Cortical neuronal loss and cerebral white matter demyelination in multiple sclerosis: a retrospective study

Abstract

BACKGROUND

Demyelination of cerebral white matter is thought to drive neuronal degeneration and permanent neurological disability in individuals with multiple sclerosis (MS). Brain magnetic resonance imaging (MRI) studies, however, support the possibility that demyelination and neuronal degeneration can occur independently. We aimed to establish whether post-mortem brains from patients with MS show pathological evidence of cortical neuronal loss that is independent of cerebral white-matter demyelination.

METHODS

Between May 1998 and November 2012, brains and spinal cords were removed from ninety- seven deceased patients with MS. Visual examination of cm-thick slices of cerebral hemispheres from these 97 cases identified 12 brains without areas of cerebral white-matter discoloration that are indicative of demyelinated lesions. These 12 cases (referred to as myelocortical MS: MCMS) were matched based upon age, MRI imaging protocol, MS disease subtype, and disease duration with 12 of the remaining cases containing cerebral white-matter discolorations/demyelinated lesions (referred to as typical MS: TMS). Demyelinated lesion area in tissue sections of cerebral white matter, spinal cord, and cerebral cortex were compared in MCMS and TMS using myelin protein immunocytochemistry. Neuronal densities in cortical layers III, V, and VI from five cortical regions not directly connected to spinal cord (cingulate gyrus, superior frontal, superior temporal, motor, and insular cortex) were also compared between the two groups and with aged- matched postmortem brains from individuals without evidence of neurological disease.

FINDINGS

In MCMS cases, histological studies performed between September 6, 2011 and February 2, 2018 detected demyelinated lesions in spinal cord and cerebral cortex, but not in cerebral white matter. Cortical demyelinated lesion area was similar between MCMS and TMS cases (median 4.45% [IQR 2.54-10.81] in MCMS vs 9.74% [1.35-19.50] in TMS; p=05512). Spinal cord demyelinated area was significantly greater in TMS cases compared to MCMS (median 3.81% [IQR 1.72-7.42] in MCMS vs 13.81% [6.51-29.01] in TMS; p=0 0083). Despite the lack of cerebral white-matter demyelination in MCMS cases, mean cortical neuronal densities were significantly decreased compared to control brains (layer III, 349.8 [SD 51.9] neurons per mm in MCMS and 419.0 [43.6] in controls, p=0 0104; layer V, 355.6 [46.5] in MCMS and 454.2 [48.3] in controls, p=0 0006; layer VI, mean 366.6 [SD 50.9] in MCMS and 458.3 [48.4] in controls, p=0 0049) and were decreased in layer V of TMS brains compared to controls (392.5 [59.0] in MCMS and 454.2 [48.3] in controls, p=0 0182).

INTERPRETATIONS

A subtype of MS, which we call myelocortical MS, is characterized by demyelination of spinal cord and cerebral cortex, but not of cerebral white matter. Cortical neuronal loss is not accompanied by cerebral white-matter demyelination and can be an independent pathological event in myelocortical MS. Future studies should investigate molecular mechanisms of primary neuronal degeneration and axonal pathology in myelocortical MS.

INTRODUCTION

Neurodegeneration is the major cause of permanent neurological disability in individuals with multiple sclerosis (MS).¹ Axons are transected during inflammatory-mediated demyelination² and many chronically demyelinated axons degenerate³. Demyelination of cerebral white matter is the pathological hallmark of MS. The incidence and dynamics of cerebral white-matter T2- weighted hyperintense lesions detected by magnetic resonance imaging (MRI) are critical for the diagnosis of MS,⁴ reflect disease activity,⁵ and demonstrate efficacy of disease-modifying therapies.⁶ However, only 55% of cerebral T2-weighted hyperintense white-matter lesions are demyelinated ^7,8 and T2-weighted cerebral white-matter lesion burden accounts for less than 30% of the variance in the rate of brain atrophy.^9-11 In the absence of specific MRI metrics for demyelination, the relationship between cerebral white-matter demyelination and neurodegeneration remains speculative. Disease mechanisms other than cerebral white-matter demyelination contribute to permanent neurological disability, including demyelination of spinal cord,³ demyelination of cerebral cortex,¹² and neurodegeneration that may be independent of demyelination.¹

While MS was historically considered to be a white matter disease, recent studies have identified cortical demyelination as prominent features of postmortem MS brains.¹ There are two major types of cortical lesions; leukocortical and subpial. Leukocortical lesions are mixed white matter/gray matter lesions that include subcortical white matter and lower layers of the cerebral cortex. Subpial lesions (the most abundant type of cortical lesions) extend from the pial surface, most often stop at cortical layers III or IV, and span several gyri. The relationship between cortical demyelination and cortical neuronal loss has been investigated and the majority of these studies report similar neuronal densities in myelinated and demyelinated cortex^12-14 (but see¹⁵).

Gray-matter atrophy is one of the best MRI predictors of neurological disability in individuals with MS,^16-19 can precede white-matter atrophy, and can occur independently of brain white- matter lesions.^{17,20, 21} Similar brain atrophy rates at initial and late stages of MS raise a question as to the role of white-matter demyelination in driving brain atrophy and neurodegeneration.²² Cortical atrophy is an early feature of MS and can be detected in individuals with very low brain white-matter lesion burden.^23-25 While MRI studies support cortical atrophy in the absence of cerebral white-matter demyelination,^{17, 20, 21} postmortem studies have not directly correlated cortical neuronal loss with cerebral white-matter demyelination.

Here, we identify a subpopulation of postmortem brains from disabled individuals with MS that do not contain demyelination of cerebral white matter. We provide a detailed description of cortical neuronal loss, cortical atrophy, cortical demyelination, spinal cord demyelination, and cerebral white-matter MRI abnormalities.

METHODS

Study Design

This study was approved by the Institutional Review Board of the Cleveland Clinic (see appendix p 1). Brains and spinal cords were obtained from 97 MS patients through the Cleveland Clinic rapid autopsy protocol. Consent for postmortem MRI and donation of brain and spinal cord tissue was obtained from living patients or family members. Prior to autopsy, brains were scanned in situ by MRI. Following removal, brains were divided into two hemispheres. One hemisphere was cut into centimeter-thick slices at autopsy and short-fixed in 4% paraformaldehyde for 2-4 days. The other hemisphere was long-fixed in 4% paraformaldehyde for 2-4 months, re-imaged in a slicing box for MRI-path analyses, and then cut into centimeter- thick slices. Centimeter-thick, short-fixed hemispheric slices from the 97 brains were examined for white-matter lesions. Twelve brains did not contain visible lesions in cerebral white matter. Pathological features of these twelve brains (referred to as myelocortical MS (MCMS) brains) and corresponding spinal cords were compared to twelve typical MS (TMS) cases. The 12 subjects with TMS were selected from the remaining brains and were matched to MCMS cases based upon age, sex, disease duration, MRI imaging protocol, and MS subtype at time of death (Table 1). Postmortem brains from 8 individuals (4 males, 4 females; median age at death=67 yrs, range=53-79) without macroscopic or microscopic indications of neurological disease were identified by a neuropathologist (SMS) from records of patients who died at Cleveland Clinic and had complete diagnostic autopsies performed with consent from the next of kin. These brains were used as controls for cortical neuronal counting. MRIs of 13 living individuals without neurological disease (7 males, 6 females; median age 51 8 yrs, range=41 7-66 8) were used as controls for cerebral atrophy measures.

Table 1:

Clinical Details from Typical and Myelocortical Multiple Sclerosis Patients

Group		MS Course	Sex	Race	Age(at death)	Disease Onset	Disease Duration (yrs)	Initial Symptoms	Other clinical Symptoms	CSFResults	Treatment	EDSS(Prior to Death)	PMI(hrs)	Cause of Death
TMS	1	SP	M	Cau	46	06/1964	36∙0	Optic neuritis at age 10	Optic neuritis, weakness, gait dysfunction, diplopia, ataxia, constipation	N/A	pulse dose steroids	7∙0	3∙00	Pneumonia
	2	SP	M	Cau	56	06/1969	32∙7	Numbness in righthand	double vision, decreased visual acuity, color saturation, fatigue, leg weakness, progressive	N/A	cyclophosph amide	9∙5	3∙05	Terminal MS, w/d of care
	3	SP	F	Cau	60	06/1973	29∙5	Relapse onset, not otherwis e defined	weakness, dysphagia, spasticity	N/A	N/A	9∙0	7∙83	Terminal MS, w/dof care
	4	SP	M	Cau	52	03/1978	25∙1	Right LEstiffness	progressive cerebellar, cognitive, spasticity and weakness, feeding tube placement	N/A	No DMT	9∙5	4∙83	Recurrent pneumonia, respiratory failure
	5	SP	F	AA	59	06/1968	37∙5	Paresthesia	weakness, spasticity, cognitive dysfunction, double vision, visionloss, vertigo, trigeminal neuralgia	No CSF	pulse dose steroids, cyclophosph amide	9∙0	5∙08	Hospice care after sepsis and poor oralintake (Terminal MS)
	6	PP	M	Cau	70	07/1990	17.1	Progressive LE weakness	double vision, vision loss, cognitive dysfunction	P: 37, OCB:Negative	pulse IV steroids, low dose methotrexate	6∙5	10∙33	Terminal MS(natural causes listed)
	7	SP	F	Cau	77	08/1973	34∙1	Unknown	LE weakness, fatigue, gait dysfunction, tremor cognitive dysfunction	N/A	IM interferon beta 1a	8∙0	5∙92	Terminal MS
	8	SP	M	Cau	63	07/1973	35∙8	Double Vision	quadriplegia	N/A	N/A	8∙0	4∙92	septic shock
	9	PP	F	Cau	51	08/1994	14∙9	Eye twitching, sensory changes	double vision, weakness, fatigue, neurogenic bladder, depression	No CSF	pulse dose steroids, IM interferon beta 1a, glatiramer acetate, methotrexate, natalizumab	7∙5	6∙92	suicide attempt, then respiratory failure
	10	PP	M	Cau	75	08/1987	23∙7	Vision loss	bowel/bladder dysfunction, UE weakness	OCB+	pulse dose steroids, low dosemethotrexate	9∙0	5∙42	sepsis, myocard ialinfarction
	11	SP	M	Cau	74	35∙5	Vision loss	weakness, incoordination, cognitive dysfunction, fatigue	No CSF	interferon beta 1b	8∙0	8∙90	choking (upper airwayobstruction)
	12	PP	F	Cau	70	07/1995	17∙4	Left UE/LE weakness	weakness, trigeminal neuralgia, seizure, vision problems, speech and swallowing difficulty	No CSF	glatirameracetate	8∙0	5∙00	Pneumonia
SUMMARY		8 SP; 4 PP	7M; 5F	11Cau; 1AA	Mean=62∙75 SD=10∙38		Mean=28∙28 SD=8∙32					Median=8∙0 IQR=7∙9, 9∙0	Median= 5∙25 IQR=4∙90, 7∙15
MCMS	1	PP	M	Cau	81	06/1968	29∙9	Progressive LE weakness	Optic neuritis, neurogenic bladder, dystaxia in right UE	P: 24, WBC: 0	No DMT	7∙5	8∙67	Terminal MS
	2	RR	M	AA	59	06/1991	10∙1	Lower backpain, progressive weakness of left LE	ED,spasticity, relapse with facial numbness	RBC: 0 WBC: 1, P:52, G: 55, OCB+	pulse dose steroids, interferon beta 1b	8∙0	20∙50	Cardiogenic shock, pseudomonas sepsis
	3	SP	F	Cau	58	07/1970	32∙7	Vision loss	gait dysfunction, weakness, urinary incontinence, fatigue, vision problems	N/A	pulse dose steroids, methotrexate, cyclophosphamide, glatirameracetate	8∙0	4∙50	Pneumonia, respiratory failure
	4	SP	M	Cau	85	06/1962	43∙0	Unsteady gait and lack of coordination	Seizure disorder, urinary incontinence, spasticity, chronic LE pain	N/A	No DMT	8∙0	10∙67	Likely sepsis (tachycardia, fever, decreased O2 saturation)
	5	SP	M	Cau	61	06/1999	6∙4	Right UE weakness	LE weakness, spasticity, bladder urgency	RBC:0, WBC:1, G:63, P:35 OCB+	IM interferon beta 1a, glatirameracetate, mitoxantrone, methotrexate	8∙0	9∙33	Pneumonia, cachexia
	6	SP	F	Cau	97	06/1962	45∙7	Vertigo/f alling	progressive gait dysfunction	WBC: 15, OCB+	No DMT	7∙5	14∙00	Multiorgan failure, CHF, kidney failure, respiratory failure
	7	RR	M	Cau	47	06/1989	19∙0	Right LE weaknes s	LE weakness, sensory loss right side, paresthesia, bowel/bladder symptoms	No CSF	pulse dosesteroids, IM interferon beta 1a, natalizumab	2∙5	5∙92
	8	SP	F	Cau	79	06/1975	33∙2	Gait disturbance	paraplegia, cognitive dysfunction, urinary frequency/urgency, spasticity	No CSF	pulse dose steroids	8∙0	5∙75	Septic shock
	9	SP	M	H/L	34	06/2000	8∙8	Numbness and paresthesia	incoordination, weakness, sensory symptoms, gait impairment, blurred vision, slurred speech, difficulty swallowing, bladder/bowel incontinence	Positive (not otherwise specified)	SC interferon beta 1a, glatirameracetate, IM interferon beta 1a	8∙0	8∙28	npneumo
	10	PP	F	Cau	73	06/1981	28∙0	Progressive gait disturbance	weakness, spasticity, bladder/bo wel involvement, fatigue, optic neuritis	No CSF	pulse dose steroids, oral steroids,	7∙5	6∙50	Respiratory failure, aspiration pneumonia
	11	SP	F	Cau	71	07/1996	13∙7	Walking and hand difficulties	weakness, bladder urgency, fatigue, paresthesia	No CSF	IMinterferon beta 1a, pulse dose steroids, cyclophosphamide, azathioprine	8∙5	7∙17	Sepsis, CHF
	12	SP	F	Cau	71	06/1975	34∙8	Weakness of the left LE	weakness in LE and in right UE, dysarthria, bladder/bowel dysfunction, headaches, fatigue	No CSF	No DMT	8∙0	7∙58	Intracran ial hemorrh age
SUMMARY		2RR; 8SP; 2PP	6M; 6F	10Cau; 1AA; 1H/L	Mean=68∙00 SD=17∙29		Mean=25∙44 SD=13∙48					Median=8∙0 IQR=7∙5, 8∙0	Median=7∙93 IQR=6∙36, 9∙67

Abbreviations: AA- African-American; Cau- Caucasian; CHF- Congestive heart failure; CSF- cerebrospinal fluid; DMT- Disease-modifying treatment; ED- erectile dysfunction; EDSS- Expanded Disability Status Scale; F- Female; G- Glucose (mg/dl); H/L- Hispanic or Latino; IM- intramuscular; IV- intravenous; LE- lower extremity; M- Male; MCMS- Myelocortical Multiple Sclerosis; N/A- Not available; OCB- Oligoclonal bands; P- Protein (mg/dl); PMI- Post-mortem Interval; PP- Primary/Progressive; RBC- red blood cells (number /microliter); RR- Relapsing/Remitting; SP- Secondary/Progressive; TMS- Typical Multiple Sclerosis; UE- Upper extremity; w/d- withdrawal; WBC- white blood cells (number/microliter)

Neuropathological Evaluation

The area of macroscopic cerebral white-matter lesions (gray/brown discoloration greater than 3 mm² ) was quantified on the posterior surface of the centimeter-thick, short-fixed MCMS and TMS hemispheric slices. A 3mm² cut-off was used to prevent misidentification of large vessels as lesions. Thirty-micron-thick, free-floating sections from four segments of spinal cord (one cervical, two thoracic, one lumbar) and five cortical areas (cingulate gyrus, superior frontal, superior temporal, motor, and insular cortex) from both groups were stained for myelin proteolipid protein (PLP) as described previously.² These spinal cord and cortical regions were selected to achieve a systematic, unbiased sampling of cord and cerebral cortex. The number and size of spinal cord and subpial cortical lesions were quantified. In both the spinal cord and the cortex, lesion area was calculated based on pixels lacking PLP staining.

Centimeter-thick, long-fixed hemispheric brain slices from 11 MCMS (the hemispheric slice from case 2, Table 1 was damaged in processing) and 12 TMS cases were embedded in paraffin, sectioned at a thickness of 7 μm, and stained with cresyl violet for neuronal counting. Hemispheres were sliced in a box containing centimeter-thick slicing slots to ensure similar slice orientation and thickness. The brain slices chosen for neuronal quantification were homologous across all patients, cut in the same orientation, and contained the same cortical regions. Neuronal densities in cortical layers III, V, and VI were determined in five cortical regions (inferior frontal gyrus, superior temporal gyrus, superior insula, inferior insula, and cingulate gyrus). These cortical regions were selected to minimize effects of spinal cord demyelination; they do not project axons to or receive axons from spinal cord. Cortical layers III, V, and VI were selected because they contain the majority of cortical projection neurons, which are likely to be affected by white-matter demyelination. Neurons were quantified using an automated ImageJ (1.47v, Java 1.6.0_24, 64 bit version) algorithm and the “Particle Analysis” tool in Fiji to identify neurons based on size (area>60 μm² ) and shape criteria. Sections cut from the same slice were stained with PLP antibodies to detect areas of demyelination. Correlations between neuronal densities and areas of cerebral white-matter and cortical demyelination were evaluated by regression analyses. Additional details of the immunocytochemistry methods are described in the appendix on p 1.

Magnetic Resonance Imaging and Image-Guided Tissue Sampling

Ten MCMS brains (cases 1 and 2, Table 1, did not have postmortem MRIs) were imaged in situ prior to autopsy using a standardized MRI acquisition protocol⁸ that revealed T2-weighted hypointensities, Tl-weighted hyperintensities, magnetization transfer ratios (MTR) and a 3D 1mm isotropic Tl-weighted gradient echo sequence for co-registration of MRI and fixed brain slices (used for MRI-path correlations).⁸ T2-weighted hyperintensities were segmented on the FLAIR image and were further classified based on their intensities in the T1-weighted and MTR images. Whole-brain parenchymal fraction, white-matter fraction, gray-matter fraction, and cortical thickness were calculated as described previously.^17,26 MRI-pathology correlations in cerebral white matter of MCMS brains were performed using the long-fixed hemisphere as described previously.⁸ A total of 31 T2T1MTR ROIs (abnormal on all three MRI sequences) were identified. Forty-six NAWM ROIs were selected from the brain slices that contained the T2T1MTR ROIs. These ROIs were removed from the brain slice, sectioned at a thickness of 30 μm, and stained with antibodies specific for myelin, axons, microglia, reactive astrocytes, and serum proteins as described previously⁸. The antibodies that were used are described in the appendix. Areas occupied by axons, myelin, and activated microglia, axonal densities, and average axonal diameters were determined as described previously^8,27.The numbers of ROIs with reactive astrocytes and detectable serum proteins were also documented. Correlations between axonal diameter and normalized T1-weighted intensities and MTR values were evaluated by regression analyses.²⁷

Statistics

Experimenters were blinded to condition for quantitative analyses of all data sets. Statistical analyses included, as appropriate: t-tests, chi-squared tests, linear mixed-effects analyses, calculation of linear correlation coefficients, analyses of variance, analysis of covariance, and Mann-Whitney U Tests for non-parametric data (for additional details regarding which tests were used for which analyses, see appendix p 1-2 and figure legends). For all statistical tests, two sided p-values of 0 05 or less were considered to be significant.

Role of funding source

The funders had no role in the experimental design, data collection, data analysis, data interpretation, manuscript preparation, or manuscript submission. The funders provided funds to complete the study, including investigator salaries and research costs. All authors had complete access to the data. All authorized submission of the manuscript. The corresponding author takes full responsibility for the submission of the manuscript for publication.

RESULTS

Autopsies were performed and brains and spinal cords were collected between May 1998 and November 2012. The study was conducted between September 6, 2011 and February 2, 2018. Clinical records and postmortem MRIs of the MCMS and TMS groups were de-identified and presented to expert neurologists (RAR, RJF, and other members of the Mellen Center for MS Treatment and Research) as well as neuroradiologists and imaging experts (SEJ, KN, EF, JC) who were blinded to the classification. Following analyses of demographic and clinical data (Table 1) and MRI images (see examples in appendix p 6), the clinical and radiological experts classified all cases as definite MS, were unable to correctly classify the cases as MCMS or TMS, and concluded that MCMS and TMS patients did not have neuromyelitis optica (NMO). Age at death and EDSS disability status were not significantly different between the TMS and MCMS groups (Table 1). Time between death and autopsy (postmortem interval) was significantly greater (p=0 0326) in the MCMS group (Table 1; Median=7 93 hr, interquartile range (IQR)=6 36, 9 67) than in the TMS group (Median=5 25 hr, IQR=4 90, 7 15).

We visually compared white-matter discolorations in centimeter-thick, short-fixed hemispheric brain slices of 12 TMS and 12 MCMS brains (see examples in figures 1A and 1B). The TMS group had a significantly higher median number of cerebral white-matter lesions per hemisphere compared to the MCMS group (p<0 0001; Table 2). The average size of individual white-matter lesions in TMS hemispheric slices was also significantly higher compared to MCMS hemispheric slices (p=0 0068; Table 2).

Figure 1: Demyelination in Myelocortical Multiple Sclerosis (MCMS) and Typical Multiple Sclerosis (TMS).

Panel A shows a centimeter-thick slice from a TMS brain that contains a large white-matter lesion (arrow). Panel B shows a centimeter-thick slice from an MCMS brain without white-matter lesions. Panel C shows a normally-myelinated spinal cord section stained for PLP. Panels D and E show similar degrees of spinal cord demyelination in tissue sections from TMS (D) and MCMS (E). Panel F shows normally- myelinated cortex stained for PLP. Panels G and H show subpial cortical lesions in TMS (G) and MCMS (H) brains. Scale bars: C-H=2 mm. Black lines separate spinal cord gray and white matter (C-E) and cortical and subcortical white matter (F-H).

Table 2.

Cerebral white matter
	MCMS	TMS
Number of lesions per hemisphere (median, IQR)	0·5 (0·0-1·0) vs. TMS: p<0.0001	10·0 (4·5-19·0)
Lesion area per hemisphere (cm2)	0·40 (0·26) vs. TMS: p =0·0068	14·14 (14·34)
Spinal cord
	MCMS	TMS
Number of segments with demyelination, (median, IQR)	2·00 (2·00-3·00) vs. TMS: p= 0.5000	2·50 (2·00-3·25)
Area occupied by demyelinated lesions per segment (%)	8·0% (8·4%) vs. TMS: p =0·0001	26·0% (22·0%)
Area occupied by demyelinated lesions per patient (median, IQR)	3·81% (1·72-7·42%) vs. TMS: p=0·0083	13·81% (6·51-29·01%)
Cerebral cortex
	MCMS	TMS
Number of regions with subpial lesions, (median, IQR)	2·00 (1·00-3·75) vs. TMS: p= 0·0163	4·00 (3·25-5·00)
Area occupied by subpial lesions per section (%)	23% (22%) vs. TMS: p= 0·5126	20% (20%)
Area occupied by subpial lesions per patient (Median, IQR)	4·45% (2·54-10·81%) vs. TMS: p= 0·5512	9·74% (1·35-19·50%)
Neuronal density per mm2 in five cortical regions
	MCMS	TMS	Control
Layer III	349·8 (51·9) vs. Control: p=0·0104 vs. TMS: p=0·2040	381·5 (58·2) vs. Control: p=0·1346	419·0 (43·6)
Layer V	355·6 (46·5) vs. Control: p=0·0006 vs. TMS: p=0·1533	392·5 (59·0) vs. Control: p=0·0182	454·2 (48·3)
Layer VI	366·6 (50·9) vs. Control: p=0·0049 vs. TMS: p=0·2220	401·7 (74·7) vs Control: p=0·0589	458·3 (48·4)
MRI
	MCMS	TMS
Total T2 lesion volume (mL)	25·32 (14·38) vs. TMS: p=0·0578	44·36 (27·66)
Total T1 lesion volume (mL)	11·81 (7·69) vs. TMS: p=0·0591	23·92 (16·44)
Total MTR lesion volume (mL)	11·20 (7·18) vs. TMS: p=0·0388	24·83 (16·43)
	MCMS	TMS	Control
Whole brain parenchymal fraction	0·80 (0·02) vs. Control: p=0·2190 vs. TMS: p=0·0006	0·75 (0·06) vs. Control: p<0·0001	0·84 (0·03)
White-matter fraction	0·32 (0·03) vs. Control: p=0·3766 vs. TMS: p=0·0211	0·30 (0·02) vs. Control: p=0·0019	0·33 (0·01)
Gray-matter fraction	0·48 (0·03) vs. Control: p=0·4194 vs. TMS: p=0·0129	0·45 (0·05) vs. Control: p=0·0013	0·51 (0·02)
Cortical thickness (mm)	2·97 (0·27) vs. Control: p<0·0001 vs. TMS: p=0·0168	2·70 (0·36) vs. Control: p<0·0001	3·75 (0·19)

Data are mean (SD) unless otherwise indicated. Control brains were not included in the study of number and size of white matter lesions per hemisphere, number of spinal cord segments with demyelination, spinal cord area occupied by demyelinated lesions, number of regions with subpial lesions, or cortical area occupied by individual subpial lesions. For between-group differences and 95% CI, see appendix.

Thirty-micron-thick tissue sections from four spinal cord segments from each TMS and MCMS case were stained with PLP antibodies and areas of demyelination were determined. Compared to normally-myelinated spinal cord (figure 1C), demyelination was prominent in TMS (figure 1D) and MCMS (figure 1E) spinal cords. The number of spinal cord segments with demyelination was similar between the TMS and MCMS groups (p=05000; Table 2). However, the average percent area occupied by demyelinated lesions in spinal cord was significantly greater in sections from individuals with TMS compared to MCMS (p=0 0001; Table 2).

We next investigated the incidence of subpial cortical demyelination in the two groups. Figure 1F shows normally-myelinated cerebral cortex. Similar to TMS cases (figure 1G), subpial cortical lesions were detected in MCMS cases (figure 1H), involved multiple gyri, and did not extend into subcortical white matter. Within the five cortical regions, TMS cases contained a significantly greater median number of regions with subpial lesions than MCMS cases (p=0 0163; Table 2). The average percentage of cortical area occupied by individual subpial lesions, however, was similar between TMS and MCMS cases (p=05126; Table 2).

Seven-micron-thick, paraffin-embedded long-fixed hemispheric sections from control, TMS, and MCMS brains were stained with cresyl violet (figure 2A). Neuronal densities were quantified in cortical layers III, V, and VI (figure 2B) in five cortical regions that are not directly connected to spinal cord (labeled in figure 2A). Compared to control cortices, neuronal densities in the five MCMS cortical regions combined were significantly decreased in all three cortical layers (Layer III: p=0 0104; Layer V: p=0 0006; Layer VI: p=0 0049; Table 2). Compared to controls, neuronal densities in TMS cortices were significantly decreased in layer V (p=0 0182; Table 2), but not in layers III (p=0-1346; Table 2) or VI (p=0 0589; Table 2). None of the three layers were significantly different between TMS and MCMS cases (Table 2). In individual cortical regions, neuronal densities were significantly decreased in 11/15 layers from MCMS cases and 5/15 layers from TMS cases when compared to control cases (appendix p 4). Only layer V of the inferior frontal gyrus was significantly decreased in MCMS compared to TMS (appendix p 4).

Figure 2: Neuronal Loss in the Absence of Cerebral White-Matter Demyelination.

Panel A shows a cresyl violet-stained hemispheric section from a TMS brain. Neuronal densities were compared in cortical layers III, V, and VI (B, yellow indicates neurons with an area greater than 60 μm²) in the five cortical areas labeled in A. Panels C and D show staining for myelin proteolipid protein and the distribution of demyelinated lesions (blue=white-matter demyelination, pink=subpial demyelination) in hemispheric sections from TMS (C) and MCMS (D). A significant correlation between reduced cortical neuronal density and increased cerebral white-matter lesion volume was found in TMS, but not in MCMS (E; Pearson’s Correlation). Dashed lines in (E) indicate 95% CI. IFG=inferior frontal gyrus, STG=superior temporal gyrus, INi=inferior insular, INs superior insula, CG=cingulate gyrus. Scale bars: B=200 μm

As implied for centimeter-thick slices of short-fixed hemispheres (figures 1A and B), white- matter demyelination was prominent in PLP-stained sections from long-fixed hemispheres in TMS cases (figure 2C; blue areas), but not in MCMS cases (figure 2D). Subpial cortical demyelination was detected in both groups (figures 2C and D; pink areas). A significant negative linear correlation between the average neuronal density over the five cortical regions and areas of brain white-matter demyelination was found in TMS cases (r= −0 6758, p=0 0159, CI= −0 90, - 017; figure 2E), but not in MCMS cases. There was no correlation between cortical neuronal density and subpial cortical demyelination in either group (appendix p 5).

Postmortem MRIs were performed in situ on TMS (figure 3A) and MCMS (figure 3B) cases. Despite the paucity of demyelinated cerebral white-matter lesions in MCMS brains, abnormalities were detected in T2-weighted, Tl-weighted, and MTR images (Table 2). Total T2- weighted and Tl-weighted lesion volumes were not significantly different between MCMS and TMS brains (Table 2). MTR lesion volumes were significantly greater in TMS compared to MCMS (p=0 0388; Table 2). The brain parenchymal fraction (Control vs TMS p<0 0001; TMS vs MCMS p=0 0006; Table 2), white-matter fraction (Control vs TMS p=0 0019; TMS vs MCMS p=0 0211; Table 2), and gray-matter fraction (Control vs TMS p=0 0013; TMS vs MCMS p=0 0129; Table 2) were significantly lower in TMS cases compared to MCMS cases and healthy controls. Compared to live controls, cortical thickness was significantly decreased in both MS groups and was thinner in TMS cases compared to MCMS cases (Control vs TMS p<0 0001; Control vs MCMS p<0 0001; TMS vs MCMS p=0 0168; Table 2). The spatial distributions of T2-weighted lesions were not significantly different in probability maps of TMS and MCMS MRIs (appendix p 6). Before correcting for multiple comparisons, 1 2% of lesion voxels showed statistically higher probability for TMS compared to MCMS; no lesion voxels showed a higher MCMS lesion probability (appendix p 6).

Figure 3: Magnetic Resonance Imaging in Myelocortical and Typical Multiple Sclerosis.

Panels A and B show representative T2-weighted, fluid-attenuated inversion recovery images from postmortem TMS (A) and MCMS (B) brains.

Thirty-one T2T1MTR ROIs and 46 NAWM ROIs were identified in MCMS MRIs and mapped onto co-registered brain tissue slices (figures 4A-D). The number of ROIs from each case is listed in Supplementary Table 2 (appendix p 7). The mean sizes of the NAWM and T2T1MTR ROIs were similar (NAWM=31 10 mm² (SD=7 90); T2T1MTR=33 20 mm² (SD=10 50) (appendix p 7). These white-matter ROIs were then blocked, sectioned, and stained for myelin, axons, activated microglia, activated astrocytes, and serum proteins. Demyelination was not detected in the 46 NAWM regions. Four of the 31 T2T1MTR ROIs contained small areas of white-matter demyelination that represented less than 5% of the ROI. The area occupied by myelin (appendix p 8; figure 4G) was 5% less in the T2T1MTR ROIs (T2T1MTR Mean=0 95, SD=0-10, p=0 0235) than in NAWM ROIs (Mean=1 00, SD=0 067). The areas occupied by axons (figures 4E and 4F) were not significantly different between the NAWM and T2T1MTR ROIs (figure 4G). Mean axonal density (NAWM Mean=1930, SD=334, T2T1MTR Mean=1527, SD=463, p=0 0001) was significantly lower and mean axonal diameter was significantly higher (NAWM Mean=194, SD=0 25, T2T1MTR Mean=2 24, SD=0 40, p<0 0001) in T2T1MTR ROIs (figure 4G). The T2T1MTR ROIs, therefore, contained a significant increase in swollen myelinated axons. While postmortem interval was greater in the MCMS group (Table 1), postmortem time did not correlate with axonal swelling in individual MCMS patients.

Figure 4: Pathological Correlates of Cerebral White-Matter Imaging Abnormalities in Myelocortical Multiple Sclerosis.

An MCMS brain slice (A) is co-registered with T1-weighted (B), T2-weighted (C), and MTR (D) images. The red circle indicates a normal-appearing white-matter (NAWM) region as defined by no imaging abnormalities. The blue circle indicates a region of interest that is abnormal in T2-weighted, T1-weighted and MTR images (T2T1MTR). Panels E and F show axonal staining in tissue sections of the NAWM (E) and T2T1MTR (F) regions. Panel G compares measurements of myelin area, axonal area, axonal numbers, and axonal diameter in 46 NAWM and 31 T2T1MTR regions. Myelin area and axon count were significantly smaller in T2T1MTR regions (Student’s t-tests), whereas axonal diameter (Mann-Whitney U Test) was significantly higher. Furthermore, average axonal diameter negatively correlates with normalized T1-weighted intensity (H) and MTR (I). Individual dot colors represent individual patients (n=10), and dashed lines indicate 95% CIs in H and I. Data in H and I were analyzed by linear models where group, age, and sex were specified as fixed factors and subject was specified as a random factor. Scale bars: E and F=20 μm

Normalized MTR values in T2T1MTR ROIs varied from 0 44 to 0 92. Normalized T1-weighted intensity varied between 061 and 1 00. There were negative linear correlations between axonal diameter and normalized T1-weighted intensity (regression coefficient= −0-124, 95% confidence interval= −0 20, −0 05, p=0 0029; figure 4H) and MTR (regression coefficient= −0 152, 95% confidence interval= −0 23, −0 08, p=0 0003; figure 4I). T2T1MTR ROIs also contained more activated microglia, reactive astrocytes, and serum proteins (appendix p 8). Two of 31 MCMS T2T1MTR ROIs contained small areas of perivascular myelin thinning (appendix p 9), suggestive of ischemic damage.

DISCUSSION

Myelocortical MS is characterized by spinal cord demyelination, subpial cortical demyelination, and an absence of cerebral white-matter demyelination. Cases of MCMS represented ~12% of a consecutive series of autopsies (12/97) from individuals with MS. Our studies highlight the concept that cerebral white-matter T2T1MTR abnormal ROIs are not always demyelinated and provide pathological evidence that cerebral white-matter demyelination and cortical neuronal degeneration can be independent events in MCMS.

Cerebral white-matter demyelination was investigated by three approaches: 1) macroscopic examination of centimeter-thick brain slices, 2) myelin staining of ROIs that have abnormal MRI intensities, and 3) myelin staining of hemispheric sections where neuronal densities were determined. All three approaches failed to detect significant cerebral white-matter demyelination in tissues from individuals with MCMS. Cerebral white-matter demyelination is not a feature of MCMS and there was no histological evidence for cerebral white-matter remyelination in MCMS.

In the absence of cerebral white matter demyelination in MCMS, cortical neuronal density (Table 2) and cortical thickness (Table 2) were significantly decreased when compared to controls. Cerebral white matter demyelination and cortical neuronal loss, therefore, are independent events in MCMS. Spinal cord demyelination is the likely cause of the severe physical disability in MCMS cases (median EDSS=8 0; IQR=7 5, 8 0; Table 1). The possibility that spinal cord demyelination influences cortical neuronal loss by trans-synaptic mechanisms seems unlikely, as TMS has 3 6 times more spinal cord demyelination, but less cortical neuronal loss than MCMS. Cortical demyelination did not correlate with cortical neuronal loss in either TMS or MCMS (appendix p 5). Collectively, our data support cortical neuronal loss and demyelination as independent events in MCMS.

In TMS, 83% of the white-matter T2T1MTR ROIs are demyelinated with reduced axonal densities and increased axonal diameters.^8,27 In these chronically-demyelinated lesions, the degree of decreased normalized Tl-weighted hypointensities and reduced MTR values correlated with increased swelling of demyelinated axons.^8,27 A similar correlation exists in MCMS ROIs, except that the swollen axons are myelinated (figures 4H and I). Tl-weighted hypointensities and MTR measures are influenced by free-water content.²⁸ Axonal swelling increases free water and is a prominent and underappreciated influence on these MRI metrics in cerebral white matter of MS. Activated microglia, astrocytosis, and serum proteins were also observed in MCMS T2T1MTR ROIs and contribute to MRI abnormalities in the absence of demyelination. Small vessel disease (SVD) can also cause white-matter MRI abnormalities. Pathological features of SVD, lacunae and frank chronic infarcts²⁹, were not detected in MCMS brain slices (figure 1B) or tissue sections (figure 2D). Two MCMS T2T1MTR ROIs had small areas of perivascular myelin thinning (appendix p 9), which is consistent with ischemic damage. SVD may be a minor contributor to MRI abnormalities in MCMS. Demyelination is not the cause of MCMS MRI lesions, but we do not know if the myelin is normal or why the MRI lesions have sharp borders.

Our data provide pathological evidence that neuronal degeneration and cerebral white-matter demyelination can be independent events in individuals with MS and identify cortical neuronal degeneration and swollen myelinated axons as attractive targets for neuroprotective therapies. MCMS should be considered as a distinct subtype of MS. The etiology and pathogenesis of T2T1MTR ROI abnormalities in MCMS are not known. It is possible that the peripheral immune system plays a causal role in MCMS cerebral white-matter abnormalities. It would be informative, therefore, to know whether anti-inflammatory therapies reduce new cerebral white- matter MRI lesions or cortical atrophy in individuals with MCMS. Identification of living individuals with MCMS will require the development of more sensitive brain imaging modalities that reliably delineate myelinated and demyelinated cerebral white matter. It is also possible that abnormalities in MCMS T2T1MTR ROIs are due to a primary neurodegeneration process that is independent of the peripheral immune system.

There are limitations to this study. The most significant issue relates to possible selection bias, which in turn limits generalization of our data to all MS patients. Research subjects in the study died from complications related to advanced MS, so it is not appropriate to conclude that the same proportion of all MS patients have MCMS as were observed in this 97-subject sample. Also, it is not possible to extrapolate pathological findings observed in this sample to earlier stages of MS. Finally, as is the case with most autopsy series, clinical annotation was derived from chart review, where data were recorded in a non-standardized fashion. In MS practice, clinical assessments do not include standardized, quantitative, comprehensive clinical data. Because of this limitation, it is possible that subtle differences in clinical manifestations between typical and myelocortical MS exist.

Evidence before this study

We searched PubMed for histopathological studies of brains from patients with multiple sclerosis (MS) that found no cerebral white-matter lesions prior to March 31, 2018 using this search strategy: “multiple sclerosis” AND (paucity OR absence OR no OR none) AND (histologic OR histological OR histology OR pathologic OR pathological OR pathology OR autopsy OR postmortem OR post-mortem) AND (cerebral OR brain) AND (white-matter OR “white matter”) AND (demyelination OR lesions) and filtering for “Humans”.

Of the 157 hits, most were studies involving magnetic resonance imaging (MRI) or spectroscopy in living patients without histological analyses. Other hits included studies of demyelinated lesions, histological studies of diseases other than MS, histological studies of MS white matter that appeared normal based upon visual inspection, review articles describing white-matter lesions, review articles of MRI of MS, articles describing the diagnostic criteria of MS, MRI techniques to detect demyelination, and various case reports and review articles of patients with other diseases in addition to MS, special MS populations, oligodendrocyte pathology, MS pathogenesis, positron emission tomography (PET imaging) of MS patients, brainstem auditory- evoked potentials in MS patients, and biomarkers of MS. None of the hits included any mention of MS patients without evidence of cerebral white-matter demyelinated lesions.

Added value of this study

Our study is the first to describe myelocortical multiple sclerosis (MCMS), a new subtype of MS characterized by demyelination of spinal cord and subpial cerebral cortex, but not of cerebral white matter. MCMS cases represented 12 of 97 autopsies. Individuals with MCMS were severely disabled. Multiple spinal cord demyelinated lesions met pathological criteria for the diagnosis of MS (separation of demyelinated lesions in space) and are the likely cause of the severe ambulatory disability in individuals with MCMS. We compared pathological and MRI characteristics of MCMS and typical MS (TMS, cases with cerebral white-matter demyelination). We show that both MCMS and TMS cases exhibit similar cerebral white-matter MRI abnormalities. Furthermore, MRI-defined lesions in MCMS cerebral white matter are associated with, and presumably due to, swelling of myelinated axons. Cortical neuronal density in MCMS brains was significantly decreased compared to aged-matched control brains and was similar to neuronal density in TMS brains. Our study, therefore, provides pathological evidence that cerebral white-matter demyelination and neurodegeneration can be independent events in MS.

Implications of all the available evidence

Identification of a new subtype of MS supports the concept that MS is a complex condition with multiple etiologies. Our findings are consistent with a neurodegenerative process in MS that is independent of demyelination. Our findings also highlight the non-specific nature of the MRI abnormalities that are traditionally assessed in MS patients and their lack of specificity for demyelination. Inclusion of MCMS patients in immunomodulatory clinical trials may also explain variable individual MRI responses to therapy.