High-Resolution Myelin Water Measurements in RatSpinal Cord

Piotr Kozlowski,1,2* Jie Liu,3 Andrew C. Yung,2 and Wolfram Tetzlaff 3,4

Multiecho imaging data were acquired at 7T from control andinjured (dorsal column transection) rat spinal cords ex vivo within-plane resolution of 61, 78, and 100 �m, and from a control ratspinal cord in vivo with in-plane resolution of 117 �m. Themyelin water maps were calculated using nonnegative least-squares (NNLS) analysis of the decay curves. For the controlcords, myelin water maps showed details of the cord morphol-ogy, and the average myelin water fraction (MWF) values inwhite matter and gray matter corresponded well with previouslypublished results and the expected amounts of myelin withinthe cord, and correlated very well with Luxol Fast Blue stain(R2 � 0.95). Myelin water maps from an injured cord showedexcellent qualitative correlation with histology. This pilot studydemonstrates that high-resolution myelin water mapping in ratspinal cord is feasible, and this technique has potential to be avaluable tool in studying white matter damage in rat models ofspinal cord injury. Magn Reson Med 59:796–802, 2008. © 2008Wiley-Liss, Inc.

Key words: myelin water; rat spinal cord; white matter; spinalcord injury

Functional loss following spinal cord injury (SCI) islargely caused by damage to the white matter (WM), as themechanical trauma to the spinal cord leads to the destruc-tion of myelinated axons in WM tracts. In addition, manyaxons that remain intact following injury lose their myelinsheaths as a result of oligodendrocytes undergoing celldeath (1). Since myelin is essential for the conduction ofnervous signals (2), the initial loss of myelin and thesubsequent demyelination process play a major role in theloss of motor and sensory function and poor recoveryfollowing SCI (3). Repair of the myelin sheaths of thesurviving fiber tracts would therefore contribute to func-tional recovery of the SCI patients. In some ways, reestab-lishing myelin on axons appears easier than regeneratingneurons and their connections (4). Therefore, significanteffort has been directed to designing restoration therapiesaiming to rebuild myelin at the injury site (5). However,

the efficacy of these strategies is difficult to assess, ascurrently there is no noninvasive technique capable oftracking myelin in spinal cord in vivo.

Myelin cannot be detected directly with MRI becausethe majority of its signal has decayed by 3 ms (6). Therehave been, however, many attempts to identify myelinindirectly by studying properties of water surroundingmyelin sheaths and its interaction with protons associatedwith proteins and lipids that form myelin sheaths. Thespin–spin relaxation of water in vivo is influenced by thelocal tissue environment. Although its complex mecha-nism is not fully understood, it is known that T2 relaxationtime in vivo depends on factors like proximity of largemacromolecules, diffusion properties of water moleculeswithin the local environment, and other processes leadingto exchange of water protons with the surrounding envi-ronment. In homogenous systems the spin–spin relaxationprocess is monoexponential, with a well-defined T2 relax-ation time. However, in a complex system, such as brain orspinal cord, multiple water compartments (e.g., myelinbilayers, intra-/extracellular space, cerebrospinal fluid) re-sult in spin–spin relaxation becoming a multiexponentialprocess giving rise to multiple T2 values (7). Quantitativeanalysis of T2 decay curves acquired with a multiechoCarr-Purcell-Meiboom-Gill (CPMG) sequence producescontinuous distributions of T2 values representing variousT2 components present in the tissue (7). Such analysis ofbrain and spinal cord tissue in vitro and in vivo identifiedthree distinctive components with T2 values of 10–20 ms,50–300 ms, and over 1 sec (8). It has been argued that thethree components represent water trapped between mye-lin bilayers, water in the intra-/extracellular compartment,and the cerebrospinal fluid, respectively (8). Indeed, his-tological analysis showed good correlation between mye-lin content and the fractional amount of water trappedbetween myelin bilayer, the so-called myelin water frac-tion (MWF), measured from the T2 decay curve (9). Thistechnique has been successfully applied to study demyeli-nation processes in multiple sclerosis (MS) patients (10).

Measuring MWF in a rat model of SCI can be an impor-tant marker of the cord pathology and can be used tomeasure the success of potential SCI therapies. Myelinwater measurements, however, require very high signal-to-noise ratio (SNR) and spatial resolution, making it chal-lenging to apply this technique in rat spinal cord. In thispilot study we investigated the feasibility of high spatialresolution myelin water measurements in control and in-jured rat spinal cord in vivo and ex vivo.

MATERIALS AND METHODSAnimal Preparation

All animal experimental procedures were carried out incompliance with the guidelines of the Canadian Council

1University of British Columbia, Department of Radiology, Vancouver, BritishColumbia, Canada.2University of British Columbia MRI Research Centre, Vancouver, BritishColumbia, Canada.3International Collaboration on Repair Discoveries, Vancouver, British Colum-bia, Canada.4University of British Columbia, Department of Zoology, Vancouver, BritishColumbia, Canada.Grant sponsor: Rick Hansen Man In Motion Research Fund.Part of this work was presented at the 15th Annual Meeting of the ISMRM inBerlin, May 2007, Abstract #319.*Correspondence to: Piotr Kozlowski, UBC MRI Research Centre, Life Sci-ences Centre, 2350 Health Sciences Mall, Vancouver, BC, V6T 1Z3 Canada.E-mail: pkozlow@interchange.ubc.caReceived 3 July 1007; revised 29 November 2007; accepted 3 December2007.DOI 10.1002/mrm.21527Published online in Wiley InterScience (www.interscience.wiley.com).

Magnetic Resonance in Medicine 59:796–802 (2008)

© 2008 Wiley-Liss, Inc. 796

for Animal Care and were approved by the institutionalAnimal Care Committee. Eleven male Sprague–Dawleyrats (250–280 g) were used in this study—one controlanimal for in vivo and 10 animals for ex vivo measure-ments. Dorsal column transection was induced in two ratsat the C4/C5 level of the cervical spinal cord as previouslydescribed (11); 18 days later, two injured and two controlrats were deeply anesthetized and perfused intracardiallywith phosphate-buffered saline for 3 min followed byfreshly hydrolyzed paraformaldehyde (4%) in 0.1 M so-dium phosphate buffer at pH 7.4. Spinal cords were thenharvested and postfixed overnight in the same fixative. Exvivo MRI experiments were carried out 1 day after theexcision and postfixation of the cords. In a separate exper-iment, six additional spinal cords were excised from thecontrol rats in a similar manner.

Following ex vivo MRI examination the spinal cordswere cryoprotected in 22% sucrose, frozen, and cut on acryostat in the transverse plane (cross sections). The sec-tions from four spinal cords were immunostained usingstandard procedures with a monoclonal antibody to mye-lin basic protein (1:100, Serotec, Bicester, UK) to stainmyelin, or a rabbit anti-neurofilament-M antibody (1:500,Chemicon, Temecula, CA) to stain axons. Primary antibod-ies were incubated overnight at 4°C, washed, then incu-bated with fluorescent-conjugated secondary antibodiesfor 2 hr at room temperature. Epifluorescence illuminationwas used to generate micrographs in a Zeiss Axioplan 2microscope. Sections from the additional six spinal cordswere stained with Luxol Fast Blue (LFB) using standardprocedures. The sections were dehydrated with alcoholand incubated in 1% LFB solution (Harleco, Philadelphia,PA) overnight at 60°C. The following day sections wererinsed with distilled water and washed with 0.05% lith-ium carbonate solution (EMD Chemicals, San Diego, CA)until a sharp contrast was obtained between the gray mat-ter (GM) and WM. Images of the LFB sections were gener-ated on a Zeiss Axioplan 2 microscope using NorthernEclipse 6.0 software (Empix Imaging, Mississauga, ON,Canada) and the optical density was measured in variousregions of interest (ROIs) in dorsal and lateral WM andGM.

MRI Experiments

All MRI experiments were carried out on a 7T animalscanner (Bruker, Germany). For in vivo measurements therat was anesthetized with isoflurane and placed in a spe-cially designed holder in a supine position over a 3-cminner-diameter (i.d.) circular surface coil that was used forpulse transmission and signal reception. For ex vivo mea-surements the excised cords were positioned in a 5-mmi.d. glass tube filled with the fixation solution. A woodensplint was attached to the cord with Teflon tape to supportthe cord. A 2-cm i.d. four-turn solenoid coil was used forpulse transmission and signal reception.

Myelin water measurements were carried out using asingle-slice, multiecho CPMG sequence (12) with 32 ech-oes, 256 � 256 matrix, TR/TE � 1500/6.673 ms, and NA �6, for a total acquisition time of 38 min. For the controlexcised cords a 1-mm slice was selected through the cer-vical spine and three field of view (FOV) values were used:

2.56, 2, and 1.56 cm, resulting in an in-plane resolution of100, 78, and 61 �m, respectively. For the excised injuredcords 1-mm slices were selected through the injury site,5 mm cranial, and 5 mm caudal to the injury site, and twoFOV values of 2.56 and 2 cm were used. For the in vivoexperiment a 1.5-mm slice was selected through the lum-bar spine and an FOV of 3 cm was used, resulting in anin-plane resolution of 117 �m.

In a separate experiment six excised spinal cords werepositioned in a 4.5-mm i.d. plastic tube filled with thefixation solution. A small plastic rod was positionedalongside the cord to prevent it from bending. A four-turn,13-mm i.d. and 20-mm long solenoid coil was used forpulse transmission and signal reception. Myelin watermeasurements were carried out on these cords using thesame parameters as described above, but with only onevalue of FOV � 2 cm.

Data Processing and Analysis

CPMG data were processed using software procedures de-veloped in-house using Igor Pro (Wave Metrics, Lake Os-wego, OR)—an interpretative data processing language. T2

distributions were calculated for each pixel from the mul-tiecho data using a regularized nonnegative least-squares(NNLS) analysis to minimize �2 plus an energy constraint.The T2 axis was partitioned into 101 logarithmicallyspaced values between 7.75 and 1500 ms. The small modelused in these analyses minimizes T2 distribution energy ofthe sum of the square of the amplitudes A(T2)2, whichresults in smooth T2 spectra. The proportion of the regu-larizer in the minimization was adjusted such that the final�2 was 2–2.5% higher than the minimum �2 (8). Myelinwater maps were generated by integrating the T2 distribu-tion between 7.75 and 20 ms and dividing the result by thetotal integral of the T2 distribution in each pixel. For theROI analyses the average signal decay curves were ex-tracted from the ROIs encompassing WM or GM and theNNLS analyses were carried out on the averaged signals.

Quantitative correlation between MRI and histology wascarried out on six excised control spinal cords. AverageMWF values were calculated from ROIs encompassinglateral and dorsal WM and GM, defined on myelin watermaps. Similar size and location ROIs were defined also onLFB histological sections and the average optical densitywas calculated from these ROIs. Correlation between MWFand LFB staining was carried out using linear regressionanalysis.

RESULTS

Figure 1 shows the first echo of the CPMG data (top left)and the reconstructed myelin water map (top right) ac-quired from a control rat spinal cord in vivo. The T2

distributions from individual pixels positioned in GM(solid line) and dorsal WM (dashed line) show, as ex-pected, three distinct water components corresponding tomyelin water (T2 � 20 ms), intra-/extracellular water(30 ms � T2 � 100 ms), and cerebrospinal fluid (T2 �1000 ms). The average MWF values are 5% in GM and24% in WM.

Myelin Water Measurements in Rat Spinal Cord 797

Figure 2 shows the first echoes of the CPMG data (top)and the reconstructed myelin water maps (bottom) ac-quired from a control excised rat spinal cord with anin-plane resolution of 100 �m (left), 78 �m (middle), and61 �m (right). The circular structure in the center of theimages (marked wood) is the wooden splint that was usedto stabilize the cord. It contains water with short T2 times,which is consistent with previous studies of wood (13).Figure 3 shows T2 distributions from individual pixelslocated in GM (solid line) and dorsal WM (dashed line)from CPMG data acquired with an in-plane resolution of100 �m (bottom), 78 �m (middle), and 61 �m (top). Theaverage MWF values in GM and WM for the three datasets

are: 18% and 39%, 12% and 33%, and 4% and 30%,respectively.

Figure 4 shows representative T2 distributions from in-dividual pixels positioned in dorsal WM (bottom graph)

FIG. 1. First echo of the CPMG data (top left) and the reconstructedmyelin water map (top right) acquired from a rat spinal cord in vivowith an in-plane resolution of 117 �m. The bottom graph shows T2

distributions from single pixel located in GM (solid line) and dorsalWM (dashed line). The average MWF values are 5% in GM and 24%in WM.

FIG. 2. First echo of the CPMG data (top) and the reconstructedmyelin water maps (bottom) acquired from an excised fixed controlrat spinal cord. The in-plane resolution is 100 �m (left), 78 �m(middle), and 61 �m (right).

FIG. 3. T2 distributions from single pixels located in GM (solid line)and dorsal WM (dashed line) calculated from CPMG data acquiredfrom an excised fixed control rat spinal cord with the in-planeresolution of 61 �m (top), 78 �m (middle), and 100 �m (bottom). Theaverage MWF values in GM and WM for the three datasets are: 18%and 39%, 12% and 33%, and 4% and 30%, respectively.

FIG. 4. Representative T2 distributions from individual pixels posi-tioned in GM (top graph) and dorsal WM (bottom graph). Distribu-tions were calculated from the CPMG data acquired in vivo (solidline) and ex vivo from one of the excised fixed control spinal cords(dashed line).

798 Kozlowski et al.

and GM (top graph). Distributions were calculated fromthe CPMG data acquired in vivo (solid line) and ex vivofrom one of the excised control spinal cords (dashed line).As expected, the in vivo distributions show longer T2

values of both short and medium component.Table 1 shows average T2 values of the short and me-

dium T2 components and MWF from ROIs encompassingdorsal WM, lateral and ventral WM, and GM calculatedfrom six excised fixed control rat spinal cords. Low valuesof standard deviations for all parameters demonstrate ex-cellent reproducibility of the myelin water measurementsin excised control spinal cords.

Figure 5 shows average MWF values versus averageoptical density of LFB staining calculated from ROIs en-compassing lateral and dorsal WM and GM. The solid lineshows the result of linear regression fit for all the datapoints, the dashed line shows the result of linear regres-sion fit for the dorsal WM, and the dotted line shows theresults of linear regression fit for the lateral and dorsalWM. Quantitative analysis showed statistically significantcorrelation between MWF and LFB for all the data points(R2 � 0.95), for dorsal WM (R2 � 0.89), and for dorsal andlateral WM (R2 � 0.48).

Figure 6 shows the first echoes of the CPMG data andreconstructed myelin water maps acquired with in-plane

resolution of 78 �m from a rat spinal cord excised 18 daysfollowing dorsal column transection injury. Images wereacquired at injury site and 5 mm cranial and 5 mm caudalto the injury site. Histological sections from the samelocations, immunostained with an antibody to myelin ba-sic protein (MBP) and an antibody to neurofilament pro-tein for axons, as well as combined MBP stain in red andaxonal stain in green from slices 5 mm cranial and 5 mmcaudal to injury, are shown at the bottom. Dorsal columntransection injury severs all axons in the gracile fascicle,which subsequently undergo degeneration cranial to theinjury site (Fig. 6). The medial parts of the cuneate fascicleundergo degeneration as well; however, since these axons

Table 1Average T2 Values of the Short and Medium T2 Components and Myelin Water Fraction from ROIs Encompassing Dorsal WM,Lateral � Ventral WM, and GM in Excised Control Rat Spinal Cords

Dorsal WM Lateral � Ventral WM GM

T2 [ms] (short T2 component) 11.52 � 1.18 10.76 � 0.76 8.93 � 1.48T2 [ms] (medium T2 component) 42.72 � 1.43 45.75 � 2.59 35.36 � 1.78Myelin water fraction 0.35 � 0.04 0.32 � 0.03 0.06 � 0.02

Numbers represent average � standard deviation (n � 6).

FIG. 5. Average MWF values versus average optical density of LFBstaining calculated from ROIs encompassing lateral (circles) anddorsal (diamonds) WM and GM (triangles) of the excised fixedcontrol spinal cords. Solid line shows the result of linear regressionfit for all the data points, dashed line shows the result of linearregression fit for the dorsal WM, and the dotted line shows theresults of linear regression fit for the lateral and dorsal WM. Quan-titative analysis showed statistically significant correlation betweenMWF and LFB for all the data points (R2 � 0.95), for dorsal WM(R2 � 0.89), and for dorsal and lateral WM (R2 � 0.48).

FIG. 6. First echoes of the CPMG data acquired with 78 �m in-plane resolution (top), reconstructed MWF maps (second from thetop), histological sections immunostained with an antibody to MBP(third from the top), and an antibody to neurofilament protein foraxons (second from the bottom), and combined MBP in red andaxonal stain in green (bottom) acquired from a rat spinal cordexcised and fixed 18 days following dorsal column transection;images were acquired at 5 mm cranial to injury (left), injury site(middle), and 5 mm caudal to injury (right). The axonal stain showsaxonal damage in the gracile fascicle cranial to injury and in thecorticospinal tract caudal to injury (lack of green stain on the com-bined MBP/axonal stain images), while both MWF and MBP mapsshow increased image intensity in this area (intensive red area in thegracile fascicle cranial to injury and in the medial parts of thecuneate fascicle caudal to injury on the combined MBP/axonal stainimages), which corresponds to myelin debris.

Myelin Water Measurements in Rat Spinal Cord 799

send collaterals for several segments caudally in additionto their main ascending axon, degenerative changes areseen cranially as well as caudally to the lesion site in thecuneate fascicle. As a result, large amounts of myelindebris were formed, which were still present 18 dayspostinjury. This debris showed as bright spots on the MBP-stained slice, as well as a high intensity area on myelinwater map (Fig. 6). Note excellent qualitative correlationbetween MBP stain and myelin water map.

DISCUSSION

The T2 distribution has been used extensively in studyingthe demyelination process in MS patients (10) and it hasbeen shown that MWF correlates well with the amount ofmyelin measured by histology (9). Myelin water mappingin animal models, however, is challenging due to oftenconflicting requirements of very high spatial resolutionand very high SNR and B1 homogeneity. Graham et al. (14)estimated that to reliably detect a short T2 component inWM using 32 echoes acquisition with a TE on the order of6.5 ms an SNR of 250 is required. Thus, few studies havereported the application of the T2 distribution in animalCNS (15,16); many have applied a nonimaging CPMGpulse sequence to acquire nonlocalized continuous T2 dis-tributions from the excised tissue. There have been severalattempts to identify multiple T2 components and measureMWF in animals (17,18) and in human (19,20) spinal cord.Although these studies were successful in identifying themyelin water component, the reconstructed myelin watermaps generated from spinal cords were of limited spatialresolution, making it difficult to depict the morphology ofthe cord. Here we present results of a pilot study thatinvestigated the feasibility of high spatial resolution mye-lin water mapping in rat spinal cord. We have shown thathigh-resolution myelin water maps can be generated fromrat spinal cord both in vivo and ex vivo.

The myelin water map generated from a rat spinal cordin vivo shows good contrast between GM and WM, andthe average T2 distributions from GM and WM (Fig. 1)compare favorably with previously published results(17,18,20). The average MWF values from in vivo GM andWM were 5% and 24%, respectively, which correspondswell with the previously reported values and expectedamounts of myelin in rat spinal cord (our recent histologymeasurements in normal rat spinal cords showed the ratioof myelin content in WM/GM equal 5). The quality of themyelin water map is largely determined by the SNR of theCPMG imaging data and B1 homogeneity. In this study weused an external surface coil with a fairly large diameter toensure reasonably homogenous B1 across the cord. As aresult, SNR within the entire cord measured on the firstecho was only 55, which limited the quality of the myelinwater map. The SNR can be significantly improved byusing a small surface RF coil surgically implanted againstthe cord (21). However, to ensure homogenous spin exci-tation an external volume RF coil actively decoupled fromthe implanted coil is needed for pulse transmission. Weare currently working on developing such a system.

Ex vivo results from the control cords show good qualitymyelin water maps demonstrating good GM/WM contrastand the characteristic butterfly shape of GM (Fig. 2). The

higher resolution myelin water maps show also contrastbetween two major WM tracts in the dorsal column (DC),the corticospinal tract (CST), and gracile fascicle. This isconsistent with the morphology of the DC, as both of thesetracts have smaller axons, and hence thinner myelinsheaths, resulting in smaller amounts of myelin and mye-lin water. The maps, however, become progressively nois-ier with the increased resolution. The SNR within the cordmeasured on the first echoes of the CPMG data was 181,117, and 88 for the 100, 78, and 61 �m in-plane resolu-tions, respectively. Representative T2 distributions (Fig. 3)show two distinct water components corresponding tomyelin water (T2 � 20 ms) and intra-/extracellular water(30 ms � T2 � 50 ms). The MWF values for the excisedcords are generally higher than for the in vivo case. This isnot surprising, as the fixation process results in a de-creased amount of water in the tissue, with the intra-/extracellular compartment showing a larger decrease thanthe myelin water (9). It has been shown that the fixationprocess also shortens T2 of both short and medium T2

components, although the shift of the medium T2 compo-nent is more significant (9,22). This can be explained bythe fact that the T2 of water is strongly influenced bycompartmentalization of cellular structure (13). Since theformalin establishes bonds between protein groups to in-crease cellular stability, the mobility of water moleculesbecomes more restricted and T2 shortens (9). We also ob-served shortening of the T2 values in fixed spinal cords ascompared to the in vivo case (Fig. 4).

It is somewhat surprising to see decreasing MWF valuesin both WM and GM with the increasing resolution. Themost likely reason for that is progressively lower SNR withthe increased resolution. In a similar experiment carriedout on a control excised spinal cord, CPMG data acquiredusing a smaller solenoid RF coil produced much higherSNR values: 546, 333, and 199 for the in-plane resolutionof 100, 78, and 61 �m, respectively (data not shown).Average MWF values calculated from these data in ROIspositioned in WM and GM showed very little variationbetween different in-plane resolutions.

Reproducibility of myelin water measurements wastested on a group of six control fixed rat spinal cords bymeasuring the average T2 values of the short and mediumT2 components and MWF from ROIs encompassing dorsalWM, lateral and ventral WM, and GM (Table 1). It isinteresting to note that all measured parameters had lowstandard deviations. Whittall et al. (8) reported that com-puter simulations indicate that for noisy CPMG decaycurves MWF is usually centered on the true value, whereasthe T2 times are more difficult to estimate by the NNLSalgorithm. Thus, standard deviations of 7–16% for theshort T2 component and 3–6% for the medium T2 compo-nent obtained in our study demonstrate high reproducibil-ity of our myelin water measurements. The accuracy ofestimating MWF will depend on the interecho spacing andthe minimum first TE of the CPMG sequence, as well as theSNR. With the initial TE and the interecho spacing of6.673 ms used in our studies and an SNR of 100 (a typicalvalue used for myelin water measurements) the number ofpoints on the decay curve above the noise level for theshort and the medium T2 components will be 7 and 28,respectively. In our reproducibility study the SNR within

800 Kozlowski et al.

the spinal cord on the first echo was in the range 235–345,thus the decay curves had 8–9 points for the short T2

component and, potentially, 33–36 points for the mediumcomponent (only 32 echoes were acquired). This suggeststhat the accuracy of identifying both T2 components in ourstudy was high, as is demonstrated by the high reproduc-ibility of the results.

Quantitative analysis showed excellent correlation be-tween MWF values and the degree of LFB staining, whichis a measure of myelin, for the control excised fixed cords(Fig. 5). Although the analysis of all data points may sug-gest bimodal distribution of the data, the analysis of thedorsal WM alone also showed a very high R2 value of 0.89,and the slopes of the fitted lines for all the data points anddorsal WM were very similar (0.006 and 0.007, respec-tively; see also Fig. 5). The R2 value calculated for thecombined dorsal and lateral WM was lower (0.48) than forthe dorsal WM alone; however, the correlation was statis-tically significant (P � 0.01). There was no significantcorrelation between MWF and the optical density of theLFB staining for the GM or the lateral WM alone. The lackof correlation for the GM alone is likely due to a smallrange of both MWF and LFB staining values (0.05–0.09and 55–67, respectively) in this region. The lack of corre-lation for the lateral WM alone and relatively weak corre-lation for the combined dorsal and lateral WM may resultfrom the fact that the ROI for these regions were definedindependently for the MRI and the histology data. Whilethe dorsal WM is a well-defined anatomical structure eas-ily identified on both datasets, the exact location and ex-tent of the ROI covering the same region of the lateral WMin both MRI images and histological sections is difficult toachieve without registering the two datasets to one an-other, which was not done in this study. Nevertheless, ourresults agree very well with the work by Laule et al. (9),who showed strong correlation between MWF and LFBstaining in formalin-fixed brain samples from MS patients(average R2 � 0.67). We recently completed a study inwhich myelin water measurements were carried out onexcised fixed spinal cords from rats that underwent dorsalcolumn transection injury. In this study we also foundstatistically significant correlation between MWF and LFBstaining with the R2 values of 0.41 and 0.77 at 3 weeks and8 weeks postinjury, respectively (data not shown, inprep.). Our results thus demonstrate that MWF is a goodmeasure of myelin in rat spinal cord ex vivo.

Myelin water maps from an injured cord (Fig. 6) showedthe morphology of the injury very well, e.g., the lack ofmyelin at the site of injury was apparent from the lack ofmyelin water in the dorsal part of the cord, while ventraland lateral WM showed largely intact on the myelin watermap. In general, qualitative analysis showed excellent cor-relation between MWF and MBP images (Fig. 6). One ofthe results of dorsal column transection injury is damageto the axons in major WM tracts, which subsequentlyundergo degeneration. As the axons distal to injury disin-tegrate, large amounts of myelin debris are formed bothcranial and caudal to the injury site, and this debris per-sists in the medial parts of the cuneate fascicle even after18 days following injury. It is well established that Walle-rian degeneration and myelin debris clearance is very slowin the CNS, and occurs over several months (see, for ex-

ample, Ref. 23). Myelin debris can be identified as hyper-intense regions on both MBP-stained slices and myelinwater maps (Fig. 6). The increased MBP staining of myelindebris can be explained by the fact that, as the myelinsheaths disintegrate, MBP antibodies gain better access toMBP molecules, resulting in the increased antibody bind-ing, and thus increased MBP staining. It is not clear, how-ever, why the myelin debris was also bright on myelinwater maps. Analysis of T2 distributions from the ROIsencompassing myelin debris and the intact ventral andlateral WM did not show any major differences in T2

values or their distributions (data not shown). One expla-nation could be that in the absence of axons myelinsheaths become looser and form onion-shaped structureswith larger distances between the myelin bilayers, result-ing in increased myelin water. It is, however, unclear whythe myelin water’s T2 values remain similar for both intactmyelin and myelin debris. More studies involving electronmicroscopy are required to explain this phenomenon.

In conclusion, we have shown that high spatial resolu-tion (100 �m or better) myelin water mapping is possiblein rat spinal cord both in vivo and ex vivo. The averageMWF values in WM and GM correspond well with previ-ously published results and the expected amounts of my-elin within the cord. Qualitative and quantitative analysisshow excellent correlation between myelin water mapsand histology. This study shows that myelin water mea-surements have the potential to be a valuable tool in study-ing WM damage in rat models of SCI.

ACKNOWLEDGMENTS

The authors thank Dr. Alex MacKay for helpful discus-sions and Ms. Clarrie Lam for excellent technical assis-tance.

REFERENCES1. Blight AR. Cellular morphology of chronic spinal cord injury in the cat:

analysis of myelinated axons by line-sampling. Neuroscience 1983;10:521–543.

2. Webster HP. Myelin injury and repair. Adv Neurol 1993;59:67–73.3. Salgado-Ceballos H, Guizar-Sahagun G, Feria-Velasco A, Grijalve I,

Espitia L, Ibarra A, Madrazo I. Spontaneus long-term remyelinationafter traumatic spinal cord injury in rats. Brain Res 1998;782:126–135.

4. Kocsis JD. Restoration of function by glial cell transplantation intodemyelinated spinal cord. J Neurotrauma 1999;16:695–703.

5. Lankford KL, Imaizumi T, Honmou O, Kocsis JD. A quantitative mor-phometric analysis of rat spinal cord remyelination following trans-plantation of allogenic Schwann cells. J Comp Neurol 2002;443:259–274.

6. Henkelman RM, Stanisz GJ, Graham SJ. Magnetization transfer in MRI:a review. NMR Biomed 2001;14:57–64.

7. Whittall KP, MacKay AL. Quantitative interpretation of NMR relax-ation data. J Magn Reson 1989;84:134–152.

8. Whittall KP, MacKay AL, Graeb DA, Nugent RA, Li DK, Paty DW. Invivo measurement of T2 distributions and water contents in normalhuman brain. Magn Reson Med 1997;37:34–43.

9. Laule C, Leung E, Lis DK, Traboulsee AL, Paty DW, MacKay AL, MooreGR. Myelin water imaging in multiple sclerosis: quantitative correla-tions with histopathology. Mult Scler 2006;12:747–753.

10. Laule C, Vavasour IM, Moore GR, Oger J, Li DK, Paty DW, MacKay AL.Water content and myelin water fraction in multiple sclerosis. A T2relaxation study. J Neurol 2004;251:284–293.

11. Chan CC, Khodarahmi K, Liu J, Sutherland D, Oschipok LW, SteevesJD, Tetzlaff W. Dose-dependent beneficial and detrimental effects ofROCK inhibitor Y27632 on axonal sprouting and functional recoveryafter rat spinal cord injury. Exp Neurol 2005;196:352–364.

Myelin Water Measurements in Rat Spinal Cord 801

12. Poon CS, Henkelman RM. Practical T2 quantitation for clinical appli-cations. J Magn Reson Imaging 1992;2:541–553.

13. Araujo CD, MacKay AL, Whittall KP, Hailey JRT. A diffusion model forspin-spin relaxation of compartmentalized water in wood. J MagnReson B 1993;101:248–261.

14. Graham SJ, Stanchev PL, Bronskill MJ. Criteria for analysis of multi-component tissue T2 relaxation data. Magn Reson Med 1996;35:370–378.

15. Gareau PJ, Rutt BK, Karlik SJ, Mitchell JR. Magnetization transfer andmulticomponent T2 relaxation measurements with histopathologiccorrelation in an experimental model of MS. J Magn Reson Imaging2000;11:586–595.

16. Stanisz GJ, Webb S, Munro CA, Pun T, Midha R. MR properties ofexcised neural tissue following experimentally induced inflammation.Magn Reson Med 2004;51:473–479.

17. Yung AC, Choo A, Liu J, Tetzlaff W, Oxland T, Kozlowski P. Measure-ments of myelin water in rat spinal cord in vivo. In: Proc 12th MeetingISMRM, Kyoto; 2004:1537.

18. Hallihan D, Yahya A, Wachowicz K, Hanstock C, Mushahwar VK,Allen PS. The observation of water compartmentalization in vivo infeline lumbar spinal cord. In: Proc 11th Meeting ISMRM, Toronto;2003:2320.

19. MacMillan EL, Madler B, Kolind SH, Yip E, MacKay AL. 3D myelinwater imaging in the spinal cord at 3.0T. In: Proc 15th Meeting ISMRM,Berlin; 2007:2331.

20. Wu Y, Alexander AL, Fleming JO, Duncan ID, Field AS. Myelin waterfraction in human cervical spinal cord in vivo. J Comput Assist Tomogr2006;30:304–306.

21. Yung AC, Kozlowski P. Signal-to-noise ratio comparison of phased-array vs. implantable coil for rat spinal cord MRI. Magn Reson Imaging2007;25:1215–1221.

22. MacKay AL, Laule C, Vavasour I, Bjarnason T, Kolind S, Madler B.Insights into brain microstructure from the T2 distribution. Magn Re-son Imaging 2006;24:515–525.

23. Stoll G, Jander S. The role of microglia and macrophages in the patho-physiology of the CNS. Prog Neurobiol 1999;58:233–247.

802 Kozlowski et al.