. Ben-Joseph O, Bader-Gofer RS, Morris PG, Bachelard HS 49:19–28 (1993) Glycerol-3-phosphate and lactate as indicators of cyto- 459. Miyake M, Morino H, Mizobuchi M, Kakimoto Y (1982) N- plasmic redox state in severe and mild hypoxia respectively; a Acetyl-L-aspartic acid, N-acetyl-alpha-L-aspartyl-L-glutamic 13C and 31P NMR study. Biochem J 291:915–919 acid and beta-citryl-L-glutamic acid in human urine. Clin Chim 443. Van Rijen PC, Luyten PR, Berkelbach van der Sprenkel JW, Acta 120:119–126 Kraaier V, van Huffelen AC, Tulleken AF, den Hollander JA 460. Ruggieri M, Tortorella C, Ceci E, Paolicelli D, Solfrizzi V, Di (1989) 1H and 31P NMR measurements of cerebral lactate, Bitonto G, Pica C, Mastrapasqua M, Livrea P, Trojano M (2011) high-energy phosphate levels and pH in humans during volun- Age-related changes of serum N-acetyl-aspartate in healthy tary hyperventilation: associated EEG, capnographic and controls. Age Ageing 40:391–395 Doppler findings. Magn Reson Med 10:182–193 461. Moffett JR, Ross B, Arun P, Madhavarao CN, Namboodiri 444. Bell JD, Brown JC, Kubal G, Sadler PJ (1988) NMR-invisible AMA (2007) N-acetylaspartate in the CNS: from neurodiag- lactate in blood plasma. FEBS Lett 235:81–86 nostics to neurobiology. Prog Neurobiol 81:89–131 445. Williams SR, Proctor E, Allen K, Gadian DG, Crockard HA 462. Beárd E, Braissant O (2010) Synthesis and transport of creatine (1988) Quantitative estimation of lactate in the brain by 1H in the CNS: importance for cerebral functions. J Neurochem NMR. Magn Reson Med 7:425–431 115:297–313 446. Graham GD, Blamire AM, Rothman DL, Brass LM, Fayad PB, 463. Kottke M, Wallimann T, Brdiczka D (1994) Dual electron Petroff OA, Prichard JW (1993) Early temporal variation of microscopic localization of mitochondrial creatine kinase in cerebral metabolites after human stroke. A proton magnetic brain mitchondria. Biochem Med Metab Biol 51:105–117 resonance spectroscopy study. Stroke 24:1891–1896 464. Nies AT, Jedlitschky G, Konig J, Herold-Mende C, Steiner HH, 447. Mangia S, Tkac I, Gruetter R, Van de Moortele PF, Maraviglia Schmitt HP, Keppler D (2004) Expression and immunolocali- B, Ugurbil K (2007) Sustained neuronal activation raises oxi- zation of the multidrug resistance proteins, Mrp1-Mrp6 dative metabolism to a new steady-state level: evidence from H- (ABCC1–ABCC6), in human brain. Neuroscience 129:349–360

123 36 Neurochem Res (2014) 39:1–36

465. Aoyama K, Suh SW, Hamby AM, Liu JL, Chan WY, Chen YM, 467. Rae C, Fekete AD, Kashem MA, Nasrallah FA, Nasrallah FA, Swanson RA (2006) Neuronal glutathione deficiency and age- Broër S (2012) Metabolism, compartmentation, transport and dependent neurodegeneration in the EAAC1 deficient mouse. production of acetate in the cortical brain tissue slice. Neuro- Nat Neurosci 9:119–126 chem Res 37:2541–2553 466. Shanker G, Allen JW, Mutkus LA, Aschner M (2001) The uptake of cysteine in cultured primary astrocytes and neurons. Brain Res 902:156–163

123 Journal of Magnetic Resonance 136, 143–151 (1999) Article ID jmre.1998.1628, available online at http://www.idealibrary.com on

Analysis of J Coupling-Induced Fat Suppression in DIET Imaging

L. A. Stables,* R. P. Kennan,† A. W. Anderson,*,† R. T. Constable,† and J. C. Gore*,† *Department of Applied Physics and †Department of Diagnostic Radiology, Yale University School of Medicine, New Haven, Connecticut 06520-8042

E-mail: [email protected]

Received November 6, 1997; revised September 25, 1998

The DIET (or dual interval echo train) sequence, a modification of THEORY the fast spin echo (FSE) sequence that selectively reduces signal from fat in MR images, has been investigated. The DIET sequence uses an Scalar J coupling can strongly influence the transverse initial echo spacing longer than that of a conventional FSE sequence, thus allowing J coupling-induced dephasing to take effect. The se- signal decay in multipulse spin echo experiments (6). In quence is evaluated theoretically, and its effectiveness on a hydrocar- hydrocarbon chains, J couplings arise largely from homo- bon (1-pentene) is demonstrated numerically using density matrix nuclear interactions between neighboring protons. In homo- calculations. The sequence is also evaluated experimentally using in nuclear coupled systems, 180° pulses do not reverse the vitro solutions and in vivo imaging. The efficacy of the sequence is phase evolution caused by coupling between spins. As a compared for different lipid chemical structures, field strengths, and result, the NMR signal is modulated in an oscillatory man- pulse sequence parameters. ©1999AcademicPress ner. For a simple doublet, when J/␦ Ӷ 1and␦␶ Ͼ 1(where Key Words: fat suppression; J coupling; density matrix; numer- J is the coupling constant, ␦ is the absolute value of the ical simulations; pulse sequence; DIET. chemical shift difference, and ␶ is the echo spacing), the echo train is modulated by cos(␲nJ␶), where n is the echo number. For more complex spectra, the summation of mod- INTRODUCTION ulated signals may appear more like a monotonically de-

creasing signal whose apparent T 2 is dominated by the The signals from fat appear relatively brighter in fast spin number and strength of the couplings present, rather than echo (FSE) images than in conventional spin echo (SE) images simply the intrinsic relaxation time (9). An equation characterizing the effect of J coupling in the with comparable T 1 and T 2 weightings (1–4). The magnitude of the fat signal generally increases as the spacing between the Carr–Purcell spin echo sequence for different multiple spin 180° pulses decreases. This pulse spacing dependence results systems has been derived using a density matrix formalism by from changes in the effects of J coupling within the hydrocar- Allerhand (6). Using this formalism, he showed that if the bon chains of lipid molecules (2, 5). As the pulse spacings pulse spacing is small, such that the products J jk␶ and ␦ jk␶ are become shorter, the effects of J coupling disappear (6), and Ӷ 1 for all j, k (where ␦ jk is the chemical shift difference hence the lipid signal is enhanced relative to the surrounding between spin groups j and k and J jk is their coupling constant), tissue. It is often desirable to reduce the signal from fat selec- then the effects of chemical shift disappear. This renders all of tively to achieve better contrast for detecting tissue abnormal- the spins magnetically equivalent and thus removes any ities. However, the performances of conventional fat saturation dephasing effects. As a result, the equation governing the techniques are often limited by the presence of B 0 and B 1 effects of J coupling in this regime reduces analytically to an inhomogeneities. Recently a pulse sequence was developed expression independent of echo time. Figure 1 shows the effect which exploits J coupling-induced dephasing to reduce the of pulse spacing on the echo train of a hypothetical strongly lipid signal (7, 8). The sequence relies on the use of a longer coupled A 3B 2 system, with J AB ϭ 6 Hz and ␦ AB ϭ 40 Hz. The echo spacing at the start of a FSE sequence. This permits J echo train was evaluated numerically using Allerhand’s for- coupling effects to reduce the signals from fat while still malism and omits intrinsic T 2 relaxation (9). The absolute allowing multiple echoes to be acquired in a relatively short value of the signal is given relative to its value immediately time. In this paper we evaluate the efﬁcacy of this technique, after the 90° pulse. When the pulse spacing, ␶, is 4 ms, the echo known as the dual interval echo train, or DIET, sequence, using train shows no modulation from J coupling, while in the ␶ ϭ theory and experiments on phantoms and report results using a 10 ms echo train, the effects of J coupling cause a smoothly DIET imaging sequence in vivo. varying signal decay. As ␶ increases to 30 ms, oscillatory

the first echo, which leads to dephasing and hence signal attenuation. A similar modification to the FSE imaging sequence was first suggested by Norris as a way of increasing the TE of an image; however, the potential for fat suppression with this sequence was not addressed (10). Figure 2 shows a schematic of the DIET sequence, where

the parameters ␶1 and ␶2 refer to the time of the initial echo and the spacing of the following echoes, respectively. The advantage of this sequence for reducing the signal from fat over other techniques such as selective saturation is that it should be spatially uniform, since it is based on J coupling effects which are not subject to ﬁeld inhomogeneities (11). However, the precise choice of timing parameters and the effectiveness of the sequence for different systems have not been previously investigated, nor has the validity of the sequence been demonstrated theoretically. The density matrix formalism described in the previous FIG. 1. The effect of J coupling on a strongly coupled A 3B 2 spin system. section can also be applied to the DIET sequence, as is shown The plot shows signal vs echo number for CPMG sequences where ␶, the in the Appendix. We demonstrate there that in the limit of very spacing between echoes, is 4, 10, or 30 ms. J AB ϭ 6 Hz, ␦ AB ϭ 40 Hz. Intrinsic

T 2 relaxation is neglected. Note that as ␶ increases, J coupling becomes more fast pulsing in the second phase of the DIET sequence, the J effective at suppressing the NMR signal. modulation of the echo train disappears. Furthermore, the fat

signal retains the suppression achieved during the ﬁrst (␶1) phase of the sequence. The pulse rates used in clinical FSE behavior enters the echo train, causing a faster initial decrease sequences fall short of the strict fast pulse requirement, which in signal intensity, as well as a nonmonotonic signal decay. is that ͉J jk͉␶ 2 and ͉␦ j͉␶ 2 are Ӷ 1 for all j and k. However, when

Using the density matrix formalism, one can show that the ␶2 is short but does not meet the fast pulse requirement, we quantities which determine the strength of J-coupled dephasing expect that the DIET sequence should show a similar ability to are the products ␦ jk␶, and J jk␶ (as opposed to the individual retain the fat suppression achieved during the initial ␶1 period. values of these parameters). Note that the J jk coupling con- We conﬁrmed the validity of this hypothesis in computer stants depend solely on the molecular properties of the lipid, simulations of the evolution of the density matrix during the whereas the ␦ jk are determined by both the lipid electronic DIET sequence. structure and the external magnetic ﬁeld.

THE DIET SEQUENCE SIMULATION AND PHANTOM STUDIES OF THE DIET SEQUENCE The DIET sequence works by exploiting the dependence of fat signal on pulse spacing. In order to suppress J-coupled Simulation Algorithm signals, the standard CPMG multiecho sequence is modiﬁed such that the spacing between the initial 90° pulse and the The signal produced by using the DIET sequence on a given ﬁrst echo is longer than the spacing between subsequent spin system was calculated by numerically solving the matrix echoes. This allows a greater mixing of coupled spins during equation from quantum mechanics (12)

FIG. 2. The DIET pulse sequence. This sequence is a modiﬁcation of the CPMG multiecho sequence, where the spacing between the initial 90° pulse and the ﬁrst echo is longer than the subsequent echo spacing. The longer initial echo time allows more J coupling-induced spin dephasing to occur. The spins remain dephased even after the faster pulsing begins. J COUPLING-INDUCED FAT SUPPRESSION IN DIET IMAGING 145

In our simulations, we studied 1-pentene (CH3CH2CH2CH ϭ

CH2) as a model for the hydrocarbon chains found in lipids. 1-Pentene was chosen because, as a 10-spin system, it was the largest hydrocarbon chain which we could simulate (MATLAB required 100 Mb of RAM to calculate its DIET signal). The J coupling values and chemical shifts for the protons in 1-pentene were ﬁrst estimated from values given in the literature for 1-hexene, 1-propene, and 1-butene (13, 14). These estimates were then ﬁne-tuned by matching the theoretical spectra they produced to a high-resolution (7 T) spectrum of 1-pentene (15).

Experimental Methods The DIET signal behavior of 1-pentene was also studied experimentally, along with olive oil and a solution of water

doped with 0.03 mM MnCl2. Olive oil was chosen for its

similarity to tissue fat, while the MnCl2 solution was chosen because it has no J coupling and its transverse relaxation time FIG. 3. Simulated DIET echo train for 1-pentene. ␶ ϭ 2 ms, ␶ ϭ 2 to 96 2 1 is similar to that of olive oil in the limit of rapid refocusing. ms. The absolute value is shown and intrinsic T 2 relaxation is neglected. The topmost curve shows the echo train when ␶1 ϭ ␶2, i.e., the CPMG version of Measurements were performed at 2.0 T on a GE Omega the sequence. In this fast pulse regime, the sole effect of increasing ␶1 is to shift imaging spectrometer. the entire echo train downward. The simulations support the hypothesis that when the pulse spacing is shortened in the second phase of the DIET sequence, the spin system retains the J coupling-induced signal suppression acquired RESULTS OF SIMULATION AND PHANTOM STUDIES before the ﬁrst echo. Figures 3 and 4 show the results of the simulated and experimental studies, respectively, of the DIET signal for ˆ ϭ ␳ ˆ ͗Ix͑t͒͘ Tr͓ ˆ ͑t͒Ix͔, [1] 1-pentene, where ␶2 (the second echo spacing) ϭ 2 ms. Echo

trains are shown for ␶1 values ranging from 2 to 96 ms. In Fig. ˆ where I x is the x component of the angular momentum oper- 3, the absolute value of the signal, S(t), is shown relative to its ator, ␳ˆ (t) is the density matrix of the system, Tr denotes the value immediately after the initial 90° RF pulse, S(0), and trace, and the angle brackets denote the expected value (12). The equation for ␳ˆ, as derived in the Appendix, is

ˆ ˆ ˆ ˆ Ϫi /2H␶2 Ϫi /2H␶2 n Ϫi /2H␶1 Ϫi /2H␶1 ␳ˆ͑␶1 ϩ n␶2͒ϰ͑e Rˆ 180xe ͒ ͑e Rˆ 180xe ͒

ˆ ˆ ˆ ˆ Ϫi /2H␶1 Ϫi /2H␶1 Ϫ1 Ϫi /2H␶2 Ϫi /2H␶2 Ϫn ϫ Iˆx͑e Rˆ 180xe ͒ ͑e Rˆ 180xe ͒ . [2]

The effect of T 2 relaxation is omitted, but can be included by multiplying the RHS of Eq. [2] by exp[Ϫ(␶ 1 ϩ n␶ 2)/T 2]. The time evolution of ␳ˆinEq.[2]wastrackedusingaprogram written in MATLAB (Mathworks, Natick, MA). The program is based on a simulation, written to calculate ␳ˆ(t)inaCPMG sequence, that has been described previously (9). The parameters needed by the program to characterize each spin system are the chemical shifts of its protons and their respective J couplings. These values are used to derive the matrix representation of the Hamiltonian operator. The program can handle spin systems of arbitrary size and complexity, provided the computer has enough memory to manipulate the resulting 2N ϫ 2N matrices, where N is the number of spins in the system. The pulse sequence is charac- FIG. 4. Experimental DIET echo train for 1-pentene. ␶2 ϭ 2 ms, ␶1 ϭ 2 terized by ␶1 and ␶2 and the length of the echo train (on which to 96 ms. The echo trains are nearly identical to those predicted numerically, there are no restrictions). All refocusing pulses are treated as ideal indicating that J coupling is responsible for the signal suppression seen here as and produce exact 180° rotations. ␶1 increases. 146 STABLES ET AL.

FIG. 5. Simulated DIET echo train for 1-pentene. ␶2 ϭ 8 ms, ␶1 ϭ 8 to 96 ms. In this regime, ␶ is not fast enough to fully suppress the effects of J 2 FIG. 6. Experimental DIET echo train for 1-pentene. ␶2 ϭ 8 ms, ␶1 ϭ 8 coupling, resulting in a greater signal modulation than was seen in Fig. 3. to 96 ms. The experimental echo trains show the same variations as those Again, as ␶ increases, the signal is more fully suppressed. 1 predicted numerically for ␶2 ϭ 8 ms; however, the amplitudes of these variations are not as pronounced.

intrinsic T 2 relaxation is omitted. In Fig. 4, the precise value of S(0) is not known; therefore the signal at t ϭ 2 ms is used as lated and experimental echo trains are qualitatively very sim- an approximation. The topmost curve of each ﬁgure shows the ilar, but the variations in the echo trains predicted numerically are greater than those seen experimentally. It may be that in the echo train when ␶1 ϭ ␶2, i.e., the CPMG version of the sequence. experimental data, the effects of J coupling are slightly mod- The simulated and experimental echo trains shown in Figs. erated by stimulated echoes or coherence transfer cross-relax- 3 and 4 are in excellent agreement, establishing that the effects ation phenomena, effects which are not accounted for in the of J coupling are responsible for the DIET signal suppression simulations. of 1-pentene. The ﬁgures show that each echo train retains the

J coupling suppression obtained during the initial ␶1 period of the sequence, and thus as ␶1 increases, the signal in the entire echo train becomes more effectively suppressed. When ␶1 ϭ 48 ms, for example, the signal decreases by over 90%. Further- more, the echo trains show behavior approaching the time independence predicted theoretically in the Appendix, suggesting that when ␶2 ϭ 2 ms, the DIET sequence is near the fast pulse limit (͉J jk͉␶ 2, ͉␦ j͉␶ 2 Ӷ 1) for the 1-pentene system.

(Simulated echo trains where ␶2 ϭ 0.5 and 1 ms show even greater time independence.) The DIET echo trains roughly match the CPMG (␶1 ϭ ␶2) echo train in slope, but their overall signal is lower. This is true even for the simulated ␶1 ϭ 96 ms echo train; however, the signal in that echo train is negative, so when the absolute value is taken, the slope of the echo train becomes positive. Figures 5 and 6 show the simulated and experimental echo trains for 1-pentene when ␶2 ϭ 8 ms. Because of the longer ␶2, the J coupling-induced signal decay is much more pronounced than when ␶ ϭ 2 ms. Just as in Figs. 3 and 4 however, 2 FIG. 7. Experimental DIET echo train for a solution of water doped with increasing ␶1 shifts the echo trains downward. Once the signal 0.03 mM MnCl2. ␶2 ϭ 2 ms, ␶1 ϭ 2 to 96 ms. Because the solution does not falls near zero, the echo train remains suppressed but its contain (nonequivalent) J-coupled protons, the echo trains show normal T 2 dependence on ␶1 becomes much less pronounced. The simu- decay and are independent of ␶1. J COUPLING-INDUCED FAT SUPPRESSION IN DIET IMAGING 147

change in the 1-pentene data versus that of the olive oil data is that the majority of the oil molecule consists of chains of

saturated neighboring methylene (CH2)groups.(1-Pen-

tene’s chemical structure is CH3(CH2)2CH ϭ CH2,whilethe chemical structure of olive oil’s principal component, oleic

acid, is CH3(CH2)7CH ϭ CH(CH2)7CO2H.) Protons within

long CH2 chains have similar chemical shifts and J couplings and are therefore nearly magnetically equivalent. Thus, a greater percentage of the alkene molecule consists of spins experiencing strong J coupling effects while the longer lipid molecule has more near-equivalent spins which experience weak J coupling effects and dilute the overall signal change. In addition, the intrinsic transverse relaxation times of the protons in these molecules may be different, and are likely shorter in the larger oil chains. Thus the relative importance of decreasing the effects of coupling may be further reduced in the oil. One aspect of Fig. 8 which is not completely understood is FIG. 8. Experimental DIET echotrain of olive oil. ␶ ϭ 2 ms, ␶ ϭ 2 to 96 2 1 the higher slope of the ␶1 ϭ 2 ms echo trains relative to that of ␶ ms. The dependence on 1 is not as dramatic as it is for 1-pentene, presumably echo trains in which ␶ is longer. One possible explanation is because the molecules in olive oil contain a greater number of nearly equiv- 1 alent protons. that the T 2 decays of 1-pentene and olive oil are multiexpo- nential, and that in each molecule, the component which decays fastest is also the component most heavily suppressed by The hypothesis that J coupling is responsible for the signal J coupling. suppression observed in Fig. 4 is further supported by the

DIET signal behavior seen with the MnCl2 solution, as shown IMAGING EXPERIMENTS

in Fig. 7. ␶2 ϭ 2 ms and again S(t ϭ 2 ms) is used to approximate S(0). The decay curves essentially overlap for all The DIET scheme shown in Fig. 2 can also be incorpo-

values of ␶1, as is to be expected for normal T 2 decay without rated into an imaging sequence by increasing the pulse J coupling. spacing before the ﬁrst phase encode step, as is illustrated

DIET echo trains (with ␶2 ϭ 2ms)areshownforoliveoil in Fig. 9. In order to evaluate the contrast derived from in Fig. 8. This system again shows signal suppression with this sequence, we compared images evaluated with the

increasing ␶1,althoughtheeffectwasnotasdramaticas DIET imaging sequence to those from a conventional with 1-pentene. At an echo time of 48 ms, the signal of the FSE sequence and a conventional SE sequence. Images

␶1 ϭ 48 ms echo train is roughly half that of the ␶1 ϭ 2ms within the thigh and the abdomen were collected using a echo train. One possible explanation for the greater signal 1.5-T GE “Signa” systems magnet. Mid-thigh images

FIG. 9. The DIET Imaging Sequence. a, b, c, and d denote gradient durations (7). 148 STABLES ET AL.

FIG. 10. Thigh images created using the (a) SE, (b) FSE, and (c) DIET imaging sequences. The TE values were 110, 112, and 108 ms, respectively. The fat signal is much less pronounced in the DIET image than in the FSE image, hence the DIET image contrast approaches that of the SE image.

were obtained using an extremity coil to contrast fat and clearly closer to that of a SE image than to that of an FSE muscle, while abdominal images were obtained with a body sequence. Table 1 shows the ratio of signal from various coil to contrast fat, muscle, and internal organs such as the tissues-to-fat signal, as measured in regions of interest from kidney. The imaging parameters were: (a) SE: TE ϭ 110 ms, both the thigh and abdominal images. These ratios conﬁrm that TR ϭ 2000 ms, and ␶ ϭ 55 ms; (b) FSE: TE ϭ 112 ms, the contrast provided by the DIET sequence approaches that of TR ϭ 3000 ms, echo train length (ETL) ϭ 16, and ␶ ϭ 14 a conventional SE sequence. ms; and (c) DIET: TE ϭ 108 ms, TR ϭ 3000 ms, ETL ϭ 16,

␶1 ϭ 24 ms, and ␶2 ϭ 14 ms. The FOV of each image was FIELD DEPENDENCE OF J COUPLING-INDUCED 20 cm. In order to make a fair comparison between images, FAT SUPPRESSION the receiver gain was held constant for all the imaging sequences. Furthermore, to keep the effective echo times For the DIET sequence to suppress fat signal adequately,

(TE) comparable between the images, the zero order phase ␶1 must be long enough such that ͉J jk͉␶ 1 or ͉␦ jk͉␶ 1 Ն 1. ␦ is encode of the DIET sequence was shifted to the seventh directly proportional to ﬁeld strength, and thus the initial echo, rather than the eighth echo, where it occurred in the pulse spacing necessary to suppress fat will become shorter

FSE sequence. as B 0 increases. In order to qualitatively evaluate the ﬁeld Figure 10 shows images of the upper thigh using the FSE, dependence of J coupling-induced fat suppression, we stud- DIET, and SE sequences. The images are thresholded to the ied the dependence of echo amplitudes on ␶ for olive oil at same levels for comparison. The contrast of the DIET image is ﬁeld strengths of 0.47 and 2.0 T using a conventional CPMG J COUPLING-INDUCED FAT SUPPRESSION IN DIET IMAGING 149

TABLE 1 the echo train without signiﬁcantly effecting its subsequent Ratios of Tissue Signal to Fat Signal behavior. Once the faster pulsing begins, the effects of J for FSE, DIET, and SE Images coupling are removed, but the echo train’s signal remains

FSE DIET SE suppressed. As discussed above, the product of the chemical shift differ- Fat 1 1 1 ence(s) within the spin system and the pulse spacing, ␦␶,isan Marrow 1.06 1.21 1.25 important parameter which determines the degree to which J Muscle 0.12 0.16 0.15 ␦␶ Kidney 0.13 0.21 0.23 coupling dephases lipid signals (6, 9). As decreases, the effects of J coupling diminish, enhancing the lipid signal. It is apparent from our phantom experiments that even at rapid FSE pulse spacings (of order 10–20 ms), the effects of J coupling are still not sequence. We recorded the echo amplitude at a fixed TE eliminated. Thus, as new FSE sequences become available where (128 ms) for a range of pulse intervals, ␶.Measurementsat the pulse spacing is decreased below the limits currently available 0.47 T were performed on a Bruker Minispec relaxometer. on commercial systems (␶ Х 12–14 ms), the effects of J coupling Figure 11 shows the field dependence of the olive oil signal. will be more fully suppressed and lipid signal will be further At both field strengths the signal is greatest at short pulse enhanced. Under these conditions the DIET sequence will be intervals (␶ Ͻ 4ms),andsimilarchangesareseeningoing particularly effective. The product ␦␶ is also smaller at lower from the limit of very rapid pulsing to slow pulsing. How- imaging field strengths, such as are found with open-bore mag- ever, as expected, at2Tthereisamorerapiddecreasein nets, resulting in a greater difference between SE and FSE fat signal as ␶ increases. signals. Once more the use of a long initial echo can greatly For most practical FSE imaging sequences at clinical field reduce these effects, though at lower field strengths it may be strengths, the pulse spacing can at best be considered to be in necessary to use longer ␶1 values (scaled inversely with field an intermediate regime where ␶ is not short enough to eliminate strength) to suppress fat signal effectively. the fat signal entirely. For a typical FSE sequence with ␶ ϭ 16 The length of delay before the first refocusing pulse will of ms, we can estimate from Fig. 11 the maximum amount of lipid course alter the sensitivity of the DIET sequence to effects suppression attainable by incorporating a longer initial echo other than J coupling, such as subject motion, flow, and into the sequence. At 2.0 T, as ␶ (or, for a DIET sequence, ␶1) diffusion through susceptibility-induced field inhomogeneities. is increased from 16 to 64 ms, the signal from olive oil is The point spread function of the image will also be altered decreased by about 25%, whereas at 0.47 T the signal is because of the delay before each echo is acquired. decreased by over 40%. It should be noted that these estimates simply serve to show the qualitative field dependence and should not be taken as maximum estimates of suppression in vivo since this depends on the exact chemical nature of the coupled lipid system.

DISCUSSION

The simulations and experiments clearly support the theoretical prediction that when the pulse spacing is shortened in the second phase of the DIET sequence, the spin system retains the J coupling suppression acquired before the first echo. In addition, the results highlight a potential advantage of the DIET sequence over conventional CPMG-based imaging sequences: as ␶2 becomes shorter, the modulation of the echo train decreases. In imaging sequences which acquire multiple phase encodes in a single TR, a smoother echo train results in a narrower and less distorted point spread function (PSF) in the phase encode direction. Another finding of the simulations and experiments is that FIG. 11. Field dependence of J coupling induced fat suppression. The intensity of the echo at TE ϭ 128 ms is plotted vs ␶ for B 0 ϭ 0.47 and 2 T. for both large and small values of ␶2, the behavior of the echo The signal in each curve is normalized relative to its value when ␶ ϭ 1 ms. The train once fast pulsing begins is in many cases remarkably chemical shift, ␦, is larger at the higher field, and thus J coupling effects will independent of ␶1. Therefore one can lower the initial signal of be seen at shorter ␶ values (i.e., when ͉␦ jk͉␶ Ն 1). 150 STABLES ET AL.

CONCLUSIONS short ␶ ␳ˆ ͑n␶͒ O¡ Iˆx. [8] We have demonstrated through density matrix theory, numerical simulations, and phantom and in vivo imaging experiments For the DIET sequence, the expression equivalent to Eq. [3] is that the DIET sequence is an effective method of reducing lipid signal. The sequence allows one to obtain spin echo-like contrast Ϫi /2Hˆ Ϫi /2Hˆ n Ϫi /2Hˆ Ϫi /2Hˆ ␳ˆ͑␶ ϩ n␶ ͒ϰ͑e ␶2Rˆ e ␶2͒ ͑e ␶1Rˆ e ␶1͒ from an FSE-type sequence. The simulations and experiments 1 2 180x 180x ˆ ˆ ˆ ˆ Ϫi /2H␶1 Ϫi /2H␶1 Ϫ1 Ϫi /2H␶2 Ϫi /2H␶2 Ϫn clearly support the theoretical prediction that in the limit of very ϫ Iˆx͑e Rˆ 180xe ͒ ͑e Rˆ 180xe ͒ . [9] fast pulsing in the second phase of the DIET sequence, the echo train becomes time independent. They also support the hypothesis that the DIET sequence should remain an effective means of fat In the limit of fast pulsing in the second phase of the sequence (i.e., J ␶ and ␦ ␶ Ӷ 1), the outer terms can be simpliﬁed suppression outside of the limit of short ␶2. ͉ jk͉ 2 ͉ j͉ 2 using a Taylor expansion, APPENDIX short ␶ Ϫi /2Hˆ Ϫi /2Hˆ 2 ␶2 ˆ ␶2 ϩ ˆ ˆ ␶ ˆ In a standard CPMG sequence, the equation for the density e R180xe O¡ ͑1 i ͸ JjkIj ⅐ Ik 2͒R180. matrix, ␳ˆ, at the nth echo is (neglecting T decay) j 2 [10]

Ϫi /2Hˆ ␶ Ϫi /2Hˆ ␶ n Ϫi /2Hˆ ␶ Ϫi /2Hˆ ␶ Ϫn ␳ˆ ͑n␶͒ϰ͑e Rˆ 180xe ͒ Iˆx͑e Rˆ 180xe ͒ , [3] However, because the ﬁrst pulse spacing is chosen to be long

enough to allow J coupling effects, i.e., ͉J jk͉␶ 1 or ͉␦ jk͉␶ 1 Ն 1, where ␶ is the spacing between 180Њx pulses, and Rˆ 180x is the one cannot ignore higher order terms when expanding ˆ ␶ ˆ rotation operator for the 180Њx pulse (6). The Hamiltonian of the exp(ϪiH 1). These higher order terms contain I zj operators, system is which do not commute with the remaining operators in Eq. [9], so the expression for ␳ˆ (␶ 1 ϩ n␶ 2) cannot be simpliﬁed. ␳ ˆ ˆ ϭ ␦ ˆ Ϫ ˆ ˆ The expression for Tr[ ˆ (t)I x] can be simpliﬁed, however, by H Ϫ͸ j Izj ͸ JjkIj ⅐ Ik , [4] j jϽk using the relation

Tr͓␳ˆIˆx͔ ϭ ͗␣j͉␳ˆIˆx͉␣j͘, [11] where ␦ j is the chemical shift of nucleus j with respect to the ͸ j average Larmor frequency of the spins, and J jk is the J coupling constant between spins j and k. Iˆ j is the angular momen- where the ␣ ͘ are an orthonormal set of basis states for the tum operator for the jth spin and Iˆ zj is its z component. If ␶ is ͉ j short enough that system. The basis states we choose to use are the eigenstates of i i exp(Ϫ2 Hˆ ␶ 1)Rˆ 180xexp(Ϫ2 Hˆ ␶ 1), which we denote ͉␰ j͘. Using Eq. [9], Eq. [11] becomes ͉Jjk␶͉, ͉␦jk␶͉ Ӷ 1, for all spin groups, j, k, [5]

n Ϫ1 i Tr͓␳ˆIˆx͔ ϭ ͗␰j͉͑Uˆ 2Rˆ Uˆ 2͒ ͉␰k͗͘␰k͉͑Uˆ 1Rˆ Uˆ 1͒Iˆx͑Uˆ 1Rˆ Uˆ 1͒ ͉␰l͘ then one can expand the time evolution operator, exp(Ϫ2 Hˆ ␶), ͸ to ﬁrst order in a Taylor series: j,k,l Ϫn ϫ ͗␰l͉͑Uˆ 2Rˆ Uˆ 2͒ Iˆx͉␰j͘, [12] i i i Ϫi /2Hˆ ␶ ˆ ˆ ˆ ˆ e Ϸ 1 Ϫ H␶ ϭ 1 ϩ ␦jIzj␶ ϩ JjkIj ⅐ Ik␶. ˆ 2 2 ͸ 2 ͸ where for brevity the subscript of R 180x has been dropped, and j jϽk ˆ ˆ i ˆ ␶ [6] U 1 and U 2 represent the time evolution operator, exp(Ϫ2 H ), at ␶ ϭ ␶1 and ␶2, respectively. Uˆ 1Rˆ Uˆ 1 is unitary, and thus has eigenvalues of the form exp(i␭ )(6). Substituting these eigen- Plugging this expansion into Eq. [3], we obtain j values and rearranging terms, we obtain

ˆ ˆ short ␶ Ϫi /2H␶ ˆ Ϫi /2H␶ ϩ ˆ ˆ ␶ ˆ ␳ˆ ϭ i͑␭kϪ␭l͒ ␰ ˆ ˆ ˆ Ϫnˆ ˆ ˆ ˆ n ␰ ␰ ˆ ␰ e R180xe O¡ ͑1 i ͸ JjkIj ⅐ Ik ͒R180. [7] Tr͓ ˆIx͔ ͸ e ͗ l͉͑U2RU2͒ Ix͑U2RU2͒ ͉ k͗͘ k͉Ix͉ l͘. j k,l [13] The expression on the RHS of Eq. [7] commutes with Iˆ x and Ϫi/2Hˆ ␶ Ϫi/2Hˆ ␶ Ϫ1 thus cancels the (e Rˆ 180xe ) terms in Eq. [3] (6). In the limit of fast pulsing in the second (␶2) phase of the DIET

Therefore, in a CPMG sequence, in the limit of 1/␶ ӷ ͉J jk͉, sequence, one can use the Taylor expansion shown in Eq. [10].

͉␦ jk͉, The terms in this expansion commute with Iˆ x so that J COUPLING-INDUCED FAT SUPPRESSION IN DIET IMAGING 151

manipulation and artifact assessment of 2D and 3D RARE se- short ␶2 Ϫn n quences, Magn. Reson. Imaging 8, 557–566 (1990). ͑Uˆ 2Rˆ Uˆ 2͒ Iˆx͑Uˆ 2Rˆ Uˆ 2͒ O¡ Iˆx. [14] 5. R. S. Hinks and D. Martin, Bright fat, fast spin echo, and CPMG, in The expression for the DIET signal therefore becomes “Proceedings, Society of Magnetic Resonance in Medicine 11th Annual Meeting,” p. 4503, Berlin, 1992. 6. A. Allerhand, Analysis of Carr–Purcell spin echo NMR experiments ␶ ϩ ␶ ␳ˆ ϭ i͑␭kϪ␭l͒ ␰ ˆ ␰ 2 S͑ 1 n 2͒ϰTr͓ ˆIx͔ ͸ e ͉͗ k͉Ix͉ l͉͘ . [15] on multiple-spin systems. I. the effect of homonuclear coupling, k,l J. Chem. Phys. 44, 1–9 (1966). 7. H. Kanazawa, H. Takai, Y. Machida, and M. Hanawa, Contrast

While the values of ␭ j and ͉␰ j͘ depend on ␶1, there is no naturalization of fast spin echo imaging: A fat reduction technique free from field inhomogeneity, in “Proc., SMR, 2nd Meeting, San dependence on ␶2 or n in Eq. [15]. We have therefore demonstrated that in the limit of very fast pulsing in the second phase Francisco, 1994,” p. 474. of the DIET sequence, the modulation of the echo train will 8. K. Butts, J. M. Pauly, G. H. Glover, and N. J. Pelc, Dual echo “DIET” disappear. The fat signal instead retains the suppression fast spin echo imaging, in “Proceedings, Society of Magnetic Res- onance, 3rd Scientific Meeting,” Nice, France, p. 651, 1995. achieved during the first (␶ ) phase of the sequence. 1 9. L. A. Stables, A. W. Anderson, and J. C. Gore, The effects of J coupling in SE and FSE MR imaging, in “Proceedings, Society of ACKNOWLEDGMENT Magnetic Resonance in Medicine, 12th Annual Meeting,” New York, p. 1267, 1993. This work was supported by Grant CA40675 from the National Institutes of 10. D. Norris, P. Bo¨rnert, T. Reese, and D. Leibfritz, On the application of Health. ultra-fast RARE experiments, Magn. Reson. Med. 27, 142–164 (1992). 11. R. T. Constable, R. C. Smith, and J. C. Gore, Coupled-spin fast REFERENCES spin-echo MR imaging, J. Magn. Reson. Imaging 3, 547–552 (1993).

1. R. S. Hinks and R. M. Henkelman, Problems with organic materials 12. C. P. Slichter, “Principles of Magnetic Resonance,” Harper and for magnetic resonance imaging phantoms, Med. Phys. 15, 61–63 Row, New York (1963). (1988). 13. A. A. Bothner-By and C. Naar-Colin, The proton magnetic reso- 2. R. T. Constable, A. W. Anderson, J. Zhong, and J. C. Gore, Factors nance spectra of olefins. I. Propene, butene-1 and hexene-1, influencing contrast in fast spin-echo MR imaging, Magn. Reson. J. Amer. Chem. Soc. 83, 231–236 (1961). Imaging 10, 497–511 (1992). 14. H. W. Quinn, J. S. McIntyre, and D. J. Peterson, Coordination 3. R. M. Henkelman, P. A. Hardy, J. E. Bishop, C. C. Poon, and D. B. compounds of olefins with anhydrous silver salts, Can. J. Chem. Plewes, Why fat is bright in RARE and fast spin echo imaging, J. 65, 2896–2910 (1965). Magn. Reson. Imaging 2, 533–540 (1992). 15. “The Aldrich Library of 13C and 1H FT NMR Spectra,” Aldrich 4. R. V. Mulkern, S. T. S. Wong, C. Winalski, and F. A. Jolesz, Contrast Chemical Co., Milwaukee, (1993). CME JOURNAL OF MAGNETIC RESONANCE IMAGING 36:1314–1328 (2012)

Review: MR Physics for Clinicians

Hyperpolarized 13C Metabolic Imaging Using Dissolution Dynamic Nuclear Polarization

Ralph E. Hurd, PhD,1* Yi-Fen Yen, PhD,1 Albert Chen, PhD,2 and Jan Henrik Ardenkjaer-Larsen, PhD3

This article is accredited as a journal-based CME activity. The peer reviewers have no relevant financial relation- If you wish to receive credit for this activity, please refer to ships. The peer review process for Journal of Magnetic the website: www.wileyhealthlearning.com Resonance Imaging is double-blinded. As such, the identi- ties of the reviewers are not disclosed in line with the ACCREDITATION AND DESIGNATION STATEMENT standard accepted practices of medical journal peer Blackwell Futura Media Services designates this journal- review. based CME activity for a maximum of 1 AMA PRA Category Conflicts of interest have been identified and resolved 1 CreditTM. Physicians should only claim credit commensu- in accordance with Blackwell Futura Media Services’s rate with the extent of their participation in the activity. Policy on Activity Disclosure and Conflict of Interest. No Blackwell Futura Media Services is accredited by the Ac- relevant financial relationships exist for any individual in creditation Council for Continuing Medical Education to control of the content and therefore there were no conflicts provide continuing medical education for physicians. to resolve.

EDUCATIONAL OBJECTIVES INSTRUCTIONS ON RECEIVING CREDIT Upon completion of this educational activity, participants For information on applicability and acceptance of CME will be better able to describe the basic physics of dissolution credit for this activity, please consult your professional dynamic nuclear polarization (dissolution-DNP), and the licensing board. impact of the resulting highly nonequilibrium spin states, on This activity is designed to be completed within an hour; the physics of magnetic resonance imaging (MRI) detection. physicians should claim only those credits that reflect the time actually spent in the activity. To successfully earn ACTIVITY DISCLOSURES credit, participants must complete the activity during the No commercial support has been accepted related to the valid credit period. development or publication of this activity. Follow these steps to earn credit: Faculty Disclosures: Log on to www.wileyhealthylearning.com The following contributors have no conflicts of interest to Read the target audience, educational objectives, and disclose: activity disclosures. Editor-in-Chief: C. Leon Partain, MD, PhD Read the article in print or online format. CME Editor: Scott B. Reeder, MD, PhD Reflect on the article. CME Committee: Scott Nagle, MD, PhD, Pratik Mukher- Access the CME Exam, and choose the best answer to jee, MD, PhD, Shreyas Vasanawala, MD, PhD, Bonnie Joe, each question. MD, PhD, Tim Leiner, MD, PhD, Sabine Weckbach, MD, Complete the required evaluation component of the Frank Korosec, PhD activity. Authors: Ralph E. Hurd, PhD, Yi-Fen Yen, PhD, Albert Chen, PhD, Jan-Henrik Ardenkjaer-Larsen, PhD This activity will be available for CME credit for twelve This manuscript underwent peer review in line with months following its publication date. At that time, it will the standards of editorial integrity and publication ethics be reviewed and potentially updated and extended for an maintained by Journal of Magnetic Resonance Imaging. additional period.

1GE Healthcare, Menlo Park, California, USA. 2GE Healthcare, Toronto, Canada. 3GE Healthcare, Denmark. *Address reprint requests to: R.E.H., GE Healthcare, 333 Ravenswood Ave., Menlo Park, CA 94025. E-mail: [email protected] Received November 15, 2011; Accepted June 10, 2012. DOI 10.1002/jmri.23753 View this article online at wileyonlinelibrary.com.

VC 2012 Wiley Periodicals, Inc. 1314 Hyperpolarized 13C MRI 1315

nuclear spins. This was a fundamental discovery This article describes the basic physics of dissolution dynamic nuclear polarization (dissolution-DNP), and the causing disbelief at the time: that heating of one spin impact of the resulting highly nonequilibrium spin states, system could lead to the cooling of another. The pre- on the physics of magnetic resonance imaging (MRI) diction by Overhauser for metals was extended to detection. The hardware requirements for clinical transla- electron spins in solution by Abragam (7), and most tion of this technology are also presented. For studies nuclear magnetic resonance (NMR) spectroscopists that allow the use of externally administered agents, hy- are today familiar with the nuclear and electronic perpolarization offers a way to overcome normal magnetic Overhauser effect. However, this effect is limited to resonance sensitivity limitations, at least for a brief T1-de- solutions where relaxation processes couple the spin pendent observation window. A 10,000–100,000-fold systems via molecular motions. Soon after, the Solid signal-to-noise advantage provides an avenue for real- Effect was described for spins in the solid state time measurement of perfusion, metabolite transport, exchange, and metabolism. The principles behind these coupled by dipolar interactions (8). Later, DNP in the measurements, as well as the choice of agent, and pro- solid state was extended mechanistically to processes gress toward the application of hyperpolarized 13C meta- involving several electron spins (thermal mixing) (9). bolic imaging in oncology, cardiology, and neurology are The theory of DNP in the solid state, however, has reviewed. failed to provide a quantitative description of the general case. In the solid state, the high electron spin Key Words: hyperpolarized 13C; DNP; metabolic imaging; pyruvate polarization is in part transferred to the nuclear spins J. Magn. Reson. Imaging 2012;36:1314–1328. by microwave irradiation close to the resonance fre- VC 2012 Wiley Periodicals, Inc. quency of the electron spin. The efficiency of this process depends on several parameters characterizing the various spin systems, but also on technical factors such as microwave frequency and power. HYPERPOLARIZED 13C magnetic resonance spectro- DNP has mainly been applied to the generation of scopic imaging (MRSI) has recently progressed beyond polarized targets for neutron scattering experiments, a substantial number of very exciting preclinical stud- and it has been demonstrated that the nuclear polar- ies, into man (1). The goal of this article is to intro- izations of 1H and 13C could be increased to almost duce the basic principles and progress that have been 100% and to 50%, respectively, in the solid state by made toward the clinical application of hyperpolarized means of DNP at low temperature (10,11). The mecha- 13C using dissolution dynamic nuclear polarization nism requires the presence of unpaired electrons (DNP). Topics include the significant contributions to (electron paramagnetic agent; EPA), which are added the science of dissolution-DNP, rapid multinuclear to the sample as, for example, an organic radical. The spectroscopic imaging methods, and animal model magnetic moment of the electron is 658 times higher work targeted at a wide variety of potential than that of the proton. This means that the electron indications. spin will reach unity polarization at a moderate mag- In MR, hyperpolarization indicates that the polariza- netic field strength and liquid helium temperature. At, tion is no longer determined by the static magnetic eg, 3.35 T and 1 K the electron spin polarization is field of the scanner. The enhanced polarization of the 98%. agent is created outside the imaging system by means of a polarizer. Hyperpolarization can be based on several principles (2–4). One such is the dissolution-DNP DNP Sample Preparation method that has been very successful over the past 5– 10 years in terms of making solutions of biologically The first step of hyperpolarizing a new molecule by interesting molecules with highly polarized nuclear the dissolution-DNP method is to add unpaired elec- spins. The method takes advantage of DNP in the tron spins to the sample. Unpaired electron spins solid state followed by rapid dissolution in a suitable with well-defined properties are most conveniently solvent (2–4). The polarization is retained almost com- provided by chemical doping. In order for the DNP pletely in the dissolution step, creating a solution process to be effective, the EPA agent must be homo- with a nonthermal nuclear polarization approaching geneously distributed within the sample. Many mole- unity. cules will be crystalline, or have a tendency to crystal- To take advantage of a hyperpolarized liquid state lize as saturated aqueous solutions. This will cause solution requires rapid transfer into the subject, as the EPA to concentrate in domains and lead to a poor illustrated in Fig. 1, followed by efficient and rapid DNP effect. To prevent this, the sample should stay 13C spectroscopic imaging sequences. amorphous when frozen to ensure homogenous distribution of the EPA. Three examples of molecules that are liquids at room temperature and stay amorphous 13 Hyperpolarization by the Dissolution-DNP Method when frozen without additives are [1- C]pyruvic acid (or any other isotopic labeling), 2-keto-[1-13C]isocap- DNP was first described theoretically by Overhauser roic acid, and bis-1,1-(hydroxymethyl)-[1-13C]cyclo- in 1953 (5), and a few months later demonstrated by propane-d8(HP001). All three molecules are liquids at Carver and Slichter (6) in metallic lithium. Overhauser room temperature and dissolve well a hydrophilic predicted that saturating the conduction electrons of EPA. For other compounds it is necessary to prevent a metal would lead to a dynamic polarization of the crystallization by mixing or dissolving the compound 1316 Hurd et al.

Figure 1. The principle of dissolution-DNP. At room temperature and, eg, 3 T the 13C nuclear spins are only weakly polarized to about 0.00025%, lower right graph. However, an electron spin has a 2700 times stronger magnetic moment and these spins are easily polarized. When the sample is cooled the electron spin polarization of the EPA can reach almost 100%. By irradiation with microwaves close to the resonance frequency of the electron spins, electron-nuclear transitions are induced, and the nuclear spin polarization will be enhanced hundred-fold by DNP, raising the 13C spins to a robust polarization of over 40%. This process is slow and takes typically 15–60 minutes. When the sample is polarized it can be dissolved in, eg, heated water or a buffer, and within seconds a room temperature solution of the hyperpolarized molecule is obtained. The hyperpolarized nuclear spins relax to thermal equilibrium (0.00025%) with a time constant (T1) of typically 40–80 seconds for carboxylic acids. Therefore the hyperpolarized sample has to be quickly carried to the scanner, injected, and imaged. in a suitable solvent such as glycerol, or dimethylsulf- two examples of this can be mentioned: The cesium 13 oxide (DMSO) can be used as solvent for the molecule salt of bicarbonate (CsH CO3 (13)), and the TRIS salt and the EPA. For in vivo studies it is necessary to be of acetate (14). Both of these salts have higher solubil- able to formulate the molecule in a concentrated form ity than their sodium counterpart. Finally, for amino in order to achieve a high concentration of the mole- acids (zwitter ions at neutral pH) it has shown that ei- cule after dissolution. To give an estimate of the ther high or low pH preparations increase the aque- requirements, a patient dose of 0.1 mmol/kg body ous solubility by reducing the charge of the molecule weight can be assumed, requiring 10 mmol of com- to a point (15) that no or little glycerol is needed to pound, or 1 g with a molecular weight of 100 g/mol. form an amorphous sample. The consequence is that 30%–50% solubility is needed in order to keep the sample size reasonable Electron Paramagnetic Agent (EPA) (see clinical polarizer description). A solvent mixture with high solubility for the molecule and EPA, pre- The source of the unpaired electron is typically an or- venting sample crystallization and with good in vivo ganic radical, but a few metal ions have been tolerance therefore has to be chosen. A good example employed successfully for DNP, Cr(V) in particular of a biologically compatible formulation is fumaric (16). The choice of EPA will depend on a number of 13 acid (eg, [1,4- C2, 2,3-D2]fumaric acid) in DMSO. factors. First, the EPA needs to be chemically stable DMSO is a widely used solvent for pharmaceuticals and dissolve readily in the matrix of interest. Second, and has a good safety proﬁle (12). It will, however, the electron paramagnetic resonance (EPR) spectrum crystallize when frozen (dry DMSO; melting point of the radical should have a width that allows DNP to 16C). However, this does not prevent the use of be effective for the nucleus of interest, ie, a line width DMSO as solvent for fumaric acid. As saturated solu- that exceeds the Larmor frequency of the nuclear spin. tion with a molarity of 3.6 mol/L or 1:1.8 by weight, In practice the above criteria mean that two classes of the solution forms an amorphous solid when frozen. EPA are available, namely nitroxides (17,18) and tri- When using DMSO as a glassing agent, care should tyls (19–21). The nitroxides belong to a class of mole- be taken to ensure the DMSO stays dry (hygroscopic), cules that have been studied extensively by EPR, and as small amounts of water will decrease solubility and which have been used for DNP for many samples. increase supersaturation. Nitroxides are characterized by having a broad EPR Another means of improving the solubility involves spectrum. The EPR line width is 4.0 per mil (%) of the changing the counterion of salts. Solubility typically EPR frequency, compared to the 1H resonance fre- increases with increasing size of the counterion, and quency, which is 1.5% of the EPR frequency. Some of Hyperpolarized 13C MRI 1317 them have reasonable chemical stability and come with different degrees of hydrophilicity. Another class of EPA with superior properties for direct polarization of low gamma nuclei such as 13C, 15N, and 2H is the trityl. These radicals have a line width that is only 0.80% (22,23) of the resonance frequency, much less than the proton resonance frequency, but perfectly matched for 13C, which has a resonance frequency which is 0.37% of the EPR frequency. The trityls also exist with a range of hydrophilicities and some of them are chemically very stable. It has been shown that gadolinium (Gd) can positively affect the solid state DNP enhancement (24). þ Other paramagnetic ions and molecules (Mn2 and

O2) can in part have the same effect. The physics is Figure 2. Longitudinal relaxation time, T1, as a function of not yet understood, but Ardenkjaer-Larsen et al (24) temperature for neat [1-13C]pyruvic acid with n and without showed that the longitudinal relaxation time of the ^ 20 mmol/L trityl. EPA is shortened by the presence of the Gd ions. The 3þ effect of adding 1–2 mmol/L Gd is a 50%–100% compound [1-13C]pyruvic acid, the 13C polarization improvement of the DNP enhancement factor. The improved from 27% at 3.35 T to 60% at 4.64 T in the effect seems to be general to most samples, but has to solid state. The literature has been limited and dis- be optimized for each sample similarly to the concen- agrees on the magnetic field dependence of DNP (32). tration of the EPA. There is no direct DNP effect of the þ On the other hand, it has been well established that Gd3 by itself under the conditions typically used. þ lowering the temperature is beneficial for DNP. Finally, Gd3 may enhance the solid state polarization by DNP, but care should be taken in avoiding accelerated relaxation in the liquid state. Free Gd ions would Dissolution and Relaxation in the Liquid State cause detrimental liquid state relaxation and pose an To make the polarized solid sample useful for in vivo in vivo safety risk. After dissolution the low concentra- imaging, it needs to be dissolved in a suitable buffer. tion of radical and chelated Gd will have a negligible Depending on the solid sample preparation the dis- effect on T1 in most cases. solution may involve neutralization of the agent with acid or base. Buffering of the solution may be DNP Instrumentation required to maintain control of pH within the physiologic range of 6.8 to 8.1. Physiological buffers such Most solid-state DNP has been performed at magnetic as Tris(hydroxymethyl)aminomethane (TRIS) or 4-(2- fields between 0.35 T (25) and 16.5 T (26,27), and at hydroxyethyl)piperazine-1-ethanesulfonate (HEPES) temperatures from a few hundred mK to room tem- are commonly used. Attention to the tonicity of the perature. At temperatures below a few Kelvin and formulation should be paid and close to isotonic is magnetic field strengths above a few Tesla, electron desired. This may mean lowering the concentration spins are almost fully polarized, and large nuclear of solutes after dissolution by dilution or adding so- polarizations can be obtained. Unlike solid-state NMR dium chloride to the dissolution medium. The disso- spectroscopy applications where nonequilibrium lution has to be efficient and fast compared to the polarizations can be regenerated by repeating the nuclear T in order to preserve the nuclear polariza- microwave irradiation and NMR acquisition, the 1 tion in this process. Formulating the solid sample as polarization generated for in vivo applications will beads or powder may improve the dissolution (in decay irreversibly after dissolution. Hence, the goal is terms of polarization and recovery of the solid sam- to generate polarizations close to unity. It is therefore ple), but understanding and optimizing the fluid dy- important to choose initial operating conditions that have been proven to provide high nuclear polarization, namics (33) as well as providing the necessary heat but are at the same time easily achievable using is essential for optimal performance of more difficult standard instrumentation. Temperatures of 1 K can agents. Relaxation during the dissolution process be achieved by pumping on liquid helium. In the origi- can depend on several factors. To minimize relaxa- nal dissolution-DNP polarizer design the liquid helium tion, dissolution is performed inside the cryostat in was supplied to the sample space through a needle the high field of the polarizer (eg, 3 T in the case of valve from the magnet cryostat, but in a recent publi- a 3.35 T polarizer), but above the liquid helium sur- cation an alternative arrangement that used a sepa- face. Any paramagnetic ions that could increase the rate helium dewar was described (28). A magnetic relaxation rate are chelated by adding, for example, field strength of 3.35 T was chosen since microwave ethylenedinitrotetraacetic acid (EDTA) to the dissolu- sources are readily available at 94 GHz for irradiation tion medium, or to the sample. To illustrate the se- of the electron spin. However, recently it has been verity of relaxation during dissolution [1-13C]pyruvic demonstrated that for both nitroxides and trityls a acid (pyruvic acid enriched with 13C to 99% in the significant improvement in polarization can be C-1, ie, carboxylic acid, position) is chosen as an obtained by increasing the magnetic field strength example. This molecule has been well studied with (29,30) or lowering the temperature (31). For the DNP and has high biological relevance. In Fig. 2 the 1318 Hurd et al.

Table 1 Dissolution-DNP Compounds (97–106) Agent Products Ref. 13 13 13 13 13 [1- C]pyruvate [1- C]lactate, [1- C]alanine, [ C] bicarbonate, CO2 41 [2-13C]pyruvate [2-13C]lactate, [2-13C]alanine, [1-13C]acetyl-carnitine, 92,95 [1-13C]citrate, [5-13C]glutamate 13 13 13 13 [1,2- C2]pyruvate [1,2- C2]lactate, [1,2- C2]alanine, [1- C]acetyl-carnitine, 100 13 13 13 13 [1- C]citrate, [5- C]glutamate, [ C]bicarbonate, CO2 13 13 13 13 13 [1- C]lactate [1- C]pyruvate, [1- C]alanine, [ C] bicarbonate, CO2 101 13 13C-bicarbonate CO2 13 13 13 [1,4- C2]fumarate [1,4- C2]malate 85 [1-13C]acetyl-methionine [1-13C]methionine 102 [2-13C]fructose [1-13C]fructose-6-phosphate 103 [5-13C]glutamine [5-13C]glutamate 107 [1-13C]ethylpyruvate [1-13C]pyruvate, [1-13C]lactate, [1-13C]alanine, 93 13 13 [ C] bicarbonate, CO2 13 [1,1’- C2]acetic anhydride Multiple depending on reactant 104 [1-13C]acetate [1-13C]acetylcarnitine 109 13C Urea None 105 bis-1,1-(hydroxymethyl)-[1-13C]cyclopropane-d8 HP001 None 108 a-keto-[1-13C]isocaproate [1-13C]leucine 83 [1-13C]dehydro ascorbic acid [1-13C]ascorbic acid 97,98 [1-13C]alanine [1-13C]lactate, [1-13C]pyruvate, [13C]bicarbonate 106

13 T1 of the C-1 of [1- C]pyruvic acid at 9.4 T is Clinical Polarizer Requirements given as a function of temperature (unpubl. data). Recently a DNP polarizer designed with sterile use It can be seen that the shortest T1 is 1.6 seconds at 0C. With the trityl radical present (20 mmol/L) intent was published (31). The key design criteria of there will be an additional (dipolar) relaxation con- this concept are: tribution from the electron spin. It can be seen 1. To provide a sterile barrier to the product that the contribution from the trityl is marginal, through a single use fluid path. but shifts the minimum to a different temperature 2. To eliminate consumption of liquid cryogens. (correlation time). According to relaxation theory the 3. To increase throughput by having four independ- minimum T1 scales with B0, which means that a ent parallel samples. minimum T1 of 0.7 seconds should be expected 4. To increase the size of the individual DNP sam- during the dissolution in the 3 T polarizer field. ples up to 2.0 mL. The data illustrate that the nuclear spin during the 5. To automate the operation and remove operator dissolution should pass through this T1 minimum variability and interventions. on a much faster time scale to avoid a loss of 6. To add noncontact quality control (QC). polarization. The example illustrates that it is not unreasonable to expect that the loss of polarization The above criteria are required for successful clini- during dissolution can be overcome, but that a fast cal translation of hyperpolarization. The future will and efficient dissolution process is needed. The se- reveal if these are adequate and can fulfill regulatory verity of the problem will depend on the target spin requirements as well as gain user acceptance. Most and sample properties, but several parameters can important, it should be emphasized that a sealed ster- be controlled, eg, the distance to other spins (label- ile fluid path for all components in the polarization ing position), the abundance of other spins (full or and dissolution process (compounding) provides a partial deuteration), and the concentration of the concept that allows filling and sealing at a manufac- EPA. turing site without the need to break this barrier at In most cases the EPA or Gd chelate do not cause any point in the process until the point of product significant relaxation after dissolution, and may also release by the QC system. Second, it has been demon- be safe to inject into animals. For preclinical imaging strated that a closed-cycle, sorption pump based cryo- it is not required to remove the EPA. The same applies genic system served by a cryo-cooler was able to to the Gd chelate in case it is used in the formulation. achieve a base temperature of less than 0.8 K. The However, the solution may undergo a filtration or lower temperature translated directly into a higher chromatography step to remove the EPA involved in solid state polarization. Thus, the authors reached a the DNP process. In case a Gd chelate has been solid state polarization of 35%–40% at 0.8 K and 3.35 added, this agent may be removed as well. The filtra- T. tion can either be in-line with the dissolution process or a subsequent step. In either case the filtration is Dissolution-DNP Agents completed in a matter of a few seconds with insignificant loss of polarization or target molecule (unpubl. Of the growing list of agents investigated in vivo using work). this method (Table 1), the most studied is Hyperpolarized 13C MRI 1319

[1-13C]pyruvate. This agent has shown great utility in cept clinical trial of hyperpolarized 13C metabolic oncology, as exemplified by studies showing correla- imaging in prostate cancer patients, a 13C transmit- tions with disease progression (34) and early response only volume coil built into a custom patient table was to therapy (35). Research in cardiology (36) and brain used in conjunction with a receive-only endorectal (37) have also shown promise. [1-13]pyruvate was also coil that contained both a 13C and a 1H element for the first agent to be used in a human study of hyper- signal reception and the system body coil was used polarized metabolic imaging (1). This molecule illus- for 1H RF transmission during 1H imaging (39). A trates a number of important features of an ideal multichannel 13C receive-only array coil suited for agent for hyperpolarized metabolic imaging. First, as other human applications has also been demon- pyruvic acid, it is a liquid at room temperature and strated recently (40). Regardless of the coil design and can directly solubilize enough EPA (15 mM trityl) for combination, the MR system needs to be configured relatively fast polarization build-up (time constant of so that the correct coils/channels are active or dis- 15 min at 1.4 K and 3.35 T), and relatively high abled during specific periods of the scans to avoid sig- polarization (>20%). The high concentration inherent nal degradation due to coupling. The gradient coils in the choice of a neat liquid (14 M for pyruvic acid) existing on all MR scanners to provide spatially vary- also yields a relatively high concentration after disso- ing magnetic fields to allow localization of RF signals lution. As a result, this agent can be injected safely at can be used for 13C imaging without any hardware 250 mM, in doses up to 0.43 mL/kg. modification. However, it is important to note that for a given magnetic field gradient the spatial variation in MNS Hardware resonance frequency experienced by the nucleus is proportional to its gyromagnetic ratio. Thus, the high- Standard clinical MR systems and coils are designed est spatial resolution achievable for 13C imaging is to transmit and receive radiofrequency (RF) signals at approximately one-fourth that of 1H under the same 1H resonance frequency only. However, multinuclear imaging conditions; the designs and implementations spectroscopy (MNS) packages are available from most of RF pulse sequences for 13C imaging need to take manufacturers of whole-body MR scanners. This this limitation into consideration. It is possible to cir- option allows the system to perform MR experiments cumvent the low gyromagnetic ratio limitation by on nonproton nuclei of interest such as 13C and 15N transferring the 13Cor15N magnetization to neighbor- (for simplicity, the remainder of this section focuses ing 1H nuclei for detection (41,42). But simultaneous on hardware required for 13C studies). An MNS pack- RF transmission at both 1H and the low g nucleus fre- age typically includes a broadband RF power ampli- quencies is required for the polarization transfer pulse fier, in addition to the standard 1H narrowband ampli- sequence, and this capability may not be available on fier, that amplifies the RF pulse waveforms to give some clinical MR systems even with MNS package them enough power through a transmit RF coil to cre- installed. ate the necessary B1 field at the resonance frequency of the nucleus of interest. This transmit RF coil could Nonrecoverable Magnetization be a dedicated coil tuned to the resonance frequency for 13C. The 13C coil can be designed to perform both The magnetization of hyperpolarized 13C substrate is RF transmission and reception for 13C, or designed to largely enhanced in the DNP polarizer. After dissolu- perform RF transmission only with a separate coil(s) tion, the liquid state polarization currently achievable for RF reception, also tuned to the 13C resonance fre- is 20% (or 200,000 ppm). Once dissolved, the hyper- 13 quency. Whether the reception of the MR signal is polarized C magnetization undergoes T1 relaxation performed by a dedicated RF receive coil(s) or by a toward thermal equilibrium in a similar physical transmit/receive coil, the signal is amplified by a pre- mechanism as water protons in the human body after amplifier prior to digitization, processing, and image an RF excitation or inversion in a typical MRI scan. reconstruction. The preamplifiers typically work at a But body protons recover to thermal equilibrium via 13 narrow range of frequencies and can be built into the T1 relaxation, whereas the hyperpolarized C sub- MR scanner or into the RF coil, with one preamplifier strate irreversibly decays into thermal equilibrium via generally required for each receive channel. Thus, to T1 relaxation. Once decayed, the 200,000 ppm hyper- perform 13C experiments, dedicated preamplifiers that polarized magnetization is not recoverable and the operate at the 13C frequency need to be either added magnetization of the 13C substrate remains at the to the system or built into the 13C coils. thermal equilibrium level, about 2.6 ppm at 3 T. In 1 Since it is desirable to perform both H anatomical addition to the loss of polarization due to T1 relaxa- imaging and hyperpolarized 13C metabolic imaging tion, RF pulses deplete polarization in a nonrecover- during the same exam without repositioning the sub- able way. Hyperpolarized gas imaging using 3He or ject, the 13C RF coil design and setup need to preserve 129Xe also utilizes nonrecoverable magnetization, sub- 1 the ability to perform H imaging with a minimal com- ject to T1 relaxation and RF depletion. Thus, sample promise of image quality. Volume coils that can oper- delivery and data acquisition in hyperpolarized imag- ate at both 1H and 13C frequencies (dual-tuned) have ing need to be sufficiently fast in order to utilize this been demonstrated for preclinical hyperpolarized 13C decaying and nonrecoverable magnetization (for imaging (34,38). The coil configuration and design [1-13C]pyruvate, most of the nonequilibrium polariza- can be further optimized for imaging a particular tion is lost within 2–3 minutes postdissolution). How- organ/anatomy. For example, in the first proof-of-con- ever, different from hyperpolarized gases, some 1320 Hurd et al.

Figure 3. Dynamic curves of 13C- pyruvate, and its metabolic products: 13C-alanine, 13C-lactate, and 13C-bicarbonate (as labeled), following an injection of hyperpolarized [1-13C]pyruvate into a rat. The dynamic signals were obtained from the individual peak height of a stack of spectra (right insert) acquired every 3 seconds from a 90-mm thick section of the rat torso. hyperpolarized 13C substrates such as [1-13C]pyru- organ of interest, and perfusion. Therefore, a nonspa- vate also undergo metabolic conversions to down- tially resolved dynamic scan of the same region (Fig. stream metabolites. Therefore, the signal-to-noise ra- 3) was often performed (in a separate bolus injection) tio (SNR) of hyperpolarized 13C in vivo depends on the prior to the imaging study to gain timing information T1 relaxation time, metabolic conversion rates, liquid from the metabolic signal–time curves (38). For a typi- state polarization, concentration, agent delivery time, cal protocol of a 16 16 matrix and 5 5 mm in- acquisition timing, and pulse sequence strategies of plane resolution, it requires 15–20 seconds to acquire utilizing the nonrecoverable magnetization. The fol- CSI of a single slice because of the long readout dura- lowing section describes the most popular data acqui- tion (to obtain adequate spectral resolution) and one sition strategies for hyperpolarized 13C imaging. TR is needed for each spatial encoding point in X and Y. Higher spatial resolution is possible in the same Pulse Sequences scan time, but requires a smaller field of view (FOV), which may result in spatial aliasing in clinical settings Hyperpolarized 13C MRI typically requires acquisition unless the 13C coil receptivity profile limits the FOV, of 13C signals from the injected metabolite and its such as is the case for surface coils or endorectal metabolic products. These 13C-labeled metabolites coils. can be observed as a spectrum of peaks at different Rapid CSI techniques have been developed to resonance frequencies. Both the spatial distribution improve sampling efficiency within the available time and temporal evolution of the metabolite signals are window. Echo-planar spectroscopic imaging (EPSI) of a strong interest for understanding the dynamic with flyback (46) or symmetric gradient waveform (47) metabolic process in vivo. Pulse sequence design pro- traverses time and one spatial frequency domain in a gressed rapidly from single-slice, single-timepoint ac- single readout period, shortening the acquisition and quisition to five-dimensional MRSI: temporal, spectral, allowing either single-timepoint 3D MRSI or time- and three spatial dimensions. Optimizing sampling ef- resolved multislice 2D MRSI (34,45) on a standard ficiency of nonrecoverable magnetization has been the clinical 3 T system (with a maximum gradient primary focus of pulse sequence development for strength of 4 G/cm and slew rate of 150 mT/m/ms). hyperpolarized 13C MRSI. Methods such as com- There is a trade-off between spectral bandwidth and pressed sensing and iterative decomposition of water spatial resolution in the design of these gradient tra- and fat with echo asymmetry and least square estima- jectories. Typically, a 5-mm resolution is achievable tion (IDEAL) have been applied to accelerate acquisi- with 500 Hz spectral bandwidth without spectral ali- tions in this context, and various RF pulse designs asing of [1-13C]pyruvate and its metabolic products have been used to optimize SNR and/or contrast-to- (except 13C bicarbonate). A similar trade-off also noise ratio (CNR). In addition, there are specialty exists for spiral CSI (48), which employs spiral read- sequences for quantitation of T2 and metabolic out gradients to sample X and Y simultaneously, and kinetics. concatenates the spiral gradients multiple times for chemical shift encoding. However, even with multiple Single Timepoint MRSI interleaves to minimize the impact of gradient slew- rate, the 2D spiral readout time can result in a spec- Early work in hyperpolarized 13C MRSI employed con- tral bandwidth that is insufficient to fully cover the centric phase encoding and variable flip angle (38,43– metabolite chemical shift range. This causes spectral 45) techniques to acquire chemical shift images (CSI) aliasing, which needs to be corrected in the image in two dimensions within a short time window that reconstruction (48,49) or otherwise, results in image coincides with the maximum 13C signals of metabolic blurring. On a clinical system, spiral CSI completes a products. For these single timepoint images, the opti- 2D MRSI of a single slice in 375 msec, a 50-fold mum acquisition depends on the bolus injection, the reduction in scan time (50) compared to the Hyperpolarized 13C MRI 1321

MRSI has been demonstrated by using spiral CSI (51,52) and compressed sensing (53,54), both yielding high-quality images and dynamic curves. Taking advantage of the considerable sparsity in hyperpolarized 13C spectra, compressed sensing pseudoran- domly undersampled spectral and X-Y spatial domains during EPSI flyback readout, yields up to a factor of 7.53 in acceleration (55) relative to the conventional 3D EPSI sequence (46) (Fig. 5). The acceleration can be used to improve spatial resolution and decrease acquisition time, or to cover a larger FOV, which will be useful for clinical applications. The trade-off of this technique is the loss of metabolite peaks with low SNR (appears to break down for SNR less than 7). This could potentially limit its applications depending on the achievable clinical SNR, which Figure 4. Comparison of spiral CSI and conventional is yet to be determined by clinical trials. FIDCSI. Images were obtained from a single slice rat kidney Another approach under development is IDEAL spi- for [1-13C]pyruvate (left), [1-13C]lactate (middle), and 13 ral CSI (56), using the iterative least-squares chemical [1- C]alanine (right). Spiral CSI (0.375 sec per slice) than FIDCSI (17 sec per slice) with similar image quality. Courtesy shift-based (LSCS) method. This technique has been of Dr. Dirk Mayer of SRI, International and Stanford used clinically for decompositions of water from fat 13 University. (57), but is also capable of decomposing multiple C- labeled chemical species (58). Spectral sampling is conventional CSI method (Fig. 4). However, for clinical accomplished by shifting the echo time (TE) from exci- applications that require a large FOV, spiral CSI ac- tation to excitation and 2D images at each TE are quisition time may increase drastically due to the acquired by using spiral gradient trajectories. IDEAL increase of interleaves required to maintain the same requires a priori information of chemical shift fre- 13 spatial resolution and spectral bandwidth. In addi- quencies of C metabolite peaks and, with such, tion, spiral CSI encodes a circular FOV and can IDEAL allows minimum numbers of excitations for become inefficient for a region of interest (ROI) with spectral decomposition, an efficient sampling strategy an asymmetric FOV. On the other hand, EPSI allows for a sparse spectrum over a wide bandwidth. There asymmetric FOV and the FOV in the one direction is no trade-off between spatial resolution and spectral encoded by the EPSI readout is virtually unlimited bandwidth and, therefore, the spatial resolution can due to the very high sampling rate available on all be as high as SNR permits. This technique has been clinical systems. demonstrated in time-resolved 2D imaging, with a potential of combining with a pulse-and-acquire FID acquisition to obtain a pseudo spectrum (59). The in- Acceleration to 5D MRSI tegrity of the pseudo spectrum obtained by using this When the information of temporal dynamics and spa- technique for quantitation purposes is under investi- tial distribution are both needed, time-resolved MRSI gation in preclinical studies. with multislice 2D or 3D volumetric coverage is a good strategy. Time-resolved metabolic data can be RF Designs to Optimize SNR used to determine rate constants (51), and signal averaging over the time course for each voxel can The signal of the injected, relatively concentrated, regain most of the SNR observed in optimized non- hyperpolarized 13C substrate is often 5 to 10 times time-resolved methods. In preclinical studies, 5D larger than the signals of its metabolic products

Figure 5. Pulse diagram of multiband and compressed sensing sequence (left), and reconstructed spectra (right) of time- resolved 3D MRSI data. Improved peak detection of the wavelet-in-time method is most evident at 30 seconds after hyperpolarized [1-13C]pyruvate injection. Courtesy of Dr. Peder Larson, Dr. Simon Hu, and Dr. Dan Vigneron, University of California at San Francisco. 1322 Hurd et al.

13 13 Figure 6. Quantitative C-lactate (middle) and C-pyruvate (right) T2 maps of a single 2 mm slice on TRAMP tumor. Tumor 13 13 appears homogeneous on proton 3D-SPGR image (left) but has heterogeneous T2 values for [1- C]lactate and [1- C]pyruvate. The estimated T2 errors range from 100 to 200 msec (not shown). [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.] initially, whereas the product signals are replenished aT2 of 1.4 seconds. The large T2 difference between during the acquisition via recirculation of the surplus tumor and normal tissues presents an opportunity for 13 C substrate. Multiband spectral-spatial RF excita- greater imaging contrast when using a T2-based tion pulses (60) use spectral selectivity to minimally sequence. A large SNR gain is also expected by using 13 excite the injected hyperpolarized C substrate while T2-based sequences as compared to T2 -based sequen- exciting the metabolic products with a larger flip angle ces. For example, most of the sequences mentioned to obtain higher SNR of the metabolites without satu- before are limited by signal loss caused by T2 decay. 13 rating the substrate magnetization prematurely. The T2 can be as long as 100–200 msec for [1- C]pyru- metabolic products are observable for a longer win- vate, but is reduced to about 25 msec for [1-13C]lac- 13 dow and with better SNR than a uniformly constant tate and [1- C]alanine due to stronger JCH coupling flip angle (60) strategy. A recent development combin- (38), and all these T2 values are shorter than the T2 ing multiband RF pulse design and compressed sens- values. T2-based sequences, such as multiecho bal- ing random sampling created a sequence for time- anced-steady-state free precession (SSFP) (68,69) and resolved 3D MRSI acquisition (55,61) with good SNR. stabilized fast spin echo (FSE)-EPSI (70), have signifi- The flip-angle of the injected 13C substrate and that of cant signal gain and are excellent for single-timepoint the products can be optimized for optimal CNR ratio MRSI. The challenge to utilize this strategy for time- for a particular organ or for disease characterization resolved MRSI lies in the strong RF depletion during (62). the echo train. Multiecho balanced-SSFP has been An alternative to spectroscopically resolving multi- demonstrated in 3D acquisition for time-resolved ple metabolites is direct imaging of each metabolite af- hyperpolarized chemical shift imaging (71), but the ter selective excitation by a spectral-spatial pulse (63). temporal resolution of 16 seconds may not be suffi- Recently, a multislice cardiac-gated sequence consist- cient for characterizing the metabolic dynamics. ing of a large flip-angle spectral-spatial excitation RF pulse with a single-shot spiral trajectory was devel- Kinetic Modeling oped for 13C imaging of cardiac metabolism (64). The sequence alternates among the chemical shift fre- Hyperpolarized pyruvate-to-lactate signal-time curves quencies corresponding to each metabolite and allows have been described by two-site exchange (35,72) for rapid imaging of each individual metabolite. models. Under saturating conditions, the apparent rate constant Kpl increases as the pyruvate dose T -Based Sequences decreases (72). The small tip-angle, pulse-and-acquire 2 dynamic curves are biased by the substrate dose, 13 13 Long T2 relaxation time of C metabolites was first bolus shape, and accumulated in flow of [1- C]lac- observed using a TRAMP (transgenic adenocarcinoma tate. These factors can be eliminated by using a satu- of the mouse prostate) tumor model (65). The T2 dif- ration recovery method (67,68), resulting in dynamic ference between tumor and normal tissue was curves that describe the instantaneous metabolic con- explored in a rat hepatocellular carcinoma (HCC) version at the local tissue level during the passage of study (66) using a single-voxel preparation pulse fol- hyperpolarized [1-13C]pyruvate. This method typically lowed by a train of spin-echoes to measure the T2 consists of multiple 90 -excitations to acquire 13 decay of the signal within the voxel. T2’s of [1- C]ala- dynamic spectroscopic images and spectrally selective nine and [1-13C]lactate were found to be longer in saturation pulses applied in between acquisition of HCC tumors (1.2 sec and 0.9 sec, respectively) than time frames to spoil the inflow of [1-13C]lactate while in normal liver (0.4 sec and 0.5 sec, respectively). preserving inflow of fresh [1-13C]pyruvate. The result- Recently, a T2 mapping sequence was developed to ing kinetic data were fully sampled at each timepoint 13 measure T2 of C-labeled metabolites pixel-by-pixel and were unbiased by the substrate dose, bolus with high resolution (67). T2 values were extracted shape, and product decay. Apparent maximal reaction from regions of interest on the T2 maps with better velocity Vmax and asymptotic conversion rate at satu- precision. Figure 6 shows the T2 map of a single-slice rated condition KM can be derived by kinetic modeling acquired through a TRAMP tumor and the tumor has of the saturation recovery dynamic curves (67). Hyperpolarized 13C MRI 1323

Exchange vs. Flux study was conducted with the primary objective to 13 13 13 assess the safety of hyperpolarized pyruvate ( C) The conversion of [1- C]pyruvate to [1- C]lactate, as injection in men with prostate cancer and intact pros- observed in hyperpolarized metabolic imaging, is a tates. The secondary objectives were to determine: 1) combination of flux (net creation of lactate) and 13 The kinetics of hyperpolarized pyruvate injection exchange ( C enrichment of the lactate pool with no delivery and metabolism throughout the prostate, and net change in concentration). In whole blood, pyru- 2) to determine the SNR for 13C pyruvate metabolites vate-lactate exchange occurs at a rate 3–5 times the and total hyperpolarized carbon (THC) in regions of rate of flux (73). The impact of lactate pool size and cancer and in surrounding benign prostate as a func- exchange was demonstrated for hyperpolarized 13 13 tion of the dose of the hyperpolarized pyruvate ( C) [1- C]pyruvate metabolic spectroscopy using cells injection. All doses were well tolerated without excep- preconditioned with unlabeled lactate (35). Under tion, and excellent CNR for [1-13C]lactate was conditions of elevated steady-state lactate (unlabeled), observed even at the lowest dose (private commun.). a large increase in hyperpolarized [1-13C]lactate was observed. Recently, the importance of exchange has Liver Metabolism and Hepatocellular Carcinoma been demonstrated in a lymphoma model, under the 13 high bolus concentration hyperpolarized [1-13C]pyru- Using hyperpolarized [1- C]pyruvate, Hu et al (77) vate and magnetization transfer technique (74). The studied liver metabolism in fasted rats and found authors of this study further concluded that steady- higher lactate-to-alanine signal ratios and lower ala- state lactate pool size is the likely limit of detection for nine signal level in the fasted rats than in free-fed [1-13C]lactate in regions-of-interest such as blood and rats. The low alanine signal is most likely due to a muscle. reduction of alanine aminotransferase (ALT) activity in The availability of the reduced form of nicotinamide fasted rat liver during gluconeogenesis. Alanine is adenine dinucleotide NADH, from sources beyond lac- also a good biomarker for HCC detection. Using 13 tate dehydrogenase (LDH) catalyzed exchange, also hyperpolarized [1- C]pyruvate, Darpolor et al (78) impacts the conversion of hyperpolarized [1-13C]pyru- found elevated alanine and lactate levels, consistent 13 with enzyme expression analysis on rat HCC tissue vate to [1- C]lactate. For example, added NADH from 13 aldolase processing of ethanol in liver (75), or from extract. Interestingly, C MRSI showed high alanine mitochondria via a reverse of the malate-aspartate signals specifically in HCC tumors, whereas it showed high lactate signals in the HCC tumors and in blood shuttle (76), have been shown to increase the flux of 13 hyperpolarized [1-13C]pyruvate to [1-13C]lactate. How- vessels. Low C-alanine signals in vessels may be ever, the balance of flux and exchange has not yet due to the much slower transport of alanine than lactate from cells to blood. Therefore, within the 1 mi- been quantitatively established, and remains to be 13 13 determined, even for normal tissues and conditions. nute of C acquisition time window, not much C-alanine signal was observed in vessels but only in HCC tumors. This is a promising technique for liver cancer APPLICATIONS diagnosis and treatment monitoring. Oncology Glioma Prostate Cancer Park et al (79) assessed the potential use of hyperpo- 13 Initial experience in hyperpolarized Cmetabolic larized 13C-pyruvate for glioma prognosis in rat mod- imaging of prostate cancer was reported by Chen et al els. The signal levels of 13C-pyruvate and its metabolic 13 (45) by injecting hyperpolarized [1- C]pyruvate into product, 13C-lactate, as well as their relative signal transgenic adenocarcinoma of mouse prostate (TRAMP) ratios were significantly higher in tumors than in nor- model. The study showed highly elevated lactate signal mal brain. The 13C-lactate signal correlated with pro- in late-stage prostate tumors. Albers et al (34) com- liferation. The different 13C metabolic profile between 13 pared hyperpolarized C metabolic imaging of prostate two different models in the study was consistent with cancer with histology. Normal mice and TRAMP of vari- their immunohistochemical data. Time-resolved 2D 13 ous histologic grades were studied. Images of C-py- MRSI was reported recently in a rat glioma model, 13 13 ruvate, C-lactate, and C-alanine were obtained by comparing metabolic conversion rates between glioma using 3D EPSI sequence in 14 seconds. The lactate and normal brain (80). In both studies, large 13C-py- signal level increases with tumor progression and cor- ruvate uptake was observed due to the disruption of relates strongly with histologic grade. the blood–brain barrier (BBB) in gliomas. For studies of 13C-pyruvate metabolism in normal rat brain, Clinical Trial where the BBB is intact, see the Neurology section below. The first clinical trial of hyperpolarized 13C-pyruvate metabolic imaging of prostate cancer patients was Lymphoma successfully conducted at the University of California in San Francisco (1). This study was a proof-of-con- Extracellular pH is known as a biomarker of intersti- cept trial entitled ‘‘A Phase 1/2a Ascending-Dose tial lactic acid production (81). Although intracellular Study to Assess the Safety and Tolerability and Imag- pH has been measured by 31P MRS (82), the lower ing Potential of Hyperpolarized Pyruvate (13C) Injec- sensitivity of 31P MRS limits its application for human tion in Subjects with Prostate Cancer.’’ This 33-patient studies with appropriate spatial resolution and 1324 Hurd et al. reasonable imaging time window. With the 5 orders of and diabetes; it also occurs during ischemia and reper- magnitude signal enhancement afforded by the DNP fusion. Since all substrates are converted to acetyl- technique, Gallagher et al (13) mapped the pH of mu- CoA prior to entering the Krebs cycle, measurement of rine lymphoma tumor by applying 13C MRSI following the metabolic fluxes of acetyl-CoA production from an injection of hyperpolarized 13C-bicarbonate. The various substrates can be used to monitor the changes pH value in each voxel was calculated using the rela- in substrate selection and utilization. Pyruvate dehy- tive signal of 13C-bicarbonate and its metabolic prod- drogenase (PDH) is the enzyme that decarboxylates the 13 uct CO2 using the Henderson-Hasselbalch equation. carbohydrate derived pyruvate to acetyl-CoA and CO2, The tumor showed lower pH than the surrounding and the control of this enzyme’s expression and activity healthy tissues. is closely tied to myocardial substrate selection, thus Another hyperpolarized 13C substrate that has been the ability of using hyperpolarized 13C pyruvate to non- tested on lymphoma is 2-keto-[1-13C]isocaproate invasively probe PDH flux is potentially a powerful (KIC). KIC is metabolized to leucine by branched diagnostic tool in cardiology. chain amino acid transferase (BCAT), a biomarker for Indeed, a number of recent reports in small and large metastasis in some tumors and a target of proto-onco- animal models have demonstrated the ability of hyper- gene c-myc. Following injection of hyperpolarized KIC, polarized 13C MR imaging and spectroscopy to charac- Karlsson et al (83) found more than a 7-fold higher terize the PDH flux noninvasively in normal hearts and signal of 13C-leucine in murine lymphoma than in hearts during ischemia-reperfusion and cardiac dis- 13 13 healthy tissue. In the same study, no C-leucine was eases (86–90). In normal hearts, the CO2 derived from observed in rat mammary adenocarcinoma. Ex vivo [1-13C]pyruvate due to cardiac PDH flux is observed 13 13 BCAT expression analysis yielded a high BCAT level mostly as C-bicarbonate (in equilibrium with CO2) in murine lymphoma and a very low BCAT level in rat signal (Fig. 7) in MR spectroscopy data (64,87,88), and mammary tumor, consistent with the hyperpolarized some [1-13C]lactate and [1-13C]alanine signals can also 13C metabolic imaging findings. be observed. In spatially resolved 13C MRI data obtained from large animal models, the substrate signal was Therapeutic Response found to be localized mostly in the cardiac chambers, while 13C-bicarbonate was localized in the myocardium Day et al (35) reported decreased flux between pyru- (Fig. 8); [1-13C]lactate signal was more diffuse and vate and lactate in lymphoma tumors when treated observed in both the blood and cardiac muscle (64,86). with etoposide and interrogated with hyperpolarized In models of ischemia and reperfusion, impaired 13 [1- C]pyruvate. The etoposide induces apoptosis and PDH flux can be observed as decreased 13C-bicarbon- loss of NADH due to activation of poly (ADP-ribose) ate signal shortly following reperfusion (86,89). Poten- polymerase (PARP) leading to reduced LDH activity. tially, the viability of the affected tissue may be This study is the benchmark to demonstrate the feasi- probed by following the recovery of the PDH flux (or 13 bility of using hyperpolarized C metabolic MR to the lack of it) post reperfusion and assessment of monitor early treatment effects. interventions targeting this metabolic pathway may A similar finding was also reported by Chen et al also benefit from this technique. Changes in PDH flux (84) in a study of treatment response on TRAMP due to diabetes have been investigated in a small ani- 13 tumors by using hyperpolarized [1- C]pyruvate. mal model (88). Very recently it has also been reported 13 Reduced C lactate to pyruvate ratio was found in that in a porcine pacing model of dilated cardiomyop- the TRAMP mice that responded to androgen depriva- athy (DCM) the disease progression can be followed tion therapy and no change was found in the ratio in noninvasively with 13C metabolic imaging using nonresponding mice. hyperpolarized [1-13C]pyruvate, and altered cardiac In a study of treatment monitoring of lymphoma PDH flux was found to be strongly associated with 13 tumors using hyperpolarized [1,4- C2]fumarate, Gal- onset of decompensated DCM (90). Monitoring cardiac lagher et al (85) found that production of substrate utilization in patients may provide valuable 13 [1,4- C2]malate from the labeled fumarate is a sensi- information regarding progression of these diseases tive marker of cellular necrosis. The conversion was and aid clinical management. 2.4-fold higher in etoposide-treated lymphoma Although most of the efforts so far in utilizing tumors, where significant levels of tumor cell necrosis hyperpolarized 13C MR metabolic imaging in cardiol- formed than in the untreated tumors. This technique ogy have been focused on probing substrate utiliza- has clinical potential for monitoring early therapeutic tion using [1-13C]pyruvate, the cardiac pH may also response. 13 13 be assessed noninvasively by the H CO3 / CO2 ratio and the Henderson-Hasselbalch equation (87,91), if 13 Cardiology sufficient SNR is obtained for the CO2 signal. Moni- toring of cardiac Krebs cycle flux in real time using Generation and utilization of adenosine triphosphate hyperpolarized [2-13C]pyruvate is also feasible since (ATP) in the heart are tightly controlled events regu- the C2 position on pyruvate is carried into the cycle lated by physiological conditions and energetic needs. through acetyl-CoA (instead of being released as 13 Normally, the heart uses fatty acids, carbohydrates, CO2), and changes of Krebs cycle flux can be and ketones as the substrates for energy production. assessed by measuring changes in the [5-13C]gluta- Altered myocardial substrate utilization is associated mate signal (92). Along with PDH flux, these addi- with diseases such as cardiomyopathy, hypertension, tional parameters obtainable by hyperpolarized 13C Hyperpolarized 13C MRI 1325

Figure 7. Cardiac-gated dynamic MRS data from pig hearts. a: Representative spectrum from the maximum bicarbonate frame in a fasted pig. b: Representative spectrum from the maximum bicarbonate frame in an oral glucose loaded pig. c: Time course of peak areas of pyruvate, bicarbonate, lactate, and alanine resonances acquired every 4 R-R intervals in the oral glucose loaded pig. The maximum bicarbonate to maximum pyruvate ratio (BPR) altered dramatically based on fed condition of the animal, due to changes in myocardial substrate utilization. Used with permission from Lau AZ, et al., Rapid multislice imaging of hyperpolarized (13)C pyruvate and bicarbonate in the heart. Magn Reson Med 2010;64:1323–1331, John Wiley & Sons.

MR provide insights into cardiac energetics and cellu- metabolic imaging has the potential to address unmet lar environment that were not previously accessible clinical needs in neurodegenerative disease, traumatic noninvasively by other imaging modalities and may brain injury, and stroke. Unfortunately, this area of become valuable clinical tools in cardiology. research trails the exciting progress that has been made in oncology and cardiology. Part of the lag in Neurology neurology may be due to the concern about the transport rate of T1-limited hyperpolarized metabolic imag- The direct quantitative measures of BBB transport, ing agents through the BBB. One strategy to overcome inﬂammation, and oxidative load with hyperpolarized the BBB transport limit explored the use of the

Figure 8. In vivo dynamic 13C MRI data acquired using a multislice respiratory-gated spiral sequence showing spatial distribution of metabolites in a short-axis view of the heart. Pyruvate volume (six slices) were acquired starting from 10 seconds after the start of [1-13C]pyruvate injection to capture the bolus through the heart (one volume of pyruvate images acquired during one respiratory cycle, 10 respiratory cycles of pyruvate data acquired, pyruvate images from peak of the bolus shown). Bicarbonate and lactate image volumes were acquired after the pyruvate bolus and were each repeated three times. The resolution of the overlaid reconstructed 13C images is 10.7 mm in-plane for bicarbonate and pyruvate and 12 mm for lactate with a 1-cm slice thickness (pyruvate images are shown with a difference scale from bicarbonate and lactate images). The scan was completed in 1 minute. Almost all the [1-13C]pyruvate signal observed was localized in the blood while 13C-bicarbonate was confined mostly in the heart muscle. Figure courtesy of Angus Z. Lau and Charles H. Cunningham of Sunnybrook Health Sciences Centre. 1326 Hurd et al. nonpolar precursor molecule, ethyl-pyruvate (93). 7. Abragam A. Overhauser effect in nonmetals. Phys Rev 1955;98: This molecule is readily taken up by the brain and 1729–1735. metabolized, but injection rate is limited, and inter- 8. Jeffries CD. Polarization of nuclei by resonance saturation in paramagnetic crystals. Phys Rev 1957;108:164–165. pretation is complicated by the rate of hydrolysis. 9. Abragam A, Goldman M. Order and disorder. Oxford, UK: Clar- However, as part of this study it was discovered that a endon Press; 1982. substantial amount of [1-13C]pyruvate does make it 10. de Boer W, Borghini M, Morimoto K, Niinikoski TO, Udo F. through normal BBB during the first passage of a Dynamic polarization of protons, deuterons, and carbon-13 nuclei: bolus, and is converted to [1-13C]lactate. Both 13C-bi- thermal contact between nuclear spins and an electron spin-spin 13 interaction reservoir. J Low Temp Phys 1974;15:249–266. carbonate and [1- C]lactate appear to be formed in 11. de Boer WD, Niinikoski TO. Dynamic proton polarization in pro- brain tissue (51,94,95). In a subsequent dynamic panediol below 0.5K. Nucl Instr Meth 1974;114:495–498. 13 metabolic imaging study, the [1- C]lactate observed 12. Lewis RJ S. Sax’s dangerous properties of industrial materials. in a brain ROI was found to arise from brain metabo- New York: John Wiley & Sons; 2000. lism, while the bulk of the [1-13C]pyruvate observed in 13. Gallagher FA, Kettunen MI, Day SE, et al. Magnetic resonance imaging of pH in vivo using hyperpolarized 13C-labelled bicar- that same ROI appeared to arise from the cerebral bonate. Nature 2008;453:940–943. blood volume (51). Since a substantial amount of 14. Jensen PR, in’t Zandt R, Karlsson M, et al. In: Proc 16th Annual 13 [1- C]pyruvate makes it across the normal BBB and Meeting ISMRM, Toronto; 2008 (abstract 892). is converted to [1-13C]lactate, it should be possible to 15. Jensen PR, Karlsson M, Meier S, Duus JØ, Lerche MH. Hyper- quantitatively measure the full range of BBB trans- polarized amino acids for in vivo assays of transaminase activ- port abnormalities. The quantitative nature of the ity. Chem Eur J 2009;15:10010–10012. 13 16. Heckmann J, Meyer W, Radtke E, Reicherz G. Electron spin res- measure of total C taken up, as well as the meta- onance and its implication of the maximum nuclear polarization 13 bolic activity, could make [1- C]pyruvate metabolic of deuterated solid target materials. Phys Rev 2006;74:Art. No. imaging an ideal tool to study the full range of dis- 134418. ease-induced disruptions in the BBB, even the subtle 17. Benjamin PS, Fuminori H, Matsumoto K-I, et al. The chemistry ones that have been reported for some nonenhancing and biology of nitroxide compounds. Free Radic Biol Med 2007; 42:1632–1650. MS lesions (96). 18. Eaton SS, Eaton GR, Berliner L. Part A: free radicals, metals, 13 Beyond [1- C]pyruvate, one of the most interesting medicine, and physiology. Part B: methodology, instrumenta- agents for the study of neurodegenerative disease may tion, and dynamics. Series: Biological Magnetic Resonance. be [1-13C]dehydroascorbic acid (DHA) (97,98). This J Biomed EPR 2005;23/24. molecule has been shown to rapidly cross the BBB 19. Thaning M. PCT patent application number WO 2006/011811 A2, priority date 30 July 2004. (99), and the conversion rate of DHA to vitamin C is 20. Andersson S, Radner F, Rydbeck A, Servin R, Wisrand L-G. expected to be a direct marker for oxidative stress United States patent number US5728370, priority date 5 Janu- (97,98). A number of other dissolution DNP agents ary 1998. have also been studied in the brain including KIC and 21. Reddy T, Iwama T, Halpern HJ, Rawal VH. General synthesis of [1-13C]acetate (see Table 1). persistent trityl radicals for EPR imaging of biological systems. J Org Chem 1998;67:4635–4639. 22. Bowman MK, Mailer C, Halpern HJ. The solution conformation of triarylmethyl radicals. J Magn Reson 2005;172:254–267. CONCLUSION 23. Hu K-N, Bajaj VS, Rosay M, Griffin RG. High-frequency dynamic Dissolution-DNP-enabled metabolic imaging is still a nuclear polarization using mixtures of TEMPO and trityl radicals. J Chem Phys 2007;126:044512. relatively young field, with active preclinical research. 24. Ardenkjaer-Larsen JH, Macholl S, Johannesson H. Dynamic This includes rapid discovery of new indications for Nuclear polarization with trityls at 1.2K. Appl Magn Reson the lead compound, hyperpolarized [1-13C]pyruvate, 2008;34:509–522. as well as a robust exploration of new agents. The 25. Duijvestijn MJ, Wind RA, Smidt J. Quantitative investigation of success of the first clinical proof-of-concept trial and the dynamic nuclear polarization effect by fixed paramagnetic centra of abundant and rare spins in solids at room tempera- the development of a clinical polarizer should enable ture. Physica B, C 1986;138:147–170. the next steps toward clinical translation of this 26. Bajaj VS, Hornstein MK, Kreischer KE, et al. 250 GHz CW gyro- technology. tron oscillator for dynamic nuclear polarization in biological solid state NMR. J Magn Reson 2007;189:251–279. 27. Hornstein MK, Bajaj VS, Kreischer KE, Griffin RG, Temkin RJ. REFERENCES CW second harmonic results at 460 GHz of a gyrotron oscillator 1. Nelson SJ, Kurhanewicz J, Vigneron DB, et al. Proof of Concept — for sensitivity enhanced NMR. The Joint 30th International Clinical Trial of Hyperpolarized C-13 in Patients with Prostate Conference of Infrared and Millimeter Waves and 13th Interna- Cancer. In: Proc 20th Annual Meeting ISMRM, Melbourne; 2012 tional Conference on Terahertz Electronics 2005;2:437–438. (abstract 274). 28. Comment A, van der Brandt B, Uffmann K, et al. Design and 2. Ardenkjaer-Larsen JH, Fridlund B, Gram A, et al. Increase in performance of a DNP prepolarizer coupled to a rodent MRI signal-to-noise ratio of > 10,000 times in liquid-state NMR. Proc scanner. Conc Magn Reson 2007;31:255–269. Natl Acad Sci U S A 2003;100:10158–10163. 29. Jannin S, Comment A, Kurdzesau F, et al. A 140 Ghz prepolar- 3. Ardenkjaer-Larsen JH, Axelsson O, Golman K, Wistrand L-G, izer for dissolution dynamic nuclear polarization. J Chem Phys Leunbach I, Petersson S. PCT patent application number 2008;128:241102. WO99/35508, priority date 5 January 1998. 30. Johannesson H, Macholl S, Ardenkjaer-Larsen JH. Dynamic 4. Wolber J, Ellner F, Fridlund B, et al. Generating highly polarized Nuclear Polarization of [1–13C]pyruvic acid at 4.6 Tesla. J Magn nuclear spins in solution using dynamic nuclear polarization. Reson 2008;197:167–175. Nucl Meth Phys Res 2004;526:173–181. 31. Ardenkjaer-Larsen JH, Leach AM, Clarke N, Urbahn J, Ander- 5. Overhauser AW. Polarization of nuclei in metals. Phys Rev 1953; son D Skloss TW. Dynamic nuclear polarization polarizer for 92:411–415. sterile use intent. NMR Biomed 2011;24:927–932. 6. Carver TR, Slichter CP. Polarization of nuclear spins in metals. 32. Goertz, S. Spintemperatur und magnetische Resonanz ver- Phys Rev 1953;92:212–213. du¨nnter elektronischer Systeme — ein Weg zur Optimierung Hyperpolarized 13C MRI 1327

polarisierter Festkorper-Targetmaterialien,€ Ruhr-Universitat€ 55. Larson PE, Hu S, Lustig M, et al. Fast dynamic 3D MR spectro- Bochum, Habilitationsschrift, April 15, 2002. scopic imaging with compressed sensing and multiband excita- 33. Jain J, Dey S, Muralidharan L, Leach AM, Ardenkjaer-Larsen tion pulses for hyperpolarized (13)C studies. Magn Reson Med JH. Jet impingment melting with vaporization: A numerical 2010;65:610–619. study. Proceedings of the ASME Summer Heat Transfer Confer- 56. Wiesinger F, Menzel MI, Weidl E, Janich M, Schwaiger M, ence, HT 2008, 559–567. Schulte RF. Minimum-norm IDEAL spiral CSI for efficient 34. Albers MJ, Bok R, Chen AP, et al. Hyperpolarized 13C lactate, hyperpolarized 13C metabolic imaging. In: Proc 18th Annual pyruvate, and alanine: noninvasive biomarkers for prostate can- Meeting ISMRM, Stockholm; 2010 (abstract 3290). cer detection and grading. Cancer Res 2008;68:8607–8615. 57. Reeder SB, Pineda AR, Wen Z, et al. Iterative decomposition of 35. Day SE, Kettunen MI, Gallagher FA, et al. Detecting tumor water and fat with echo asymmetry and least-squares estima- response to treatment using hyperpolarized 13C magnetic reso- tion (IDEAL): application with fast spin-echo imaging. Magn nance imaging and spectroscopy. Nat Med 2007;13:1382–1387. Reson Med 2005;54:636–644. 36. Schroeder MA, Clarke K, Neubauer S, Tyler DJ. Hyperpolarized 58. Reeder SB, Brittain JH, Grist TM, Yen YF. Least-squares chemi- magnetic resonance: a novel technique for the in vivo assessment cal shift separation for (13)C metabolic imaging. J Magn Reson of cardiovascular disease. Circulation 2011;124:1580–1594. Imaging 2007;26:1145–1152. 37. Hurd RE, Yen YF, Mayer D, et al. Metabolic imaging in the anes- 59. Wiesinger F, Weidl E, Menzel MI, et al. IDEAL spiral CSI for thetized rat brain using hyperpolarized [1–13C] pyruvate and dynamic metabolic MR imaging of hyperpolarized [1–13C]pyru- [1–13C] ethyl pyruvate. Magn Reson Med 2010;63:1137–1143. vate. Magn Reson Med 2011;25:142–151. 38. Kohler SJ, Yen Y, Wolber J, et al. In vivo 13 carbon metabolic 60. Larson PE, Kerr AB, Chen AP, et al. Multiband excitation pulses imaging at 3T with hyperpolarized 13C-1-pyruvate. Magn Reson for hyperpolarized 13C dynamic chemical-shift imaging. J Magn Med 2007;58:65–69. Reson 2008;194:121–127. 39. Tropp J, Calderon P, Carvajal L, et al. An endorectal dual fre- 61. Larson PE, Bok R, Kerr AB, et al. Investigation of tumor hyper- quency 13C-1H receive only probe for operation at 3.0 Tesla. In: polarized [1–13C]-pyruvate dynamics using time-resolved multi- Proc 14th Annual Meeting ISMRM, Seattle; 2006 (abstract 2594). band RF excitation echo-planar MRSI. Magn Reson Med 2010; 40. Tropp J, Lupo JM, Chen A, et al. Multi-channel metabolic imag- 63:582–591. ing, with SENSE reconstruction, of hyperpolarized [1-(13)C] 62. Josan S, Hurd RE, Kerr AB, et al. Effects of RF excitation pyruvate in a live rat at 3.0 Tesla on a clinical MR scanner. scheme on signal-to-noise-ratio and apparent rate constant esti- J Magn Reson 2011;208:171–177. mation in dynamic volumetric imaging of hyperpolarized [1– 41. Barb AW, Hekmatyar SK, Glushka JN, Prestegard JH. Exchange 13C]-pyruvate. In: Proc 19th Annual Meeting ISMRM, Montreal; facilitated indirect detection of hyperpolarized 15ND2-amido- 2011 (abstract 3528). glutamine. J Magn Reson 2011;212:304–310. 63. Cunningham CH, Chen AP, Lustig M, et al. Pulse sequence for 42. Harris T, Giraudeau P, Frydman L. Kinetics from indirectly dynamic volumetric imaging of hyperpolarized metabolic prod- detected hyperpolarized NMR spectroscopy by using spatially ucts. J Magn Reson 2008;193:139–146. selective coherence transfers. Chemistry 2010;17:697–703. 64. Lau AZ, Chen AP, Ghugre NR, et al. Rapid multislice imaging of 43. Golman K, in’t Zandt R, Thaning M. Real-time metabolic imag- hyperpolarized 13C pyruvate and bicarbonate in the heart. ing. Proc Natl Acad Sci U S A 2006;103:11270–11275. Magn Reson Med 2010;64:1323–1331. 44. Golman K, in’t Zandt R, Lerche M, Pehrson R, Ardenkjaer- 65. Yen YF, Le Roux P, Bok R, et al. Apparent T2 of 13C-labeled Larsen JH. Metabolic imaging by hyperpolarized 13C magnetic metabolites in vivo. In: Proc 16th Annual Meeting ISMRM, To- resonance imaging for in vivo tumor diagnosis. Cancer Res ronto; 2008 (abstract 1747). 2006;66:10855–10860. 66. Yen YF, Le Roux P, Mayer D, et al. T(2) relaxation times of (13)C 45. Chen AP, Albers MJ, Cunningham CH, et al. Hyperpolarized C- metabolites in a rat hepatocellular carcinoma model measured 13 spectroscopic imaging of the TRAMP mouse at 3T-initial ex- in vivo using (13)C-MRS of hyperpolarized [1-(13)C]pyruvate. perience. Magn Reson Med 2007;58:1099–1106. NMR Biomed 2010;23:414–423. 46. Cunningham CH, Chen AP, Albers MJ, et al. Double spin-echo 67. Yen YF, Josan S, Senadheera L, et al. Hyperpolarized 13C T2 sequence for rapid spectroscopic imaging of hyperpolarized 13C. Mapping. Experimental NMR Conference. Pacific Grove, CA; J Magn Reson 2007;187:357–362. 2011 (abstract 138). 47. Yen YF, Kohler SJ, Chen AP, et al. Imaging considerations for in 68. Leupold J, Mansson S, Petersson JS, Hennig J, Wieben O. Fast vivo 13C metabolic mapping using hyperpolarized 13C-pyru- multiecho balanced SSFP metabolite mapping of (1)H and vate. Magn Reson Med 2009;62:1–10. hyperpolarized (13)C compounds. MAGMA 2009;22:251–256. 48. Mayer D, Levin YS, Hurd RE, Glover GH, Spielman DM. Fast 69. Leupold J, Wieben O, Mansson S, et al. Fast chemical shift map- metabolic imaging of systems with sparse spectra: application ping with multiecho balanced SSFP. MAGMA 2006;19:267–273. for hyperpolarized 13C imaging. Magn Reson Med 2006;56: 70. Yen Y-F, Le Roux P, Bok R, et al. Signal enhancement in low- 932–937. dose hyperpolarized 13C imaging using multi-slice FSEPSI 49. Levin YS, Mayer D, Yen YF, Hurd RE, Spielman DM. Optimiza- sequence. In: Proc 16th Annual Meeting ISMRM, Toronto; 2008 tion of fast spiral chemical shift imaging using least squares (abstract 890). reconstruction: application for hyperpolarized (13)C metabolic 71. Perman WH, Bhattacharya P, Leupold J, et al. Fast volumetric imaging. Magn Reson Med 2007;58:245–252. spatial-spectral MR imaging of hyperpolarized 13C-labeled com- 50. Mayer D, Yen YF, Levin YS, et al. In vivo application of sub-sec- pounds using multiple echo 3D bSSFP. Magn Reson Imaging ond spiral chemical shift imaging (CSI) to hyperpolarized 2010;28:459–465. 13C metabolic imaging: comparison with phase-encoded CSI. 72. Zierhut ML, Yen YF, Chen AP, et al. Kinetic modeling of hyperpo- J Magn Reson 2010;204:340–345. larized 13C1-pyruvate metabolism in normal rats and TRAMP 51. Hurd RE, Yen YF, Tropp J, Pfefferbaum A, Spielman DM, Mayer mice. J Magn Reson 2010;202:85–92. D. Cerebral dynamics and metabolism of hyperpolarized 73. Romijn JA, Chinkes DL, Schwarz JM, Wolfe RR. Lactate-pyru- [1-(13)C]pyruvate using time-resolved MR spectroscopic imag- vate interconversion in blood: implications for in vivo tracer ing. J Cereb Blood Flow Metab 2010;30:1734–1741. studies. Am J Physiol 1994;266(3 Pt 1):E334–340. 52. Mayer D, Yen YF, Tropp J, Pfefferbaum A, Hurd RE, Spielman 74. Kettunen MI, Hu DE, Witney TH, et al. Magnetization transfer DM. Application of subsecond spiral chemical shift imaging to measurements of exchange between hyperpolarized [1–13C]py- real-time multislice metabolic imaging of the rat in vivo after ruvate and [1–13C]lactate in a murine lymphoma. Magn Reson injection of hyperpolarized 13C1-pyruvate. Magn Reson Med Med 2010;63:872–880. 2009;62:557–564. 75. Spielman DM, Mayer D, Yen YF, Tropp J, Hurd RE, Pfefferbaum 53. Lustig M, Donoho D, Pauly JM. Sparse MRI: the application of A. In vivo measurement of ethanol metabolism in the rat liver compressed sensing for rapid MR imaging. Magn Reson Med using magnetic resonance spectroscopy of hyperpolarized 2007;58:1182–1195. [1–13C]pyruvate. Magn Reson Med 2009;62:307–313. 54. Hu S, Lustig M, Chen AP, et al. Compressed sensing for resolu- 76. Moreno KX, Sabelhaus S, Merritt ME, Sherry D, Malloy CR. tion enhancement of hyperpolarized 13C flyback 3D-MRSI. Malate-aspartate shuttle reversal allows for lactate concentra- J Magn Reson 2008;192:258–264. tion increases upon rapid changes in 13C pyruvate 1328 Hurd et al.

concentration. In: Proc 18th Annual Meeting ISMRM, Stock- 93. Hurd RE, Yen YF, Mayer D, et al. Metabolic imaging in the anes- holm; 2010 (abstract 1329). thetized rat brain using hyperpolarized [1–13C] pyruvate and 77. Hu S, Chen AP, Zierhut ML, et al. In vivo carbon-13 dynamic [1–13C] ethyl pyruvate. Magn Reson Med 2010;63:1137–1143. MRS and MRSI of normal and fasted rat liver with hyperpolar- 94. Mayer D, Yen YF, Takahashi A, et al. Dynamic and high-resolu- ized 13C-pyruvate. Mol Imaging Biol 2009;11:399–407. tion metabolic imaging of hyperpolarized [1-(13) C]-pyruvate in 78. Darpolor MM, Yen YF, Chua MS, et al. In vivo MRSI of hyperpo- the rat brain using a high-performance gradient insert. Magn larized [1-(13) C]pyruvate metabolism in rat hepatocellular car- Reson Med 2011;65:1228–1233. cinoma. NMR Biomed 2011;24:506–513. 95. Marjanska M, Iltis I, Shestov AA, et al. In vivo 13C spectroscopy 79. Park I, Larson PE, Zierhut ML, et al. Hyperpolarized 13C mag- in the rat brain using hyperpolarized [1-(13)C]pyruvate and [2- netic resonance metabolic imaging: application to brain tumors. (13)C]pyruvate. J Magn Reson 2010;206:210–218. Neuro Oncol 2010;12:133–144. 96. Soon D, Tozer D, Altmann D, Tofts P. Quantification of subtle 80. Park J, Josan S, Jang D, et al. Metabolic kinetics of a glioma blood-brain barrier disruption in non-enhancing lesions in mul- model using hyperpolarized 13C magnetic resonance spectro- tiple sclerosis: a study of disease and lesion subtypes. Multiple scopic imaging. In: Proc 19th Annual Meeting ISMRM, Montreal; Sclerosis 2007;13:884–894. 2011 (abstract 3517). 97. Bohndiek SE, Kettunen MI, Hu DE, et al. Hyperpolarized [1- 81. Gillies RJ, Barry JA, Ross BD. In vitro and in vivo 13C and 31P (13)c]-ascorbic and dehydroascorbic Acid: vitamin C as a probe NMR analyses of phosphocholine metabolism in rat glioma cells. for imaging redox status in vivo. J Am Chem Soc 2011;133: Magn Reson Med 1994;32:310–318. 11795–11801. 82. Stubbs M, Bhujwalla Z, Tozer G, et al. An assessment of 31P MRS as a method of measuring pH in rat tumours. NMR Biomed 98. Keshari KR, Kurhanewicz J, Bok R, Larson PEZ, Vigneron DB, 1992;5:351–359. Wilson DM. Hyperpolarized 13C dehydroascorbate as an endog- 83. Karlsson M, Jensen PR, in ’t Zandt R, et al. Imaging of branched enous redox sensor for in vivo metabolic imaging. Proc Natl chain amino acid metabolism in tumors with hyperpolarized Acad Sci U S A 2011;29:1035–1040. 13C ketoisocaproate. Int J Cancer 2010;127:729–736. 99. Huang J, Agus DB, Winfree CJ, et al. Dehydroascorbic acid, a 84. Chen AP, Bok R, Zhang V, et al. Serial hyperpolarized 13C 3D- blood-brain barrier transportable form of vitamin C, mediates MRSI following therapy in a mouse model of prostate cancer. In: potent cerebroprotection in experimental stroke. Proc Natl Acad Proc 16th Annual Meeting ISMRM, Toronto; 2008 (abstract Sci U S A 2001;98:11720–11724. 888). 100. Chen A, Hurd R, Schroeder MA, et al. Simultaneous investiga- 85. Gallagher FA, Kettunen MI, Hu DE, et al. Production of hyper- tion of cardiac pyruvate dehydrogenase flux, Krebs cycle metab- polarized [1,4–13C2]malate from [1,4–13C2]fumarate is a olism and intracellular pH using hyperpolarized [1,2–13C2] marker of cell necrosis and treatment response in tumors. Proc pyruvate. NMR Biomed 2012;25:305–311. Natl Acad Sci U S A 2009;106:19801–19806. 101. Chen AP, Kurhanewicz J, Bok R, et al. Feasibility of using hyper- 86. Golman K, Petersson JS, Magnusson P, et al. Cardiac metabo- polarized [1–13C]lactate as a substrate for in vivo metabolic 13C lism measured noninvasively by hyperpolarized 13C MRI. Magn MRSI studies. Magn Reson Imaging 2008;26:721–726. Reson Med 2008;59:1005–1013. 102. Chen AP, Hurd RE, Gu YP, Wilson DM, Cunningham CH. (13)C 87. Merritt ME, Harrison C, Storey C, Jeffrey FM, Sherry AD, Malloy MR reporter probe system using dynamic nuclear polarization. CR. Hyperpolarized 13C allows a direct measure of flux through NMR Biomed 2011;24:514–520. a single enzyme-catalyzed step by NMR. Proc Natl Acad Sci U S 103. Keshari KR, Wilson DM, Chen AP, et al. Hyperpolarized [2–13C]- A 2007;104:19773–19777. fructose: a hemiketal DNP substrate for in vivo metabolic imag- 88. Schroeder MA, Cochlin LE, Heather LC, Clark K, Radda GK, ing. J Am Chem Soc 2009;131:17591–17596. Tyler DJ. In vivo assessment of pyruvate dehydrogenase flux in 104. Wilson DM, Hurd RE, Keshari K, et al. Generation of hyper- the heart using hyperpolarized carbon-13 magnetic resonance. polarized substrates by secondary labeling with [1,1–13C] Proc Natl Acad Sci U S A 2008;105:12051–12056. acetic anhydride. Proc Natl Acad Sci USA 2009;106: 89. Merritt ME, Harrison C, Storey C, Sherry AD, Malloy CR. Inhi- 5503–5507. bition of carbohydrate oxidation during the first minute of 105. Golman K, Ardenkjaer-Larsen JH, Petersson JS, Mansson S, reperfusion after brief ischemia: NMR detection of hyperpolar- Leunbach I. Molecular imaging with endogenous substances. ized 13CO2 and H13CO3. Magn Reson Med 2008;60: Proc Natl Acad Sci U S A 2003;100:10435–10439. 1029–1036. 90. Schroeder MA, Lau AZ, Chen AP, et al. Hyperpolarized 13C 106. Jensen PR, Karlsson M, Meier S, Duus JO, Lerche MH. Hyper- magnetic resonance imaging and spectroscopy uniquely reveal polarized amino acids for in vivo assays of transaminase activ- early and late onset metabolic changes in the failing heart. In: ity. Chemistry 2009;15:10010–10012. Proc 19th Annual Meeting ISMRM, Montreal; 2011 (abstract 107. Gallagher FA, Kettunen MI, Day SE, et al. Detection of tumor 21). glutamate metabolism in vivo using (13)C magnetic resonance 91. Schroeder MA, Swietach P, Atherton HJ, et al. Measuring intra- spectroscopy and hyperpolarized [1-(13)C]glutamate. Magn cellular pH in the heart using hyperpolarized carbon dioxide Reson Med 2011;66:18–23. and bicarbonate: a 13C and 31P magnetic resonance spectros- 108. Johansson E, Mansson S, Wirestam R, et al. Cerebral perfusion copy study. Cardiovasc Res 2010;86:82–91. assessment by bolus tracking using hyperpolarized 13C. Magn 92. Schroeder MA, Atherton HJ, Ball DR, et al. Real-time assess- Reson Med 2004;51:464–472. ment of Krebs cycle metabolism using hyperpolarized 109. Comment A, Uffmann K, Jannin A, et al. In vivo DNP-enhanced 13C magnetic resonance spectroscopy. FASEB J 2009;23: 13C labeled acetate brain studies in a 9.4T animal scanner. In: 2529–2538. Proc 15th Annual Meeting ISMRM, Berlin; 2007 (abstract 369). Special Issue Review Article

Received: 10 December 2010, Revised: 9 June 2011, Accepted: 14 June 2011, Published online in Wiley Online Library: 2011

(wileyonlinelibrary.com) DOI: 10.1002/nbm.1772 13C MRS studies of neuroenergetics and neurotransmitter cycling in humans Douglas L. Rothmana,b*, Henk M. De Feytera, Robin A. de Graafa,b, Graeme F . Masona,c and Kevin L. Beharc

In the last 25 years, 13 C MRS has been established as the only noninvasive method for the measurement of glutamate neurotransmission and cell-specific neuroenergetics. Although technically and experimentally challenging, 13 C MRS has already provided important new information on the relationship between neuroenergetics and neuronal function, the energy cost of brain function, the high neuronal activity in the resting brain state and how neuroenergetics and neurotransmitter cycling are altered in neurological and psychiatric disease. In this article, the current state of 13 C MRS as it is applied to the study of neuroenergetics and neurotransmitter cycling in humans is reviewed. The focus is predominantly on recent findings in humans regarding metabolic pathways, applications to clinical research and the technical status of the method. Results from in vivo 13 C MRS studies in animals are discussed from the standpoint of the validation of MRS measurements of neuroenergetics and neurotransmitter cycling, and where they have helped to identify key questions to address in human research. Controversies concerning the relationship between neuroenergetics and neurotransmitter cycling and factors having an impact on the accurate determination of fluxes through mathematical modeling are addressed. We further touch upon different 13C-labeled substrates used to study brain metabolism, before reviewing a number of human brain diseases investigated using 13 C MRS. Future technological developments are discussed that will help to overcome the limitations of 13 C MRS, with special attention given to recent developments in hyperpolarized 13 C MRS. Copyright © 2011 John Wiley & Sons, Ltd.

Keywords: 13C; MRS; brain; metabolism; neurotransmitter cycling; human; neuroenergetics

INTRODUCTION astrocytes to provide substrates for the synthesis of glutamate lost during neurotransmission (6). The neuron then converts glu- 13 tamine to glutamate via phosphate-activated glutaminase. The In this article, we review the current state of C MRS as it is used complete series of steps from neuronal glutamate release to to study neuroenergetics and neurotransmitter cycling in humans. the resynthesis of glutamate from glutamine is called the ’gluta- We focus primarily on the present status of the measurement mate/glutamine cycle’. (pathways and spatial resolution) and recent findings, leaving Figure 2 shows the spectra obtained at 4 T from human brain descriptions of the experimental protocols and methodology to localized to the midline occipital/parietal lobe during the infu- other articles in this issue. We finish by reviewing the results of 13 1 13 sion of three different isotopically labeled substrates: 99% the initial applications of C MRS and H– C MRS to the study of human disease and potential improvements in the sensitivity, cost and ease of performance of studies. * Correspondence to: D. L. Rothman, Magnetic Resonance Research Center, 13 C MRS is presently the only method that provides noninvasive Departments of Diagnostic Radiology and Biomedical Engineering, Yale measurements of neuroenergetics and neurotransmitter cycling in University School of Medicine, 300 Cedar Street, PO Box 208043, New Haven, the human brain. The ability to use 13C MRS to study cell-specific CT 06520–8043, USA. E-mail: [email protected] neuroenergetics and neurotransmitter cycling is a result of the localization of key enzymes and metabolite pools in neurons a D. L. Rothman, H. M. De Feyter, R. A. de Graaf, G. F. Mason and glia, and the specificity of 13 C-labeled precursors to specific Department of Diagnostic Radiology, Magnetic Resonance Research Center, cell types. Figure 1 shows a diagram of neuronal and astrocyte (a Yale University School of Medicine, New Haven, CT, USA type of glial cell) cell metabolism and the interplay of neuronal b D. L. Rothman, R. A. de Graaf and astrocyte metabolism via the glutamate/glutamine cycle. Department of Biomedical Engineering, Yale University, New Haven, CT, USA Both neurons and astrocytes can take up glucose and oxidize it in their mitochondria via the tricarboxylic acid (TCA) cycle. As dis- c G. F. Mason, K. L. Behar cussed below, neurons and astrocytes can oxidize lactate and b- Department of Psychiatry, Magnetic Resonance Research Center, Yale Univer- sity School of Medicine, New Haven, CT, USA hydroxybutyrate in addition to glucose, and astrocytes can also oxidize acetate and fatty acids. Excitatory glutamatergic neurons, Abbreviations used: AV, arteriovenous; CMRglc(ox), cerebral metabolic rate of which account for over 80% of the neurons and synapses in the glucose oxidation; GABA, g-aminobutyric acid; PC, pyruvate carboxylase; PET, cerebral cortex (1), release glutamate as a neurotransmitter, most positron emission tomography; RF, radiofrequency; TCA, tricarboxylic acid; of which is taken up by astrocytes (2,3) and converted to gluta- Vcyc, glutamate/glutamine cycle rate; VGln, glutamine synthesis rate; VPC, rate of anaplerosis through pyruvate carboxylase; VTCAa, astroglial TCA cycle rate;

mine or oxidized (4,5). Neurons lack the enzymes required for 943 VTCAn, neuronal TCA cycle rate; Vx, mitochondrial a-ketoglutarate to cytosolic the de novo synthesis of glutamate, and therefore depend on glutamate exchange rate.

enriched [1-13C]glucose, [3-13C]lactate and [2-13C]acetate. It Glutamate neurotransmitter cycling is the main pathway of shows labeling in the brain pools of aspartate, g-aminobutyric cerebral cortex glutamine synthesis acid (GABA), glutamine and glutamate. The brain pools of glutamate, GABA and glutamine have been shown to be localized Although the metabolic pathways of glial glutamate uptake and within glutamatergic neurons, GABAergic neurons and glia, re- the glutamate/glutamine cycle are well established from 14C ra- spectively (under nonpathological conditions). By following the diotracer and cellular studies, they were not considered to be rel- flow of 13 C label from glucose, acetate and other precursors into evant to neuroenergetics prior to in vivo studies using MRS these metabolites, MRS in combination with metabolic modeling (4,12,13). Because the neurotransmitter glutamate was shown allows the measurement of the TCA cycle rate in glutamatergic to be packaged in small vesicles, the predominant concept arose neurons, GABAergic neurons and glia, as well as glutamate and of a small, nonmetabolic ’transmitter’ pool which did not interact GABA neurotransmitter cycles between neurons and astrocytes with the large ’metabolic’ glutamate pool (14,15). This concept (7,8). When expressed as total substrate oxidation, the rates de- was brought into question by one of the first 13C MRS studies termined by MRS are found to be in excellent agreement with of human brain (16), which found a high rate of glutamine label- earlier methods, including positron emission tomography (PET) ing from [1-13C]glucose in the human occipital/parietal lobe. At (7,9–11). However, in vivo MRS is unique among other techniques the time of the study, it was unclear whether this high labeling in its measurement of cell type-specific energetics and neuronal/ was a result of the glutamate/glutamine cycle or of glutamine glial neurotransmitter cycles. synthesis to remove ammonia from the brain, the latter believed to be its major role (17). As pointed out by Sibson et al. (18) in 1997, these pathways could be distinguished because glutamine that leaves the brain must be replaced by anaplerosis, which occurs in glial cells. Furthermore, because of mass balance con- STUDIES IN ANIMAL AND CELL MODELS OF straints, the glutamine synthesized for this purpose must match fl THE GLUTAMATE/GLUTAMINE CYCLE AND the ef ux of glutamine and uptake of ammonia and CO2 by the NEURONAL AND GLIAL ENERGETICS brain, as measured by the arteriovenous (AV) difference (5,18). To distinguish these possibilities, glutamine synthesis was mea- Although this article primarily focuses on human studies, we sured in rat cortex during hyperammonemia. When blood am- briefly review some relevant investigations in animal and cell monia levels were extrapolated to a physiologically normal and models which have helped to validate 13 C MRS measurements low value, anaplerosis was found to account for approximately of neuroenergetics and neurotransmitter cycling, as well as to 20% of glutamine synthesis (18). Measurements of anaplerotic identify key questions to address in human research. glutamine synthesis using precursors that label this pathway

Figure 1. Diagram of the glutamate (Glu)/glutamine (Gln) cycle (Vcycle). The diagram shows the metabolic pathways within glutamatergic neurons and surrounding astroglial cells. Glucose and lactate enter both the glial (VTCAa) and neuronal (VTCAn) tricarboxylic acid (TCA) cycles via pyruvate dehydrogenase (Vpdh), b-hydroxybutyrate (b-HB) is directly incorporated into the neuronal and astroglial TCA cycles, and acetate is near-exclusively incorporated into the glial TCA cycle. Neuronal Glu that is released via neurotransmission is taken up by astroglial cells and converted by Gln synthetase to Gln at a rate proportional to the Glu/Gln cycle. The synthesis of Gln is believed to be exclusively within astroglia and other glial cells. In addition to neurotransmitter cycling, Gln may be synthesized de novo starting with the pyruvate carboxylase (PC) reaction (VPC). Gln synthesized via PC can replace neurotransmitter Glu oxidized in the astrocyte or elsewhere (and be recycled back to the neuron) or leave the brain (Vefﬂux) to remove ammonia and maintain the nitrogen balance (5,17,26). To measure the rates of these pathways, 13 C-labeled substrates are used and the ﬂow of 13 C isotope into Glu and Gln is measured using 13 C MRS. For detailed descriptions of how these pathways are tracked using 13 C MRS, and how isotopically labeled sub-

944 strates and rates are calculated by metabolic modeling, see refs. (24–26,44,45,47,143). AcCoA, acetyl-CoA; Asp, aspartate; Glut 1, glucose transporter 1; a-KG, a-ketoglutarate; Lac, lactate; MCT1, monocarboxylate transporter 1; OAA, oxaloacetate; Pyr, pyruvate.

Glu C4 Glu C3

Glu C2 Glu C4,3

Gln C2

Asp C2 NAA C3 Gln C4 Gln C3 Asp C3 Glu C3,4

GABA C2 [1-13C]-glucose

[3-13C]-lactate

[2-13C]-acetate

60 55 50 45 40 35 30 25 20 ppm Figure 2. Localized 13 C MR spectra acquired at 4 T from the midline occipital/parietal lobe of a volunteer infused with 13C-labeled glucose, lactate or acetate. Top spectrum: acquired during the last 18 min of a 2-h [1-13 C]glucose infusion. Middle spectrum: acquired during the last 18 min of a 2-h [3-13 C] 13 13 lactate infusion ([Lac]Plasma ~ 1.5 mmol/L; C fractional enrichment ~ 30%). Bottom spectrum: acquired during the last 18 min of a 2-h [2- C]acetate infusion. Spectra are scaled to N-acetylaspartate (NAA) C3 to exhibit the differences in 13 C fractional enrichment reached for glutamate (Glu), glutamine (Gln) and aspartate (Asp). The highest fractional enrichment is attained with glucose as label precursor. For glucose or lactate as precursor, the majority of labeling appears in glutamate C4, consistent with the majority of brain metabolism of these substrates occurring in the neurons which contain the majority of glutamate under normal conditions (25). In contrast, label from [2-13 C]acetate is highly enriched in glutamine C4, consistent with the localization of acetate metabolism in the astrocyte tricarboxylic acid (TCA) cycle, as shown in Fig. 1, resulting in preferential labeling of glutamine C4. GABA, g-aminobutyric acid.

13 15 14 directly ([2- C]glucose, N-ammonia and CO2) have found with a slope of 0.89 Æ 0.06. [We note that, in the article by Sibson that approximately 80% of glutamine synthesis is devoted to et al. (29), 0.5VTCAn was taken to be equivalent to neuronal glu- neurotransmitter cycling (18–22). The analysis of 13 C-labeled ex- cose oxidation, which would be the case under the hyperglycemic tracellular glutamate measured by microdialysis and mass spec- conditions of the study when glucose is the only net fuel for neuro- trometry led to a similar conclusion that neurotransmitter cycling nal oxidation (28).] Furthermore, comparison of the intercept at is the major source of glutamine (23). Similar conclusions have isoelectricity versus the values of VTCAn and Vcyc in the awake state been obtained from studies performed in human brain using confirms that over 80% of neuronal oxidative ATP production is 13 13 13 [1- C]glucose, [2- C]acetate and [2- C]glucose as labeled sub- coupled to neuronal signaling (as measured by Vcyc), even in the strates (24–27). absence of stimulation. Thus, Vcyc in the awake state is close to the rate of neuronal glucose oxidation and constitutes a major metabolic flux. As described below, recent 13 C MRS results from The glutamate/glutamine cycle is a major metabolic pathway human cerebral cortex are consistent with this finding. and is directly coupled to neuroenergetics There are several molecular models that have been proposed To determine the relationship between the glutamate/glutamine to explain the near-stoichiometric relationship between changes cycle and cerebral cortex neuroenergetics, 13 C MRS was used to in the flux of neuronal glucose oxidation and the glutamate/ measure the relationship between neuronal glucose oxidation glutamine cycle. The relationship has been shown to be consis- and the glutamate/glutamine cycle rate (Vcyc) in rat cerebral cor- tent with a model in which the energy for taking up the neuro- tex (28). Cortical activity was modulated through anesthesia transmitter glutamate into the astrocyte is provided using glyco- ranging from an isoelectric electroencephalograph to higher lytic ATP production from glucose or glycogen (4,30,31). electroencephalograph activity at two lower doses of anesthesia. Alternatively, it has been shown that the observed slope can With increasing electrical activity, Vcyc and the rate of neuronal be explained by the redox shuttling requirements between the glucose oxidation via the TCA cycle [expressed as CMRglc(ox) in neuronal cytosol and mitochondria in order to oxidize glutamine Sibson et al. (29)] increased linearly with a slope of 1.0 Æ 0.1. Sub- taken back up from glial cells (4). The determination of the sequent studies have confirmed this relationship [see Hyder et al. actual molecular mechanism may have important implications (11) for a review]. Figure 3 shows a plot of measurements in the for the understanding of brain disease as any dysfunction in this rat cerebral cortex of the VTCAn plotted versus Vcyc taken from 11 mechanism would have a severe impact on the ability to sustain published research articles at different levels of brain electrical neurotransmission. In addition, the measured coupling between activity ranging from isoelectricity to awake. As reported origi- neurotransmission and neuroenergetics has provided key data

nally by Sibson et al. (29), the relationship between Vcyc and for detailed models of the energy budget of the brain for the 945 0.5VTCAn (where VTCAn is the neuronal TCA cycle rate) is linear support of signaling and information transfer (32–35).

under normal conditions in rodents and humans in vivo (5,20,21,25,41).

Controversies regarding the relationship between neuronal energetics and the glutamate/glutamine cycle Despite the good agreement between ﬁndings by different investigators (see Fig. 3 and Table 1), there is considerable con- troversy regarding the relationship between the neuronal TCA cycle and Vcyc. The two major areas of debate are the proposed molecular mechanism and the slope of the relationship as a result of questions regarding the accuracy of measurement of VTCAn and Vcyc (4,30,32,43). In the sections below, the controversies over the accuracy of determination of VTCAn and Vcyc are discussed.

Figure 3. Approximately 1 : 1 relationship between the neuronal tricar- The fraction of glutamine synthesis as a result of anaplerosis via boxylic acid (TCA) cycle (0.5VTCAn) and the glutamate/glutamine cycle pyruvate carboxylase (PC) as opposed to neurotransmitter cycling (Vcyc) with increasing electrical activity in the rat cerebral cortex. The plot shows the mean values of 0.5VTCAn [equivalent to CMRglc(ox)N in Sibson As shown in Fig. 1, in addition to neurotransmitter cycling, gluta- et al. (29)] plotted versus Vcyc reported from 11 published studies at activ- mine synthesis is used to replace glutamine that leaves the brain ity levels ranging from awake to isoelectricity (20,22,28,36,39,59,127, for ammonia detoxification and for the maintenance of nitrogen 2 150–153). Regression analysis yields a slope of 0.89 (R = 0.92) and an in- balance across the blood–brain barrier (5,17). Glutamine synthe- tercept of 0.5VTCAn of 0.09 at isoelectricity (Vcyc ~ 0), values similar to sis used to maintain nitrogen balance derives from anaplerosis those found in the original 1998 study by Sibson et al. (29). In the case via PC which, under normal conditions, appears to exclusively of ref. (39), for both the anesthetized and awake states, values of V TCAn take place in astrocytes (and glial cells in general). The contro- were calculated from the time constants reported for glutamate turnover during a glucose infusion. The ratio of glutamate to glutamine steady- versy regarding anaplerosis has largely been settled as a result state fractional enrichment during [2-13 C]acetate infusion was used to of a convergence of results in animal models (5,11,19,20,23) fi ’ 13 calculate Vcyc using the equation described in Lebon et al. (25). and similar ndings in the human brain (24,26) (see In vivo C MRS studies of human brain’ section below). There is now general consensus that, in rat cerebral cortex, a minimum of 80% of glutamine synthesis supports neurotransmitter cycling under Validation of measurements of neurotransmitter cycling and normal conditions and between 70 and 90% in humans. Of the anaplerosis using alternative substrates and labeling remaining 20% of glutamine synthesized by anaplerosis, the strategies majority is probably also used to replete neurotransmitter glutamate that is oxidized in glial cells (see below) as opposed to am- A strategy that has been employed to validate glutamate/gluta- monia detoxification (5,26). mine cycle measurement is the use of alternative 13C- and 15N- labeled substrates that are incorporated specifically in astrocytes. The effect of a slow rate of mitochondrial glutamate exchange (V ) The subsequent flow of label from astrocyte glutamine (or other x, on the calculation of V potential trafficking substrates) into neuronal glutamate then TCAn provides an independent measure of Vcyc from studies using Another factor that may influence the measured relationship is [1-13C]glucose, which labels the neuronal glutamate pool to a the rate of exchange of label from mitochondrial a-ketoglutarate greater degree than the glial pool. Animal studies employing this to cytosolic glutamate, often termed ’Vx’ in the literature. If this strategy have used as label sources 15N-ammonia (27), [2-13C] exchange is not much faster than the TCA cycle rate, it may 14 13 glucose (5), CO2 (19,20) and [2- C]acetate (36–40), and have cause an underestimation of VTCAn, unless it is incorporated in found results consistent with a high rate of glutamate neurometabolic modeling through measurement of the C2 and/or C3 transmitter trafficking. Similar results have been reported in positions of glutamate (24,44). Interestingly, it has been noted studies of human subjects using [2-13C]acetate (25,41) and that, independent of whether multiple glutamate positions are [2-13C]glucose as tracers (26). In addition, the analysis of the measured, very little difference is found in the calculated values 15 fl N labeling of glutamine obtained from extracellular uid also of VTCAn or Vcyc in studies of human brain [see Hyder et al. (11)]. supported the majority of glutamine being derived from gluta- On the basis of numerical simulations, it has been shown that fl 13 mate neurotransmitter cycling (23). In principle, extracellular uid the accuracy of measurement of Vx in the brain using C MRS measurement is a more direct assessment than whole-tissue MRS is low (44,45), which most probably explains the extended con- measurement of the pools relevant to neurotransmission. troversy on this issue. Recently, Yang et al. (46), using an elegant In addition to the glutamate/glutamine cycle, there are several saturation transfer method, have shown that Vx is, at minimum, other potential pathways of neurotransmitter repletion by astro- several times greater than VTCAn in the brain, explaining why it cytes that use TCA cycle intermediates to return carbon to the does not have an impact on measurements of VTCAn. neurons (25,42). However, the agreement found between the In addition to the consistency of results between different stud- rates determined for the glutamate/glutamine cycle measured ies and laboratories, VTCAn values measured in rat and human by [1-13C]glucose and the rates measured using glial-specific cerebral cortex are consistent with results previously published 14

946 substrates (which take into account all cycling pathways) sug- using C-2-deoxyglucose autoradiography, AV difference and gests that the glutamate/glutamine cycle is the major pathway PET (7,11). Recently, two studies have compared directly PET

Table 1. Experimental mean and standard deviation (SD) of Vcyc, VTCAn, VTCAa, VPC, Vcyc/VGln and Vcyc/0.5VTCAn in the resting awake human midline occipital/parietal lobe from 13 C and 1H–13C MRS studies. As in the rat brain, the majority of human cortical glutamine synthesis is a result of the glutamate/glutamine cycle as shown by the average value of Vcyc/VGln of 0.83. Similarly, the ratio of Vcyc/0.5VTCAn of 0.75 is consistent with the coupling between Vcyc and VTCAn measured in the rat cerebral cortex, and indicates that the majority of neuronal energy production in the resting awake human brain is likely to be devoted to the support of neuronal activity. Based on the relative values of VTCAa, VTCAn and VPC, approximately 20% of brain ATP production occurs in astrocytes

Reference VTCAtotal VTCAn VTCAa Vcyc VPC Vcyc/0.5VTCAn Vcyc/VGln (45)b 0.73 (58) 0.74 0.32 0.08 0.86 0.80 (27) 0.77 0.71 0.06 0.32 0.04 0.90 0.95 (57)c 0.66 (84)b 0.83 (75) 0.70 0.13 (25) 0.12a 0.28a 0.78 (148) 0.75 0.29 (24) 0.63 0.57 0.06 0.17 0.09 0.60 0.65 (26) 0.72 0.02 0.73 0.93 (41) 0.65 0.53 0.13 0.16 0.64 (140) 0.09 (149)b 0.79 Mean Æ SD 0.69 Æ 0.06 0.70 Æ 0.09 0.09 Æ 0.04 0.26 Æ 0.08 0.06 Æ 0.03 0.75 Æ 0.12 0.83 Æ 0.14

TCA, tricarboxylic acid; Vcyc, glutamate/glutamine cycle rate; VGln, glutamine synthesis rate; VPC, rate of anaplerosis through pyruvate carboxylase; VTCAa, astroglial TCA cycle rate; VTCAn, neuronal TCA cycle rate; VTCAtotal, sum of neuronal and astroglial TCA cycle rate. a Measured steady state ratios converted to rates using the value of VTCAn from Shen et al. (27). b One-compartment model for the neuron used. We assume that the derived TCA rate most closely reﬂects VTCAn. cAverage of white and gray matter rates which were measured separately.

measurements in nonhuman primates with 13 C MRS and found replacement by anaplerosis must occur in order to maintain con- excellent agreement between the total rate of glucose metabo- stant levels of TCA cycle intermediates and glutamate-derived lism measured (9,10). neurotransmitters, as the transport of the necessary five-carbon precursors from blood is comparatively minimal. This possibility is seen by several researchers as being in contradiction with Effects of isotopic dilution of glutamine on the calculated rate of the measured relationship between cycling and energetics (48). neurotransmitter cycling (Vcyc) However, as described previously, the presence of glutamate ox- Shestov et al. (44) published the results of simulation studies in idation has no impact on the MRS measurement of the gluta- fi which they reported much poorer precision when measuring Vcyc mate/glutamine cycle using a glial-speci c labeled precursor, using a [1-13C]glucose precursor than reported in experimental such as [2-13C]acetate, because the oxidized glutamate is papers. Shen et al. (47) were able to explain this discrepancy by replaced by de novo glutamine synthesis, which is cycled back showing that, when isotopic dilution of glutamine was taken into to the neuron (5,25,26). When [1-13C]glucose is used as the pre- account, as had been performed in most previous experimental cursor, replacement of oxidized glutamate will be included in studies, but not in the simulations by Shestov et al. (44), the the- the anaplerotic contribution to glutamine synthesis, unless it is oretical precision of the Vcyc measurement was similar to that distinguished from glutamine synthesis related to ammonia reported experimentally. A similar conclusion regarding the im- removal (detoxification) on the basis of other measurements. portance of including glutamine dilution in metabolic modeling was arrived at by Oz et al. (20) when comparing rates calculated 13 14 13 using [1- C]glucose and C-CO2 as precursors. We note that, IN VIVO C MRS STUDIES OF HUMAN BRAIN when astrocyte-specific labels, such as [2-13C]acetate, are used 13 13 1 13 to determine Vcyc, alone or in combination with [1- C]glucose, In the sections below, we review the results from C and H– C considerably higher precision for the measurement of Vcyc is MRS studies of human brain, focusing primarily on studies that obtained (38,41). report metabolic rates or labeling. Following initial studies of the animal brain in the 1980s (49–51), the availability of high- field human MR systems and improvements in B shimming Oxidation of the neurotransmitter glutamate (52) led to the first 13 C and 1H–13 C MRS studies of humans Glutamate released by neurons and taken up by astroglia can (16,53–55). Although the use of 13 C MRS in humans has been rel- undergo oxidation as well as conversion to glutamine. The oxi- atively limited (see ’Future prospects for 13 C MRS studies in 13

dized glutamate is then replaced by anaplerosis and cycled back humans’ section below), C MRS has already made major contri- 947 to the neuron as glutamine (4). This scenario of oxidation and butions to the understanding of human brain energetics and

neurotransmitter cycling and how alterations in these pathways synthesis occurs via the glutamate/glutamine cycle versus ana- may contribute to a range of human diseases. plerosis via PC, several studies in the human brain have measured this rate. As shown in Table 1, on the basis of these 13 measurements, the fraction of glutamine synthesis via the gluta- C MRS measurements of the rate of the neuronal (VTCAn) mate/glutamine cycle (V /V , where V is the glutamine syn- and astrocyte (VTCAa) TCA cycle and Vcyc in human brain cyc Gln Gln thesis rate) has been reported to range between 0.65 and 0.93 in Table 1 shows a compilation of studies from different groups human cerebral cortex, with a mean value of 0.83 Æ 0.14 measuring VTCAn, VTCAa, VPC (rate of anaplerosis through pyruvate (24,26,27). A complication when using the [1-13 C]glucose precur- carboxylase) and Vcyc in the human midline occipital lobe. As can sor to measure this ratio is that label enters the inner positions of be seen, there is very good agreement in the rates derived by glutamate via both pyruvate dehydrogenase and PC. To elimi- different studies, with most of the difference explainable by nate the complications arising from [1-13 C]glucose as precursor, different volume fractions of white matter [which has an approx- Mason et al. (26) measured 13 C incorporation into glutamate and imately three to four times lower rate of VTCAn than gray matter glutamine in the midline occipital/parietal lobe of human volun- (56,57)]. As discussed above, VTCAn values are similar whether teers from infused [2-13 C] glucose, which labels the C2 and C3 modeling uses the C4, C3 and C2 positions of glutamate and glu- positions of glutamine and glutamate primarily via PC. Labeling tamine or just the C4 positions. For the measurement of VTCAa, in glutamate C4 was used to assess the rate of pyruvate recycling there is good agreement between three independent labeling strat- 13 13 13 coupled to glutamate oxidation. Metabolic modeling of the la- egies using [1- C]glucose, [2- C]acetate and C-bicarbonate beling data indicated that the PC flux (V ) ranges from 6 to fl PC as precursors. In the human studies, VTCAn largely re ects gluta- 10% of the rate of glutamine synthesis (0.02–0.03 mmol/g/min). matergic neurons, as it is derived from the fitting of the isotopic Comparison of the measurements of VPC in humans to date with turnover of the large glutamate pool. Although the rate of the AV difference measurements of human brain glutamine efflux GABAergic neuron TCA cycle has not yet been determined, label- fl 13 suggests that the majority of the PC ux is used to replace gluta- ing of GABA during the infusion of [1- C]glucose has been mate lost by oxidation in the glia and possibly elsewhere, and reported at 4 T (24,58). Results in animal models suggest that therefore can be considered to support neurotransmitter cycling on the order of 10% of the energy consumption of the cerebral (26). cortex may be caused by GABAergic neurons (22,59).

Relationship between neuronal energetics and the glutamate/ Studies of substrate oxidation and transport glutamine cycle in human brain and implications for resting Because of its ability to distinguish a substrate from its metabolic brain functional activity products, 13 C MRS can be used to independently assess both the 13 parameters of transport and rates of metabolism. Furthermore, it C MRS studies by several groups have found a ratio of VTCAn to can be uniquely used to determine cell type-specific metabolism. Vcyc that is highly consistent with the predictions based on stud- Glucose has long been known to be the primary fuel for brain ies performed in the rat. Table 1 shows the ratio of Vcyc/0.5VTCAn derived in human studies. The values range from 0.6 to 0.9 with metabolism (28). However, the brain can also consume alterna- an average of 0.75 Æ 0.12. This average is similar to the value of tive substrates, including acetate, b-hydroxybutyrate and lactate. fi b-Hydroxybutyrate is a particularly important substrate during Vcyc/0.5VTCAn predicted from animal studies using the best t of development and under conditions of fasting, where AV differ- the relationship between Vcyc and VTCAn in Fig. 2. The variation in the ratio measured in humans is largely a result of variation ence methods have revealed its capacity to supply up to 60% of the fuel oxidized by the brain (62). Similar high rates of utiliza- in the measurement of VPC. However, on the basis of AV mea- tion of lactate have been reported under conditions of elevated surements in humans, the majority of VPC represents replacement of oxidized glutamate, and therefore reflects glutamine plasma lactate, such as during exercise (63,64). Although the synthesis that is cycled back to the neuron (26). Overall, these total usage and oxidation of these substrates have been determined for humans by AV difference methods and, to some ex- studies show that functional neuronal activity is extremely high 13 in the awake resting human brain and, if there is a similar rela- tent, by PET, the C MRS studies reviewed below have fi fi tionship as in the rat, accounts for the majority of neuronal glu- provided the rst information on the cell type speci city of sub- cose consumption. Furthermore, the energy devoted to neuronal strate usage, as well as revealing new insights into the blood– activity in the resting state is much higher than the changes in brain transport of these substrates. activity that occur during standard activation paradigms, such fi as cognitive challenges and visual stimulation (60). These nd- Glucose ings have contributed to the recent surge in interest in the study 13 of resting brain activity by functional MRI and other methods, As a result of the commercial availability of [1- C]glucose and the high rate of brain glucose metabolism, the majority of meta- and form part of the basis of several theories of resting brain 13 function (60,61). bolic C MRS studies have focused on the measurement of neuronal and glial glucose metabolism [see ’13C MRS measurements of the rate of the neuronal (V ) and astrocyte (V ) TCA cycle Measurements of the rate of glutamine synthesis via PC in TCAn TCAa and V in human brain’ section above]. 13C MRS has been used human cerebral cortex cyc to measure glucose transport parameters (apparent Michaelis– As a result of the astrocyte localization of PC, anaplerosis occurs Menten constant of transport and maximal unidirectional trans- only in the glia and is used both to synthesize glutamine that port rate) in the human midline occipital/parietal lobe (54). Sub- leaves the brain to maintain nitrogen balance and to replace sequent studies of brain glucose transport in humans have used 1

948 released neurotransmitter glutamate oxidized in the astrocytes. H MRS for higher sensitivity. These studies have provided sup- In order to address the question of what fraction of glutamine port for the blood–brain barrier being the rate-limiting step for

wileyonlinelibrary.com/journal/nbm Copyright © 2011 John Wiley & Sons, Ltd. NMR Biomed. 2011; 24: 943–957 13C MRS STUDIES OF HUMAN NEUROENERGETICS AND NEUROTRANSMITTER CYCLING human brain glucose transport (65), which is best described by a brain oxidative metabolism in nonfasted volunteers, consistent reversible, Michaelis–Menten transport model [as opposed to with earlier reports using AV difference methods (28). the nonreversible model used previously to interpret most PET glucose studies (66,67)] (68,69). In this model, glucose transport Lactate is described by transporters with reversible Michaelis–Menten kinetics across the membranes of the capillary endothelial cells Studies over the last decade have provided evidence that lactate that make up the blood–brain barrier (66,67). On the basis of in may be an important metabolic fuel for the brain, including pro- vivo kinetic and isolated transporter studies, the glial and neuro- posals that an astrocyte-to-neuron lactate shuttle may exist in nal glucose transporters have a relatively much higher activity order to provide neurons with fuel during periods of enhanced and can, to a first order, be neglected in the kinetic modeling. activity (30,32). In addition, AV difference and PET studies of These results have been used to develop and test kinetic models humans after blood lactate elevation by exercise have reported of glucose transport and metabolism (68), and may have potential that elevated lactate can provide a significant fraction of oxida- value for the assessment of whether impaired substrate delivery tive fuel to the brain (63,64). To determine the conditions under may have an impact on brain function in disease (70). which plasma lactate might contribute as a significant fuel for human brain energy metabolism, Boumezbeur et al. (78) infused [3-13 C]lactate and measured, by 13C MRS, the entry and utiliza- Acetate tion of lactate in the midline occipital/parietal cortex of healthy Studies in animal models and neural cell culture have found that human volunteers. During the 2-h infusion study, 13C incorpora- acetate is almost exclusively transported into and metabolized tion in the amino acid pools of glutamate and glutamine was by the astroglia (37,71–74). Lebon et al. (25) studied healthy hu- measured every 5 min. With a plasma concentration of lactate man subjects infused intravenously with [2-13 C]acetate whilst in the 0.8–2.8 mmol/L range, the brain tissue lactate concentra- monitoring 13C labeling in the midline occipital/parietal lobe tion was assessed, as was the fractional contribution of lactate with 13C MRS. The concentration of brain acetate was approxi- to brain energy metabolism. By fitting the measured relationship mately 10-fold lower than the plasma concentration, indicating between the unidirectional lactate influx and the plasma and that acetate transport was primarily unidirectional. Analysis of brain lactate concentrations, the lactate transport constants the steady-state 13C labeling pattern of glutamine and gluta- were calculated using a model in which the rate-limiting step mate, as shown in Fig. 2, as well as the kinetics of glutamate was assumed to be, on the basis of previous work, lactate trans- and glutamine labeling was consistent with acetate metabolism port at the blood–brain barrier. The transporters at the blood– localized to glial cells (25). Furthermore, the steady-state labeling brain barrier were modeled as reversible Michaelis–Menten patterns were in agreement with findings from [1-13 C]glucose of transporters, similar to that performed with glucose transport. a high rate of glutamate/glutamine cycling (25,41). Similar con- The results showed that, in the physiological range of plasma lac- clusions were obtained by Blüml et al. (75) using [1-13C]acetate tate concentration, the unidirectional rate of transport and con- as a precursor. Although normal plasma levels of acetate in centration of brain lactate increased linearly with plasma humans are relatively low, the levels can be elevated to approx- concentration. The maximum potential contribution of plasma imately 1 mM or more by alcohol consumption, becoming a lactate to brain metabolism was 10% for a basal plasma lactate major source of oxidative energy for the astrocyte (25,41,75,76). concentration of approximately 1.0 mM, and possibly as much as 60% at supraphysiological plasma lactate concentrations when the transporters are saturated (assuming that lactate oxi- -Hydroxybutyrate b dation is limited only by transport). Based on the similarity of b-Hydroxybutyrate is a substrate critical for brain function during the steady-state pattern of 13C labeling, as shown in Fig. 2, it fasting. b-Hydroxybutyrate enters the brain via facilitated diffu- was concluded that the relative consumption of plasma lactate sion using a monocarboxylate carrier at the blood–brain barrier between neurons and astrocytes is similar to that of glucose (70). Although brain b-hydroxybutyrate consumption had been (78). The calculation of the lactate metabolic capacity is in good studied extensively using AV difference methods, it was not agreement with recent AV difference studies using isotopically known whether b-hydroxybutyrate was preferentially consumed labeled lactate as a tracer, further confirming the potential impor- in neurons or astrocytes, or whether the blood–brain barrier was tance of plasma lactate as a substrate for brain metabolism (63). limiting for its metabolism (28). To answer these questions, localized 13C MRS measurements were performed during an infusion Use of 1H–13C MRS to study the energetics of brain tissue of [2,4-13C]b-hydroxybutyrate in healthy subjects, and the entry types (white/gray matter) and sensory stimulation and metabolism of this compound were measured in the medial occipital/parietal cortex (77). During the 2-h infusion study, 13C As described in the ’Future prospects for 13C MRS studies in label was detected in the b-hydroxybutyrate resonance positions humans’ section below, the sensitivity and spatial resolution lim- at a time resolution of 5 min and in the amino acid pools of glu- itations of 13C MRS can be partially overcome by using the more tamate, glutamine and aspartate. The pattern of 13C labeling at sensitive 1H nucleus to measure the 13C enrichment of bound the steady-state period (60–120 min) was very different from carbon atoms. These inverse MRS methods take advantage of that resulting from infusions of 13 C-acetate, but was broadly sim- the J coupling between 1H and 13C nuclei (51,79). By combining ilar to that of [1-13 C]glucose, indicating a predominant neuronal 1H–13C MRS with spectroscopic imaging and gray/white matter consumption of this substrate. The cortical b-hydroxybutyrate segmentation, Pan et al. (57) reported VTCAn in gray and white concentration (0.18 mM) was much lower than in plasma matter, finding an approximately four times higher rate of me- (2 mM), indicating that transport across the blood–brain barrier tabolism in gray matter. The higher resolution of 1H–13C MRS

limits brain b-hydroxybutyrate metabolism. The consumption has been used to address the question of whether glycolytic or 949 of b-hydroxybutyrate accounted for approximately 6% of total oxidative ATP production provides most of the incremental

energy for brain function during activation (11,80–83). Using this lactate in tumors. Recently, 13C MRS has been used to measure technique, Chen et al. (84) reported a 25% increase in VTCAn dur- lactate turnover in a human brain tumor (106). ing visual stimulation. This finding is consistent with recent cali- brated functional MRI measurements of the cerebral metabolic Application of 13C MRS to study hepatic encephalopathy and rate of oxygen consumption (85–87), and indicates that ATP pro- genetic diseases of ammonia metabolism duced oxidatively is the major source of energy for the incremental neuronal activity during sensory stimulation. Studies by Ross and coworkers (23,99) and other laboratories (107,108) have established 1H MRS measurements of glutamine and glutamate as one of the best ways to assess the severity of 13 Application of C MRS to study neuroenergetics and hepatic encephalopathy, a disease of the brain caused by neurotransmission in human brain disease chronic exposure to elevated ammonia in the blood as a result of liver failure. Animal studies using conventional methods, as The important role of neuroenergetics and the glutamate/GABA 13 15 neurotransmitter cycles in the pathogenesis of brain disease is well as C and N MRS, showed that ammonia led to increased being increasingly recognized. Changes in the concentrations anaplerosis and glutamine synthesis in astrocytes, as well as to of GABA, glutamate and glutamine have been measured by 1H disruption of the glutamate/glutamine cycle (5,17,27,109–112). MRS in a wide range of neurological and psychiatric diseases, To test whether similar metabolic alterations were present in – humans, Blüml et al. (113) studied patients with diagnosed he- including aging (88), depression and related disorders (89 94), 13 epilepsy (95–97), genetic disorders of metabolism (98), hepatic patic encephalopathy during the infusion of [1- C]glucose at encephalopathy (99) and neurodegenerative disorders (100). 1.5 T. The studies found disrupted neuroenergetics with increas- However, concentration measurements, although informative, ing disease severity, including evidence of impairment of the fi fl glutamate/glutamine cycle. More recently, Gropman et al. (114) are not speci c for the alterations in metabolic uxes or cellular 13 neurochemical distributions that may have led to these changes. have used C MRS to demonstrate abnormalities in glutamate fi 13C MRS allows the measurement of metabolic rates in humans, metabolism in patients with ornithine transcarbamylase de ciency. and may potentially be of great value in studying the pathogen- 13 esis and treatment of brain disease. However, its application to Application of C MRS to study Alzheimer’s disease and healthy disease has been limited by several factors, including the techni- aging 13 cal complexity of conducting C MRS experiments, which is ex- Mitochondrial dysfunction has been implicated in the loss of acerbated by the lack of study support with most clinical brain function in neurodegenerative disease and normal aging scanners. Furthermore, there is the need for a support team to (115). Studies using PET have found decreased rates of brain perform the infusion and analysis of substrates labeled with oxygen consumption and glucose consumption in Alzheimer’s the 13C isotope. Moreover, further problems include the rela- 13 disease and in healthy aging (116,117). In a pioneering study tively low sensitivity and volume resolution of C MRS com- ’ 1 on Alzheimer s disease, Lin et al. (118) infused two patients with pared with H MRS (and especially MRI), the availability and [1-13C]glucose and found a reduction in 13C labeling of C4 gluta- cost of 13C-labeled substrates, radiofrequency (RF) heating 13 mate, consistent with an impairment in the TCA cycle. Recently, concerns with the need to decouple the J interaction of C reso- Boumezbeur et al. (41), using combined infusions of [1-13C]glu- nances with bound protons, and the need to perform sophisti- cose and [2-13C]acetate with 13C MRS, compared a healthy group cated kinetic analysis to extract metabolic rate information of elderly subjects with young adult controls. The elderly sub- from the MRS data [see Ross et al. (101) for an informative review jects showed decreased VTCAn and Vcyc, together with increased on clinical 13C MRS studies in humans and the obstacles in VTCAa, changes which were independent of the relatively small performing them]. In the ’Future prospects for 13C MRS studies age-dependent loss of brain tissue volume. The decrease in VTCAn in humans’ section below, we speculate on how some of these correlated highly with decreases in N-acetylaspartate and gluta- barriers may be overcome through a range of developments, in- mate concentrations (Fig. 4), indicating that the reduced meta- cluding hyperpolarized 13C. Despite these challenges, the initial 13 bolic rates were associated with cellular level changes as opposed applications of C MRS in studies of human brain disease and to differences in sensory input. These findings are consistent with dysfunction have been highly informative, and are reviewed the theory that mitochondria lose oxidative capacity with advancing fl brie y below. age, leading to a loss of brain function. Overall, the ability to study aging and associated dementias with 13C MRS provides a unique Application of 13C MRS to study stroke and tumors opportunity to study the role of mitochondria in the pathogene- 13 sis process and how this process can be slowed or ceased The initial application of C MRS to clinical brain disease was to through treatment. assess the source of chronic lactate elevation in stroke. In this study, 1H MRS was used to measure lactate 13C fractional enrich- Application of 13C MRS to study the complications of type 1 ment in an infarct during the infusion of [1-13 C]glucose. 1H MRS diabetes was used because of its several-fold higher sensitivity for the measurement of 13C enrichment than direct 13C MRS (102). Lac- A major complication in insulin therapy for type 1 diabetes is tate was found to rapidly incorporate label from glucose, indicat- hypoglycemia, the frequency of which is worsened by the phe- ing active metabolic production long after the initial infarct nomenon of hypoglycemia unawareness (119,120). Among the event (103). Subsequent 1H MRS and histological assessment in- theories to explain this phenomenon is that repeated hypoglyce- dicated that most of the lactate elevation in chronic stroke was mic episodes in patients with type 1 diabetes may lead to meta- probably a result of macrophage metabolic activity and infiltra- bolic adaptations that allow improved function during periods of 13

950 tion (104). A similar strategy was used in animal models of brain mild hypoglycemia. Using [2- C]acetate as a tracer, Mason et al. cancer by Terpstra et al. (105) to assess the metabolic source of (76) tested the hypothesis that patients with type 1 diabetes

hypoglycemia leads to elevated glycogen synthesis on restora- tion of normal glucose levels. Future studies may be able to determine whether these alternative fuel sources can account fully for the cortical component of hypoglycemia unawareness.

Application of 13C MRS to study epilepsy Epilepsy has been studied extensively by 1H MRS and there is considerable evidence from PET, 1H MRS and 31P MRS of hypo- metabolism in brain regions affected by epilepsy that may be secondary to a failure in neuroenergetics (122). Furthermore, in medial temporal lobe epilepsy, chronically elevated extracellular glutamate has been found by microdialysis in epileptogenic scle- Figure 4.ComparisonofVTCAn versus glutamate (Glu) and N-acetylaspartate (NAA) concentration in the midline occipital/parietal lobe of healthy rotic tissue, possibly contributing to the hyperexcited state of the elderly subjects. The results show a strong correlation between the rate tissue (123). In order to test whether the chronically elevated of the neuronal tricarboxylic acid (TCA) cycle and the concentrations extracellular glutamate was a result of an impairment in glial glu- Glu and NAA, both of which have been associated with cellular dysfunc- tamate uptake and cycling, Petroff et al. (124) obtained neurosur- tion and chronically reduced mitochondrial activity in other studies. Pear- gical specimens from patients infused intravenously with [2-13C] son correlation coefﬁcients are shown in the top left-hand corners of (a) glucose, and analyzed the labeling ex vivo using high-resolution and (b). Filled circles, values measured for the individual elderly subjects 13C MRS. In the epileptogenic tissue from the hippocampus with (n = 7); open circles, average values for the respective metabolite concen- sclerosis and glial proliferation, there was marked impairment in trations from a young cohort (n = 7). Fluxes and metabolite concentra- glutamate/glutamine cycling compared with more histologically tions are expressed as mmol/g/min and mmol/g, respectively. normal tissue (124). A subsequent study showed that this impairment in glutamate/glutamine cycling may be secondary to reduced activity in glutamine synthetase, establishing this step have upregulated blood–brain transport and metabolism of in the pathophysiology and as a potential therapeutic target monocarboxylic acids (e.g., acetate, b-hydroxybutyrate and (125). lactate), supporting heightened brain function during hypoglycemia. In their study of the medial occipital/parietal cortex, Application of 13C MRS to study pediatric disease patients with type 1 diabetes showed increased metabolism of acetate, which was most probably secondary to increased ace- The increasing concerns over radiation dosages in PET scans of tate transport. An example of the higher 13C labeling attained pediatric patients have provided additional motivation for the in patients with type 1 diabetes is shown in Fig. 5. A blood con- noninvasive application of 13C MRS in this vulnerable group. centration of 1 mM acetate was found to provide, on average, Studies of infants and young children with 13C MRS, however, approximately20% of astrocyte oxidative needs in control sub- are complicated by the need for extended infusion times, in- jects and approximately 35% in patients with type 1 diabetes. creasing the time spent in the magnet. To assess the feasibility Oz et al. (121) used 13C MRS to show that brain glycogen may of pediatric studies, Blüml et al. (126) performed 13C MRS on 17 be a signiﬁcant fuel source during hypoglycemia, and that children and pediatric patients receiving [1-13C]glucose either

Figure 5. Comparison of steady-state 13 C MRS spectra during [2-13 C]acetate infusion of a patient with type 1 diabetes and a healthy control subject. Brain 13 C MRS spectra were averaged over the final 45 min of a hypoglycemic period during infusion of [2-13 C]acetate. The patient with diabetes (top fi spectrum) showed signi cantly greater labeling in glutamate (Glu) and glutamine (Gln) C4 than the control (bottom spectrum). The acetate C2 signal 951 was also greater in the patient with diabetes. Other resonances labeled in the figure include g-aminobutyric acid (GABA) C2 and Glu C3.

orally or intravenously. They observed marked differences in 13C Improvements in shimming and reduction in RF heating to labeling patterns in premature brain and pediatric patients with allow multi-volume 13C MRS leukodystrophies and mitochondrial disorders. This study To date, the majority of human brain 13C MRS studies have been demonstrates the significant potential of 13C MRS applications performed in the midline occipital or occipital/parietal lobe, to pediatric disease, particularly with the improvements in sensi- largely because of the relative ease of shimming to improve B tivity discussed below. homogeneity in this region and the distance from the eyes, which are believed to be more sensitive than the brain to heating from the decoupling B1 field. Over the last decade, limita- 13 FUTURE PROSPECTS FOR C MRS STUDIES IN tions of shimming have been greatly reduced as a result of HUMANS improvements in shim coil strengths and advanced field mapping and shim calculation and updating methods, allowing As discussed above, there are several major challenges that must well-shimmed spectra to be obtained from multiple volumes fi be met for 13C MRS to become a routine tool in the study of within the human brain even at ultrahigh elds (52,132–134). fl A continuing limitation is the heating that results from the ap- human brain disease and treatment. Below, we brie y discuss 13 these limitations, and how they may be overcome through fu- plied RF energy to the protons bound to C needed to decouple ture technological developments. the J interaction. It has been shown that there are theoretical limits on the minimum decoupling power (135) which, even at 4 T, are close to the allowable power deposition limits mandated by the US Food and Drug Administration (136). Although Improvements in the sensitivity and spatial resolution of 13C advances in RF coil design have allowed human brain 13C studies MRS measurements to be performed safely even up to 4 T (137,138), concerns remain about RF heating of the eyes, which may be more vulnerable The primary limitation for the study of human brain disease by than the brain because of areas of restricted circulation. A re- 13C MRS is its low sensitivity, with the typical volume resolution cently developed alternative approach to circumvent RF heating being on the order of 25–100 cm3. Substantial improvements is to observe 13C labeling of the carboxyl groups of glutamate have been achieved by detecting 13C labeling indirectly via the and glutamine which require no (or low-power) decoupling J scalar coupling to bound protons (55,103), enhancing the spa- (139,140). This approach takes advantage of the turnover kinetics tial resolution to several cubic centimeters obtained at 4 T of glutamate C5 from exogenous [2-13 C]glucose, which is (57,84). However, because of the limited spectral resolution of 1H MRS, only labeling of glutamate C4, the combined resonances of glutamate and glutamine C3, and lactate C3 have been reported, limiting the rates that can be measured to the neuronal TCA cycle or, in the case of elevated lactate, to glycolysis. With the advent of ultrahigh-field human MRS systems (7 T and above), in principle, it should be possible to measure resonances of glutamine and GABA, as has been demonstrated in animal studies (59,127,128), although the increased heating associated with decoupling at higher fields may limit this application. An alternative possibility, which would retain the high spectral resolution and information of direct 13C MRS, and provide much higher spatial resolution, is the use of hyperpolarized 13C MRS, and there have been several promising initial reports in animal models (129–131). The major limitations in the measurement of the rates of the pathways discussed in this article are that there are several enzymatic steps between the precursor (e.g. acetate, lactate, glucose) and enzymatic reactions in the pathways of 13 interest. For example, to measure Vcyc from hyperpolarized C- acetate, the acetate must first be transported into glial cells and then converted successively to acetyl-CoA, citrate, isocitrate, a-ketoglutarate and glutamate, before being converted by glutamine synthetase into glutamine. Based on the concentration of the precursor pools and the metabolic rate of acetate metabolism, it would take on the order of 1–2 min for the immediate 13 precursor glial glutamate to be labeled sufficiently to measure Figure 6. C MRS time course spectra of glutamate (Glu), glutamine (Gln) and aspartate (Asp) turnover detected in the occipital lobe the glutamine synthesis rate. Measurement of the glial and neu- 13 ronal TCA cycle, and anaplerosis, may be feasible by the direct ex- during the intravenous infusion of [2- C]glucose. Lorentzian–Gaussian transformation (Lorentzian broadening (LB) = 3 Hz, Gaussian broadening amination of TCA cycle intermediates (e.g. citrate) from precursors À 13 13 (GB) = 0.3) was applied. The time-averaged decoupling power was such as [2- C]acetate and [2- C]lactate, which label TCA cycle 1.46 W. Each spectrum corresponds to 8.5 min of signal averaging intermediates in one or two enzymatic steps. An alternative ap- (128 scans). Glu C5 (182.0 ppm) and C1 (175.4 ppm), Gln C5 (178.5 ppm) proach is to hyperpolarize directly TCA cycle intermediates and C1 (174.9 ppm), Asp C4 (178.3 ppm) and C1 (175.0 ppm) and

952 themselves, the feasibility of which has been demonstrated in N-acetylaspartate (NAA) C5 (174.3 ppm) were detected. No baseline extracts (131). corrections were made. Figure adapted from Li et al. (141).

wileyonlinelibrary.com/journal/nbm Copyright © 2011 John Wiley & Sons, Ltd. NMR Biomed. 2011; 24: 943–957 13C MRS STUDIES OF HUMAN NEUROENERGETICS AND NEUROTRANSMITTER CYCLING identical to the turnover of glutamate C4 from exogenous 13C have the potential to greatly increase the sensitivity of the [1-13 C]glucose (141). The carboxylic carbons are only coupled method, leading to the possibility of using 13C MRS for metabolic to protons via very weak long-range 1H–13C scalar couplings, imaging of the human brain. These technological developments, so that they can be effectively decoupled at low RF power. An together with further improvements in 13C infusion protocols to additional advantage of this strategy is the lack of contamination minimize patient time in the magnet, have the potential to from subcutaneous lipids, because there are no overlapping fat greatly expedite clinical and research studies. signals in the vicinity of the glutamate C5 and glutamine C5 peaks. An example showing the feasibility of this strategy at 3 T is the work of Li et al. (141) (Fig. 6). High-quality spectra can be Acknowledgements obtained with a maximum regional power deposition in the This work was supported by grants R01-DK027121, P30- brain below 2 W/kg, several times below the US Food and Drug NS052519, R01-EB000473, R01-MH95104, R01-DA021785 R21- 1 Administration limit, even using a H resonator to deliver the AA018210, R21-AA019803 from NIH and 10A087 from AICR. We fi RF decoupling eld (142). The ability to deliver RF decoupling thank Dr Jun Shen for providing the figure 6. from a volume coil further opens up the possibility of multi-volume whole-brain 13C MRS. An alternative approach to reduce RF heating would be to use REFERENCES hyperpolarized 13C MRS without decoupling. The higher sensitiv- 13 1. Shephard GM. The Synaptic Organization of the Brain. Oxford Uni- ity of hyperpolarized C would, in principle, make decoupling versity Press, Oxford; 1994. unnecessary. However, as a result of the blood–brain barrier 2. Rothstein JD, Dykes-Hoberg M, Pardo CA, Bristol LA, Jin L, Kuncl RW, restricting the ability to isotopically label brain metabolites from Kanai Y, Hediger MA, Wang Y, Schielke JP, Welty DF. Knockout of glu- plasma, 13C-labeled substrate decoupling may be necessary tamate transporters reveals a major role for astroglial transport in exci- 13 totoxicity and clearance of glutamate. Neuron, 1996; 16(3): 675–686. even for hyperpolarized C MRS applications. 3. Bergles DE, Diamond JS, Jahr CE. Clearance of glutamate inside the synapse and beyond. Curr. Opin. Neurobiol. 1999; 9(3): 293–298. Improvements in 13C infusion protocols 4. Hertz L, Peng L, Dienel GA. Energy metabolism in astrocytes: high rate of oxidative metabolism and spatiotemporal dependence on A third major limitation in 13C MRS studies is the requirement for glycolysis/glycogenolysis. J. Cereb. Blood Flow Metab. 2007; 27(2): a continuous infusion of the isotopically labeled substrate with 219–249. venous sampling for fractional enrichment determination. This 5. Sibson NR, Mason GF, Shen J, Cline GW, Herskovits AZ, Wall JE, Behar KL, Rothman DL, Shulman RG. In vivo (13)C NMR measurement infusion must occur over a time period (typically 2 h) to capture of neurotransmitter glutamate cycling, anaplerosis and TCA cycle flux in sufficient kinetic information from spectral time courses for ab- rat brain during [2-13 C] glucose infusion. J. Neurochem. 2001; 76(4): solute rate estimation using metabolic modeling (24,45,143). 975–989. However, several studies have shown that it is possible to obtain 6. Yu AC, Drejer J, Hertz L, Schousboe A. Pyruvate carboxylase activity considerable information on metabolism using simplified infu- in primary cultures of astrocytes and neurons. J. Neurochem. 1983; 41(5): 1484–1487. sion schemes or oral ingestion (144–147). Given that the majority 7. Rothman DL, Behar KL, Hyder F, Shulman RG. In vivo NMR studies of of information on absolute and relative rates is derived from the the glutamate neurotransmitter flux and neuroenergetics: implica- early and steady-state portions of the time course, a significant tions for brain function. Annu. Rev. Physiol. 2003; 65: 401–427. reduction in the time required for a subject to be in the scanner 8. Mason GF, Rothman DL. Graded image segmentation of brain tissue in the presence of inhomogeneous radio frequency fields. may be possible (143). Magn. Reson. Imaging, 2002; 20(5): 431–436. 9. Boumezbeur F, Besret L, Valette J, Gregoire M-C, Delzescaux T, Maroy R, Vaufrey F, Gervais P, Hantraye P, Bloch G, Lebon V. Glycol- CONCLUSIONS ysis versus TCA cycle in the primate brain as measured by combining 18F-FDG PET and 13C-NMR. J. Cereb. Blood Flow Metab. 2005; 25 13 Work over the past two decades has established C MRS studies (11): 1418–1423. of the brain in animal models and humans as the only noninva- 10. Chaumeil MM, Valette J, Guillermier M, Brouillet E, Boumezbeur F, sive method for the measurement of neuronal and glial energy Herard A-S, Bloch G, Hantraye P, Lebon V. Multimodal neuroima- metabolism and glutamate and GABA neurotransmitter cycling. ging provides a highly consistent picture of energy metabolism, validating 31P MRS for measuring brain ATP synthesis. Proc. Natl. Although some debate regarding metabolic modeling remains, Acad. Sci. USA, 2009; 106(10): 3988–3993. much of our present knowledge of the brain energy budget 11. Hyder F, Patel AB, Gjedde A, Rothman DL, Behar KL, Shulman RG. (glutamatergic neurons, GABAergic neurons, astrocytes) and Neuronal–glial glucose oxidation and glutamatergic–GABAergic the relationship between neuroenergetics and neurotransmis- function. J. Cereb. Blood Flow Metab. 2006; 26(7): 865–877. sion has been obtained from in vivo 13C MRS studies. 13C MRS 12. Van den Berg CJ, Krzalić L, Mela P, Waelsch H. Compartmentation of glutamate metabolism in brain. Evidence for the existence of two differ- studies have also played an important role in delineating how alent tricarboxylic acid cycles in brain. Biochem. J. 1969; 113(2): 281–290. ternative substrates, such as acetate, ketone bodies and lactate, 13. Hertz L. Functional interactions between neurons and astrocytes I. support neuronal and astrocyte energetics. The application of Turnover and metabolism of putative amino acid transmitters. 13C MRS to the study of human disease faces considerable obsta- Prog. Neurobiol. 1979; 13(3): 277–323. 13 14. Maycox PR, Hell JW, Jahn R. Amino acid neurotransmission: spot- cles, particularly the cost, the need for C substrate infusions light on synaptic vesicles. Trends Neurosci. 1990; 13(3): 83–87. and monitoring, decoupling heating and the low sensitivity of 15. Nicholls D, Attwell D. The release and uptake of excitatory amino 13C MRS and lack of technical capability of most MR systems. De- acids. Trends Pharmacol. Sci. 1990; 11(11): 462–468. spite these difficulties, 13C MRS has been successfully applied to 16. Gruetter R, Novotny EJ, Boulware SD, Mason GF, Rothman DL, 13 the study of a variety of neurological and psychiatric diseases, as Shulman GI, Prichard JW, Shulman RG. Localized C NMR spectroscopy in the human brain of amino acid labeling from D-[1-13 C]glu- well as diabetes and healthy aging. The increasing availability of cose. J. Neurochem. 1994; 63(4): 1377–1385. fi high- eld MR magnets, which allow higher sensitivity indirect 17. Cooper AJ, Plum F. Biochemistry and physiology of brain ammonia. 953 detection methods, and the development of hyperpolarized Physiol. Rev. 1987; 67(2): 440–519.

18. Sibson NR, Dhankhar A, Mason GF, Behar KL, Rothman DL, Shulman 40. Patel AB, de Graaf RA, Rothman DL, Behar KL, Mason GF. Evaluation RG. In vivo 13C NMR measurements of cerebral glutamine synthesis of cerebral acetate transport and metabolic rates in the rat brain in as evidence for glutamate–glutamine cycling. Proc. Natl. Acad. Sci. vivo using 1H–[13C]-NMR. J. Cereb. Blood Flow Metab. 2010; 30(6): USA, 1997; 94(6): 2699–2704. 1200–1213. 19. Lieth E, LaNoue KF, Berkich DA, Xu B, Ratz M, Taylor C, Hutson SM. 41. Boumezbeur F, Mason GF, de Graaf RA, Behar KL, Cline GW, Nitrogen shuttling between neurons and glial cells during gluta- Shulman GI, Rothman DL, Petersen KF. Altered brain mitochondrial mate synthesis. J. Neurochem. 2001; 76(6): 1712–1723. metabolism in healthy aging as assessed by in vivo magnetic resonance 20. Oz G, Berkich DA, Henry P-G, Xu Y, LaNoue K, Hutson SM, Gruetter R. spectroscopy. J. Cereb. Blood Flow Metab. 2010; 30(1): 211–221. Neuroglial metabolism in the awake rat brain: CO2 fixation increases 42. Maciejewski PK, Rothman DL. Proposed cycles for functional gluta- with brain activity. J. Neurosci. 2004; 24(50): 11 273–11 279. mate trafficking in synaptic neurotransmission. Neurochem. Int. 21. Shen J, Sibson NR, Cline G, Behar KL, Rothman DL, Shulman RG. 2008; 52(4–5): 809–825. 15 N-NMR spectroscopy studies of ammonia transport and gluta- 43. DiNuzzo M, Mangia S, Maraviglia B, Giove F. Changes in glucose up- mine synthesis in the hyperammonemic rat brain. Dev. Neurosci. take rather than lactate shuttle take center stage in subserving neu- 1998; 20(4–5): 434–443. roenergetics: evidence from mathematical modeling. J. Cereb. 22. Patel AB, de Graaf RA, Mason GF, Rothman DL, Shulman RG, Behar KL. Blood Flow Metab. 2010; 30(3): 586–602. The contribution of GABA to glutamate/glutamine cycling and en- 44. Shestov AA, Valette J, Uğurbil K, Henry P-G. On the reliability of (13)C ergy metabolism in the rat cortex in vivo. Proc. Natl. Acad. Sci. metabolic modeling with two-compartment neuronal–glial models. USA, 2005; 102(15): 5588–5593. J. Neurosci. Res. 2007; 85(15): 3294–3303. 23. Kanamori K, Kondrat RW, Ross BD. 13C enrichment of extracellular 45. Mason GF, Gruetter R, Rothman DL, Behar KL, Shulman RG, neurotransmitter glutamate in rat brain – combined mass spectrom- Novotny EJ. Simultaneous determination of the rates of the TCA cy- etry and NMR studies of neurotransmitter turnover and uptake into cle, glucose utilization, alpha-ketoglutarate/glutamate exchange, glia in vivo. Cell Mol. Biol. (Noisy-le-grand), 2003; 49(5): 819–836. and glutamine synthesis in human brain by NMR. J. Cereb. Blood 24. Gruetter R, Seaquist ER, Ugurbil K. A mathematical model of com- Flow Metab. 1995; 15(1): 12–25. partmentalized neurotransmitter metabolism in the human brain. 46. Yang J, Xu S, Shen J. Fast isotopic exchange between mitochondria 13 Am. J. Physiol. Endocrinol. Metab. 2001; 281(1): E100–E112. and cytosol in brain revealed by relayed C magnetization transfer 25. Lebon V, Petersen KF, Cline GW, Shen J, Mason GF, Dufour S, Behar KL, spectroscopy. J. Cereb. Blood Flow Metab. 2009; 29(4): 661–669. Shulman GI, Rothman DL. Astroglial contribution to brain energy 47. Shen J, Rothman D, Behar K, Xu S. Determination of the glutamate– 13 fl metabolism in humans revealed by C nuclear magnetic reso- glutamine cycling ux using two-compartment dynamic metabolic nance spectroscopy: elucidation of the dominant pathway for neu- modeling is sensitive to astroglial dilution. J. Cereb. Blood Flow rotransmitter glutamate repletion and measurement of astrocytic Metab. 2008; 29: 108–118. oxidative metabolism. J. Neurosci. 2002; 22(5): 1523–1531. 48. McKenna MC. The glutamate–glutamine cycle is not stoichiometric: 26. Mason GF, Petersen KF, de Graaf RA, Shulman GI, Rothman DL. Mea- fates of glutamate in brain. J. Neurosci. Res. 2007; 85(15): 3347–3358. surements of the anaplerotic rate in the human cerebral cortex us- 49. Behar KL, Petroff OA, Prichard JW, Alger JR, Shulman RG. Detection of 13 13 13 metabolites in rabbit brain by 13C NMR spectroscopy following ad- ing C magnetic resonance spectroscopy and [1- C] and [2- C] 13 glucose. J. Neurochem. 2007; 100(1): 73–86. ministration of [1- C]glucose. Magn. Reson. Med. 1986; 3(6): 911–920. fl 27. Shen J, Petersen KF, Behar KL, Brown P, Nixon TW, Mason GF, Petroff 50. Fitzpatrick SM, Hetherington HP, Behar KL, Shulman RG. The ux from glucose to glutamate in the rat brain in vivo as determined OA, Shulman GI, Shulman RG, Rothman DL. Determination of the 1 13 rate of the glutamate/glutamine cycle in the human brain by in vivo by H-observed, C-edited NMR spectroscopy. J. Cereb. Blood Flow 13C NMR. Proc. Natl. Acad. Sci. USA, 1999; 96(14): 8235–8240. Metab. 1990; 10(2): 170–179. 51. Rothman DL, Behar KL, HetheringtonHP, den Hollander JA, Bendall MR, 28. Siesjo BK. Brain Energy Metabolism. Wiley, Chichester, New York; 1978. 1 13 29. Sibson NR, Dhankhar A, Mason GF, Rothman DL, Behar KL, Shulman RG. Petroff OA, Shulman RG. H-Observe/ C-decouple spectroscopic Stoichiometric coupling of brain glucose metabolism and glutamatergic measurements of lactate and glutamate in the rat brain in vivo. – neuronal activity. Proc. Natl. Acad. Sci. USA, 1998; 95(1): 316–321. Proc. Natl. Acad. Sci. USA, 1985; 82(6): 1633 1637. 52. Gruetter R. Automatic, localized in vivo adjustment of all first- and 30. Magistretti PJ, Pellerin L, Rothman DL, Shulman RG. Energy on de- second-order shim coils. Magn. Reson. Med. 1993; 29(6): 804–811. mand. Science, 1999; 283(5401): 496–497. 53. Beckmann N, Turkalj I, Seelig J, Keller U. 13C NMR for the assessment 31. Shulman RG, Hyder F, Rothman DL. Cerebral energetics and the gly- of human brain glucose metabolism in vivo. Biochemistry, 1991; 30(26): cogen shunt: neurochemical basis of functional imaging. Proc. Natl. 6362–6366. Acad. Sci. USA, 2001; 98(11): 6417–6422. 54. Gruetter R, Novotny EJ, Boulware SD, Rothman DL, Mason GF, 32. Jolivet R, Magistretti PJ, Weber B. Deciphering neuron–glia com- Shulman GI, Shulman RG, Tamborlane WV. Direct measurement of partmentalization in cortical energy metabolism. Front Neuroener- brain glucose concentrations in humans by 13C NMR spectroscopy. getics, 2009; 1: 4. Proc. Natl. Acad. Sci. USA, 1992; 89(3): 1109–1112. 33. Occhipinti R, Somersalo E, Calvetti D. Astrocytes as the glucose 55. Rothman DL, Novotny EJ, Shulman GI, Howseman AM, Petroff OA, shunt for glutamatergic neurons at high activity: an in silico study. Mason G, Nixon T, Hanstock CC, Prichard JW, Shulman RG. 1H– J. Neurophysiol. 2009; 101(5): 2528–2538. [13C] NMR measurements of [4-13 C]glutamate turnover in human 34. Strelnikov K. Neuroimaging and neuroenergetics: brain activations brain. Proc. Natl. Acad. Sci. USA, 1992; 89(20): 9603–9606. fl as information-driven reorganization of energy ows. Brain Cogn. 56. Mason GF, Pan JW, Chu WJ, Newcomer BR, Zhang Y, Orr R, 2010; 72(3): 449–456. Hetherington HP. Measurement of the tricarboxylic acid cycle rate in 35. Attwell D, Laughlin SB. An energy budget for signaling in the grey mat- human grey and white matter in vivo by 1H–[13C] magnetic resonance ter of the brain. J. Cereb. Blood Flow Metab. 2001; 21(10): 1133–1145. spectroscopy at 4.1T. J. Cereb. Blood Flow Metab. 1999; 19(11): 36. Chowdhury GMI, Patel AB, Mason GF, Rothman DL, Behar KL. Gluta- 1179–1188. matergic and GABAergic neurotransmitter cycling and energy me- 57. Pan JW, Stein DT, Telang F, Lee JH, Shen J, Brown P, Cline G, Mason GF, tabolism in rat cerebral cortex during postnatal development. J. Shulman GI, Rothman DL, Hetherington HP. Spectroscopic imaging Cereb. Blood Flow Metab. 2007; 27(12): 1895–1907. of glutamate C4 turnover in human brain. Magn. Reson. Med. 2000; 37. Hassel B, Bachelard H, Jones P, Fonnum F, Sonnewald U. Trafficking 44(5): 673–679. of amino acids between neurons and glia in vivo. Effects of inhibi- 58. Gruetter R, Seaquist ER, Kim S, Ugurbil K. Localized in vivo 13C-NMR tion of glial metabolism by fluoroacetate. J. Cereb. Blood Flow of glutamate metabolism in the human brain: initial results at 4 Metab. 1997; 17(11): 1230–1238. tesla. Dev. Neurosci. 1998; 20(4–5): 380–388. 38. Patel AB, Chowdhury GMI, de Graaf RA, Rothman DL, Shulman RG, 59. van Eijsden P, Behar KL, Mason GF, Braun KPJ, de Graaf RA. In vivo Behar KL. Cerebral pyruvate carboxylase flux is unaltered during neurochemical profiling of rat brain by 1H–[13C] NMR spectroscopy: bicuculline-seizures. J. Neurosci. Res. 2005; 79(1–2): 128–138. cerebral energetics and glutamatergic/GABAergic neurotransmis- 39. Serres S, Raffard G, Franconi J-M, Merle M. Close coupling between sion. J. Neurochem. 2010; 112(1): 24–33. astrocytic and neuronal metabolisms to fulfill anaplerotic and en- 60. Shulman RG, Rothman DL. Interpreting functional imaging studies

954 ergy needs in the rat brain. J. Cereb. Blood Flow Metab. 2008; 28(4): in terms of neurotransmitter cycling. Proc. Natl. Acad. Sci. USA, 712–724. 1998; 95(20): 11 993–11 998.

61. Raichle ME. A paradigm shift in functional brain imaging. J. Neu- NMR in human visual cortex during physiologic stimulation. Proc. rosci. 2009; 29(41): 12 729–12 734. Natl. Acad. Sci. USA, 1991; 88(13): 5829–5831. 62. Owen OE, Morgan AP, Kemp HG, Sullivan JM, Herrera MG, Cahill GF. 84. Chen W, Zhu XH, Gruetter R, Seaquist ER, Adriany G, Ugurbil K. Brain metabolism during fasting. J. Clin. Invest. 1967; 46(10): Study of tricarboxylic acid cycle flux changes in human visual cortex 1589–1595. during hemifield visual stimulation using (1)H–[(13)C] MRS and 63. van Hall G, Strømstad M, Rasmussen P, Jans O, Zaar M, Gam C, fMRI. Magn. Reson. Med. 2001; 45(3): 349–355. Quistorff B, Secher NH, Nielsen HB. Blood lactate is an important en- 85. Hoge RD, Atkinson J, Gill B, Crelier GR, Marrett S, Pike GB. Linear ergy source for the human brain. J. Cereb. Blood Flow Metab. 2009; coupling between cerebral blood flow and oxygen consumption 29(6): 1121–1129. in activated human cortex. Proc. Natl. Acad. Sci. USA, 1999; 96(16): 64. Smith D, Pernet A, Hallett WA, Bingham E, Marsden PK, Amiel SA. 9403–9408. Lactate: a preferred fuel for human brain metabolism in vivo. J. 86. Kim SG, Rostrup E, Larsson HB, Ogawa S, Paulson OB. Determination Cereb. Blood Flow Metab. 2003; 23(6): 658–664. of relative CMRO2 from CBF and BOLD changes: significant increase 65. Gruetter R, Novotny EJ, Boulware SD, Rothman DL, Shulman RG. 1H of oxygen consumption rate during visual stimulation. Magn. NMR studies of glucose transport in the human brain. J. Cereb. Reson. Med. 1999; 41(6): 1152–1161. Blood Flow Metab. 1996; 16(3): 427–438. 87. Marrett S, Gjedde A. Changes of blood flow and oxygen consump- 66. Blomqvist G, Gjedde A, Gutniak M, Grill V, Widén L, Stone-Elander S, tion in visual cortex of living humans. Adv. Exp. Med. Biol. 1997; Hellstrand E. Facilitated transport of glucose from blood to brain in 413: 205–208. man and the effect of moderate hypoglycaemia on cerebral glu- 88. Haga KK, Khor YP, Farrall A, Wardlaw JM. A systematic review of 1 cose utilization. Eur. J. Nucl. Med. 1991; 18(10): 834–837. brain metabolite changes, measured with H magnetic resonance 67. Heiss WD, Pawlik G, Herholz K, Wagner R, Göldner H, Wienhard K. spectroscopy, in healthy aging. Neurobiol. Aging, 2009; 30(3): Regional kinetic constants and cerebral metabolic rate for glucose 353–363. in normal human volunteers determined by dynamic positron 89. Dager SR, Friedman SD, Parow A, Demopulos C, Stoll AL, Lyoo IK, emission tomography of [18F]-2-fluoro-2-deoxy-D-glucose. J. Cereb. Dunner DL, Renshaw PF. Brain metabolic alterations in medica- Blood Flow Metab. 1984; 4(2): 212–223. tion-free patients with bipolar disorder. Arch. Gen. Psychiatry, 68. de Graaf RA, Pan JW, Telang F, Lee JH, Brown P, Novotny EJ, 2004; 61(5): 450–458. Hetherington HP, Rothman DL. Differentiation of glucose transport 90. Sanacora G, Gueorguieva R, Epperson CN, Wu Y-T, Appel M, in human brain gray and white matter. J. Cereb. Blood Flow Metab. Rothman DL, Krystal JH, Mason GF. Subtype-specific alterations of 2001; 21(5): 483–492. gamma-aminobutyric acid and glutamate in patients with major 69. Gruetter R, Ugurbil K, Seaquist ER. Steady-state cerebral glucose depression. Arch. Gen. Psychiatry, 2004; 61(7): 705–713. concentrations and transport in the human brain. J. Neurochem. 91. Gruber S, Frey R, Mlynárik V, Stadlbauer A, Heiden A, Kasper S, 1998; 70(1): 397–408. Kemp GJ, Moser E. Quantification of metabolic differences in the 1 70. Simpson IA, Carruthers A, Vannucci SJ. Supply and demand in cere- frontal brain of depressive patients and controls obtained by H- bral energy metabolism: the role of nutrient transporters. J. Cereb. MRS at 3 Tesla. Invest. Radiol. 2003; 38(7): 403–408. Blood Flow Metab. 2007; 27(11): 1766–1791. 92. Epperson CN, Haga K, Mason GF, Sellers E, Gueorguieva R, Zhang W, 71. Badar-Goffer RS, Bachelard HS, Morris PG. Cerebral metabolism of Weiss E, Rothman DL, Krystal JH. Cortical gamma-aminobutyric acid acetate and glucose studied by 13C-n.m.r. spectroscopy. A tech- levels across the menstrual cycle in healthy women and those with pre- nique for investigating metabolic compartmentation in the brain. menstrual dysphoric disorder: a proton magnetic resonance spectros- Biochem. J. 1990; 266(1): 133–139. copy study. Arch. Gen. Psychiatry, 2002; 59(9): 851–858. 72. Cerdán S, Künnecke B, Seelig J. Cerebral metabolism of [1,2-13C2]ac- 93. Goddard AW, Mason GF, Almai A, Rothman DL, Behar KL, Petroff OA, 13 Charney DS, Krystal JH. Reductions in occipital cortex GABA levels in etate as detected by in vivo and in vitro C NMR. J. Biol. Chem. 1 1990; 265(22): 12 916–12 926. panic disorder detected with H-magnetic resonance spectroscopy. 13 – 73. Künnecke B, Cerdan S. Multilabeled C substrates as probes in Arch. Gen. Psychiatry, 2001; 58(6): 556 561. in vivo 13C and 1H NMR spectroscopy. NMR Biomed. 1989; 2(5–6): 94. Hasler G, van der Veen JW, Tumonis T, Meyers N, Shen J, Drevets WC. 274–277. Reduced prefrontal glutamate/glutamine and gamma-aminobutyric 74. Waniewski RA, Martin DL. Preferential utilization of acetate by astro- acid levels in major depression determined using proton mag- cytes is attributable to transport. J. Neurosci. 1998; 18(14): 5225–5233. netic resonance spectroscopy. Arch. Gen. Psychiatry, 2007; 64(2): 193–200. 75. Blüml S, Moreno-Torres A, Shic F, Nguy C-H, Ross BD. Tricarboxylic 95. Kuzniecky R, Ho S, Pan J, Martin R, Gilliam F, Faught E, Hetherington H. acid cycle of glia in the in vivo human brain. NMR Biomed. 2002; Modulation of cerebral GABA by topiramate, lamotrigine, and gaba- 15(1): 1–5. pentin in healthy adults. Neurology, 2002; 58(3): 368–372. 76. Mason GF, Petersen KF, Lebon V, Rothman DL, Shulman GI. In- 96. Novotny EJ, Hyder F, Shevell M, Rothman DL. GABA changes with creased brain monocarboxylic acid transport and utilization in type vigabatrin in the developing human brain. Epilepsia, 1999; 40(4): 1 diabetes. Diabetes, 2006; 55(4): 929–934. 462–466. 77. Pan JW, de Graaf RA, Petersen KF, Shulman GI, Hetherington HP, 13 97. Petroff OA, Mattson RH, Rothman DL. Proton MRS: GABA and gluta- Rothman DL. [2,4- C2]-beta-Hydroxybutyrate metabolism in hu- mate. Adv. Neurol. 2000; 83: 261–271. man brain. J. Cereb. Blood Flow Metab. 2002; 22(7): 890–898. 98. Novotny EJ, Fulbright RK, Pearl PL, Gibson KM, Rothman DL. Mag- 78. Boumezbeur F, Petersen KF, Cline GW, Mason GF, Behar KL, netic resonance spectroscopy of neurotransmitters in human brain. Shulman GI, Rothman DL. The contribution of blood lactate to brain – 13 Ann. Neurol. 2003; 54 (Suppl 6): S25 S31. energy metabolism in humans measured by dynamic C nuclear 99. Ross BD, Jacobson S, Villamil F, Korula J, Kreis R, Ernst T, Shonk T, magnetic resonance spectroscopy. J. Neurosci. 2010; 30(42): Moats RA. Subclinical hepatic encephalopathy: proton MR spectro- 13 983–13 991. scopic abnormalities. Radiology, 1994; 193(2): 457–463. 79. de Graaf RA. In vivo NMR Spectroscopy. Principles and Techniques: 100. Martin WRW. MR spectroscopy in neurodegenerative disease. Mol. Wiley, Chichester, England; 2007. Imaging Biol. 2007; 9(4): 196–203. 80. Fox PT, Raichle ME, Mintun MA, Dence C. Nonoxidative glucose 101. Ross B, Lin A, Harris K, Bhattacharya P, Schweinsburg B. Clinical expe- consumption during focal physiologic neural activity. Science, rience with 13C MRS in vivo. NMR Biomed. 2003; 16(6–7): 358–369. 1988; 241(4864): 462–464. 102. Novotny EJ, Ogino T, Rothman DL, Petroff OA, Prichard JW, 81. Dienel GA, Hertz L. Glucose and lactate metabolism during brain ac- Shulman RG. Direct carbon versus proton heteronuclear editing of tivation. J. Neurosci. Res. 2001; 66(5): 824–838. 2-13C ethanol in rabbit brain in vivo: a sensitivity comparison. Magn. 82. Mangia S, Tkác I, Gruetter R, Van De Moortele P-F, Maraviglia B, Reson. Med. 1990; 16(3): 431–443. Uğurbil K. Sustained neuronal activation raises oxidative metabo- 103. Rothman DL, Howseman AM, Graham GD, Petroff OA, Lantos G, lism to a new steady-state level: evidence from 1H NMR spectros- Fayad PB, Brass LM, Shulman GI, Shulman RG, Prichard JW. Local- copy in the human visual cortex. J. Cereb. Blood Flow Metab. ized proton NMR observation of [3-13 C]lactate in stroke after 2007; 27(5): 1055–1063. [1-13 C]glucose infusion. Magn. Reson. Med. 1991; 21(2): 302–307.

83. Prichard J, Rothman D, Novotny E, Petroff O, Kuwabara T, Avison M, 104. Petroff OA, Graham GD, Blamire AM, al-Rayess M, Rothman DL, 955 Howseman A, Hanstock C, Shulman R. Lactate rise detected by 1H Fayad PB, Brass LM, Shulman RG, Prichard JW. Spectroscopic

imaging of stroke in humans: histopathology correlates of spectral possible mechanism for raised extracellular glutamate in mesial changes. Neurology, 1992; 42(7): 1349–1354. temporal lobe epilepsy. Lancet, 2004; 363(9402): 28–37. 105. Terpstra M, Gruetter R, High WB, Mescher M, DelaBarre L, Merkle H, 126. Blüml S, Moreno A, Hwang JH, Ross BD. 1-(13)C glucose magnetic Garwood M. Lactate turnover in rat glioma measured by in vivo nu- resonance spectroscopy of pediatric and adult brain disorders. clear magnetic resonance spectroscopy. Cancer Res. 1998; 58(22): NMR Biomed. 2001; 14(1): 19–32. 5083–5088. 127. de Graaf RA, Mason GF, Patel AB, Rothman DL, Behar KL. Regional 106. Wijnen JP, van der Graaf M, Scheenen TWJ, Klomp DWJ, de Galan BE, glucose metabolism and glutamatergic neurotransmission in rat Idema AJS, Heerschap A. In vivo 13 C magnetic resonance spectros- brain in vivo. Proc. Natl. Acad. Sci. USA, 2004; 101(34): 12 700–12 705. copy of a human brain tumor after application of 13C-1-enriched 128. Pfeuffer J, Tkác I, Choi IY, Merkle H, Ugurbil K, Garwood M, Gruetter R. glucose. Magn. Reson. Imaging, 2010; 28(5): 690–697. Localized in vivo 1H NMR detection of neurotransmitter labeling in 107. Butterworth RF. Portal-systemic encephalopathy: a disorder of rat brain during infusion of [1-13C] D-glucose. Magn. Reson. Med. neuron-astrocytic metabolic trafficking. Dev. Neurosci. 1993; 15 1999; 41(6): 1077–1083. (3–5): 313–319. 129. Cudalbu C, Comment A, Kurdzesau F, van Heeswijk RB, Uffmann K, 108. Behar KL, Rothman DL, Petersen KF, Hooten M, Delaney R, Petroff OA, Jannin S, Denisov V, Kirik D, Gruetter R. Feasibility of in vivo 15 N Shulman GI, Navarro V, Petrakis IL, Charney DS, Krystal JH. Prelimi- MRS detection of hyperpolarized 15N labeled choline in rats. Phys. nary evidence of low cortical GABA levels in localized 1H-MR spec- Chem. Chem. Phys. 2010; 12(22): 5818–5823. tra of alcohol-dependent and hepatic encephalopathy patients. 130. Marjańska M, Iltis I, Shestov AA, Deelchand DK, Nelson C, Uğurbil K, Am. J. Psychiatry, 1999; 156(6): 952–954. Henry P-G. In vivo 13 C spectroscopy in the rat brain using hyperpo- 109. Fitzpatrick SM, Hetherington HP, Behar KL, Shulman RG. Effects of larized [1-(13)C]pyruvate and [2-(13)C]pyruvate. J. Magn. Reson. acute hyperammonemia on cerebral amino acid metabolism and 2010; 206(2): 210–218. 1 31 pHi in vivo, measured by H and P nuclear magnetic resonance. 131. Bhattacharya P, Chekmenev EY, Perman WH, Harris KC, Lin AP, J. Neurochem. 1989; 52(3): 741–749. Norton VA, Tan CT, Ross BD, Weitekamp DP. Towards hyperpolar- 1 15 110. Kanamori K, Ross BD. Glial alkalinization detected in vivo by H– N ized (13)C-succinate imaging of brain cancer. J. Magn. Reson. heteronuclear multiple-quantum coherence-transfer NMR in severely 2007; 186(1): 150–155. hyperammonemic rat. J. Neurochem. 1997; 68(3): 1209–1220. 132. Hetherington HP, Chu W-J, Gonen O, Pan JW. Robust fully auto- 111. Behar KL, Fitzpatrick SM. Effects of hypercarbia and porta-caval mated shimming of the human brain for high-field 1H spectro- shunting on amino acids and high energy phosphates of the rat – 1 31 scopic imaging. Magn. Reson. Med. 2006; 56(1): 26 33. brain: a H and P NMR study. In: Butterworth RF, Layrargues GP 133. Juchem C, Nixon TW, McIntyre S, Rothman DL, de Graaf RA. Mag- (eds.). Hepatic Encephalopathy, Pathophysiology and Treatment. netic field homogenization of the human prefrontal cortex with a set Humana Press: Clifton, NJ; 1989, pp. 189–200. of localized electrical coils. Magn. Reson. Med. 2010; 63(1): 171–180. 112. Fitzpatrick SM, Behar KL, Shulman RG. In vivo NMR spectroscopy 134. Koch KM, Sacolick LI, Nixon TW, McIntyre S, Rothman DL, de Graaf RA. studies of cerebral metabolism in rats after portal–caval shunting. Dynamically shimmed multivoxel 1H magnetic resonance spectros- In: Butterworth RF, Layrargues GP (eds.). Hepatic Encephalopathy, copy and multislice magnetic resonance spectroscopic imaging of Pathophysiology and Treatment. Humana Press: Clifton, NJ; 1989, the human brain. Magn. Reson. Med. 2007; 57(3): 587–591. pp. 177–187. 13 135. de Graaf RA. Theoretical and experimental evaluation of broadband 113. Blüml S, Moreno-Torres A, Ross BD. [1- C]glucose MRS in chronic decoupling techniques for in vivo nuclear magnetic resonance hepatic encephalopathy in man. Magn. Reson. Med. 2001; 45(6): spectroscopy. Magn. Reson. Med. 2005; 53(6): 1297–1306. 981–993. 136. Collins CM, Liu W, Wang J, Gruetter R, Vaughan JT, Ugurbil K, 114. Gropman AL, Sailasuta N, Harris KC, Abulseoud O, Ross BD. Orni- Smith MB. Temperature and SAR calculations for a human head thine transcarbamylase deficiency with persistent abnormality in within volume and surface coils at 64 and 300 MHz. J. Magn. Reson. cerebral glutamate metabolism in adults. Radiology, 2009; 252(3): Imaging, 2004; 19(5): 650–656. 833–841. 137. Adriany G, Gruetter R. A half-volume coil for efficient proton decou- 115. Lin MT, Beal MF. Mitochondrial dysfunction and oxidative stress in pling in humans at 4 tesla. J. Magn. Reson. 1997; 125(1): 178–184. neurodegenerative diseases. Nature, 2006; 443(7113): 787–795. 138. Klomp DWJ, Renema WKJ, van der Graaf M, de Galan BE, 116. Kalpouzos G, Chételat G, Baron J-C, Landeau B, Mevel K, Godeau C, Kentgens APM, Heerschap A. Sensitivity-enhanced 13C MR spectros- Barré L, Constans J-M, Viader F, Eustache F, Desgranges B. Voxel- copy of the human brain at 3 Tesla. Magn. Reson. Med. 2006; 55(2): based mapping of brain gray matter volume and glucose metabolism fi 271–278. pro les in normal aging. Neurobiol. Aging, 2009; 30(1): 112–124. 13 117. Rapoport SI. Functional brain imaging in the resting state and dur- 139. Li S, Yang J, Shen J. Novel strategy for cerebral C MRS using very ing activation in Alzheimer’s disease. Implications for disease low RF power for proton decoupling. Magn. Reson. Med. 2007; 57(2): – mechanisms involving oxidative phosphorylation. Ann. NY Acad. 265 271. Sci. 1999; 893: 138–153. 140. Sailasuta N, Abulseoud O, Harris KC, Ross BD. Glial dysfunction in 118. Lin AP, Shic F, Enriquez C, Ross BD. Reduced glutamate neurotrans- abstinent methamphetamine abusers. J. Cereb. Blood Flow Metab. mission in patients with Alzheimer’s disease – an in vivo (13)C mag- 2010; 30(5): 950–960. netic resonance spectroscopy study. MAGMA, 2003; 16(1): 29–42. 141. Li S, Zhang Y, Wang S, Yang J, Ferraris Araneta M, Farris A, Johnson C, Fox S, Innis R, Shen J. In vivo 13C magnetic resonance spectroscopy 119. Cryer PE. Diverse causes of hypoglycemia-associated autonomic 13 failure in diabetes. N. Engl. J. Med. 2004; 350(22): 2272–2279. of human brain on a clinical 3 T scanner using [2- C]glucose infu- 120. Cryer PE. Banting Lecture. Hypoglycemia: the limiting factor in the sion and low-power stochastic decoupling. Magn. Reson. Med. management of IDDM. Diabetes, 1994; 43(11): 1378–1389. 2009; 62(3): 565–573. 142. Li S, Zhang Y, Wang S, Araneta MF, Johnson CS, Xiang Y, Innis RB, 121. Oz G, Kumar A, Rao JP, Kodl CT, Chow L, Eberly LE, Seaquist ER. Hu- 13 man brain glycogen metabolism during and after hypoglycemia. Shen J. CMRSofoccipitalandfrontallobesat3Tusingavolumecoil Diabetes, 2009; 58(9): 1978–1985. for stochastic proton decoupling. NMR Biomed. 2010; 23(8): 977–985. 122. Pan JW, Williamson A, Cavus I, Hetherington HP, Zaveri H, Petroff OAC, 143. Mason GF, Rothman DL. Basic principles of metabolic modeling of Spencer DD. Neurometabolism in human epilepsy. Epilepsia, 2008; NMR (13)C isotopic turnover to determine rates of brain metabo- 49 (Suppl 3): 31–41. lism in vivo. Metab. Eng. 2004; 6(1): 75–84. 123. Cavus I, Kasoff WS, Cassaday MP, Jacob R, Gueorguieva R, Sherwin RS, 144. Moreno A, Blüml S, Hwang JH, Ross BD. Alternative 1-(13)C glucose Krystal JH, Spencer DD, Abi-Saab WM. Extracellular metabolites in infusion protocols for clinical (13)C MRS examinations of the brain. the cortex and hippocampus of epileptic patients. Ann. Neurol. Magn. Reson. Med. 2001; 46(1): 39–48. 2005; 57(2): 226–235. 145. Mason GF, Falk Petersen K, de Graaf RA, Kanamatsu T, Otsuki T, 124. Petroff OAC, Errante LD, Rothman DL, Kim JH, Spencer DD. Gluta- Shulman GI, Rothman DL. A comparison of (13)C NMR measure- mate–glutamine cycling in the epileptic human hippocampus. ments of the rates of glutamine synthesis and the tricarboxylic acid Epilepsia, 2002; 43(7): 703–710. cycle during oral and intravenous administration of [1-(13)C]glu- 125. Eid T, Thomas MJ, Spencer DD, Rundén-Pran E, Lai JCK, Malthankar GV, cose. Brain Res. Brain Res. Protoc. 2003; 10(3): 181–190.

956 Kim JH, Danbolt NC, Ottersen OP, de Lanerolle NC. Loss of gluta- 146. van de Ven KCC, van der Graaf M, Tack CJJ, Klomp DWJ, Heerschap A, 13 mine synthetase in the human epileptogenic hippocampus: de Galan BE. Optimized [1-(13)C]glucose infusion protocol for C

magnetic resonance spectroscopy at 3T of human brain glucose 150. Patel AB, de Graaf RA, Mason GF, Kanamatsu T, Rothman DL, metabolism under euglycemic and hypoglycemic conditions. J. Shulman RG, Behar KL. Glutamatergic neurotransmission and neu- Neurosci. Methods, 2010; 186(1): 68–71. ronal glucose oxidation are coupled during intense neuronal activa- 147. Sailasuta N, Tran TT, Harris KC, Ross BD. Swift Acetate Glial Assay tion. J. Cereb. Blood Flow Metab. 2004; 24(9): 972–985. (SAGA): an accelerated human (13)C MRS brain exam for clinical di- 151. Choi I-Y, Lei H, Gruetter R. Effect of deep pentobarbital anesthesia agnostic use. J. Magn. Reson. 2010; 207(2): 352–355. on neurotransmitter metabolism in vivo: on the correlation of total 148. Chhina N, Kuestermann E, Halliday J, Simpson LJ, Macdonald IA, glucose consumption with glutamatergic action. J. Cereb. Blood Bachelard HS, Morris PG. Measurement of human tricarboxylic Flow Metab. 2002; 22(11): 1343–1351. acid cycle rates during visual activation by (13)C magnetic reso- 152. Wang J, Jiang L, Jiang Y, Ma X, Chowdhury GMI, Mason GF. Regional nance spectroscopy. J. Neurosci. Res. 2001; 66(5): 737–746. metabolite levels and turnover in the awake rat brain under the 149. Henry P-G, Criego AB, Kumar A, Seaquist ER. Measurement of cere- inﬂuence of nicotine. J. Neurochem. 2010; 113(6): 1447–1458. bral oxidative glucose consumption in patients with type 1 diabetes 153. Yang J, Shen J. In vivo evidence for reduced cortical glutamate– mellitus and hypoglycemia unawareness using (13)C nuclear magnetic glutamine cycling in rats treated with the antidepressant/antipanic resonance spectroscopy. Metab. Clin. Exp. 2010; 59(1): 100–106. drug phenelzine. Neuroscience, 2005; 135(3): 927–937. 957