Progress in Nuclear Magnetic Resonance Spectroscopy 79 (2014) 14–47
Contents lists available at ScienceDirect
Progress in Nuclear Magnetic Resonance Spectroscopy
journal homepage: www.elsevier.com/locate/pnmrs
Review Sodium MRI: Methods and applications ⇑ Guillaume Madelin a, Jae-Seung Lee a,b, Ravinder R. Regatte a, Alexej Jerschow b, a New York University Langone Medical Center, Department of Radiology, Center for Biomedical Imaging, New York, NY 10016, USA b Chemistry Department, New York University, 100 Washington Square East, New York, NY 10003, USA
Edited by David Neuhaus and David Gadian article info abstract
Article history: Sodium NMR spectroscopy and MRI have become popular in recent years through the increased availabil- Received 12 February 2014 ity of high-field MRI scanners, advanced scanner hardware and improved methodology. Sodium MRI is Accepted 12 February 2014 being evaluated for stroke and tumor detection, for breast cancer studies, and for the assessment of Available online 7 March 2014 osteoarthritis and muscle and kidney functions, to name just a few. In this article, we aim to present an up-to-date review of the theoretical background, the methodology, the challenges, limitations, and Keywords: current and potential new applications of sodium MRI. Sodium MRI Ó 2014 Elsevier B.V. All rights reserved. Quadrupolar coupling Relaxation In vivo MRI Naþ–Kþ-ATPase Spherical tensors
Contents
1. Background...... 15 1.1. Introduction...... 15 1.2. Sodium in biological tissues ...... 16 1.3. Beginnings of biological sodium NMR ...... 16 1.4. Sodium NMR properties ...... 16 2. Theoretical formalism and pulse sequences ...... 17 2.1. Tensors and hamiltonians...... 17 2.1.1. Irreducible spherical tensor operators (ISTOs) ...... 17 2.1.2. The Zeeman term and the equilibrium density matrix ...... 18
2.1.3. The quadrupolar Hamiltonian ðHQÞ...... 18 2.1.4. The effect of a residual quadrupolar interaction ...... 18 2.2. Evolution and conversion between tensors ...... 19 2.3. Full solution of the evolution equations ...... 20 2.4. Biexponential decay of the free induction decay (FID) ...... 21 2.5. TQF buildup behavior ...... 21 2.6. Relaxation behavior of populations, third-rank tensor buildup after inversion pulse – IR-TQF experiment ...... 21 2.7. Analysis of biexponential decays and optimal settings...... 21 2.8. Notes on phase cycling and pulse flip angles ...... 21 2.9. Triexponential relaxation ...... 22 3. Summary of motional regimes and their effects on spectra ...... 22
3.1. Isotropic motion with motional narrowing ðx0sc 1Þ ...... 23 3.2. Isotropic motion in the intermediate to slow motion regime ðx0sc 1Þ; ðx0sc > 1Þ ...... 23
3.3. Partial alignment and slow motion ðx0sc > 1Þ ...... 23
⇑ Corresponding author. E-mail address: [email protected] (A. Jerschow). http://dx.doi.org/10.1016/j.pnmrs.2014.02.001 0079-6565/Ó 2014 Elsevier B.V. All rights reserved. G. Madelin et al. / Progress in Nuclear Magnetic Resonance Spectroscopy 79 (2014) 14–47 15
3.4. Compartment models ...... 23 3.5. Discrete Exchange Model (DEM) ...... 23 3.6. Berendsen-Edzes Model (BEM)...... 23 3.7. Other models ...... 24 4. Pulse sequences for compartment and fluid suppression ...... 24 4.1. Techniques based on shift reagents ...... 24 4.2. Diffusion ...... 24 4.3. Inversion recovery (IR) ...... 25 4.4. Multiple quantum filters...... 26 4.5. Optimal control pulse shape design...... 26 5. MRI implementation ...... 27 5.1. Image readout sequences ...... 27 5.2. Sodium quantification...... 27 6. Biomedical applications ...... 27 6.1. Brain...... 27 6.1.1. Stroke ...... 28 6.1.2. Tumors ...... 29 6.1.3. Multiple sclerosis ...... 30 6.1.4. Alzheimer’s disease (AD) ...... 32 6.1.5. Huntington’s disease ...... 32 6.2. Breast ...... 32 6.3. Heart ...... 32 6.4. Skeletal muscle ...... 33 6.4.1. Diabetes...... 33 6.4.2. Muscular channelopathy ...... 33 6.4.3. Myotonic dystrophy ...... 33 6.4.4. Hypertension...... 33 6.5. Cartilage...... 34 6.5.1. Osteoarthritis ...... 34 6.5.2. Cartilage repair ...... 35 6.5.3. Intervertebral disk ...... 35 6.6. Abdomen ...... 35 6.6.1. Kidney ...... 36 6.6.2. Prostate ...... 36 6.6.3. Uterus ...... 36 6.7. Whole body ...... 37 6.8. Cancer and chemotherapeutic response ...... 37 6.8.1. Chemotherapy assessment...... 37 6.8.2. Tumor resistance to therapy ...... 37 7. Conclusion ...... 38 Acknowledgements ...... 38 Appendix A.A.1. Subspace independent of quadrupolar coupling ...... 38 A.1.1. m ¼ 3 subspaces ...... 38 A.1.2. m ¼ 0 subspace...... 39 A.1.3. Example: inversion recovery ...... 39 A.2. Subspace depending on quadrupolar coupling ...... 40 A.2.1. m ¼ 1 subspaces ...... 40 A.3. Example: FID when fQ ¼ 0 ...... 41 A.4. m ¼ 2 subspaces ...... 41 A.5. Summary of decay constants ...... 41 A.6. Triple quantum filter coherence pathways, flip angle, and phase dependence...... 41 A.7. Double quantum filter coherence pathways, flip angle, and phase dependence...... 42 A.8. Summary of pulse phase choices in MQF sequences ...... 43 References ...... 43
1. Background example, in cartilage and in intervertebral disk tissue, sodium ions balance the negative charges of glycosaminoglycans, which 1.1. Introduction leads to an uptake of water into the tissue. The measurement of the sodium concentration directly in the tissues of interest Sodium is the most abundant cation present in the human or in vivo hence is of great interest for providing additional bio- body, and performs a number of vital body functions. The balance chemical information about both normal and abnormal body between intra- and extra-cellular sodium concentration is partic- functions. ularly crucial for insuring the proper functioning of a cell, and Although proton MRI has been highly successful, very often the disturbances of this balance are often signs of disorders (e.g. in information from standard MRI cannot provide direct biochemical stroke or in tumors). In addition, sodium takes part in such pro- markers for cell integrity and tissue viability, or for following cesses as nerve signal transmission and muscle action. A further changes in tissue viability upon treatment. Sodium MRI could pro- frequent function is the regulation of osmotic pressure. For vide some of this complementary information in a quantitative and 16 G. Madelin et al. / Progress in Nuclear Magnetic Resonance Spectroscopy 79 (2014) 14–47 non-invasive manner. A fact often overlooked is that 23Na also yields the second strongest nuclear magnetic resonance (NMR) signal among all nuclei present in biological tissues, after proton 1H spins. Due to the increase of the available magnetic fields for MRI scanners (1.5 T, 3 T, 7 T, 9.4 T), improved hardware capabilities such as strong gradient strengths with high slew rates, and new double-tuned radiofrequency (RF) coils, sodium MRI is now possi- ble within reasonable measurement times ( 10–15 min) with a resolution of a few millimeters and has already been applied in vivo in many human organs such as brain, cartilage, kidneys, heart, as well as in muscle and breast. The majority of current sodium MRI applications can be under- stood without much sophisticated theory. On the other hand, con- trast and quantification can in many situations be significantly affected by the underlying spin dynamics which are governed either by a residual or a fluctuating quadrupolar interaction, lead- + + ing to line splittings in the former, and to biexponential relaxation Fig. 1. Schematics of the sodium–potassium pump (Na /K -ATPase). in the latter case. In tissues, a complex mix of the two cases often persists, including the confounding factors of inhomogeneities and The Na+/K+-ATPase (also called sodium–potassium pump, or exchange. Taking full stock of all underlying phenomena and the just sodium pump) is present within the membrane of every ani- information content available from advanced methods both mal cell [19,20]. It is a plasma membrane-associated protein com- ex vivo and in vivo can allow one to extract further tissue parame- plex that is expressed in most eukaryotic cells, whose main ters and provide opportunities for new imaging contrast. function is to maintain the sodium and potassium gradients across In this review article we focus on regimes ranging from liquids the membrane. It therefore participates in the resting potential of to semi-solids, thereby specifically excluding applications in the the cell, by pumping three sodium ions out of the cell while pump- solid state. For applications in solids, we wish to refer the reader ing two potassium ions into the cell (see Fig. 1). to other review articles that cover those fields, their methodology The ion transport is performed against the electrochemical Na+ and applications [1–6]. + and K gradients existing across the cell membrane and therefore 23 Also, we wish to emphasize that Na NMR and MRI applica- requires energy provided by adenosine triphosphate (ATP) hydro- tions in semi-solids are certainly not limited to health-related lysis. This large electrochemical gradient is essential to protect fields, and studies have appeared in fields related to materials the cell from bursting as a result of osmotic swelling and also cre- science as well [7,8]. In this review article, however, we put an ates a potential that is used for transmitting nerve impulses and for emphasis on health-related and physiologically relevant pumping ions (such as protons, calcium, chloride, and phosphate) applications. and metabolites and nutrients (such as glucose and amino acids) A number of excellent review articles focusing on 23Na NMR in or neurotransmitters (such as glutamate) across the cell mem- semi-solids and 23Na MRI should be highlighted here as well brane. ATPase activity and ion transport are intimately linked [9–16]. + + and are two aspects of the same function. Regulation of Na /K - ATPase therefore plays a key role in the etiology of some patholog- ical processes. For example, when the demand for ATP exceeds its 1.2. Sodium in biological tissues + + production, the ATP supply for the Na /K -ATPase will be insuffi- cient to maintain the low intra-cellular sodium concentration Sodium is a vital component of the human body. It is an impor- and thus an increase of intra-cellular sodium concentration can tant electrolyte that helps maintain the homeostasis of the organ- be observed [18,19]. ism through osmoregulation (maintaining blood and body fluid volume) and pH regulation [17]. It is also involved in cell physiol- ogy through the regulation of the transmembrane electrochemical 1.3. Beginnings of biological sodium NMR gradient, thus partaking in heart activity, in the transmission of nerve impulses, and in muscle contractions. Sodium concentra- Biological tissues were already investigated with sodium NMR tions are very sensitive to changes in the metabolic state of tissues spectroscopy in the early 1970s [21,22] and with sodium MRI in and to the integrity of cell membranes. the early 1980s [23–25], first on animals in vivo and then on hu- The intra-cellular fraction makes up approximately 80% of the man brain [26] and human heart and abdomen [27]. Sodium MRI tissue volume with a sodium concentration of 10–15 mM, and was thereafter applied to brain tumor and ischemia detection in the extra-cellular volume fraction (including the vascular compart- the late 1980s [28]. In the 1990s there was an increase of interest ment) accounts for the rest, with a sodium concentration of 140– in sodium MRI, due to the increase of the magnetic fields in scan- 150 mM. Cells in healthy tissues maintain this large sodium con- ners, improvements of electronics and RF coils, and new rapid se- centration gradient between the intra-cellular and extra-cellular quences that allowed one to acquire sodium images within a few compartments across the cell membrane, and any impairment of minutes with millimeter resolution [29]. New contrasts such as tri- the energy metabolism or disruption of the cell membrane integ- ple quantum filtering [30–32] provided further promise of tissue rity leads to an increase of the intra-cellular sodium concentration. discrimination and characterization. This trend continued and The sodium flux in and out of cells can occur by several mecha- intensified through the 2000s until today. nisms. Examples of these include voltage- and ligand-gated Na+ channels, Na+/Ca+ exchangers (NCX), Na+/H+ exchangers (NHE), 1.4. Sodium NMR properties + + + Na /bicarbonate (HCO3 ) cotransporters, Na /K /2Cl cotransport- ers, Na+/Mg+ exchangers and most importantly the Na+/K+-ATPase The NMR-active isotope of sodium is 23Na with close to 100% [18]. natural abundance. Its Larmor frequency is approximately 5% G. Madelin et al. / Progress in Nuclear Magnetic Resonance Spectroscopy 79 (2014) 14–47 17
Table 1 selectively detected with multiple quantum filtered (MQF) se- Ranges of sodium concentrations and relaxation times in some human tissues in vivo quences [35], thereby providing additional opportunities for the (T and the fast and slow components of T , i.e. T , and T , respectively). These 1 2 2;fast 2;slow identification of intra-cellular sodium. ranges are based on Refs. [11,36–42] and references therein. WM – white matter, GM – grey matter, CSF – cerebrospinal fluid.
+ 2. Theoretical formalism and pulse sequences Tissue [Na ] (mM) T1 (ms) T2;fast (ms) T2;slow (ms)
Brain 2.1. Tensors and hamiltonians WM 20–60 15–35 0.8–3 15–30 GM 30–70 15–35 0.8–3 15–30 CSF 140–150 50–55 – 55–65 2.1.1. Irreducible spherical tensor operators (ISTOs) Cartilage 250–350 15–25 0.5–2.5 10–30 The ensemble spin-states of a spin 3/2 system can be described Blood 140–150 20–40 2–3 12–20 by a 4 4 density matrix Muscle 15–30 12–25 1.5–2.5 15–30 0 1 q11 q12 q13 q14 B C X B q q q q C B 21 22 23 24 C q ¼ðqijÞ¼ qijjiihjj¼@ A; q31 q32 q33 q34 13 23 ij larger than the one of C, and 26% of the proton frequency. Na q q q q has a nuclear spin of 3/2, and hence exhibits a nuclear quadrupolar 41 42 43 44 moment. The NMR sensitivity of sodium is 9.2% of the proton sen- where the spin-states jii2fj1i; j2i; j3i; j4ig correspond to the eigen- sitivity and the concentration in vivo is approximately 2000 times states of the z-component of the spin angular momentum operator, 3 1 1 3 lower than the water proton concentration. As a consequence, so- Iz, with the magnetic quantum numbers 2 ; 2 ; 2 ; 2, respectively. 1 dium MRI has on average a signal-to-noise (SNR) ratio which is While in spin-2 NMR it is common to use the Cartesian tensor oper- 3000–20,000 times lower than the proton MRI SNR (depending ators or product operators [43–45], irreducible spherical tensor on its concentration and visibility in organs or tissues). operators (ISTOs) are more suitable for expanding the matrix into The nuclear quadrupolar moment Q interacts with the electric individual components for a spin P 1=2. The ISTOs for the spin
field gradients (EFG) generated by the electronic distribution 3=2 case are made up of a set of 16 operators Tlm with rank around the nucleus [3,33,34]. In liquids this interaction is often l ¼ 0; 1; 2; 3 and order m ¼ l; ...; þl [46–48], allowing a complete averaged to zero. In the intermediate (semi-solid) regime, e.g. in and orthogonal expansion of the form biological tissues, the quadrupolar interaction results in a biexpo- X c T ; 1 nential relaxation behavior, and with anisotropic motion, line split- q ¼ lm lm ð Þ l;m tings may appear. The dominance of the quadrupolar relaxation mechanism for NMR signals can allow a sensitive characterization with clm being complex numbers. Several tensor normalizations are of the molecular environment of the sodium ions. A short T2 com- in use, which differ by a pre-factor which depends on l and the spin ponent T2;fast ¼ 0:5–5 ms generally contributes to 60% of the signal, number I. The most common normalization is the one listed in and a long component T 15–30 ms corresponds to 40% of the 2;slow ¼ [46,47]p,ffiffiffi which provides the spin-independent relationship signal. In order to detect both T2 components, imaging techniques I ¼ 2T1 1, and also allows one to couple tensors using Cle- with ultrashort echo times (UTE) of less than 0.5 ms are required. bsch–Gordan factors without additional conversion factors. With Typical ranges of sodium concentrations and relaxation times in this convention, the ISTOs have the matrix representations as listed some human tissues in vivo are given in Table 1. It has been shown in Table 2. The ISTOs can also be written in the form of linear that in the slow-motion regime, where biexponential relaxation is combinations of Cartesian or raising and lowering operators (Iz observed, multiple quantum coherences (MQC) can be created and and I ¼ Ix iIy) as shown in Table 3.
Table 2 m y Matrix expressions of the irreducible spherical tensors Tl;m in the jii basis. The expressions for Tl; m are found by Tl; m ¼ð 1Þ Tl;m, where y means hermitian conjugation (i.e. simple transposition in the case of real entries of these matrices). 0 1 1000 B C B 0100C T ¼ B C 0;0 @ 0010A 0001 0 qffiffi 1 0 1 3 00 0 0 3 00 2 B 2 ffiffiffi C B 1 C B p C B 0 2 00C B 00 2 0 C T ¼ B C T ¼ B qffiffi C 1;0 @ 1 A 1;1 B C 00 2 0 @ 3 A 00 0 2 00 0 3 2 00 0 0 0 qffiffi 1 3 000 B 2 C 0 pffiffiffi 1 0 pffiffiffi 1 B qffiffi C 0 3 00 00 3 0 B 3 C B C B pffiffiffi C B 0 2 00C B C B C B qffiffi C B 0000ffiffiffi C B 00 0 3 C T2;0 ¼ B C T2;1 ¼ @ p A T2;2 ¼ @ A B 3 C 0003 00 0 0 B 00 2 0 C @ qffiffi A 0000 00 0 0 3 00 0 2 0 1 0 1 3ffiffiffiffi 3ffiffiffiffi 0 1 0 1 p 000 0 p 00 3 3ffiffi 2 10 10 000 p B C B qffiffiffiffi C 00 2 0 2 B 9ffiffiffiffi C B C B C B C B 0 p 00C B 0033 0 C B 000 3 C B 000 0C T ¼ B 2 10 C T ¼ B 10 C T ¼ B 2 C T ¼ B C 3;0 B 9ffiffiffiffi C 3;1 B C 3;2 @ A 3;3 B C 00p 0 3 000 0 @ 000 0A @ 2 10 A @ 00 0 pffiffiffiffi A 10 000 p3ffiffiffiffi 000 0 000 0 2 10 00 0 0 18 G. Madelin et al. / Progress in Nuclear Magnetic Resonance Spectroscopy 79 (2014) 14–47
Table 3 jV zzj P jV yyj P jjV xx , and to define V zz ¼ eq (where e is the unit Relationship between irreducible spherical tensor (ISTO) and Cartesian operators electric charge, and q the field gradient per unit charge) and to de- (spin-independent). fine the anisotropy parameter ISTO Cartesian decomposition Description V xx V yy 6 T00 1 Identity g ¼ ; 0 g 1: ð6Þ V zz T10 Iz Longitudinal magnetization 1ffiffi T1 1 p I Rank 1 SQC 2 The quadrupolar interaction is then represented by the quadru- T20 p1ffiffi 3I2 IIðÞþ 1 Quadrupolar order polar Hamiltonian 6 z T 1 Rank 2 SQC 2 1 2 ½ Iz; I þ X2 2 M T2 2 1 I Rank 2 DQC H 2 Q ¼ xQ ðÞ 1 F2; mT2m; ð7Þ 3 T30 p1ffiffiffiffi Octupolar order m¼ 2 5Iz ðÞ3IIðÞ þ 1 1 Iz 10 qffiffiffiffihi T3 1 1 3 3 1 Rank 3 SQC 4 10 5Iz IIðÞ þ 1 2 ; I where the quadrupolar coupling angular frequency is qffiffihi þ T3 2 1 3 2 Rank 3 DQC 2 2 4 Iz; I e qQ þ xQ ¼ ; ð8Þ 3 T3 3 p1ffiffi I Rank 3 TQC 6h 2 2
SQC, DQC and TQC stand for single, double, and triple quantum coherences. with Q the nuclear quadrupole moment. The spatial tensors Flm in The anticommutator for the operators A and B is defined as ½A; B ¼ AB þ BA. the principal axis frame (PAS) of Vab take the form þ rffiffiffi 2.1.2. The Zeeman term and the equilibrium density matrix PAS 3 F2;0 ¼ ; The Zeeman Hamiltonian in angular frequency units (or units of 2 PAS ð9Þ J/h ) is given by F2; 1 ¼ 0; H PAS g z ¼ x0Iz ¼ x0T10; ð2Þ F ¼ : 2; 2 2 with x ¼ c B the Larmor angular frequency, c ¼ 70:808493 0 Na 0 Na The dependence of the interaction on orientation of the PAS 106 rad T 1 s 1 the sodium gyromagnetic ratio, and B the magnetic 0 with respect to the magnetic field can be described via field. The chemical shift range of 23Na is very small, and its contri- X bution is therefore ignored in this equation and for much of the rest l PAS Flm ¼ Dm0;mða; b; cÞFlm0 ð10Þ of this article (a notable exception is the discussion of dynamic fre- m0;m quency shifts and paramagnetic shifts which will be mentioned l using the Wigner rotation matrices ðD 0 Þ for the Euler angles later). m ;m a; b; c [49]. The Zeeman interaction is by far the most dominant interaction in most 23Na NMR applications, In this case, one can write the equi- 2.1.4. The effect of a residual quadrupolar interaction librium density matrix via the Boltzmann distribution as We may consider two cases here, with very similar outcomes:
expðh x0Iz=kTÞ 1 (1) HZ HQ and no motion or (2) partial averaging and downscal- qeq ¼ 1 þ h x0Iz=4kT; ð3Þ Z 4 ing of the quadrupolar interaction such that HZ hHQ i. In both cases one can truncate the quadrupolar Hamiltonian to the secular where Z ¼ Trfexpðh x0Iz=kTÞg; 1 is the identity matrix, and the last component, step is made by use of the high-temperature approximation. Since rffiffiffi 1 does not lead to any observables, and does not participate in hi H 3 xQ 2 the evolution, one typically omits this term and writes Q ¼ xQ T20 ¼ 3I IðI þ 1Þ : ð11Þ 2 2 z qeq ¼ aIz; ð4Þ This interaction then predicts the appearance of three reso-
nance lines with distance 3xQ =2p between them as shown sche- where the Boltzmann factor a ¼ h cB0=4kT is likewise frequently matically in Fig. 2 type-a. The relative intensity of the lines is omitted. 3:4:3, which originates from the observable signal as determined by TrfIy qg. The relative ratio then amounts to the relative values 2.1.3. The quadrupolar Hamiltonian ðH Þ Q of the squares of the off-diagonal elements of the T tensor. The quadrupolar interaction arises from the electrostatic energy 1; 1 The outer transitions are typically called the satellite transitions, term involving the charge distribution of the nucleus and the elec- and the inner one is called the central transition. tric field gradient around the nucleus [3,33,34]. The quadrupolar In the second case, where the quadrupolar coupling is the result term is the first non-vanishing term in a series expansion, with of some downscaled coupling via reorientation motion, x would higher order terms (hexadecapole) being negligibly small. Q represent the temporally and spatially averaged coupling constant Since the nuclear quadrupole moment can be assumed to be hx F i in Eq. (11). Strictly speaking, x would also contain the constant, the nuclear quadrupole interaction is determined by Q 2;0 qffiffiffiffiffiffiffiffiffiffiffiffiffiQ g2 the orientation, magnitude and temporal average of the electric term from the anisotropy factor, 1 þ 3 , but since this factor field gradient (EFG) generated by the electronic configuration of can rarely be measured independently in motionally averaged sys- the molecular environment surrounding the nucleus. If Vðx; y; zÞ tems, we will not carry it along here. is the electrostatic potential produced by the electrons at the point In solids, situations may arise when the secular approximation ðx; y; zÞ, the EFG can be described by the tensor Vab with the is no longer justified. It is common to include the second order per- components: turbation term in the overall Hamiltonian. We neglect this compo- 2 nent here, as this case is not relevant to tissues or semi-solids as @ V V ¼ ; ðwith a; b ¼ x; y; or zÞ: ð5Þ considered here. ab @a@b Fig. 2 shows different motional regimes that can be observed,
If we choose the principal axes of the symmetric tensor Vab as and their ensuing spectra. The effect described here corresponds coordinate axes, the cross-terms vanish, i.e. V xy ¼ V xz ¼ V yz ¼ 0. It to the ‘type-a’ category as classified by Rooney and Springer [12– is common to label the remaining components such that 14]. ‘Type-b’ spectra arise when the sample is inhomogeneous G. Madelin et al. / Progress in Nuclear Magnetic Resonance Spectroscopy 79 (2014) 14–47 19 Wiley-Blackwell 1991 Ó
Fig. 2. Typical energy levels and corresponding NMR spectra of sodium nuclei in different environments. Four types of motionally narrowed SQ spectra are possible (a, b, c, and d). The type-a (crystal-like) and type-b (powder-like) spectra are shown of Na+ in aqueous suspensions of oriented and unoriented dodecyl sulfate micelles, respectively. The type-c (homogeneous, biexponential, super-lorentzian-like) spectrum is that of Na+ in an aqueous solution that has a high concentration of micelle-solubilized gramicidin channels. The type-d (liquid-like) spectrum is from NaCl in H2O. We can make the correspondence xQ ¼ 3xQ in type-a. Reproduced with permission from Ref. [12]. and contains regions with a distribution of quadrupolar couplings, 2. The quadrupolar coupling Hamiltonian can only change the leading to a distribution of level shifts as indicated in Fig. 2. The ef- rank l. fect of a fluctuating quadrupolar interaction is illustrated with the 3. Quadrupolar relaxation can only change the rank l. ‘type-c’ regime and leads to biexponential relaxation as discussed later. Finally, if the fluctuations are very fast, or in the absence of After a 90y pulse acting on equilibrium magnetization, one ob- a quadrupolar interaction one recovers the spectrum of ‘type-d’ tains I p1ffiffi T T . A constant residual quadrupolar cou- q ¼ x ¼ 2 ð 1; 1 1;1Þ where all transitions are equivalent. In such a case, the spin transi- pling (type-a) would lead to evolution within the following tions do not differ from those of a spin 1/2 nucleus. subspaces
2.2. Evolution and conversion between tensors fT1;1; T2;1; T3;1g and fT1; 1; T2; 1; T3; 1gð13Þ
due to the commutation relationships In this section we first provide a general overview of the effects ffiffiffi and conversions that typically occur in experiments, and provide a p ½T2;0; T1; 1 ¼ 3T2; 1; full solution later. The evolution of a 23Na nuclear spin can be de- pffiffiffi scribed by the master equation given as 6 3 4ffiffiffi ½T2;0; T2; 1 ¼ T1; 1 þ p T3; 1; 5 5 d b H b th qðtÞ¼ i½ ; qðtÞ CfqðtÞ q g; ð12Þ 3ffiffiffi dt ½T2;0; T3; 1 ¼ p T2; 1: ð14Þ 5 where th is the density operator in the thermal equilibrium state, q b H the spin Hamiltonian, and Cb the relaxation superoperator. On the other hand, if only quadrupolar relaxation were active, The total Hamiltonian H in the rotating frame can for most the double-commutators ½ T2;0; ½ T2;0; q , ½ T2; 1; ½ T2; 1; q , and 23 ½ T2; 2; ½ T2; 2; q in the Redfield relaxation expressions that make practical purposes in Na NMR of tissues and liquids be assumed b up C can only connect operators within the following subspaces to be equal to HQ (one may wish to allow for inhomogeneous broadening, in which case a distribution of HZ terms would be fT1;1; T3;1g; fT1; 1; T3; 1g; fT1;0; T3;0g: ð15Þ added).
In order to follow the sequence of events in experiments at high Hence, in isotropic liquids only odd rank tensors T3 1 can be formed magnetic fields, one can follow these simple rules: [3,48]. under the influence of quadrupolar relaxation. In anisotropic media, where the residual quadrupolar interaction does not average to
1. A non-selective radiofrequency (RF) pulse changes the coher- zero, rank 2 tensors T2 1 can be also formed. ence order m within the limits of jjm 6 l. The amplitude induced A multiple-quantum filtered (MQF) experiment can be used to by the rotation is given by the corresponding Wigner rotation distinguish between the two cases [35,50–52]. For example, the
element, so a rotation with flip angle h and phase / of Tl;m to T2; 1 term generated in the first case can be converted to a 0 l Tl;m0 gives rise to a factor exp ðÞ i/ðm mÞ dm;m0 ðhÞ, where double-quantum coherence (DQC) term T2; 2 by another pulse, l dm;m0 ðhÞ is the reduced Wigner rotation matrix element. while this term is not available due to relaxation alone. The 20 G. Madelin et al. / Progress in Nuclear Magnetic Resonance Spectroscopy 79 (2014) 14–47
2.3. Full solution of the evolution equations
In this section we provide the complete solution of the evolu- tion equation, Eq. (12), based on residual quadrupolar coupling of the form of Eq. (11) and a fluctuating quadrupolar interaction of the form of Eq. (7) with the definitions in Eqs. (8) and (9) as the relaxation mechanism. Using the Redfield formalism, one can write the action of the relaxation superoperator as
ÈÉX2 ÂÃÂÃ b th m th C q q ¼ A ð 1Þ T2;m; T2; m; q q ðÞJðmxÞ iKðmxÞ ; m¼ 2 ð16Þ
where JðxÞ is the spectral density function, and iKðxÞ is the imag- inary portion arising in the Fourier Transform from the integration of the autocorrelation function from 0 to 1 (rather than from 1 to 1) [33], and A a constant used for conversion between different conventions as explained below. The K term gives rise to a small fre- quency shift, called the dynamic frequency shift. For the typical case of exponential autocorrelation functions, the relationship
KðxÞ¼xscJðxÞ holds. The shift is rarely observable [57], since it Fig. 3. Multiple quantum filter (MQF) pulse sequence. For the triple quantum filter is small (typically within the linewidth of the transitions [58]). (TQF), h1 ¼ h2 ¼ h3 ¼ 90 , and a typical phase cycle is /1 ¼ 30 ; 90 ; 150 ; 210 ; We therefore neglect it for the rest of this treatment. 270 ; 330 ; /2 ¼ /1; /3 ¼ 90 , /acq ¼ 0; 180 . For the double quantum filter (DQF), The convention used in our article is characterized by A ¼ 1 and h h h 90 , and a typical phase cycle is / 0 ; 90 ; 180 ; 270 , 1 ¼ 2 ¼ 3 ¼ 1 ¼ has the advantage that all quadrupolar-coupling-related expres- /2 ¼ /1 90 ; /3 ¼ 90 ; /acq ¼ 0; 180 . For DQF with magic angle pulses (DQF- sions are included in JðxÞ. As a result, the expressions as listed be- MA), h1 ¼ 90 ; h2 ¼ h3 ¼ 54:7 , and the phase cycling is the same as for DQF. For the IR-TQF sequence, h1 ¼ 180 ; h2 ¼ h3 ¼ 90 , and no additional 180 pulse is used. The low can be used directly for all fluctuation regimes, including phase cycle is /1 ¼ 30 ; 90 ; 150 ; 210 ; 270 ; 330 ; /2 ¼ /1 90 ; /3 ¼ 90 , direct fluctuations of electric field gradients, for example due to /acq ¼ 0; 180 . The delay d is kept as short as possible to avoid signal decay. These ion solvation or due to motion near a polyelectrolyte. For ease of minimal phase cycles could be augmented by an independent 2-step cycle on the comparison between different conventions, we list here the follow- first pulse (with an accompanying 0; 180 cycle on the receiver), and a 2-step cycle on the 180 pulse, giving a 12 or a 24-step cycle for the TQF/IR-TQF experiments, ing definitions of the spectral density functions and conversion fac- and a 8 or 16-step cycle for the DQF. tors A for an isotropic tumbling model: Convention I; ð17Þ A ¼ 1; ð18Þ corresponding pulse sequence is shown in Fig. 3, along with the 2 2 different phase cycles that would be used for the experiments. ð2pÞ C sc JðxÞ¼ Q : ð19Þ One can also separate the regimes to some degree using a triple- 2 60 1 þðxscÞ quantum filtered (TQF) experiment by a judicious choice of pulse flip angles, as discussed in more detail below. Here CQ is the ‘quadrupole coupling constant’ and has the usual def- 2 Since multiple quantum coherences (MQCs) cannot be detected inition qCQffiffiffiffiffiffiffiffiffiffiffiffiffi¼ e qQ=h (as noted before, the term from the anisotropy directly, the final RF pulse converts the MQCs into single quantum g2 factor, 1 þ 3 , would be contained in here as well, which we drop coherences (SQCs), T ; T ; T , which then evolve under the ac- 1 1 2 1 3 1 for simplicity). Convention I is the one used in this article, as well as, tion of relaxation and the residual quadrupolar interaction (RQI) for example in Refs. [10,59–61]. into detectable SQC (T ) during the acquisition time t .A 1; 1 acq Other conventions are in use, of which two popular ones are 180 RF pulse can be applied in the middle of the preparation time s in order to refocus any chemical shifts, or more typically mag- Convention II; ð20Þ netic field inhomogeneities. Since MQFs are very sensitive to RF 1 A ¼ ; ð21Þ imperfections, it has been shown that filters without this refocus- 3 ing pulse generate a better signal and are more robust to RF 2 2 ð2pÞ C sc inhomogeneities for in vivo MQF MRI [53]. Recently, procedures Q JðxÞ¼ 2 ð22Þ 20 1 þðxscÞ have been published for avoiding B0 inhomogeneities by separately acquiring and storing the individual phase cycling steps and used for example in Refs. [48,62], and subsequent post-processing before co-adding the individual sig- nals [54–56]. Specific absorption rate (SAR) considerations also fa- Convention III; ð23Þ vor the sequence without the 180° pulse. The formation of DQC 1 ð2pÞ2 A ¼ C2 ; ð24Þ and TQC is detected separately by choosing the appropriate RF 3 40 Q pulse flip angles and phase cycling. Typical phase cycles and 2sc choices of RF pulse flip angles are given in the caption of Fig. 3. JðxÞ¼ 2 ; ð25Þ 1 þðxscÞ DQF can detect the contribution of two tensors, T2 2; T3 2, due to the slow motion regime or/and RQI in anisotropic media. If which is used for example in [35,63]. Note that for C given in Refs. h1 ¼ h2 ¼ h3 ¼ 90 , contributions from both T2 1 and T3 1 can be [35,63], the relationship C ¼ 3A holds. detected. If h1 ¼ 90 and h2 ¼ h3 ¼ 54:7 (magic angle), only the Closed solutions can be obtained by using the tensor operator contribution of T2 2 is detected, which arises only from the aniso- basis or the single transition basis. There is no obvious advantage tropic RQI. Only T3 3 due to the slow motion regime can be of one vs. the other. Eventually, it is most useful to represent the detected by the TQF sequence. final evolution expressions in the form of spherical tensor G. Madelin et al. / Progress in Nuclear Magnetic Resonance Spectroscopy 79 (2014) 14–47 21
Table 4 2.6. Relaxation behavior of populations, third-rank tensor buildup Conversion amplitudes in the equation of motion, Eq. (12), from a tensor T to a l2 ;m after inversion pulse – IR-TQF experiment tensor Tl1 ;m. The expressions for the negative orders m are identical except for complex conjugation. For brevity, we use the notation Jm ¼ Jðmx0Þ here. One can observe the relaxation behavior of the Iz-term by fol- m l1 l2 Amplitude lowing the populations after a 180° inversion pulse as a function of 011 6 ðJ þ 4J Þ the delay . The Zeeman term also has a biexponential relaxation 5 q1 ffiffi 2 s 013 18 2 behavior according to Eq. (A.11): 5 5ðJ1 J2Þ
022 6ðJ1 þ J2Þ R s R s 6Jðx Þs 6Jð2x Þs qffiffi e 1;fast þ 4e 1;slow ¼ e 0 þ 4e 0 : ð28Þ 031 2 4 5ðJ1 J2Þ 033 6 In an IR-TQF (Fig. 3) experiment, one first generates the non- 5 ð4J1 þ J2Þ equilibrium state q ¼ Iz ¼ T10 by a 180° pulse and then the 111 3 ð3J þ 5J þ 2J Þ 5 pffiffiffi0 1 2 third-rank component builds up as a function of s, which can sub- 112 9 i 2x 5 qffiffi Q sequently be filtered out by a TQF element. The T term builds up 113 30 9 3ðJ J Þ 5 5 0 2 according to Eq. (A.11) with an amplitude proportional to 121 p3ffiffi ix 2 Q R s R s 6Jðx Þs 6Jð2x Þs e 1;fast e 1;slow e 0 e 0 : 29 122 3ðqJ0 ffiffiffiffiþ J1 þ 2J2Þ ¼ ð Þ 123 3 3i xQ qffiffi10 The buildup of the third rank component can again be thought 131 2 3ðJ J Þ of as arising from the faster relaxation of the populations in the q5ffiffi 0 2 132 6 outer energy levels due to larger fluctuations. 2i 5xQ 133 3 5 ð2J0 þ 5J1 þ 3J2Þ 2.7. Analysis of biexponential decays and optimal settings 222 3ðJ þ 2J þ J Þ p0 ffiffiffi 1 2 3 223 i 3xQ 2pffiffiffi The buildup of the third-rank component as measured by the 232 2i 3xQ TQF and IR-TQF experiments can be analyzed as illustrated in 233 3ðJ0 þ J2Þ Fig. 4. Fig. 4b illustrates the optimal setting for the delay s based 333 3ðJ1 þ J2Þ on the two relaxation rates R2 and R1 in the curve. Fig. 4c can be used to determine the maximum achievable signal at the optimal delay setting. It is clear that this signal goes to zero as the relaxa- transformations. The results are easy to interconvert from one tion rates become equal to each other, indicating that no third-rank basis to another. The right hand side of Eq. (12) can be evaluated components were built up. Fig. 4d shows how the maximum in the tensor basis as summarized in Table 4. Note, that these achievable signal depends on the ratios Jðx0Þ=Jð0Þ and results are completely equivalent to those given in other works Jð2x0Þ=Jð0Þ in the TQF experiment. The largest amplitudes are [48,62] when considering that there Convention II for scaling achieved in the cases where the ratios go to zero (for example in JðxÞ, and normalized tensors withÂÃ symmetric and antisymmetric the very slow motion regime, as discussed below). linear combinations (T s p1ffiffi T T , and T a p1ffiffi ½T lmð Þ¼ 2 lm þ l; m lmð Þ¼ 2 lm Tl; m ) were used. From these expressions, one can then derive full 2.8. Notes on phase cycling and pulse flip angles analytical solutions to the evolution for a number of special cases. The complete solution is given in Appendix A, and we highlight One can distinguish between the TQF, IR-TQF, and DQF experi- some of the main results in the text below. ments on the basis of a suitable combination of flip angles with appropriate phase cycles [15,64,65]. At the core of the selection 2.4. Biexponential decay of the free induction decay (FID) process is a 6-step cycle (with 60° intervals) for the first two pulses in the TQF/IR-TQF experiments (selecting triple quantum coher- The free-induction decay after single pulse excitation in the ab- ences), and a 4-step cycle with 90 intervals for the DQF experi- sence of quadrupolar coupling (fQ ¼ 0) can be found from Eq. ment (selecting double quantum coherences), as illustrated in (A.29) to be biexponential Fig. 3. For the TQF experiments, the flip angles of the second and third 2 R t 3 R t 2 ½3Jðx Þþ3Jð2x Þ t 3 ½3Jð0Þþ3Jðx Þ t e 2;slow þ e 2;fast ¼ e 0 0 þ e 0 : ð26Þ pulses should be 90° to maximize the signal intensities (see a com- 5 5 5 5 plete list of the flip-angle dependencies of different pathways in The origin of the biexponential relaxation can be understood as Appendix A). For the DQF experiment, suppose that the coherence arising from the faster relaxation of the satellite transitions pathways involving the second-rank tensors T2; 1 need to be exclu- compared to the central transition, because they experience larger sively selected, the existence of which will show whether there is a fluctuations from the instantaneous quadrupolar interactions. residual quadrupolar coupling or not. By setting the flip angle of either the second or the third pulse (h2 and/or h3) to the magic angle, the coherence pathways involving the third-rank tensors 2.5. TQF buildup behavior T3; 2 will be suppressed, since they have a sin h2 sin h3ð1 2 2 3 cos h2Þð1 3 cos h3Þ dependence (see Appendix A). By changing According to Eq. (A.29), a free evolution delay s after a single both flip angles from 90° to the magic angle, the signal from the pulse gives rise to the appearance of a third-rank tensor with an coherence pathways involving the second-rank tensors T2; 1 is amplitude proportional to reduced by sin h2 sin h3. Such a sequence is also typically called