Evidence for widespread axonal damage at the earliest clinical stage of multiple sclerosis

Transcription

1 DOI: /brain/awg038 Brain (2003), 126, 433±437 Evidence for widespread axonal damage at the earliest clinical stage of multiple sclerosis M. Filippi, 1 M. Bozzali, 1 M. Rovaris, 1 O. Gonen, 5 C. Kesavadas, 3 A. Ghezzi, 4 V. Martinelli, 2 R. I. Grossman, 5 G. Scotti, 3 G. Comi, 2 and A. Falini 3 1 Neuroimaging Research Unit, Department of Neuroscience Departments of 2 Neurology and 3 Neuroradiology, Scienti c Institute and University Ospedale San Raffaele, Milan 4 Multiple Sclerosis Center, Ospedale di Gallarate, Gallarate, Italy and 5 Department of Radiology, New York University School of Medicine, New York, NY, USA Summary Although axonal pathology is recognized as one of the major pathological features of multiple sclerosis, it is less clear how early in its course it occurs and how it correlates with MRI-visible lesion loads. To assess this early axonal pathology, we quanti ed the concentration of whole-brain N-acetylaspartate (WBNAA) in a group of patients at the earliest clinical stage of the disease and compared the results with those from healthy controls. Conventional brain MRI and WBNAA using unlocalized proton magnetic resonance spectroscopy were obtained from 31 patients at presentation with clinically isolated syndromes suggestive of multiple sclerosis and paraclinical evidence of dissemination in space, and from 16 matched controls. An additional conventional MRI scan was obtained in all patients 4±6 months later to detect dissemination of lesions in time. Correspondence to: Dr Massimo Filippi, Neuroimaging Research Unit, Department of Neuroscience, Scienti c Institute and University, Ospedale San Raffaele, Via Olgettina 60, Milan, Italy m. lippi@hsr.it The mean WBNAA concentration was signi cantly lower in patients compared with the controls (P < ). It was not signi cantly different between patients with and without enhancing lesions at the baseline MRI or between patients with and without lesion dissemination in time. No correlation was found between WBNAA concentrations and lesion volumes. Widespread axonal pathology, largely independent of MRI-visible in ammation and too extensive to be completely reversible, occurs in patients even at the earliest clinical stage of multiple sclerosis. This nding lessens the validity of the current concept that the axonal pathology of multiple sclerosis is the end-stage result of repeated in ammatory events, and argues strongly in favour of early neuroprotective intervention. Keywords: multiple sclerosis; magnetic resonance imaging; magnetic resonance spectroscopy; whole-brain N-acetylaspartate; axonal pathology Abbreviations: EDSS = Expanded Disability Status Scale; Gd = gadolinium; 1 H-MRS = proton MR spectroscopy; MR = magnetic resonance; NAA = N-acetylaspartate; PD = proton-density; SE = spin-echo; SIENA = structural image evaluation, using normalization, of atrophy; VOI = volume of interest; WBNAA = whole-brain N-acetylaspartate Introduction Axonal pathology has been known to occur in multiple sclerosis from the earliest post-mortem descriptions of the disease (Charcot, 1868). Recent pathological studies have con rmed that irreversible axonal loss is a prominent feature of the disease (Ferguson et al., 1997; Trapp et al., 1998; van Waesberghe et al., 1999; Evangelou et al., 2000; Peterson et al., 2001), and the extensive in vivo application of quantitative magnetic resonance (MR) techniques has led to a general appreciation that axonal loss contributes signi cantly to the development of ` xed' disability in multiple ã Guarantors of Brain 2003 sclerosis patients (Davie et al., 1995; De Stefano et al., 1995, 1998; Trapp et al., 1999; Filippi, 2001). However, it is less clear how early in the course of the disease this axonal damage occurs and how it relates to the MRI-visible lesion load. Proton magnetic resonance spectroscopy ( 1 H-MRS) enables us to quantify axonal injury and loss through the measurement of changes in the signal intensity of N-acetylaspartate (NAA) (Davie et al., 1995; De Stefano et al., 1995, 1998), a metabolite localized almost exclusively

2 434 M. Filippi et al. to neurons and neuronal processes (Simmons et al., 1991). Localized 1 H-MRS studies have shown reduced NAA in T 2 -visible lesions (Arnold et al., 1994; Davie et al., 1994, 1997; De Stefano et al., 1995; Narayana et al., 1998), normalappearing white matter (NAWM) (Davie et al., 1995; De Stefano et al., 1998; Fu et al., 1998; Narayana et al., 1998; Kapeller et al., 2001) and grey matter (Kapeller et al., 2001) of patients with clinically de nite multiple sclerosis. Unfortunately, current 1 H-MRS is restricted to small volumes of interest (VOI), which typically represent <100 ml of the brain. Since multiple sclerosis is a diffuse disease of the CNS (Trapp et al., 1999; Filippi, 2001), estimates of the extent of axonal pathology from such measurements must rely on the unproven assumption that changes in NAA in the VOI accurately re ect the status of the entire brain. The volume restriction of localized 1 H-MRS has recently been overcome with a new unlocalized 1 H-MRS sequence which makes it possible to quantify the concentration of NAA from the whole brain (WBNAA) (Gonen et al., 1998, 2000), providing an overall assessment of the total load of the disease both within and outside MRI-visible lesions. To gain insight into the nature of axonal pathology in multiple sclerosis, we quanti ed WBNAA from a group of patients at the earliest clinical stage of multiple sclerosis (McDonald et al., 2001) and correlated it with the MRI-visible lesion loads. Material and methods Patients We studied 31 consecutive patients (23 women, eight men, mean age 28.7 years, SD 6.0 years) at presentation with clinically isolated syndromes suggestive of multiple sclerosis and paraclinical evidence of dissemination in space (McDonald et al., 2001). All patients were assessed within 3 months from the onset of the clinical symptoms (mean 26 days). The presenting symptoms were optic neuritis, brainstem syndromes and spinal cord syndromes in 10, 12 and nine patients, respectively. All alternative neurological diseases were excluded by appropriate investigations (McDonald et al., 2001). Oligoclonal bands were found in the CSF of 26 patients. Before MR acquisition, but on the same day, each patient underwent a full neurological assessment performed by a single observer who was blind to the MRI and WBNAA results. Disability was rated using the Expanded Disability Status Scale (EDSS) score (Kurtzke, 1983), which had a median of 0.0 (range 0.0±2.0) in this cohort. None of the patients had been treated with corticosteroids within the previous month. To demonstrate the presence of new lesions, and hence the dissemination in time of the disease (McDonald et al., 2001), all patients were followed up both clinically and with a second MRI 4±6 months after the rst. Sixteen healthy volunteers (12 women, four men; mean age 29.7 years; SD 4.5 years), with no history of neurological dysfunction and a normal neurological examination, served as controls. We obtained approval from The Fondazione San Raffael Ethics Committee and written informed consent from each subject before study initiation. MR acquisition First session 1 H-MRS and MRI were performed using a 1.5 T scanner (Vision; Siemens, Erlangen, Germany). The following sequences were collected from all subjects during a single MR session: (i) WBNAA 1 H-MRS based on the method described by Gonen et al. (2000); (ii) dual-echo turbo spinecho (SE) [repetition time (TR) 3300 ms; rst echo time (TE) 16 ms; second echo TE 98 ms; echo-train length 5; 24 contiguous 5 mm thick axial slices with a matrix and a mm 2 eld of view]; (iii) T 1 -weighted conventional SE (TR 768 ms, TE 14 ms, 24 contiguous 5 mm thick axial slices with a matrix and a mm 2 eld of view) before and after the administration of 0.1 mmol/kg of gadolinium (Gd). Second session Patients were repositioned using ad hoc guidelines (Miller et al., 1991). Dual-echo turbo SE and T 1 -weighted conventional SE before and after administration of 0.1 mmol/kg of Gd were obtained in a single session using the same acquisition procedures as those used previously. MRI analysis Dual-echo hard copies were revised in random order by two observers who did not know who the scans belonged to. Lesions were identi ed and marked by consensus on the proton-density (PD) and post-contrast T 1 -weighted scans. T 2 -weighted and precontrast T 1 -weighted images were always used to increase con dence in lesion identi cation. The number and location of T 2 -hyperintense and Gd-enhancing lesions were evaluated, and the ful lment of MRI criteria for multiple sclerosis diagnosis was assessed (McDonald et al., 2001). Then all images were transferred to an of ine workstation (Sun Sparcstation; Sun Microsystems, Mountain View, CA, USA) for lesion volume assessment by a single observer, again unaware of who the scans belonged to. This procedure used a segmentation technique based on signal intensity thresholding and characterized by high intra-rater reproducibility (Rovaris et al., 1997). In patients, changes in normalized brain volume over the follow-up period were also measured using a fully automated and accurate method, SIENA (structural image evaluation, using normalization, of atrophy) (Smith et al., 2001). SIENA performs segmentation of brain from nonbrain tissue in the head, estimates the outer skull surface (for two time points), and uses these results to register the two

3 Axonal damage in early multiple sclerosis 435 images while correcting for imaging geometry changes. Then the registered segmented brain images are used to estimate percentage changes in normalized brain volume (Smith et al., 2001). WBNAA 1 H-MRS analysis The 1 H-MRS data from each subject and from the reference phantom were transferred to the workstation and processed of ine with our custom software (IDL; Research Systems, Boulder, CO, USA). The NAA peak area was integrated by two operators unaware of the subject's identity (Gonen et al., 1998, 2000). It was converted into an absolute amount (in mmol) by scaling against the area of the signal from the reference phantom which contained 15 mmol NAA (5 mm) in water (Gonen et al., 1998, 2000). To correct for the considerable natural interindividual variation in brain size, the absolute amount of NAA from each individual was divided by their brain volume. For this purpose, absolute rather than normalized brain volumes were needed. Brain volumes were measured using a seed-growing technique based on signal intensity thresholding, as described extensively elsewhere (Rovaris et al., 2000). This yielded an absolute WBNAA concentration, in mm, which could be compared cross-sectionally (Gonen et al., 2000). Statistical analysis Differences in WBNAA concentrations between individual groups of subjects were assessed using the Mann±Whitney test. To correct for multiple comparisons, only P values <0.01 were considered signi cant (Bonferroni correction). Univariate correlations were assessed using the Spearman rank correlation coef cient. Results No abnormalities were seen on any of the conventional MRI scans from the controls. By de nition, all patients had paraclinical evidence of dissemination in space; in 19 this was demonstrated by MRI alone and in the remaining 12 by the combination of MRI ndings and the presence of oligoclonal bands in the CSF (McDonald et al., 2001). One or more enhancing lesions were seen in eight patients (seven had 1 and one had 7 enhancing lesions). From the clinical and MRI follow-up data, 16 patients met McDonald's criteria for lesion dissemination in time (McDonald et al., 2001). The average WBNAA concentration was mm (standard error 0.58 mm) in controls and mm (standard error 0.43 mm) in patients (P < ). Speci cally, the patients showed, on average, a 22.3% decrease in NAA concentration in the whole-brain parenchyma. No signi cant difference was found (P = 0.17) in average WBNAA concentration between patients with dissemination in space shown by MRI alone (mean 9.82 mm, standard error 0.57 mm) and those in whom a combination of MRI and CSF ndings was needed to demonstrate dissemination in space (McDonald et al., 2001) (mean mm, standard error 0.61 mm). Average WBNAA concentration also did not differ signi cantly (P = 0.27) between patients with (mean 9.32 mm, standard error 0.67 mm) and without (mean mm, standard error 0.53 mm) enhancing lesions on the baseline MRI. No correlation was found between WBNAA concentrations and lesion volume (WBNAA versus T 2 lesion volume, r = ±0.18; WBNAA versus T 1 -hypointense lesion volume, r = 0.05). The mean brain volume and the average WBNAA concentrations did not differ (P = 0.57) between patients ful lling McDonald's criteria at follow-up (mean 9.78 mm, standard error 0.59 mm) and those who did not (mean mm, standard error 0.62 mm). Between the rst and the second MRI sessions, a mean 0.27% (standard error 0.16%) decrease in normalized brain volume was observed in the patients. Considering the length of the follow-up (average 4.7 months), this gure corresponds to an average annualized percentage decrease in normalized brain volume of 0.69%. Discussion The NAA loss indicates that axonal pathology is an early event in the course of multiple sclerosis and ts with the demonstration that brain atrophy occurs at a relatively fast rate in patients with early multiple sclerosis (Rudick et al, 1999; Brex et al., 2000). This observation is corroborated by the nding of an average annualized percentage decrease of normalized brain volume of about 0.7%, a gure that closely matches those reported in patients with more advanced disease (Filippi and Grossman, 2002). Admittedly, not all the patients in this study met McDonald's criteria for an early diagnosis of multiple sclerosis (McDonald et al., 2001). Nevertheless, alternative diagnoses were carefully excluded and all had clinical pictures suggestive of multiple sclerosis at presentation. They were also selected for having a dissemination in space at presentation, i.e. MRI and/or CSF ndings consistent with a diagnosis of multiple sclerosis (McDonald et al., 2001). As a consequence, we believe that all of them are at high risk of ultimately progressing to clinically de nite multiple sclerosis (Brex et al., 2002). With this in mind, the rst issue to be addressed is how much of the observed WBNAA decrease is irreversible. Partially reversible NAA reductions in T 2 -visible lesions (Davie et al., 1994; De Stefano et al., 1995; Narayana et al., 1998) and NAWM (De Stefano et al., 1999) of multiple sclerosis patients have been described and attributed to possible recovery from sublethal axonal injury following resolution of in ammatory changes. Consequently, the reduced WBNAA concentrations in these patients could be interpreted as a transient functional axonal impairment secondary to in ammatory changes associated with the recent clinical attack. Indeed, we found that patients with at least one enhancing lesion, indicative of increased blood±brain barrier permeability and associated in ammation (Katz et al., 1993), had a lower average WBNAA than those with an inactive

4 436 M. Filippi et al. MRI scan. Nevertheless, the average WBNAA reduction in the latter group of patients was still remarkable (19.9%) compared with the matched controls. Although a certain amount of in ammation is known to go undetected on conventional MRI with the standard Gd dose (Filippi et al., 1998), we believe it is unlikely that this `occult' in ammatory component could lead to such a substantial WBNAA decrease. Therefore, we interpret such `NAA loss without MRI-visible in ammation' as an indication that at least a considerable portion of the axonal pathology is permanent, even this early in the clinical course of the disease. The next question relates to the relative contributions of MRI-visible lesions and occult damage in the normalappearing tissue to the reduced WBNAA concentrations in these patients. Since average T 2 lesion volume at this early stage of the disease is negligible (<1% fraction of the brain volume; on average, 2.6 versus 1165 ml) the decrease of >20% in WBNAA in the patients must be accounted for by widespread axonal damage in the NAWM alone or in association with neuronal injury in the grey matter. Our results disagree with those of a previous study (Brex et al., 1999), which was unable to detect a signi cant reduction in NAA concentration in the NAWM from patients with clinically isolated syndromes suggestive of multiple sclerosis. This discrepancy is likely to result from the fact that Brex et al. (1999) did not select patients for having a dissemination in space at presentation (McDonald et al., 2001) and, most importantly, because their NAA concentrations were based on VOIs with a mean volume of 2.64 ml, which were positioned in various white matter areas according to the distribution of T 2 -visible lesions. On the contrary, our interpretation of the results of the present study is in agreement with the demonstration of reduced central brain NAA/creatine (Cr) resonance intensity ratios in patients with early and non-disabling relapsing±remitting multiple sclerosis (De Stefano et al., 2001). Interestingly, in the study by De Stefano et al. (2001), NAA/Cr ratios were derived from VOIs much larger than those used by Brex et al. (1999), and T 2 -visible lesions accounted for only ~6% of the VOI. This interpretation is also consistent with post-mortem data showing decreased axonal density in the NAWM (Trapp et al., 1998; van Waesberghe et al., 1999; Evangelou et al., 2000) and the presence of cortical lesions associated with apoptotic loss of neurons, and axonal and dendritic transections (Peterson et al., 2001) in patients with long-standing multiple sclerosis. The novelty of this study is to show that axonal/neuronal damage can occur very early in the course of multiple sclerosis, when de nitive histopathological data are unlikely ever to be obtained. The lack of correlation between lesion volumes and WBNAA loss also suggests that neuronal/ axonal damage in the normal-appearing tissue at this earliest stage of multiple sclerosis occurs by mechanisms mostly independent of the axonal transection known to occur within in amed white-matter multiple sclerosis lesions (Ferguson et al., 1997; Trapp et al., 1998). In conclusion, this study suggests that signi cant irreversible axonal damage occurs in patients at the earliest stage of multiple sclerosis. This, together with the recognition that MRI-detectable in ammation is only indirectly linked to the neuronal/axonal damage in these patients should prompt the development and use of treatments capable of neuroprotective action. Acknowledgements This study was supported by a grant from the Italian Ministry of Health (contract ICS030.5/RF00.79), an unrestricted educational grant from Teva Pharmaceutical and NIH grants NS33385, NS37739 and NS References Arnold DL, Riess GT, Matthews PM, Francis GS, Collins DL, Wolfson C, et al. Use of proton magnetic resonance spectroscopy for monitoring disease progression in multiple sclerosis. Ann Neurol 1994; 36: 76±82. Brex PA, Gomez-Anson B, Parker GJ, Molyneux PD, Miszkiel KA, Barker GJ, et al. Proton MR spectroscopy in clinically isolated syndromes suggestive of multiple sclerosis. J Neurol Sci 1999; 166: 16±22. Brex PA, Jenkins R, Fox NC, Crum WR, O'Riordan JI, Plant GT, et al. Detection of ventricular enlargement in patients at the earliest clinical stage of MS. Neurology 2000; 54: 1689±91. Brex PA, Ciccarelli O, O'Riordan JI, Sailer M, Thompson AJ, Miller DH. A longitudinal study of abnormalities on MRI and disability from multiple sclerosis. New Engl J Med 2002; 346: 158±64. Charcot JM. Histologie de la scleârose en plaque. Gaz Hop (Paris) 1868; 141: 554±8. Davie CA, Hawkins CP, Barker GJ, Brennan A, Tofts PS, Miller DH, et al. Serial proton magnetic resonance spectroscopy in acute multiple sclerosis lesions. Brain 1994; 117: 49±58. Davie CA, Barker GJ, Webb S, Tofts PS, Thompson AJ, Harding AE, et al. Persistent functional de cit in multiple sclerosis and autosomal dominant cerebellar ataxia is associated with axon loss. Brain 1995; 118: 1583±92. Davie CA, Barker GJ, Thompson AJ, Tofts PS, McDonald WI, Miller DH. 1H magnetic resonance spectroscopy of chronic cerebral white matter lesions and normal appearing white matter in multiple sclerosis. J Neurol Neurosurg Psychiatry 1997; 63: 736±42. De Stefano N, Matthews PM, Antel J, Preul M, Francis GS, Arnold DL. Chemical pathology of acute demyelinating lesions and its correlation with disability. Ann Neurol 1995; 38: 901±9. De Stefano N, Matthews PM, Fu L, Narayanan S, Stanley J, Francis GS, et al. Axonal damage correlates with disability in patients with relapsing-remitting multiple sclerosis. Results of a longitudinal magnetic resonance spectroscopy study. Brain 1998; 121: 1469±77. De Stefano N, Narayanan S, Matthews PM, Francis GS, Antel JP, Arnold DL. In vivo evidence for axonal dysfunction remote from

5 Axonal damage in early multiple sclerosis 437 focal cerebral demyelination of the type seen in multiple sclerosis. Brain 1999; 122: 1933±9. De Stefano N, Narayanan S, Francis GS, Arnaoutelis R, Tartaglia MC, Antel JP, et al. Evidence of axonal damage in the early stages of multiple sclerosis and its relevance to disability. Arch Neurol 2001; 58: 65±70. Evangelou N, Konz D, Esiri MM, Smith S, Palace J, Matthews PM. Regional axonal loss in the corpus callosum correlates with cerebral white matter lesion volume and distribution in multiple sclerosis. Brain 2000; 123: 1845±9. Ferguson B, Matyszak MK, Esiri MM, Perry VH. Axonal damage in acute multiple sclerosis lesions. Brain 1997; 120: 393±9. Filippi M. Linking structural, metabolic and functional changes in multiple sclerosis. [Review]. Eur J Neurol 2001; 8: 291±7. Filippi M, Grossman RI. MRI techniques to monitor MS evolution. The present and the future. [Review]. Neurology 2002; 58: 1147±53. Filippi M, Rovaris M, Capra R, Gasperini C, Yousry TA, Sormani MP, et al. A multi-centre longitudinal study comparing the sensitivity of monthly MRI after standard and triple dose gadolinium-dtpa for monitoring disease activity in multiple sclerosis. Implications for phase II clinical trials. Brain 1998; 121: 2011±20. Fu L, Matthwes PM, De Stefano N, Worsley KJ, Narayanan S, Francis GS, et al. Imaging axonal damage of normal-appearing white matter in multiple sclerosis. Brain 1998; 121: 103±13. Gonen O, Viswanathan AK, Catalaa I, Babb J, Udupa J, Grossman RI. Total brain N-acetylaspartate concentration in normal, agegrouped females: quantitation with non-echo proton NMR spectroscopy. Magn Reson Med 1998; 40: 684±9. Gonen O, Catalaa I, Babb JS, Ge Y, Mannon LJ, Kolson DL, et al. Total brain N-acetylaspartate. A new measure of disease load in MS. Neurology 2000; 54: 15±9. Kapeller P, McLean MA, Grif n CM, Chard D, Parker GJ, Barker GJ, et al. Preliminary evidence for neuronal damage in cortical grey matter and in normal appearing white matter in short duration relapsing-remitting multiple sclerosis: a quantitative MR spectroscopic imaging study. J Neurol 2001; 248: 131±8. Katz D, Taubenberger JK, Cannella B, McFarlin DE, Raine CS, McFarland HF. Correlation between magnetic resonance imaging ndings and lesion development in chronic, active multiple sclerosis. Ann Neurol 1993; 34: 661±9. Kurtzke JF. Rating neurologic impairment in multiple sclerosis: an expanded disability status scale (EDSS). Neurology 1983; 33: 1444±52. McDonald WI, Compston A, Edan G, Goodkin D, Hartung HP, Lublin FD, et al. Recommended diagnostic criteria for multiple sclerosis: guidelines from the international panel on the diagnosis of multiple sclerosis. Ann Neurol 2001; 50: 121±7. Miller DH, Barkof F, Berry I, Kappos L, Scotti G, Thompson AJ. Magnetic resonance imaging in the monitoring treatment of multiple sclerosis: concerted action guidelines. J Neurol Neurosurg Psychiatry 1991; 54: 683±8. Narayana PA, Doyle TJ, Lai D, Wolinsky JS. Serial proton magnetic resonance spectroscopic imaging, contrast-enhanced magnetic resonance imaging, and quantitative lesion volumetry in multiple sclerosis. Ann Neurol 1998; 43: 56±71. Peterson JW, Bo L, Mork S, Chang A, Trapp BD. Transected neurites, apoptotic neurons, and reduced in ammation in cortical multiple sclerosis lesions. Ann Neurol 2001; 50: 389±400. Rovaris M, Filippi M, Calori G, Rodegher M, Campi A, Colombo B, et al. Intra-observer reproducibility in measuring new putative MR markers of demyelination and axonal loss in multiple sclerosis: a comparison with conventional T2-weighted images. J Neurol 1997; 244: 266±70. Rovaris M, Inglese M, van Schijndel RA, Sormani MP, Rodegher M, Comi G, et al. Sensitivity and reproducibility of volume change measurements of different brain portions on magnetic resonance imaging in patients with multiple sclerosis. J Neurol 2000; 247: 960±5. Rudick RA, Fisher E, Lee JC, Simon J, Jacobs L. Use of the brain parenchymal fraction to measure whole brain atrophy in relapsingremitting MS. Multiple Sclerosis Collaborative Research Group. Neurology 1999; 53: 1698±704. Simmons ML, Frondoza CG, Coyle JT. Immunocytochemical localization of N-acetyl aspartate with monoclonal antibodies. Neuroscience 1991; 45: 37±45. Smith SM, De Stefano N, Jenkinson M, Matthews PM. Normalized accurate measurement of longitudinal brain change. J Comput Assist Tomogr 2001; 25: 466±75. Trapp BD, Peterson J, Ransohoff RM, Rudick R, Mork S, Bo L. Axonal transection in the lesions of multiple sclerosis. New Engl J Med 1998; 338: 278±85. Trapp BD, Ransohoff R, Rudick R. Axonal pathology in multiple sclerosis: relationship to neurologic disability. [Review]. Curr Opin Neurol 1999; 12: 295±302. van Waesberghe JH, Kamphorst W, De Groot CJ, van Walderveen MA, Castelijns JA, Ravid R, et al. Axonal loss in multiple sclerosis lesions: magnetic resonance imaging insights into substrates of disability. Ann Neurol 1999; 46: 747±54. Received August 6, Accepted September 9, 2002