Layer-specific variation of iron content in as a source of MRI contrast

Masaki Fukunagaa,1, Tie-Qiang Lia, Peter van Gelderena, Jacco A. de Zwarta, Karin Shmuelia, Bing Yaoa, Jongho Leea, Dragan Maricb, Maria A. Aronovac, Guofeng Zhangc, Richard D. Leapmanc, John F. Schenckd, Hellmut Merklea, and Jeff H. Duyna

aAdvanced MRI, Laboratory of Functional and Molecular Imaging, National Institute of Neurological Disorders and Stroke, National Institutes of Health, Bethesda, MD 20892; bLaboratory of Neurophysiology, National Institute of Neurological Disorders and Stroke, National Institutes of Health, Bethesda, MD 20892; cLaboratory of Bioengineering and Physical Science, National Institute of Biomedical Imaging and Bioengineering, National Institutes of Health, Bethesda, MD 20892; and dGeneral Electric Global Research Center, Niskayuna, NY 12309

Edited* by Leslie G. Ungerleider, National Institute of Mental Health, Bethesda, MD, and approved January 4, 2010 (received for review September 28, 2009) Recent advances in high-field MRI have dramatically improved the and is, together with the resonance frequency, particularly sen- visualization of human brain anatomy in vivo. Most notably, in sitive to spatial magnetic susceptibility variations. Most notably, cortical gray matter, strong contrast variations have been observed contrast based on resonance frequency shifts of tissue, derived that appear to reflect the local laminar architecture. This contrast from the phase of gradient-echo signals, has allowed the robust has been attributed to subtle variations in the magnetic properties visualization of laminar structure in a number of neocortical of brain tissue, possibly reflecting varying iron and content. regions, including primary visual and motor cortices (17) (Fig. 1). To establish the origin of this contrast, MRI data from postmortem Despite this progress, the origin of cortical laminar contrast brain samples were compared with electron microscopy and observed with high-field MRI remains unclear. What causes the histological staining for iron and myelin. The results show that iron magnetic properties to vary over cortical layers? Both iron and is distributed over laminae in a pattern that is suggestive of each myelin have been suggested to alter the magnetic susceptibility region’s myeloarchitecture and forms the dominant source of the (17, 19, 20), but experimental confirmation is lacking. In occipital observed MRI contrast. cortex, the laminar pattern observed with MRI is consistent with the known myelin distribution (17), which would support a myelin | ferritin | laminar variation | brain structure | magnetic susceptibility myelin-dominated contrast mechanism. However, the suscepti- bility differences seen in this region are opposite to what is uch of the human cerebral cortex is organized in a network expected for myelin-rich regions based on the contrast differ- Mof functionally specialized regions. This understanding ences between and cortex and therefore argue forms the basis of studies that attempt to expose the brain’s inner against the notion of a myelin-dominated contrast mechanism workings using modern functional imaging techniques such as (17). Conversely, if iron was the main contributor to the observed PET and MRI (1, 2). Some of this functional specialization is contrast, this would suggest the intriguing possibility of a laminar reflected in the local cortical architecture, and may be observed variation of iron, which, like the myelin distribution, may be in postmortem brain samples from the laminar and columnar indicative of cortical functional specialization. variation in a number of tissue properties including cell shape To investigate the possibility of a laminar variation of iron, we (3), myelin content (4), metabolic state (5), neurochemical pro- used a variety of imaging methods, including MRI, histochemical fl file (6), and receptor density (7). and immuno uorescence staining, and electron microscopy. We One of the most striking examples of this structure–function demonstrate that, in the human occipital cortex, (i) there is relationship is in the primary (V1, or Brodmann indeed a laminar variation of iron that resembles the frequency fi ii area 17), which is involved in the early stages of visual infor- contrast seen in recent high- eld MRI studies; ( ) this iron is a iii mation processing. Visual area V1 can be readily identified with dominant contributor to R2* and frequency contrast; and ( ) postmortem tissue analysis based on the prominence of myeli- much of this iron appears to be stored within ferritin. nated fibers in layer IVb, known as the line of Gennari. Many Results other brain areas show characteristic myelination patterns (8, 9), and systematic analyses of this type of structural information may Cortical Iron Distribution Is Consistent with MRI Contrast. Although provide important clues about the brain’s functional organ- it has been recognized that iron is responsible for much of the ization. Whole-brain postmortem studies of cellular distributions MRI susceptibility-based contrast in iron-rich brain regions such as the subcortical nuclei (19–22), very little is known about the and myelin content in human brain sections at high resolution ’ have previously been described (8, 10). element s distribution in the cortex or its possible role in the A number of attempts have been made to reveal laminar laminar variation of MRI contrast. To explore the potential cortical structure in vivo with structural MRI, by exploiting contribution of iron in MRI, correlative MRI with histochemical contrast based on the altered density and NMR relaxation times staining was performed on postmortem human brain tissue. Both of water protons in myelin-rich environments (11–15). Although iron and myelin staining were performed on a coronal slab, these studies have had some success, their impact has not been sectioned from the containing a substantial portion widespread as a result of the rather limited contrast and reso- lution available with conventional MRI. Nevertheless, much Author contributions: M.F., T.-Q.L., and J.H.D. designed research; M.F., T.-Q.L., P.v.G., J.A.d. progress has been made over the last few years. Recent devel- Z., K.S., B.Y., J.L., D.M., M.A.A., G.Z., R.D.L., J.S., H.M., and J.H.D. performed research; P.v.G. opments in high-field MRI (≥7 T) have increased spatial reso- contributed new reagents/analytic tools; M.F., T.-Q.L., and J.H.D. analyzed data; and M.F. lution to less than 300 μm (16–18), and improved the ability to and J.H.D. wrote the paper. detect subtle variations in the magnetic susceptibility of tissue. The authors declare no conflict of interest. These variations are reflected in a number of intrinsic MRI *This Direct Submission article had a prearranged editor. parameters, including the longitudinal relaxation rate R1, the 1To whom correspondence should be addressed. E-mail: [email protected]. transverse relaxation rates R2 and R2*, and the NMR resonance This article contains supporting information online at www.pnas.org/cgi/content/full/ frequency. R2* indicates the reversible transverse relaxation rate 0911177107/DCSupplemental.

3834–3839 | PNAS | February 23, 2010 | vol. 107 | no. 8 www.pnas.org/cgi/doi/10.1073/pnas.0911177107 Downloaded by guest on September 27, 2021 AB myelin sheets surrounding nerve fibers (26, 27). In fact, higher magnification views in histochemical stains (Fig. S2) suggest that R2* weighted much of the iron is associated with axonal fibers. Comparison of iron and myelin stains obtained in this study suggests that this colocalization occurs not only in white matter fibers, but also in intracortical fibers. This is consistent with earlier findings (25) and furthermore suggests that myelin-associated iron is the dominant source of intracortical magnetic susceptibility–based contrast. To confirm that the observed iron is indicative of ferritin, we frequency -8 Hz 8 Hz performed double immunofluorescence staining with antibodies to myelin basic protein (MBP) and ferritin storage protein. The results (Fig. 3) clearly show a widespread colocalization of ferritin and myelinated fibers in both gray and white matter similar to what was observed for iron distribution patterns using histochemical stain.

Ferritin Iron Content Is Sufficient to Explain MRI Contrast. Although the data presented here are consistent with the notion that ferritin Fig. 1. Example of in vivo MRI contrast in the occipital lobe. (A) Localizer is the primary source of the observed intracortical MRI contrast, it image shows region of interest for (B). (B)R2* weighted magnitude and frequency images. The frequency image reflecting resonance frequency remains unclear whether the magnetic susceptibility of ferritin is shifts. Intracortical contrast is particularly strong in the frequency image, sufficient to explain the MRI data. To clarify this, one would need revealing the centrally located Line of Gennari (arrows). to quantify both the concentration as well as the average iron loading of ferritin storage proteins. In previous work, the average gray matter content of ferritin-bound iron has been reported to be of V1 as well as parts of the secondary visual cortex (V2), and a in the range of 30 to 50 μg/g tissue (wet weight) (28), but infor- section of the superior parietal cortex, respectively. mation about the laminar distribution is lacking. As can be seen from the results (Fig. 2 and Fig. S1), MRI con- To investigate laminar differences, we performed transmission trast and iron staining show strong similarities. Areas of increased EM (TEM) on brain samples obtained from regions within the line R2* relaxation in the white matter show increased iron staining. In of Gennari and more superficial layers (layers II–III; Fig. 4). Data NEUROSCIENCE the gray matter, a laminar distribution of iron is observed with a from 100 fields (700 × 700 × 100 nm) showed average iron particle pattern that resembles the MRI contrast. Moreover, the laminar concentrations of 351 ± 27 and 115 ± 51 (mean ± SD) for these two iron staining pattern exhibits strong similarities with that of myelin. regions, respectively. Furthermore, analysis of electron energy loss In the MRI of visual cortex tissue, much of the cortex surrounding spectra (EELS) suggested an average loading factor of 1,740 ± 580 fi the calcarine ssure shows a central stripe resembling the line of Fe atoms per particle (n = 15; Figs. S3 and S4), which is within the Gennari in striate cortex (i.e., V1). Furthermore, iron is also range of literature values for samples of spleen and liver ferritin increased in subregions of the deeper cortical layers and in cortical particles (29–31). Based on these average values, one can calculate gray matter around the V1/V2 border. This apparent corre- the laminar iron content to average 50 ± 4 and 16 ± 7 μg iron per spondence between MRI and histology was also found in the gram tissue (wet weight) in the line of Gennari and superficial superior parietal cortex. These results suggest the possibility that Materials and Methods fl layers, respectively ( ). Estimates based on not only the distribution of myelin but also that of iron may re ect theory and previous experimental work (Materials and Methods) the functional subdivision of the . indicate that this variation can lead to frequency and R2* shifts in the line of Gennari of up to 5 and 13 Hz, respectively, which are Iron Colocalizes with Intracortical Fibers. The similarity between the sufficient to explain the in vivo results (17) and those obtained in cortical iron and myelin distribution seen in the histochemical tissue samples (as detailed later). stains could reflect a microscopic colocalization at the cellular and molecular levels. Earlier work has found that, in white matter, Reduction of MRI Contrast After Chemical Extraction of Tissue Iron. much of the iron is found in the form of ferritin particles dispersed fi – Having identi ed iron as a plausible source of intracortical MRI in the inner and outer loops of myelin (23 25). This ferritin is contrast from correlative analysis, can we establish a direct link? thought to constitute a storage pool that primarily serves the iron One way to address this question is to selectively manipulate tissue needs of oligodendrocytes for the production and maintenance of iron content and observe the effects on MRI contrast. Using a biochemical process to extract iron from postmortem tissue, Schenck and coworkers previously found that iron substantially myelin iron R2* 0 Hz 100 Hz affects R2* relaxation in much of the brain (22). Applying their protocol to primary visual cortex, we observed an almost complete V2 * * * extinction of intracortical R2* and frequency contrasts (Fig. 5 and V1 Table 1): these went from 13.1/0.80 Hz before extraction to 2.7/ 0.02 Hz, respectively, after extraction. This finding solidifies the notion that intracortical contrast is predominantly caused by var- * * * V2 iations in tissue iron content. Note that some R2* and frequency contrast remains after iron extraction, which is particularly the Fig. 2. Comparison of histochemical myelin and iron staining with MRI R2* case for gray–white matter contrast. One explanation would be an data in the visual cortex. The myelin stain shows the characteristic density ineffective iron extraction. However this was not evident from the increase in the line of Gennari in the pericalcarine cortex (solid arrow). The comparison of iron stains of tissue without and with extraction, V1/V2 boundaries are indicated by asterisks. A dashed line represents the fi which showed a complete elimination of intracortical iron varia- calcarine ssure. The distribution of intracortical iron mimics that of myelin, – with elevated iron in the line of Gennari. The MRI data show a striking tions by extraction and a small remaining gray white matter dif-

similarity with the iron stain, with increased R2* in the line of Gennari (solid ference (Fig. S5). Therefore, it appears more likely that the arrow), and in the deeper layers (open arrow), and subcortical white matter remaining R2* and frequency contrast is primarily caused by other in area V2 (arrowhead). factors such as variations in tissue myelin content, and it is possible

Fukunaga et al. PNAS | February 23, 2010 | vol. 107 | no. 8 | 3835 Downloaded by guest on September 27, 2021 A myelin ferritin composite B myelin ferritin composite sGM

sGM r=0.68 Gennari

Gennari

dGM r=0.63 dGM

r=0.77 WM

WM r=0.38

Fig. 3. Cellular colocalization of myelin and ferritin in the primary visual cortex. (A) Immunofluorescence costaining of MBP and ferritin protein shows strong colocalization in intracortical fibers on composite images (yellow in composite). The dashed lines represent gray/white matter boundaries. (Scale bar: 500 μm.) (B) Enlarged areas of superficial layers in the gray matter (sGM), the line of Gennari, deeper layers in the gray matter (dGM), and white matter (WM) all show widespread colocalization of MBP and ferritin. (Scale bar: 50 μm.) Image correlation between myelin and ferritin stains is indicated by r values.

that iron does not generally dominate the contrast outside the tomical marker for the myeloarchitecture of the human cerebral cerebral cortex. cortex. Third, it suggests that high-field susceptibility-based MRI Are the reductions in R2* with iron extraction in the various contrast may be used to reveal the local cortical architecture brain regions consistent with their putative iron concentrations? based on iron distribution, and may therefore provide improved Based on previous work correlating iron histology and MRI in contrast over conventional T1-weighted techniques, which are the basal ganglia (32), the observed reductions in R2* values primarily sensitized to myelin variations. (ranging from 11.5 Hz in gray matter to 21.9 Hz in Gennari The importance of iron for MRI magnetic susceptibility–based region) correspond to an iron content ranging from 30 to 63 μg/g contrast (including R2*, R2, and frequency shift) has been pre- (tissue wet weight) (32). This range is consistent with the pre- viously suggested for iron-rich gray matter areas such as the basal viously established iron concentration in human occipital cortex ganglia and for selected cortical regions (reviewed in ref. 20) and of 45.5 ± 6.7 μg/g (28), further lending support for an iron- was confirmed with an iron extraction study (22). However, con- dominated contrast mechanism. tributions from other sources have prevented the quantification of tissue iron content directly from the MRI measures. Most notably, Discussion the relationship between MRI frequency shift and iron content The experiments described in this work establish a direct link was found to be ambiguous within white matter (19). Some of the between intracortical iron variations and susceptibility-based iron-induced shifts may be offset by an opposing (i.e., diamagnetic) contrast observed in high-field MRI. This is important in several susceptibility shift originating from lipids in myelin (17, 19), which ways. First, it helps interpret the recent advances made in are abundant in white matter (33). A diamagnetic susceptibility detecting the laminar architecture of human brain in vivo with shift induced by myelin is consistent with the apparent frequency high field MRI. Second, it suggests iron as an additional ana- contrast reversal observed in the myelin-rich Gennari region after

A B 15 CK Fe L2,3 ) 4 10

5 Counts (x 10

0 300 400 500 600 700 800 Energy Loss (eV) C

Fig. 4. Distribution of individual ferritin particles measured by EM. (A) Sample bright-field TEM image shows scattered foci of reduced signal intensity, presumably originating from ferritin particles (arrow). (B) EELS spectrum of a single particle confirms iron as source of TEM contrast. (C) STEM-EELS spec- troscopic image (3.5 nm/pixel) of iron suggests particle sizes of 1 to 2 pixels, consistent with the range of 3 to 8 nm reported for the ferritin core size (13). (Scale bars: 100 nm in A,50nminC.)

3836 | www.pnas.org/cgi/doi/10.1073/pnas.0911177107 Fukunaga et al. Downloaded by guest on September 27, 2021 R2*-W R2* Frequency ability for the myelination processes during development and/or myelin maintenance and repair afterward (38). Interestingly, the MRI frequency data obtained here and in previous work (17) suggest that iron and myelin do not colocalize at constant proportions throughout the brain. That would lead to Pre-extraction identical frequency shifts in subcortical white matter and the line of Gennari (relative to surrounding gray matter), inconsistent with the in vivo (Fig. 1) and ex vivo (Fig. 5 and Table 1) MRI obser- vations. Rather, intracortical fibers appear to have a dis- proportionately high iron content. Imperfect ferritin–myelin Post-extraction 0 Hz 80 Hz -5 Hz 5 Hz colocalization is also seen in the iron-poor patches in subcortical white matter (Fig. S2), the origin of which is still poorly understood. Fig. 5. Effect of iron extraction on MRI contrast in postmortem brain tissue. Investigations into the precise anatomical relationship between – Iron extraction strongly reduces intracortical magnetic susceptibility based ferritin and myelin distributions across the brain, and their alter- contrast as evidenced from R2*-weighted, R2*, and frequency shift images. The frequency images show subtle signs of contrast reversal with extraction, ations in demyelinating diseases such as multiple sclerosis may suggesting a small, opposing frequency shift caused by the remaining provide important clues about the biological role of ferritin. In fact, myelin. The circular area of susceptibility shift at right bottom of post- preliminary high-field MRI studies have already found substantial extraction images is caused by an entrapped air bubble. contrast variations in white matter regions (16, 17), allowing visu- alization of the major fiber bundles in normal brain, and lesions in diseases such as multiple sclerosis and brain tumors (39). The iron extraction (Fig. 5), and would explain the low contrast contribution of iron and myelin to contrast in these tissues is cur- between white matter and cerebrospinal fluid observed in previous rently being investigated in this and other laboratories. work (17). If confirmed, this would open the way to quantify iron content in both white matter and gray matter by combining R2*, Materials and Methods R2, and frequency data, because these measures are differentially Human Volunteer and Brain Tissue Samples. The MRI data presented in Fig. 1 affected by iron and myelin content. For example, iron and myelin were obtained from a normal human volunteer under an institutional review board–approved protocol after obtaining informed consent. Human both increase R2 and R2*, whereas they have opposing effects on postmortem brain samples were obtained from a 60-y-old male subject the resonance frequency (paramagnetic vs. diamagnetic shift, NEUROSCIENCE (occipital lobe) and a 27-year-old female subject (parietal lobe), each with no respectively). Furthermore, similar to contrast caused by deoxy- history of neurological disease. The brain tissue was fixed and preserved in hemoglobin (34, 35), R2,R2*, and frequency depend differently on 10% neutral buffered formalin for approximately 6 months. The post- the spatial distribution of iron and myelin. Accurate quantification mortem interval before fixation was less than 24 h. of iron will require a better understanding of these interrelation- ships. It may also require prior removal of tissue geometry effects Iron and Myelin Staining. For iron staining, paraffinized brain tissue slabs were on frequency data by deconvolution methods (20, 36). cut into 20-μm sections with a Vibratome sectioning system (Vibratome). The combined results of fluorescence antibody staining, the Deparaffinized sections (20 μm) were stained for iron using Perls stain (40) with iron extraction experiment, and quantitative scanning trans- intensification of the reaction product with 3,3′-diaminobenzidine (DAB) as mission electron microscopy (STEM)–EELS indicate that much described previously (41). To further improve the detection sensitivity, two modifications suggested by Smith et al. (42) were adopted, namely (i): of the intracortical frequency contrast seen ex vivo is caused by increasing the concentrations of potassium ferrocyanide and hydrochloric acid iron in ferritin. For in vivo MRI, an additional contribution could and (ii) lengthening the incubation time and increasing the temperature. The come from deoxyhemoglobin in blood, although theory and procedure involved treating the sections with a freshly prepared 1:1 mixture preliminary experimental data suggest that this contribution is of 5% potassium ferrocyanide and 5% HCl for 2 h at 37 °C. The sections were negligible (17, 37). then washed with PBS solution three times for 10 min each. Subsequently, the The current finding of a laminar variation of iron is intriguing sections were incubated in 0.5 mg/mL DAB for 10 min followed by immersion and merits further discussion. In both gray and white matter, both in a mixture of 0.5 mg/mL DAB and 0.01% H2O2 for 10 min in the dark. Sections total iron (Fig. 2) and ferritin (Fig. 3) showed substantial coloc- were then washed with PBS solution three times for 10 min each, dehydrated in graded alcohol, air-dried, and mounted. Control sections were incubated alization with myelin. This is consistent with earlier observations fl with potassium ferrocyanide solution, but the hydrochloric acid was replaced and has been suggested to re ect the role of iron in the synthesis of with distilled water. No staining was observed in control sections. cholesterol and lipids, which are key components of the myelin For Luxol fast blue myelin staining, sections were incubated overnight at sheets surrounding nerve fibers (26, 38). Storage of iron in ferritin 56 °C in 0.1% Luxol fast blue solution (Luxol fast blue MBS 0.1 mg, 95% ethyl deposits close to myelination sites would ensure abundant avail- alcohol 100 mL, glacial acetic acid 0.5 mL). To differentiate the contrast

Table 1. Effect of iron extraction on MRI indicators of tissue magnetic susceptibility Gray matter Gray matter (inside Gray matter difference (inside Gennari) (outside Gennari) White matter − outside Gennari)

Time point Frequency R2* Frequency R2* Frequency R2* Frequency R2*

Preextraction 0.82 39.6 0.02 26.5 −0.66 46.0 0.80 13.1 Postextraction −0.28 17.7 −0.30 15.0 0.10 29.6 0.02 2.7 Difference – −21.9 – −11.5 – −16.4 −0.78 −10.4

Both R2* and relative frequency are strongly affected by iron extraction, with the strongest fractional changes observed in the line of Gennari. The contrast in both R2* and frequency between the line of Gennari and the surrounding gray matter is almost completely removed by iron extraction. All frequencies were measured relative to the water surrounding the tissue sample. As no global reference for water frequency was measured in pre- and postextraction experiments, direct comparison of frequencies for both measurements were not possible and therefore not included. All values in Hertz.

Fukunaga et al. PNAS | February 23, 2010 | vol. 107 | no. 8 | 3837 Downloaded by guest on September 27, 2021 between white and gray matter, slides were immersed in 0.05% lithium for collection semiangle β; σFeðβ; ΔFeÞ is the partial iron L2,3 inelastic scat- carbonate solution for tens of seconds. The slides were examined on an tering cross section; and σC ðβ; ΔC Þ is the partial carbon K inelastic scattering Eclipse E400 microscope (Nikon) with ×2 and ×10 objective lenses. cross section.

The collection semiangle β was 20 mrad, ΔFe was set to 10 eV to include the

Tissue Iron Extraction. To determine the contribution of iron to intracortical L3 white-line resonance, and ΔC was set to 50 eV. A value for the scattering contrast, we used an iron extraction method described previously (22). Iron cross section σFeðβ; ΔFeÞ was estimated by considering the fraction of the iron

extraction was accomplished by reductive dissolution through the combined L2,3 signal in the near-edge window relative to a 100 eV window, which was use of the reducing agent 2 mM sodium dithionite and the 1 mM iron obtained using an experimentally determined iron L2,3 spectrum (43). chelating agent desferrioxamine. The iron extraction procedure lasted for Inelastic scattering cross sections inside the 100-eV window at the iron L2,3 7 d and the chemical solutions were replaced daily. The high-resolution MRI edge and inside a 50-eV window at the carbon K edge were determined scan (as detailed later) of formalin-fixed brain tissue samples of the primary from the Gatan EELS analysis software. The value of σC ðβ; ΔC Þ was estimated visual cortex was performed before and after extraction of iron. The image −25 2 −26 as 6.19 × 10 m , and the value for σFeðβ; ΔFeÞ was estimated as 8.47 × 10 plane was perpendicular to the main magnetic field. m2. The numbers of carbon atoms contained in the analyzed regions of the ferritin particles were determined from the estimated specimen thickness Immunofluorescence Staining of Ferritin and Myelin. The sections were incu- (approximately 0.5 inelastic mean free paths), the density of carbon, and the bated for 1 h at room temperature with 1/100 rabbit IgG anti-ferritin (Sigma) area analyzed. There were approximately 105 carbon atoms in each 50-nm2 and 1 μg/mL chicken IgY anti-MBP (Aves Labs). The primary immunoreactions region containing a ferritin particle. were visualized using 1 h incubation at room temperature in the presence of The tissue iron concentration was estimated by multiplying the concen- 1 μg/mL donkey anti-rabbit IgG–Alexa Fluor 488 (Invitrogen) and goat anti- chicken IgY–Alexa Fluor 647 (Invitrogen). Subsequently, the sections were tration of particles (i.e., number of ferritin particles per unit volume) with the scanned across multiple adjacent fields using an Axiovert 200M fluorescence average loading of iron atoms per ferritin particle NFE . microscope (Carl Zeiss) through a ×20 objective with Volocity Acquisition software (Improvision). Scanned images were then stitched together to form MRI Scanning. All MRI scanning was performed on a Signa 7.0-T whole-body a larger field of view image using IDL 7.0 (ITT Visual Information Solutions) MRI scanner (GE Healthcare). The in vivo data were acquired as described software. Subsequently, the Pearson product–moment correlation coef- previously (17). MRI scanning of postmortem brain tissue for comparison ficient between MBP and ferritin stains was calculated for the tissue fields with iron histochemical staining was done using an eight-channel receive- shown in Fig. 3B. only detector array designed for imaging of tissue slabs. A high resolution 3D multiecho gradient echo acquisition was performed with the following EM. Thin trapezoidal brain slices containing superficial cortical layers and the parameters: repetition time, 55 ms; echo time, 17.5/30.9/44.2 ms; flip angle, line of Gennari were cut to lengths of 4 to 5 mm and fixed in a mixture of 10°; field of view, 156 × 156 × 30 mm3; matrix size, 1,024 × 1,024 × 200 (150 2.5% paraformaldehyde and 2% glutaraldehyde in PBS solution for 2 h μm isotropic voxel size). The small tissue samples used for the iron extraction followed by an extensive wash in PBS solution. The samples were dehydrated experiment were scanned using a dedicated, home-built four-channel in a series of ethanol solutions in water (30%, 50%, 75%, and 95%, for 10 min receive-only detector array. High-resolution 2D multiecho gradient echo each) followed by 100% ethanol for 45 min with three changes. They were acquisition was performed with the following parameters: repetition time, then infiltrated with Epon-Araldite (Ted Pella) for 2 d with 30% Epon-Araldite 1.7 s; echo time, 16.2/33.3 ms; flip angle, 60°; slice thickness, 0.5 mm; 33 in ethanol for 2 h, 50% for 4 h, 75% for 16 h, and 100% for 1 d with two slices; field of view, 40 × 40 mm2; matrix size, 512 × 512; and repetition time, changes. Samples were polymerized at 60 °C for 2 d. Ultrathin sections of 2.0 s; echo time, 15.1/32.0 ms; flip angle, 60°; slice thickness, 0.5 mm; field of thickness 100 ± 10 nm were cut using an EM UC6 Ultramicrotome (Leica) and view, 65 × 65 mm2; matrix size, 512 × 512. Data processing, including image fi collected on copper grids covered with a Formvar/carbon support lm. reconstruction, calculation of magnitude R2*-weighted images and R2* fi TEM images of unstained sections were recorded with energy ltering and maps was performed with IDL 7.0 software. Images were reconstructed − − μ at a defocus of 4to 6 m using a Tecnai TF30 electron microscope (FEI) using a phase-sensitive noise-weighted channel combination (44), and the fi equipped with a GIF Tridiem imaging lter (Gatan), and operating at an frequency image was analyzed after macroscopic background phase accelerating voltage of 300 kV. Iron particles were counted in 100 randomly removal as described previously (17). selected images obtained from the line of Gennari and 100 images from control regions in the more superficial layers (layers II–III). The full width of Estimation of Susceptibility Effects Originating from Ferritin Iron. The each square field was 700 nm. Approximately 40% of the areas from the line dependence of R * relaxation rate on tissue iron content was estimated from of Gennari contained more than five particles, and approximately 15% of 2 previous work, which established a linear dependence with a slope of 0.05 the areas from the control region contained more than five particles. Hz/T/ppm iron (i.e., μg/g wet weight) (32). A 34 μg/g iron concentration To determine the iron loading of the particles, EELS were recorded across difference in the line of Gennari (relative to superficial layers), found with selected fields using a HB501 scanning transmission electron microscope (VG) the TEM experiments, would correspond to an R * shift (at 7.0 T) of 13 Hz. operated at an accelerating voltage of 100 kV, and equipped with an ENFINA 2 An estimate for the associated resonance frequency shift was derived from electron energy loss spectrometer (Gatan). From these data, Fe L2,3 maps were extracted from three STEM-EELS spectrum-images that were acquired previous calculations (45) of ferritin susceptibility shifts (from water) based using Digital Micrograph software (Gatan). on a number of assumptions, including an average magnetic moment of 3.78 Bohr magnetons per iron atom: Δχ≈1:3·10 − 9·c, with c the iron con- The number of Fe atoms NFE in each iron particle analyzed (n = 15) was μ Δ calculated according to the following: centration in g/g. Associated frequency shifts f are geometry and orien- fi    tation-dependent (46), and are maximal for in nitely long structures parallel fi Δ ≈ γ Δχ= ¼ : γ · − 9· IFeðβ; ΔFeÞ σCðβ; ΔCÞ to the eld: f 2πB0 3 0 432πB0 10 c. For water protons at 7.0 T and c NFe ¼ Nc (Eq. 1) of 34 μg/g, this leads to a maximal 5 Hz frequency shift in the line of Gennari. ICðβ; ΔCÞ σFeðβ; ΔFeÞ ACKNOWLEDGMENTS. Ken M. Fish (General Electric Global Research Center), Here, NC is the number of carbon atoms in the analyzed region that includes ðβ; Δ Þ Kant M. Matsuda [National Cancer Institute/National Institutes of Health the ferritin particle; IFe Fe is the integrated iron L2,3 edge signal for (NIH)], and Eiji Matsuura [National Institute of Neurological Disorders and Δ β energy window Fe above the iron L2,3 edge, and for collection semiangle Stroke (NINDS)/NIH] are acknowledged for help with the iron extraction, defined by the spectrometer entrance aperture; IC ðβ; ΔC Þ is the integrated providing tissue samples, and histochemical staining respectively. This research carbon K edge signal for energy window ΔC above the carbon K edge and was supported by the Intramural Research Program of the NINDS/NIH.

1. Raichle ME (1981) Measurement of local cerebral blood flow and metabolism in man 5. Wong-Riley MT, et al. (1993) Cytochrome oxidase in the human visual cortex: with positron emission tomography. Fed Proc 40:2331–2334. distribution in the developing and the adult brain. Vis Neurosci 10:41–58. 2. Ogawa S, Lee TM, Kay AR, Tank DW (1990) Brain magnetic resonance imaging with 6. Hockfield S, Tootell RB, Zaremba S (1990) Molecular differences among neurons contrast dependent on blood oxygenation. Proc Natl Acad Sci USA 87:9868–9872. reveal an organization of human visual cortex. Proc Natl Acad Sci USA 87:3027–3031. 3. Brodmann K (1909) Vergleichende Lokalisationlehre der Grosshirnrinde (Johann 7. Zilles K, Schleicher A, Rath M, Glaser T, Traber J (1986) Quantitative autoradiography Ambrosius Barth Verlag, Leipzig, Germany). of transmitter binding sites with an image analyzer. J Neurosci Methods 18:207–220. 4. Vogt O (1911) Die Myeloarchitektonik des Isocortex parietalis. J Psychol Neurol 15: 8. Annese J, Pitiot A, Dinov ID, Toga AW (2004) A myelo-architectonic method for the 221–232. structural classification of cortical areas. Neuroimage 21:15–26.

3838 | www.pnas.org/cgi/doi/10.1073/pnas.0911177107 Fukunaga et al. Downloaded by guest on September 27, 2021 9. Vogt C, Vogt O (1919) Allgemeinere Ergebenisse unserer Hirnforschung. J Psychol 29. Bulte JW, et al. (1994) Magnetoferritin: characterization of a novel superparamagnetic Neurol 25:279–461. MR contrast agent. J Magn Reson Imaging 4:497–505. 10. Schleicher A, et al. (2005) Quantitative architectural analysis: a new approach to 30. Gossuin Y, Roch A, Muller RN, Gillis P (2000) Relaxation induced by ferritin and cortical mapping. Anat Embryol (Berl) 210:373–386. ferritin-like magnetic particles: the role of proton exchange. Magn Reson Med 43: 11. Clark VP, Courchesne E, Grafe M (1992) In vivo myeloarchitectonic analysis of human 237–243. striate and extrastriate cortex using magnetic resonance imaging. Cereb Cortex 2: 31. Gossuin Y, Muller RN, Gillis P, Bartel L (2005) Relaxivities of human liver and spleen – 417 424. ferritin. Magn Reson Imaging 23:1001–1004. 12. Bridge H, Clare S (2006) High-resolution MRI: in vivo histology? Philos Trans R Soc 32. Yao B, et al. (2009) Susceptibility contrast in high field MRI of human brain as a Lond B Biol Sci 361:137–146. function of tissue iron content. Neuroimage 44:1259–1266. 13. Barbier EL, et al. (2002) Imaging cortical anatomy by high-resolution MR at 3.0T: 33. O’Brien JS, Sampson EL (1965) Lipid composition of the normal human brain: gray detection of the stripe of Gennari in visual area 17. Magn Reson Med 48:735–738. matter, white matter, and myelin. J Lipid Res 6:537–544. 14. Augustinack JC, et al. (2005) Detection of entorhinal layer II using 7Tesla [corrected] 34. Boxerman JL, Hamberg LM, Rosen BR, Weisskoff RM (1995) MR contrast due to magnetic resonance imaging. Ann Neurol 57:489–494. intravascular magnetic susceptibility perturbations. Magn Reson Med 34:555–566. 15. Walters NB, et al. (2003) In vivo identification of human cortical areas using high- 35. Yablonskiy DA, Haacke EM (1994) Theory of NMR signal behavior in magnetically resolution MRI: an approach to cerebral structure-function correlation. Proc Natl Acad – Sci USA 100:2981–2986. inhomogeneous tissues: the static dephasing regime. Magn Reson Med 32:749 763. 16. Li TQ, et al. (2006) Extensive heterogeneity in white matter intensity in high- 36. Shmueli K, et al. (2009) Magnetic susceptibility mapping of brain tissue in vivo using – resolution T2*-weighted MRI of the human brain at 7.0 T. Neuroimage 32:1032–1040. MRI phase data. Magn Reson Med 62:1510 1522. 17. Duyn JH, et al. (2007) High-field MRI of brain cortical substructure based on signal 37. Lee J, Hirano Y, Fukunaga M, Silva AC, Duyn JH (2010) On the contribution of deoxy- phase. Proc Natl Acad Sci USA 104:11796–11801. hemoglobin to MRI gray-white matter phase contrast at high field. Neuroimage 49: 18. Duyn J, Koretsky AP (2008) Magnetic resonance imaging of neural circuits. Nat Clin 193–198. Pract Cardiovasc Med 5 (Suppl 2):S71–S78. 38. Cheepsunthorn P, Palmer C, Connor JR (1998) Cellular distribution of ferritin subunits 19. Ogg RJ, Langston JW, Haacke EM, Steen RG, Taylor JS (1999) The correlation between in postnatal rat brain. J Comp Neurol 400:73–86. phase shifts in gradient-echo MR images and regional brain iron concentration. Magn 39. Hammond KE, et al. (2008) Development of a robust method for generating 7.0 T Reson Imaging 17:1141–1148. multichannel phase images of the brain with application to normal volunteers and 20. Haacke EM, et al. (2005) Imaging iron stores in the brain using magnetic resonance patients with neurological diseases. Neuroimage 39:1682–1692. – imaging. Magn Reson Imaging 23:1 25. 40. Perls M (1867) Nachweis von Eisenoxyd in gewissen Pigmenten. Virchows Arch 39: – 21. Drayer B, et al. (1986) MRI of brain iron. AJR Am J Roentgenol 147:103 110. 42–48. fi 22. Schenck JF, et al. (2006) High- eld magnetic resonance imaging of brain iron in 41. Nguyen-Legros J, Bizot J, Bolesse M, Pulicani JP (1980) [“Diaminobenzidine black” as a Alzheimer disease. Top Magn Reson Imaging 17:41–50. new histochemical demonstration of exogenous iron (author’s transl)]. Histochemistry 23. LeVine SM (1991) Oligodendrocytes and myelin sheaths in normal, quaking and 66:239–244. shiverer brains are enriched in iron. J Neurosci Res 29:413–419. 42. Smith MA, Harris PL, Sayre LM, Perry G (1997) Iron accumulation in Alzheimer disease 24. Morris CM, Candy JM, Oakley AE, Bloxham CA, Edwardson JA (1992) Histochemical is a source of redox-generated free radicals. Proc Natl Acad Sci USA 94:9866–9868. distribution of non-haem iron in the human brain. Acta Anat (Basel) 144:235–257. 43. Leapman RD (2003) Detecting single atoms of calcium and iron in biological structures 25. Connor JR, Menzies SL, St Martin SM, Mufson EJ (1990) Cellular distribution of NEUROSCIENCE – transferrin, ferritin, and iron in normal and aged human brains. J Neurosci Res 27: by electron energy-loss spectrum-imaging. J Microsc 210:5 15. 595–611. 44. de Zwart JA, et al. (2004) Signal-to-noise ratio and parallel imaging performance of a – 26. Connor JR, Menzies SL (1996) Relationship of iron to oligodendrocytes and myelination. 16-channel receive-only brain coil array at 3.0 Tesla. Magn Reson Med 51:22 26. Glia 17:83–93. 45. Schenck JF (1996) The role of magnetic susceptibility in magnetic resonance imaging: 27. Todorich B, Pasquini JM, Garcia CI, Paez PM, Connor JR (2008) Oligodendrocytes and MRI magnetic compatibility of the first and second kinds. Med Phys 23:815–850. myelination: the role of iron. Glia. 57:467–478. 46. Chu SC, Xu Y, Balschi JA, Springer CS, Jr. (1990) Bulk magnetic susceptibility shifts in 28. Hallgren B, Sourander P (1958) The effect of age on the non-haemin iron in the NMR studies of compartmentalized samples: use of paramagnetic reagents. Magn human brain. J Neurochem 3:41–51. Reson Med 13:239–262.

Fukunaga et al. PNAS | February 23, 2010 | vol. 107 | no. 8 | 3839 Downloaded by guest on September 27, 2021