A Magnetization Transfer Imaging Study of Normal-Appearing White Matter in Multiple Sclerosis

Pathologic ^[1,2] and biochemical ^[2-4] postmortem studies have revealed subtle changes in the normal- appearing white matter (NAWM) of multiple sclerosis (MS) patients. More recently, several quantitative magnetic resonance (MR) studies demonstrated such abnormalities in vivo by magnetization decay analysis, ^[5-9] MR proton spectroscopy, ^[10] and magnetization transfer imaging (MTI) ^[11,12].

These MR techniques all provide indirect information about the amount of destruction within brain tissues by measuring different variables possibly involving the most disabling aspects of MS (ie, demyelination and axonal loss). MTI is based on the interactions between protons in a relatively free state and those in a restricted motion state (corresponding to water molecules in close association with macromolecules). Magnetization transfer contrast is obtained by applying a selective irradiation of the latter pool of protons by a radiofrequency centered well beyond or above the frequency of the free water (off-resonance irradiation). This off-resonance irradiation saturates the energy level of macromolecular and bound protons, thus inducing energy transfer to the mobile protons and, consequently, reducing their signal intensity. The difference of signal intensity without and with magnetization transfer can be easily and reliably measured ^[11]. It indirectly reflects the capacity of the matrix molecules to exchange energy with the surrounding water molecules, thus indicating the integrity or the destruction of the matrix. In the brain, the main macromolecular matrix consists of myelin and other cell membranes, so MTI seems to be very promising for monitoring the evolution of MS.

There are some discrepancies between the extent of lesions seen on conventional MRI and the clinical status of the patients ^[13]. Several biologic explanations and methodologic limitations may account for this lack of correlation, as discussed elsewhere ^[13-15]. One of these possible explanations is that the NAWM might be more diffusely and severely affected in more disabled patients, ^[16] contributing, along with other factors, to determining the disability in MS. Previous MR studies evaluating this issue have provided conflicting results. Several authors found no difference in the relaxation times (RTs) of NAWM of MS patients with different disease courses ^[5,9] and disabilities, ^[8] but Thompson et al ^[16] demonstrated higher T₁ values for the NAWM of patients with secondary progressive MS than for the NAWM of patients with primary progressive MS, and Dousset et al ^[11] found lower MT ratios (MTRs) for the NAWM of patients with chronic-progressive MS than for those with relapsing-remitting MS. Finally, a recent pixel-by-pixel RT study ^[17] found a significant proportion of the NAWM in MS to contain multiple discrete small areas, often of one or two pixels, suggesting again a possible role of the "invisible" lesion load in the evolution of MS.

This study had three major aims. First, to confirm the potential of MTI for detecting subtle changes in the NAWM of MS. Second, to evaluate whether such changes are more striking in the NAWM adjacent to the lesions. Third, to evaluate whether the NAWM of more disabled patients is more severely affected, providing insight into the nature of disability in MS.

Methods. Patients. Patients with clinically definite MS, ^[18] having either relapsing-remitting MS or chronic-progressive MS, were considered for the study. Patients with relapsing-remitting MS were defined as those having a clinical course characterized by acute exacerbations followed by complete or incomplete recoveries (ie, leaving a mild or moderate residual disability) and separated in time by periods of relative disease stability. Patients with chronic-progressive MS were those who presented a progression of neurologic deficits without periods of significant remission for at least 6 months before entering the study. At the time MTI was performed, a detailed clinical history was obtained and a full neurologic examination was done. Disability was assessed by the Expanded Disability Status Scale ^[19].

MR. MR was performed on a 1.5-tesla machine. Proton density and T sub 2-weighted spin-echo (SE) images of the brain were acquired (TR/TE = 2,000/40,80; slice thickness = 5 mm with an interslice gap of 0.5 mm; matrix = 192 x 256) to detect "isolated" MS lesions. T₁-weighted SE images (TR/TE = 600/17; slice thickness = 5 mm with an interslice gap of 0.5 mm) were obtained after gadolinium-DTPA injection (0.1 mmol/kg) to evaluate whether such lesions were enhancing or not. Quantitative, semiautomated computer assessment of the lesion load was performed using 5-mm contiguous axial 2,000/50 SE slices. The technique, which is fully described elsewhere, ^[20] involved simple intensity threshold segmentation, first to isolate the brain from the surrounding tissues, then to identify the MS white matter lesions. A manual review of the computed lesion load was performed for each assessment to correct any error in lesion detection. MTI studies were performed prior to the injection of gadolinium-DTPA by obtaining 2-D gradient-echo images (TR/TE = 600/12; slice thickness = 5 mm with an interslice gap of 2 mm; matrix = 192 x 256) with and without a saturation pulse. The saturation pulse had the following parameters: off-resonance gaussian RF pulse centered 1.5 kHz below the water frequency, with a duration of 16.384 msec, a bandwidth of 250 Hz, and a power intensity of 3.4 x 10 sup - 6 tesla. The magnetization transfer was quantified as a percentage of signal loss according to the following equation: MTR = (S₀ -S_s)/S₀ x 100, in which S₀ is the mean signal intensity for a given region without the saturation pulse and S_s is the mean signal intensity for the same region when the saturation pulse is applied. The MTRs of all the isolated unenhancing lesions present in each patient and of the adjacent NAWM were measured. Isolated lesions were those that were surrounded on the same slice and in at least the two adjacent ones by macroscopically normal white matter in order to avoid partial volume effects from other lesions. Regions of interest (ROI) of 2 mm in diameter were placed in the center of the lesion, and three others were placed in contiguous NAWM areas, starting from the external edge of the lesions and progressing toward the cortical gray matter (first, second, and third NAWM ROI). In addition, for each patient, MTRs were calculated for two regions of NAWM in the frontal lobes. In the same areas, MTRs were also calculated in 10 age- and sex-matched healthy volunteers. All the MR measurements were performed concordantly by three of us (M.F., A.C., and C.B.), none of whom knew to which group the patient belonged.

Statistical analysis. The data were analyzed statistically by the Student's t test or one-way ANOVA. When the data sets were not distributed normally, the Mann-Whitney test was used.

Results. Clinical data. Twenty patients had relapsing-remitting MS and seven had chronic- progressive MS. The two groups did not differ in age and disease duration, although patients with chronic-progressive MS were significantly more disabled (p < 0.0001) (table 1).

MRI lesion load. The median lesion load was 4,165 mm³ (range = 575 to 12,605 mm³) for patients with chronic-progressive MS and 7,995 mm³ (range = 540 to 31,175 mm³) for those with relapsing-remitting MS. The difference was not statistically significant.

MTI. Frontal NAWM of patients had lower mean MTRs than the frontal white matter of the controls (mean MTR +-\SD = 49.8 +-\1.6% for patients versus 50.8 +-\1.7% for controls, p = 0.02).

Forty-one isolated lesions were studied (30 from patients with relapsing-remitting MS and 11 from patients with chronic-progressive MS). Their mean MTR +-\SD was 40.7 +-\4.1% and was significantly lower than those of all the NAWM regions evaluated (p = 0.0001). For the group as a whole, the mean MTR +-\SD of NAWM adjacent to lesions increased progressively from the first NAWM ROI (48.1 +-\3.0%) to the second NAWM ROI (48.8 +-\2.3%) and the third NAWM ROI (49.5 +-\2.1%) (p = 0.04). At post-hoc analysis, the MTRs obtained for the first NAWM ROI were significantly lower than those of the third NAWM ROI (t = 2.4, df = 80, p = 0.02); the mean difference between the MTRs of the third NAWM ROI and those of the first NAWM ROI was 1.4% (range = -8 to +9%). This pattern of increased MTRs in NAWM moving away from the lesions was found to apply to patients with chronic-progressive MS (t = 2.1, df = 20, p = 0.05) but not to patients with relapsing-remitting MS (t = 1.6, df = 58, p = 0.1) (table 2).

The MTRs of lesions, frontal NAWM, and third NAWM ROI were similar for patients with chronic-progressive and relapsing-remitting MS, whereas significant differences were found in the first (p = 0.01) and the second (p = 0.02) NAWM ROIs, with the MTRs being lower in patients with chronic- progressive MS (table 2).

Discussion. We performed this study to evaluate the role of subtle changes in the NAWM in influencing the natural history of MS. We had three main questions: (1) Has MTI the potential to detect such changes in vivo? (2) If so, are these changes more frequent and severe around lesions visible on conventional MRI? (3) And, if so, are these abnormalities detectable in patients with different patterns of disease evolution, which determine different degrees of disability?

(1) MTI of the frontal NAWM. Previous studies have already demonstrated that MTI is useful for evaluation of the pathologic features of both chronic ^[11,12] and new active ^[21,22] MS lesions and that the average MTR of all the brain lesions is correlated with the degree of disability in patients with MS ^[12]. However, the results obtained for the NAWM are still conflicting ^[11,12]. Dousset et al ^[11] reported that the MTR of the NAWM in MS patients was significantly lower, particularly in patients with chronic-progressive MS, than that of the white matter of controls, whereas this difference was not found by Gass et al ^[12]. There are several possible explanations for this discrepancy, such as the MR methodology, the extent of white matter disease, and the clinical characteristics of the patients. However, another factor might be responsible for these conflicting results. Barbosa et al ^[17] suggested in a pixel-by-pixel RT study that microscopic changes in the NAWM of MS do not occur randomly and diffusely but are related to the presence of small, discrete abnormalities, often only one to two pixels in size. If so, when NAWM is studied, the MTR might vary significantly in relation to the number of pixels that are involved by the pathologic process within the ROI. This suggestion seems to be confirmed by an RT study ^[8] and a spectroscopic study ^[23] that found some samples of NAWM in MS patients to be normal and by a histologic study in which 28% of specimens were normal ^[2].

Because of the above, we decided to study only small areas of NAWM in the frontal lobes. Frontal lobes are relatively spared by the pathologic process in MS and therefore the risk of partial volume effects is markedly reduced. In addition, the analysis of the NAWM far from visible lesions was restricted to the frontal lobe, since MT is known to change in different brain regions ^[12] depending on the extent of myelination, the varying amounts of bound water in myelin, or the density of nerve fibers. Thus, we decided to avoid this additional confounding variable.

With all these precautions, we found lower mean MTRs for the frontal NAWM of MS patients than for the frontal white matter of the controls. The pathologic nature of this reduction in MT is still unknown. It might derive in part from small lesions that are beyond the resolution of the system ^[17] and in part from the widespread abnormalities described by pathologic studies ^[2]. These abnormalities, which include diffuse astrocytic hyperplasia, patchy edema, and perivascular cellular infiltration, ^[1,2] can modify the relative proportions of mobile and immobile protons, thus inducing changes in MT. In this respect, a postmortem study demonstrated water content in the brain NAWM of MS patients to be 4% greater than that of normal controls ^[24]. Moreover, Arstila et al ^[25] described abnormally thin myelin in biopsies from NAWM of MS patients.

(2) NAWM adjacent to the lesions. MTRs of the NAWM were lowest around the lesions visible on conventional MRI and increased progressively moving away from the lesions to values similar to those in the NAWM of the frontal lobes.

Since we considered only NAWM around lesions that were not surrounded on the same slice and in at least the two adjacent slices by other visible lesions, partial volume effects from other lesions can be reasonably excluded. This is even more evident when the small size of the ROIs (2 mm in diameter) is compared with the larger thickness of the slices (5 mm) and of the interslice gap (2 mm).

Therefore, two other explanations seem to be more likely. First, there might be a partial volume effect from the same lesion surrounded by the NAWM studied. Even if no abnormalities at all were visible on conventional MRI around these lesions, microscopically, the shapes of such lesions are not so well defined ^[26]. Second, the presence of small lesions or widespread abnormalities might be more frequent and severe in NAWM closer to lesions, thus causing a "penumbra" around the visible lesions.

These explanations are not mutually exclusive, and whatever the explanation, this study confirms that there is an invisible lesion load in patients with MS. If this finding is combined with those of previous reports indicating that many lesions detected pathologically in the corpus callosum, infratentorial regions, and especially the cortical gray matter are not seen on conventional MRI ^[27,28] and that NAWM contains scattered small areas with prolonged RT, located especially on or near the boundary between gray and white matter, ^[17] the extent of such an invisible lesion load is important and might at least partially account for some of the discrepancies between the clinical and MRI findings in MS reported by several studies using conventional MRI ^[13-15]. We also found that patients with relapsing-remitting MS had visible lesion volumes similar to patients with chronic-progressive MS, despite their being significantly less disabled.

(3) NAWM in different clinical MS subgroups. Our study also indicates that the abnormalities of the NAWM adjacent to the lesions are more striking in patients with chronic-progressive MS than in those with relapsing-remitting MS. The former group has significantly lower MTRs in the first and second NAWM ROIs, suggesting a larger involvement of the white matter in the more disabled patients, even though white matter appears normal on conventional MRI.

A recent longitudinal MRI study demonstrated that new lesion formation is correlated with changes in disability in relapsing-remitting MS, although it does not correlate in patients with chronic-progressive MS, raising the possibility that other mechanisms might be related to the development of the disability in such patients ^[29]. One possible mechanism might be the presence of repeated or continuous activity in chronic lesions that could lead to greater degrees of axonal loss or demyelination in such lesions and might be responsible for a diffusion of the pathologic process into the macroscopically NAWM. This observation needs confirmation, especially since we studied isolated lesions to exclude partial volume effects from surrounding lesions. This is an artificial situation in MS in which the lesions tend to be confluent.

In conclusion, this study indicates that MTI has the potential to detect subtle changes of the NAWM in MS and that such changes are more severe around visible lesions. These NAWM abnormalities seem greater in more disabled patients and therefore may influence the natural history of the disease. If confirmed by further studies, the role of MTI in monitoring the efficacy of treatment in MS should be considered.

Acknowledgments

We would like to thank Dr. Mark A. Horsfield, Department of Medical Physics, Leicester Royal Infirmary, Leicester, UK; Drs. David H. Miller and Gareth J. Barker, NMR Research Group, Institute of Neurology, London, for providing the software for the semi-automated quantitative assessment of lesion load; and Mr. Clodoaldo Pereira for his skillful technical assistance.