Spinal Cord Neuronal Pathology in Multiple Sclerosis

Abstract

The objective of this study was to assess neuronal pathology in the spinal cord in multiple sclerosis (MS), both within myelinated and demyelinated tissue. Autopsy material was obtained from 38 MS cases and 21 controls. Transverse sections were taken from three spinal cord levels and stained using Luxol Fast Blue/Cresyl Violet and myelin protein immunohistochemistry. Measurements of neuronal number and size were made for all neurons within the anterior horns of the gray matter. Neurons were classified as motoneurons or interneurons according to size criteria. In comparison with controls, both motoneuron and interneuron number were reduced in MS cases at the upper cervical (interneuron P = 0.0549; motoneuron P = 0.0073) and upper thoracic (interneuron P = 0.0507; motoneuron P = 0.0144), but not the lumbar level. Interneuron cross‐sectional area was reduced in MS cases at all levels (upper cervical, P = 0.0000; upper thoracic, P = 0.0002; lumbar, P = 0.0337). Neuronal loss appears to be predominantly related to local gray matter plaques, whereas interneuron atrophy occurs in both myelinated and demyelinated areas.

INTRODUCTION

Neuronal pathology may contribute to permanent disability in multiple sclerosis (MS), along with demyelination and axonal loss. A small number of post‐mortem studies—examining small numbers of MS cases (n < 10)—provide evidence of neuronal pathology in the cerebral cortex and thalamus in MS 6, 11, 36, 37. Further evidence of neuronal injury in MS comes from magnetic resonance spectroscopy studies demonstrating reductions in cortical and thalamic N‐acetylaspartate levels in MS 5, 6.

The mechanisms that drive this neuronal pathology are unclear. Neuronal injury may occur as a direct consequence of gray matter demyelination; reductions in neuronal density, neuronal atrophy and apoptotic neurons have been demonstrated within cortical plaques 29, 36, 37. Alternatively, injury may occur subsequent to axonal damage in distant white matter lesions. Neuronal degeneration may occur secondary to axonal transection (retrograde degeneration), following interruption of their afferent connections (anterograde transneuronal degeneration), or after the death of the neurons upon which they project (retrograde transneuronal degeneration) (7).

The spinal cord is a predilection site for demyelination in MS (16) and a site of substantial axonal loss 4, 8. However, no study has quantified neuronal pathology in this clinically eloquent region. In this large post‐mortem study we investigated neuronal pathology in the spinal cord in MS, comparing the number and size of motoneurons and interneurons between 38 MS cases and 21 controls. We assess the extent of neuronal pathology, both within myelinated and demyelinated gray matter.

METHODS

Clinical material

Human autopsy material was obtained from 37 pathologically confirmed cases of MS and 22 controls. This material was derived from the neuropathology department, Oxford Radcliffe NHS Trust. The cases were selected at random from a collection of 55 MS cases and 33 controls studied previously 8, 12, 15. The MS patients (18 male, 19 female) were aged 32–83 years (mean 57.8 years) with disease durations of 2–43 years (mean 17 years). The controls (10 male, 12 female), aged 31–81 years (mean 56.9 years), had no clinical or pathological evidence of spinal cord disease. The local research ethics committee approved the study.

For each of the MS and control cases formalin‐fixed paraffin‐embedded transverse sections were taken from the upper cervical, upper thoracic and lumbar levels of the spinal cord. Sections were cut at a microtome setting of 15 µm and stained for Nissl substance and myelin with Luxol Fast Blue Cresyl Violet(24). Adjacent sections, cut at 10 µm, were stained for proteolipid protein as described previously (16).

Neuronal counts

Cases were coded to blind the observer to the clinical information. Neuronal counts were restricted to the anterior horns of the gray matter, defined—for the purposes of this study—as the area ventral to the anterior border of the gray matter commissure (Figure 1). The entire anterior horn was systematically viewed using a video camera (JVC KY‐F55B 3‐CCD; JVC UK Ltd., London, UK) linked to microscope (40×, Leitz Dialux 20EB; Leica Microsystems, Wetzler, Germany).

Figure 1.

Paraffin sections from the upper thoracic level of the spinal cord stained with anti‐Proteolipid Protein antibody (A), used to assess the extent of gray matter demyelination, and Luxol Fast Blue Cresyl Violet (B–H), used to perform the neuronal counts. Our assessment was restricted to the anterior horns of the gray matter, defined as the area ventral to the gray matter commissure (B–D). E, F—Measurements of neuronal number and cross‐sectional area (panel F, shaded) were made for all neurons in which the nucleolus was visible within the plane of section. G, H—Neurons were classified as motoneurons or interneurons according to size criteria; Neurons with a maximum diameter of at least 30 μm and a minimum diameter of at least 13.5 μm were classified as motoneurons (G); all other neurons were classified as interneurons (H). Scale bars (E–H) represent 20 μm.

The Nissl stain readily differentiates between neurons and non‐neuronal cell populations when viewed at high magnification. Neurons are larger with a well‐defined nucleolus and a cell body rich in endoplasmic reticulum, whereas neuroglia are smaller with denser nuclei, less conspicuous nucleoli and a greater nucleus to cytoplasm ratio. Neurons were photographed and counted if the nucleolus was visible in the plane of the section. Neuronal counts were performed on both anterior horns. The neuronal bodies and nucleoli were outlined manually using image analysis software (‘AnalySIS Pro’ running SIS software, Olympus UK Ltd., Southall, UK). The cross‐sectional area of the neuron was determined as was the maximum and minimum neuronal diameter (for application of the Abercrombie correction). To evaluate intra‐observer reproducibility, neuronal size measurements were made on two separate occasions, 2 months apart, using a randomly selected tissue section (coefficient of variation = 6.7%; inter‐observer coefficient of variation = 6.5%).

Neurons were classified as motoneurons or interneurons according to size criteria; those with a maximum diameter of at least 30 µm and a minimum diameter of at least 13.5 µm were classified as motoneurons; all other neurons were classified as interneurons. Counts were performed on both anterior horns of each section. Raw counts were converted using the Abercrombie formula. Section thickness was measured at a workstation using a video camera (JVC TK‐1380, JVC UK Ltd., London, UK) attached to a microscope (Olympus BX50 using ×100 oil‐immersion lens, Olympus UK Ltd., Southall, UK) with a motorized stage (Prior, Fulbourm, UK) that controlled movements in the z‐axis. This system was connected to a PC running CAST software (Computer Assisted Stereological Toolbox, Olympus, Albertslund, Denmark).

A note on the Abercrombie correction

When autopsy material is sectioned, cut structures appear on multiple sections. Simple profile counts therefore overestimate the number of objects in a structure. The Abercrombie method (1) attempts to correct the overestimates generated by “raw” counts of cellular profiles. The Abercrombie formula states N = n · t/(t + H) where n is the counted number of objects in the sampled portion; N is the estimate of true number of objects; t is the section thickness; and H is the mean height of the object (i.e. nucleolus), which we estimate by measuring the profile diameters in the xy plane.

Assessment of demyelination

The extent of demyelination within the anterior horn (expressed as the proportion of the total area) was assessed using the proteolipid protein‐stained sections. The influence of myelin loss on neuronal pathology was assessed in the MS cases by comparing anterior horns that were completely myelinated (n = 81 hemisections, from 29 MS cases) with those that were completely demyelinated (n = 33 hemisections, from 13 MS cases).

Statistics

Non‐parametric statistical tests (unpaired Wilcoxon rank sum test) were used to compare neuronal numbers between MS cases and controls, and between the myelinated and demyelinated regions within the MS group. Unpaired t–tests were used to make comparisons in neuronal sizes. Multiple regression analyses were used to examine the influence of age, gender, disease state and—in MS cases—disease duration and the extent of demyelination on neuronal size and number (Stata version 9; StataCorp, College Station, TX, USA).

RESULTS

Neuronal counts

Interneuron and motoneuron number is reduced in MS cases

Total neuron count was reduced in the MS cases in comparison with controls at the upper cervical (P = 0.0024) and upper thoracic (P = 0.0145), but not the lumbar (P = 0.7158) level. Similarly, in comparison with the control cases, motoneuron number was reduced in the MS cases at the upper cervical (P = 0.0073) and upper thoracic levels (P = 0.0144), whereas there was a trend towards a reduction in interneuron number at these levels (upper cervical P = 0.0549; upper thoracic P = 0.0507). Neither interneuron or motoneuron number was reduced at the lumbar level (interneuron P = 0.6081; motoneuron P = 0.7931) (Figure 2). In keeping with these results multiple regression analysis suggests that—controlling for age and gender—disease state had a significant influence on total neuronal number at the upper thoracic level (P = 0.010, with MS cases showing a 30.3% reduction in neuronal number). There was a trend towards significance at the upper cervical level (P = 0.069; 23.8% reduction in MS cases), but not the lumbar level (P = 0.632). Neuronal number at the upper cervical and upper thoracic levels did not appear to be influenced by disease duration in the MS cases (upper cervical, P = 0.5460; upper thoracic P = 0.8090).

Figure 2.

Bar charts of the number of interneurons and motoneurons at the upper cervical (A), upper thoracic (B) and lumbar (C) levels of the spinal cord. Values represent mean ± standard error.

Neuronal loss is greater in plaques than in myelinated gray matter

Total neuronal count was reduced in the gray matter plaques in comparison with controls at the upper cervical (P = 0.0018) and upper thoracic (P = 0.0043) levels. Similarly interneuron count was reduced in the gray matter plaques in comparison with controls at the upper cervical (P = 0.0262) and upper thoracic (P = 0.0099) levels. Motoneuron count was reduced in the gray matter plaques at the upper cervical (P = 0.0021), but not the upper thoracic (P = 0.1326) level. We were unable to adequately examine the influence of demyelination on neuronal number in the lumbar cord as only three hemisections were completely demyelinated at this level.

Total neuronal count was not reduced in the myelinated gray matter of MS cases in comparison with controls at any of the cord levels (upper cervical, P = 0.2530; upper thoracic, P = 0.0797; lumbar, P = 0.7799). Similarly interneuron count was not reduced in the myelinated gray matter of MS cases in comparison with controls at any of the cord levels (upper cervical, P = 0.9317; upper thoracic, P = 0.2079; lumbar, P = 0.8459). Motoneuron count was reduced in the myelinated gray matter of MS cases in comparison with controls at the upper thoracic (P = 0.0190) but not the upper cervical (P = 0.1039) or lumbar (P = 0.7706) levels.

Within the MS cases, the total neuronal count was reduced within the gray matter plaques in comparison with the myelinated gray matter at the upper cervical level (P = 0.0504, trend) but not at the upper thoracic (P = 0.2130) level. There was a trend towards a reduction in interneuron count within the gray matter plaques at the upper cervical (P = 0.0808) and upper thoracic (P = 0.0915) level. Motoneuron count was reduced within gray matter plaques at the upper cervical (P = 0.0276) but not the upper thoracic (P = 0.4093) level.

Measurements of neuronal size

Interneuron cross‐sectional area is reduced in MS cases in comparison with controls

The mean interneuron cross‐sectional area was smaller in MS cases in comparison with the controls at all three cord levels (upper cervical, 147.4 µm² in MS cases versus 185.1 µm² in controls, P = 0.0000; upper thoracic, 147.2 µm² vs 162.9 µm², P = 0.0002; lumbar, 186.8 µm² vs 193.9 µm², P = 0.0337) (Figure 3).

Figure 3.

Bar charts of the cross‐sectional areas of interneurons (A) and motoneurons (B) at different levels of the spinal cord. UC = upper cervical; UT = upper thoracic; Lum = lumbar. Values represent mean ± standard error.

There was no significant difference in motoneuron size between MS cases and controls in the upper cervical (P = 0.7416; 655.1 µm² in MS cases vs 645.9 µm² in controls) and upper thoracic (P = 0.6426; 583.7 µm² in MS cases vs 570.9 µm² in controls) levels. However the mean motoneuron size was greater in MS cases than controls at the lumbar level (P = 0.0059; 1001.0 µm² vs 908.4 µm²) (Figure 3).

Multiple regression analysis also indicated that disease state had a significant influence on neuronal size, with interneurons being smaller in the MS cases at the upper cervical and upper thoracic levels (upper cervical, P = 0.000; upper thoracic, P = 0.000; lumbar, P = 0.404) and motoneurons being larger in the MS cases at the lumbar level (upper cervical, P = 0.461; upper thoracic, P = 0.706; lumbar, P = 0.016). Disease duration had no influence on interneuron size at the upper cervical or lumbar levels, but did appear to influence interneuron size at the upper thoracic level (upper cervical, P = 0.362; upper thoracic P = 0.023; lumbar P = 0.537).

Interneuron cross‐sectional area is reduced in both myelinated and demyelinated gray matter

Mean interneuron size was reduced in the gray matter plaques in comparison with controls at the upper cervical (122.8 µm² in gray matter plaques vs 185.1 µm² in controls, P = 0.0000), upper thoracic (139.2 µm² vs 162.9 µm², P = 0.0004) and lumbar (173.6 µm² vs 193.9 µm², P = 0.0041) levels. Motoneuron size was not significantly different between the gray matter plaques and the controls at any of the cord levels (upper cervical, 633.7 µm² in gray matter plaques vs 645.9 µm² in controls, P = 0.7555; upper thoracic, 531.9 µm² vs 570.9 µm², P = 0.3199; lumbar, 993.4 µm² vs 908.4 µm², P = 0.2958).

Interneurons within the myelinated gray matter of MS cases were significantly smaller than in controls at the upper cervical (158.0 µm² in the myelinated gray matter of MS cases vs 185.1 µm² in controls, P = 0.0000) and upper thoracic (149.8 µm² vs 162.9 µm², P = 0.0033), but not the lumbar level (193.2 µm² vs 193.9 µm², P = 0.8694). Motoneuron size was not significantly different between myelinated gray matter in MS cases and controls in the upper cord (upper cervical, 665.0 µm² in the myelinated gray matter of MS cases vs 645.9 µm² in controls, P = 0.5847; upper thoracic, 598.6 µm² vs 570.9 µm², P = 0.3817), but was increased in the MS cases at the lumbar level (1023.4 µm² vs 908.4 µm², P = 0.0025).

Within the MS cases, interneuron size was reduced within the gray matter plaques in comparison with the myelinated gray matter at the upper cervical (122.8 µm² in gray matter plaques vs 158.0 µm² in myelinated gray matter, P = 0.0000) and lumbar (173.6 µm² vs 193.2 µm², P = 0.0148) levels. There was no significant difference at the upper thoracic level (139.2 µm² vs 149.8 µm², P = 0.1268). There was no significant difference in motoneuron size between gray matter plaques and myelinated gray matter in the MS cases (upper cervical, P = 0.4415; upper thoracic, P = 0.1144; lumbar, P = 0.7358).

DISCUSSION

To our knowledge, this is the first study to quantify neuronal pathology in the spinal cord in MS. We have studied material from 37 MS cases and 22 controls, performing counts on Nissl‐stained sections, adjusting the “raw” counts using the Abercrombie correction, to avoid the bias of over‐counting larger objects. There is much debate as to the relative advantages and disadvantages of profile‐based counts in comparison with stereological‐based methods 2, 17. When applied appropriately, profile‐based counts, used in conjunction with the Abercrombie correction, can provide extremely accurate estimates of neuronal number (17). We report absolute counts, rather than neuronal densities, counting all neurons in the anterior horn, regardless of the anterior horn volume. Cell density measurements can be difficult to interpret if changes in the reference space may occur in the disease state (28) (i.e. neuronal density may be influenced by oedema, gliosis, etc., in addition to neuronal loss per se).

We observe substantial reductions in neuronal number in the MS cases in comparison with controls. Consistent with two recent studies suggesting neuronal loss in the cerebral cortex occurs predominantly within cortical plaques 36, 37, we find that neuronal counts in demyelinated tissue are reduced in comparison with both control tissue and with the myelinated gray matter of MS cases. These changes only reach statistical significance in the cervical and thoracic cord. The apparent preservation of neuronal number in the lumbar cord may reflect a number of factors. First, within the lumbar region, relatively few sections demonstrate demyelination of the entire anterior horn (i.e. when comparing neuronal counts between gray matter plaques and myelinated gray matter we consider only anterior horns that are either completely demyelinated or completely myelinated). Second, there is a high degree of interindividual and interlevel variability in neuronal number in the lumbar cord (19), so modest reductions in neuronal numbers may go undetected.

The mechanism of neuronal loss within MS plaques is poorly understood. Demyelination may itself lead to neuronal injury, depriving the neuron of trophic support (10). Alternatively, a number of components of the inflammatory milieu are potential mediators of neuronal injury: lymphocyte‐mediated neurotoxicity has been reported in a number of settings, via a number of mechanisms 3, 26, 27, whereas antibodies to various neuronal antigens have also been reported in MS 23, 30, 33, although it is unclear if they contribute to neuronal injury. Neurons may also be susceptible to bystander injury secondary to cytokines (9), chemokines (34), reactive nitrogen and oxygen species (21), glutamate and proteases (32).

Axonal loss in MS almost exclusively affects small caliber fibers (11). We do not observe such a striking size‐related susceptibility to neuronal injury, with similar proportions of interneurons and motoneurons being lost. Focal muscular atrophy is a relatively uncommon but well recognized feature of MS, estimated to occur in 6%–7% of patients 13, 22. Given the magnitude of the anterior horn cell loss that we observed, it is perhaps surprising that we do not observe the classical signs of lower motoneuron pathology in the clinical setting more frequently. However, it is estimated that 50% of motoneurons are lost in motoneuron disease before muscle wasting and fasiculations are evident clinically (18).

Despite extensive axonal loss in the corticospinal tracts in MS (8) we find little evidence for neuronal loss in the myelinated gray matter of the spinal cord, suggesting that transneuronal degeneration is not a prominent mechanism of neuronal death in this region. It is hypothesized that susceptibility to transneuronal degeneration is dependant on the extent of collateral connectivity, with neurons possessing relatively few sources of synaptic contact being particularly vulnerable (14). In contrast many spinal cord neurons are involved in complex neuronal networks; such an abundance of synaptic connections may be sufficient to maintain neuronal survival following removal of descending afferents 20, 25, 35.

There appears to be marked regional heterogeneity both in the extent of neuronal pathology in MS and in the mechanisms driving these changes, although there are obvious limitations in comparing studies using different MS cohorts, different counting techniques etc. Cifelli et al reported neuronal loss in the thalamus (35% reduction in MS patients relative to controls) (6), whereas we observed substantial neuronal loss in the spinal cord (24% reduction in the upper cervical cord, 30% reduction in the upper thoracic cord). In contrast, Wegner et al reported a 10% reduction in neuronal number within cortical lesions, with no evidence of neuronal loss within the myelinated cortex (37). Such variability may reflect regional differences in: (i) the extent of gray matter demyelination; and (ii) susceptibility to retrograde and transneuronal degeneration, which varies considerably between neuronal populations (7). Therefore, neuronal loss in the spinal cord and the cerebral cortex appears to occur predominantly—if not exclusively—within local gray matter plaques, whereas it is likely that transneuronal degeneration is more prominent in the thalamus. We also note that despite substantial neuronal loss, the spinal cord gray matter volume is well‐preserved in the MS cases (15). Conversely, atrophy of the cerebral cortex occurs despite no change in neuronal density (37), highlighting the complex interaction between tissue loss and atrophy in MS. Presumably loss of neurons—and myelin—within the spinal cord is countered by some other process such as oedema, gliosis or expansion of the extracellular matrix, thus maintaining gray matter volume.

In addition to reductions in neuronal numbers in the MS cases, we find significant differences in neuronal sizes between MS cases and controls. The average interneuron cross‐sectional area is significantly reduced in the MS cases, suggesting there is shrinkage of these cells. Although these changes appear to be most marked within demyelinated areas (e.g. interneurons within gray matter plaques in the upper cervical cord show a 34% reduction in size in comparison with controls), reductions in interneuron size are also observed within myelinated tissue. Again the mechanisms behind these changes are unclear; interneuron shrinkage in the myelinated gray matter of the spinal cord may reflect transneuronal damage following axonal injury in the white matter tracts (11).

There is some evidence that motoneuron size is increased in the MS cases in comparison with controls. The difference only reaches statistical significance in the lumbar region. The significance of this finding is unclear; changes in motoneuron size are particularly difficult to interpret in the context of neuronal loss. An apparent increase in motoneuron size could result from: (i) pathological swelling of motoneurons; (ii) a preferential loss of smaller motoneurons resulting in an increase in the average size of remaining motoneurons; or even (iii) perikaryal shrinkage so that small motoneurons are misclassified as interneurons.

A number of limitations to our work warrant further discussion. First, the use of myelin‐stained sections will interfere with blinding. Second, we have used size criteria to divide neurons into presumed interneurons and presumed motoneurons. We are not aware of an immunohistochemical marker that is specific to either neuronal subtype and which can be reliably applied to formalin‐fixed material, particularly when fixation times are long. In contrast, the Nissl stain is extremely robust when used on the archival material. However, there is some overlap in size between the two neuronal populations (31) and, if there is perikaryal atrophy of motoneurons in the diseased state, a proportion of pathologically shrunken neurons may be misclassified as interneurons. Third, considerable dimensional changes occur in tissue post‐mortem as a result of fixation and embedding. Although post‐mortem delay and fixation time are likely to influence the degree of shrinkage (28), we are unable to control for these variables, because of a lack of detailed information. It is also possible that shrinkage differs between the “healthy” and diseased states. Finally, because of a lack of detailed clinical information (e.g. MS subtype, Expanded Disability Status Scale), we are unable to investigate the functional significance of neuronal pathology.

CONCLUSIONS

To our knowledge this study is the largest autopsy examination of neuronal pathology in MS. We find that neuronal loss in the spinal cord appears to occur predominantly within gray matter plaques, with both motoneurons and interneurons being affected. Although there is little evidence of neuronal loss within the myelinated gray matter, we do observe changes in neuronal size, indicating that pathology is not exclusive to plaques. Further work is required to elucidate the mechanisms of this neuronal injury. Additional studies are also required to establish whether particular subpopulations of interneurons are more susceptible to neuronal injury than others.