Corpus Callosum Axonal Injury in Multiple Sclerosis Measured by Proton Magnetic Resonance Spectroscopic Imaging FREE

The development of noninvasive imaging modalities for monitoring disease progression and evaluating response to therapy has become critical for the treatment of patients with multiple sclerosis (MS). While conventional magnetic resonance images (MRIs) provide important information, there remains a need for more sensitive and specific markers of the biological effects of MS. This is particularly true for following up patients with early-stage disease and evaluating differences in factors associated with the various subtypes of MS. Proton magnetic resonance spectroscopic imaging (¹HMRSI) is a technique that can be combined with conventional MRI that enables the measurement and quantification of the spatial distribution of brain metabolites such as choline-containing compounds (Cho), creatine-phosphocreatine (Cr), and N-acetylaspartate (NAA).¹ N-acetylaspartate is found mainly in neurons and axons of the mature brain² and its intensity level relative to that of Cr has been proposed as an index of axonal damage.³ Strong motivation for using the NAA/Cr ratio as a marker of disease status in patients with MS has come from observations that axonal loss and/or dysfunction in normal-appearing white matter (NAWM) correlate with clinical disability^{4 - 6} as measured by the Expanded Disability Status Scale.⁷ Although some studies have used single-voxel magnetic resonance spectroscopy (MRS) for studying specific regions of interest (ROIs) in patients with MS, ¹HMRSI provides a spatial array of spectral data and is, therefore, particularly valuable for examining variations in metabolite levels in different regions of the brain. Previous studies with ¹HMRSI have shown that decreases in the NAA/Cr ratio extend beyond the border of visible lesions^{4 ,8} as well as within the lesions.^{9 - 12} To be a valuable marker for axonal damage, the NAA/Cr ratio needs to be sensitive to early-stage disease. Although previous studies have shown no significant reduction in the level of NAA in NAWM for patients with clinically isolated syndromes who are at risk of developing MS,¹³ the results have been mixed for patients with relapsing-remitting MS (RRMS).^{5 ,14} It is unclear whether such differences are because of a lack of sensitivity in the techniques being used or because of biological variability in patients with similar disease status.

It is well known that MS lesions have a tendency to cluster in periventricular white matter. This suggests that alterations in the NAA/Cr ratio may also be nonuniformly distributed in NAWM and that regional variations in metabolite levels may be of interest for identifying in vivo measurements of tissue damage. Recent histopathologic studies of the corpus callosum (CC) have reported a significant loss of the total number of axons and the axonal density of fibers relative to nondiseased brain in patients with RRMS and patients with secondary progressive MS (SPMS).¹⁵ If this differential can also be detected using ¹HMRSI based on the values of the NAA/Cr ratio, it would represent a noninvasive marker for assessment of early-stage disease and evaluation of response to therapy. This article tested the hypothesis that changes in metabolite ratios in the normal-appearing CC of patients with MS are larger than the changes in other regions of NAWM centered on the CC.

Twenty-four patients with MS from a large cohort followed up at the Multiple Sclerosis Center, University of California, San Francisco, were considered for this study based on the absence of T1- and T2-weighted visible MRI abnormalities in the CC, confirmed by an experienced MS neurologist (D.P.) (12 patients with RRMS and 12 patients with clinically definite SPMS as defined by Poser et al¹⁶ criteria). Fifteen healthy control subjects were examined using the same MR protocol. All subjects gave their informed written consent.

Magnetic resonance data were acquired using a 1.5-T clinical scanner equipped with a quadrature head coil. Each MRI examination included oblique T2-weighted fast-spin echo (repetition time/echo time [TR/TE], 2000/90 ms; acquisition matrix, 256 × 256 pixels; field of view, 240 × 240 mm²; and 16 contiguous 5-mm-thick sections), axial T2-weighted (TR/TE, 2500/80 ms; acquisition matrix, 192 × 256 pixels; field of view, 180 × 240 mm²; and 48 interleaved 3-mm-thick sections), and axial T1-weighted 3-dimensional spoiled gradient echo (SPGR) (TR/TE, 27/6 ms; flip angle, 40°; acquisition matrix, 192 × 256 × 124 pixels; and field of view, 180 × 240 × 186 mm³) images. The T2-weighted fast-spin echo images were used as a reference for the ¹HMRSI acquisition. Two-dimensional chemical shifting imaging was applied with point-resolved spectroscopy volume selection and 1.5-cc nominal spatial resolution using a commercially available pulse sequence (GE Medical System, Milwaukee, Wis). The point-resolved spectroscopy volume was positioned to cover a slab of approximately 160 cc centered at the middle of the CC (central brain [CB]). The 2-dimensional chemical shifting imaging parameters were as follows: TR/TE, 1000/144 ms; phase encoding matrix, 24 × 24 pixels; field of view, 240×240 mm²; and 15-mm-thick sections. Automatic shimming and water suppression were applied as part of the data acquisition sequence.

After each examination, the images and raw spectra data were transferred to a SUN Ultra 10 workstation (Sun Microsystems, Calif) for postprocessing. The processing algorithms for the ¹HMRSI data were developed in-house and have been described previously.^{17 - 18} The signals were then quantified by both peak height and integrated peak area. Although both measurements were available, the levels of intensity of NAA, Cho, and Cr were calculated from peak height for the current study because it had been observed that the variation in metabolite ratios obtained from peak areas was larger than for the height. It was anticipated that any difference in line width due to variations in shimming would affect all resonances equally.¹⁸ Voxels corresponding to the anatomical ROIs were determined by resampling the corresponding masks according to the point-resolved spectroscopic selection and chemical shifting imaging phase encoding using in-house software described previously.¹⁸

The axial T1-weighted SPGR images were resampled to create high-resolution sagittal images to manually segment out the CC and exclude visible MS lesions. The ROIs corresponding to the CC were drawn in a conservative manner since it was impossible to find sharp boundaries where it merged into white matter. The CC ROIs were saved as a 3-dimensional axial mask image and resampled to correspond to the T2-weighted fast-spin echo volume image set (NAWM_CC). The point-resolved spectroscopic selected volume was composed mainly of white matter with some partial voluming of gray matter and MS lesions. Tissue with the intensity characteristics of NAWM was segmented based on the high-resolution T1-weighted SPGR images with the exclusion of the ROIs corresponding to the CC and of regions that had intensity characteristics of cerebral spinal fluid and MS lesions on corresponding T2-weighted images (NAWM_CB−CC). Masks with constant values in the NAWM_CC and NAWM_CB−CC were generated for analysis of the corresponding spectral intensities. Figure 1 shows the point-resolved spectroscopic selected volume, NAWM_CB−CC and NAWM_CC ROIs for a control and for a patient with SPMS.

The NAWM_CB−CC voxels were determined by including only those that had more than 90% overlap with the region corresponding to NAWM for all subjects. To avoid variations in intensity because of voxels at the edge of the point-resolved spectroscopic selected volume, only the central 8×10 voxels of the chemical shifting imaging array were used to represent those metabolite levels. The NAWM_CC voxels were identified by determining which had more than 20% overlap with the CC region for all subjects. Owing to the limited spatial resolution of the ¹HMRSI used in the current study, some degree of partial voluming with gray matter would be expected for certain NAWM_CC voxels, particularly for patients with callosal atrophy. Findings from a previous study from our laboratory showed that NAA and Cr levels were proportionally higher in the cortical gray matter than white matter.¹⁹ Based on this finding, we believe that the degree of partial voluming with gray matter in the NAWM_CC voxels would not adversely affect the NAA/Cr ratio for all subjects.

Regions of interest corresponding to T1 lesions were drawn based on semiautomated threshold with manual editing on the axial T1-weighted 3-dimensional SPGR volume by an experienced MS neurologist (D.P.) including hypointense relative to the white matter as well as isointense to the gray matter. The axial T2-weighted anatomical images acquired in the same examination were used to confirm that each T1 lesion was associated with a region of T2 hyperintensity.

Statistical analyses were performed using standard least square mean tests with age adjustment to consider the relatively younger mean age of the controls and patients with RRMS with respect to the patients with SPMS in this study. The results were reported as least square mean (SE) unless otherwise noted. The nonparametric Spearman method was used for MR modalities for correlation tests. Statistical significance was set at P≤.05.

The mean age was 45.1 years for all patients and 43.2 years for controls. The mean age, disease duration, Expanded Disability Status Scale score, and T1 lesion load for all subjects are listed in Table 1.

To avoid the effect of differential coil loading on the signal from different subjects, variations in metabolite ratios in the entire point-resolved spectroscopic volume (CB region), NAWM_CB−CC, and NAWM_CC ROIs were compared in this study. Note that the CC and NAWM ROIs considered in this study were chosen to exclude visible T1 (3-dimensional SPGR) and T2 (SE) lesions. This had the effect of reducing the mean number of voxels in the NAWM_CB−CC ROIs from 14.3 in the controls to 10.3 in the the patients with RRMS and 8.3 in the patients with SPMS. Metabolite levels from the NAWM_CB-CC voxels from 1 patient with RRMS and 4 patients with SPMS were excluded from the analysis because there were fewer than 3 voxels satisfying the definition of having more than 90% overlap with the corresponding ROIs. The mean numbers of voxels in the NAWM_CC ROIs were 6.2 in the controls, 6.5 in the patients with RRMS, and 5.5 in the patients with SPMS.

There was a significant reduction in the NAA/Cr ratio for the patients with SPMS (P = .01) compared with the controls (Table 2). There was no difference in the mean value for the patients with RRMS. However, if we consider the metabolite levels from the entire CB rather than just the NAWM, the results are confused by the inability to differentiate between metabolite levels in gray matter, white matter, or lesions and to account for the effects of atrophy. The NAA/Cr ratio was slightly reduced but did not reach significance for the patients with SPMS (P = .35) compared with the controls in the NAWM_CB−CC ROIs (Table 3). There was no difference in the mean value for patients RRMS in the same region.

In healthy controls, it was observed that there were significantly higher NAA/Cr (P<.001), NAA/Cho (P = .03), and Cho/Cr (P = .05) ratios in the NAWM_CCcompared with the NAWM_CB−CC. There were highly significant reductions of the NAA/Cr ratio derived from the NAWM_CC for both patient groups compared with the controls (P = .001 for patients with RRMS and P<.001 for patients with SPMS), and reduction in the Cho/Cr ratio for patients with RRMS (P = .003) and patients with SPMS with marginal significance (P = .09) compared with the controls (Table 3). Figure 2 shows an example of corresponding spectra voxels from the white matter and CC for a control and a patient with SPMS.

We also examined differences in metabolite ratios between NAWM_CC and NAWM_CB−CC voxels within each patient to determine whether these observations were consistent within individual subjects or were owing to overall trends between the population means. The mean of the differences in the NAA/Cr ratio between the NAWM_CC and NAWM_CB−CC is positive in the controls (0.28) but significantly smaller in patients with MS (−0.06 for patients with RRMS [P = .002] and 0.02 for patients with SPMS [P = .04]). There was a similar finding for the differences in the Cho/Cr ratios (0.09 for controls vs − 0.12 for patients with RRMS [P = .001] and −0.08 for patients with SPMS [P = .02]). This suggests that the levels of NAA and Cho are both selectively reduced in the NAWM_CC compared with other regions of NAWM_CB−CC.

There was a significant correlation between T1 lesion load and the NAA/Cr ratio in the entire CB (r = −0.46, P = .03) for all patients combined. Similar correlation was observed in the NAWM_CC region (r = −0.53, P = .01) but not in the NAWM_CB−CC (r = −0.40, P = .10).

This study has addressed variations in metabolite levels in the CC compared with other regions of white matter for the healthy control and MS population. The first finding was that significantly higher NAA/Cr, NAA/Cho, and Cho/Cr ratios were observed in the CC compared with other regions of white matter from the healthy controls. This finding is consistent with previous results from the literature and is thought to be due to the increased axonal density in the highly ordered region of the brain.²⁰ The second finding was that both patients groups showed a significant decrease of the NAA/Cr ratio in the normal-appearing CC compared with both controls and their own NAWM. Although results from our research group showed a small reduction of the NAA/Cr ratio in a small group of patients with early-stage RRMS with a mean disease duration of 5 years, this was not statistically significant in whole supratentorial and CB regions (P = .36 and P = .33, respectively).²¹ The approach suggested in our study, which analyzes differences in metabolite levels in the CC compared with levels from NAWM in the same individual, may assist in reducing the effects of biological variability and, hence, provide a more sensitive assay of axonal damage. There was a trend to a reduced NAA/Cr ratio in voxels from NAWM for patients with SPMS but not for patients with RRMS. In conjunction with findings from our current NAWM_CC, this suggests that the differences in the ability to detect changes in the NAA/Cr ratio in previous studies may be explained by variations in the region of the brain studied and not to intrinsic differences between populations. This may indicate that the normal-appearing CC is a more sensitive region for detecting axonal damage owing to densely packed unidirectional long fiber that may accentuate the reduction in the NAA/Cr ratio compared with other regions of white matter.

Pelletier et al²² showed that patients with early RRMS with mild disability had significant callosal atrophy and its relationship with brain T2 lesion volume and functional transfer indexes. These findings support our hypothesis that demyelinating white matter lesions induce axonal loss and wallerian degeneration that can be detected in the CC. In our study, however, we investigated damages in the NAWM regions caused by remote distant MS lesions, not MS plaque in the CC. A moderate but significant correlation between T1 lesions and the NAA/Cr ratio was observed in the normal-appearing CC for all patients with MS. This may be explained by the transection of axons in lesions causing distant axonal damage and/or dysfunction that are expressed in the normal-appearing CC.^{15 ,23} Such a moderate correlation coefficient can be explained by the fact that some transected axons in lesions simply do not cross the CC. The lack of correlation for the same parameters in the NAWM outside the CC may support our findings that improved sensitivity to detect metabolite changes induced by distant MS lesions was observed in the CC.

Our study also showed that there was reduction of the Cho/Cr ratio within the CC for patients with MS compared with healthy controls. These results may suggest some degree of decreased cellularity.²⁴ The stability of the Cho/Cr ratio in the NAWM voxels is consistent with previous studies.^{14 ,25} The similarity between the NAA/Cho ratio in the CC for the healthy controls and all subgroups of patients with MS is most likely explained by a concomitant reduction in the levels of both NAA and Cho.

We should mention that the point-resolved spectroscopic volume acquired with the 2-dimensional ¹HMRSI sequence used in our study did not always cover all of the CC. This means that the number of voxels used to calculate the metabolite levels and, thus, the reliability of the estimated metabolite ratios could be increased by the use of a 3-dimensional ¹HMRSI technique. Although we are technically capable of performing such data acquisitions, we designed the current study to include the commercially available 2-dimensional ¹HMRSI sequence because we wanted to develop a robust method that could potentially be applied routinely at multiple institutions. Another conservative aspect of our study was the decision to manually segment the CC rather than letting a clinician visually identify the voxels corresponding to this region. This helped us to be objective in interpreting our results, but when we looked at the voxels that were selected using our segmentation procedure, it was clear that manual selection by an individual who was familiar with neuroanatomy would probably have given similar selection of voxels. As the availability of clinical packages for obtaining and analyzing 3-dimensional ¹HMRSI increases it should be possible to simplify the analysis so that metabolite ratios from the cc can be obtained directly on the scanner console and the statistical power of the test can be improved by including a larger number of voxels for each patient. This should provide a sensitive method for routinely obtaining a quantitative measure of axonal damage in patients with MS that can be applied at multiple institutions and used for following disease progression or response to therapy.

This study demonstrates that it is possible to measure significant axonal loss and/or dysfunction in the normal-appearing CC in the brain of patients with MS noninvasively using a commercially available 2-dimensional ¹HMRSI sequence. Significant reductions of the NAA/Cr and Cho/Cr ratios were observed in the normal-appearing CC region compared with the other NAWM regions. This suggests that the CC may be a sensitive location for depicting axonal injury in patients with MS, especially in the early stage of the disease. The ¹HMRSI techniques that were used can be expected to provide the basis for a widely available and robust measure of axonal damage.

Correspondence: Joonmi Oh, PhD, Magnetic Resonance Science Center, Department of Radiology, University of California, San Francisco, 185 Berry St, Suite 350, San Francisco, CA 94107 (joonmi@mrsc.ucsf.edu).

Accepted for publication February 25, 2004.

Author contributions: Study concept and design (Drs Oh, Pelletier, and Nelson); acquisition of data (Drs Oh, Pelletier, and Nelson); analysis and interpretation of data (Drs Oh, Pelletier, and Nelson); drafting of the manuscript (Dr Oh); critical revision of the manuscript for important intellectual content (Drs Oh, Pelletier, and Nelson); statistical expertise (Dr Oh); obtained funding (Drs Pelletier and Nelson); administrative, technical, and material support (Drs Oh, Pelletier, and Nelson); study supervision (Drs Pelletier and Nelson).

This study was supported in part by grants RG2655B6/1 from the National Multiple Sclerosis Society, New York, NY, and R01 NS39529, from the National Institutes of Health, Bethesda, Md.

Dr Oh is a National Multiple Sclerosis Society Postdoctoral Fellowship awardee.