Chapter 2 Hallmarks of spinal cord pathology
Kreiter, D., Postma, A. A., Hupperts, R., & Gerlach, O. (2023). Hallmarks of spinal cord pathology in multiple sclerosis. Journal of the neurological sciences, 122846.
Abstract
A disparity exists between spinal cord and brain involvement in multiple sclerosis (MS), each independently contributing to disability. Underlying differences between brain and cord are not just anatomical in nature (volume, white/grey matter organization, vascularization), but also in barrier functions (differences in function and composition of the blood-spinal cord barrier compared to blood-brain barrier) and possibly in repair mechanisms. Also, immunological phenotypes seem to influence localization of inflammatory activity. Whereas the brain has gained a lot of attention in MS research, the spinal cord lagged behind. Advanced imaging techniques and biomarkers are steadily improving and providing us with tools to uncover the mechanisms of spinal cord pathology in MS. In the present review, we elaborate on the underlying anatomical and physiological factors driving differences between brain and cord involvement in MS and review current literature on pathophysiology of spinal cord involvement in MS and the observed differences to brain involvement.
2.1 Introduction
In multiple sclerosis (MS) inflammatory demyelination and neurodegeneration occur in both brain and spinal cord [48, 49]. Around 80% of MS patients develop spinal cord lesions [50, 51] which, more often than brain lesions, are symptomatic and can cause significant impairment such as disturbances in ambulation, coordination, bladder and bowel function.
Whereas the brain has gained a lot of attention in MS research, the spinal cord lagged behind. One of the main reasons is that the spinal cord is a challenging structure to image, due to the small diameter, physiological motion and larger variation in susceptibility in cord regions (e.g. air-tissue interfaces, bone, cartilage). As a result of several advancements in sequences, coil-design and (physiological) motion correction techniques in recent years, spinal cord MRI including advanced techniques have become increasingly feasible [52, 53]. Consequently, with the growing amount of studies focused on the MS spinal cord, we have come to know more about the similarities and differences in the manifestation of MS pathology between the different central nervous system (CNS) regions.
Most importantly, it has become clear that a disparity exists between spinal cord and brain involvement. Patients can have a high brain lesion burden, with no or minimal spinal cord involvement, or vice versa. New cord lesions can occur without evidence for cerebral disease activity and studies highlight that spinal cord and brain abnormalities (both lesion formation and atrophy) correlate poorly. This suggests that MS pathology in these CNS regions progresses independently [54–56].
Despite the prognostic importance of spinal cord lesions, cord outcome measures have scarcely been applied in MS treatment trials, leaving a gap in knowledge on the effect of disease-modifying treatments (DMTs) on spinal cord MS pathology. If the MS pathophysiological process in the spinal cord is somewhat different from the brain, this has potentially important implications for our therapeutic strategies. Therefore, in this review, we aim to (i) elaborate on the underlying anatomical and physiological factors driving differences between brain and cord involvement in MS and (ii) review current literature on the pathophysiology of spinal cord involvement in MS and the observed differences to brain involvement.
2.2 Determinants of regional differences
Anatomical
The human spinal cord, with an approximate total volume of 20mL (16mL white matter, 4mL grey matter), is much smaller than the brain parenchyma which has a volume of roughly 1.5L (0.6L white matter, 0.9L grey matter) [57–59]. Additionally, the brain grey matter (GM) lays at the subpial cerebrospinal fluid (CSF) surface, while for the cord this is white matter (WM). Likewise, the spinal cord GM interfaces with the central canal and the brain WM with the ventricular system. This is important because a possible relation exists between CSF interfaces and lesion localization [55].
With the perivenous organization of MS lesions being a well-established phenomenon [7], differences in cerebral and spinal vasculature also needs to be considered. In MS, it is thought that primed lympho- and monocytes extravasate on the venular side of the microcirculation, forming inflammatory perivascular cuffs [60]. The valveless, richly-anastomosed, spinal venous system is made up of the intrinsic (intramedullary veins), extrinsic (cord surface), extradural (epidural plexus) and external system (the structures outside the epidural space). The venous drainage within the cord (intrinsic system, see figure 2.1) consists of radially oriented peripheral veins, which mostly start in the deep grey matter and run through the white matter to the surface. Since these peripheral radial veins collect blood from the central grey matter as well as the white matter, their caliber often equals or exceeds that of the central sulcal veins. The central sulcal veins do not branch before they reach the white commissure (nerve fiber bundles crossing the midline of the cord)[61]. The intrinsic venous system is mostly horizontally oriented, which is the probable reason that MS cord lesions do not exceed more than two vertebral levels.
In contrast to the brain, for the spinal cord, evidence for a topographic relation of focal lesions to the spinal venous system is more modest. In 1950, neuropathologist Torben Fog performed a histopathological study of eight spinal cords from MS patients and concluded that “topographic distribution of plaques may easily be related to the spinal veins and their drainage territories” also finding that there was “good correlation between the form of the plaques and the shape of the venous pattern” [62]. Later, a pathological study of 18 MS spinal cords, agreed with Fog’s findings in regard to the venous relation and even stated that “the fact that many plaques arise on the course of a small vein is now too well-established to need reiteration” [63]. More recent pathological studies also suggest that the spinal cord lesion distribution follows the venous system network [64].
The radiological correlate of the perivenous organization, the “central vein sign”, can be well visualized with the aid of MR susceptibility-weighted imaging (SWI) for the brain. For the spinal cord this is more difficult, as spinal sulcal and radial veins are only around 0.1-0.2mm in diameter and physiological motion and larger variation in susceptibility challenging cord MRI imaging. One recent preliminary report (n=20) has shown that visualization of central veins in upper cord MS lesions is feasible. The investigators applied high-resolution SWI imaging of the upper cord and 40% of the participants with upper cord lesions had at least one central vein [65].
![Figure 2.1 - Schematic axial representation of the intrinsic and (part of) the extrinsic venous system of the spinal cord, showing peripheral (radial) centrifugal veins and central (sulcal) veins, with MS plaques depicted in frequently reported locations in the dorsal and lateral columns with involvement of white as well as grey matter, based on descriptions from [61, 63, 64]. * anastomosis between posterior and ventral median veins, which can commonly be present.](img/venous_anatomy_rev.png)
Lastly, the lymphatic system is essential in interstitial waste clearance and immune cell trafficking. However, as the CNS lacks a classical lymphatic drainage system, the mechanisms that regulate the entry and exit of immune cells in the CNS remain poorly understood [66, 67]. It is hypothesized that interstitial fluid from within the CNS parenchyma is collected by the glymphatic system (system of perivascular channels formed by astroglial cells) to the CSF and to a meningeal lymphatic system. This meningeal lymphatic system carries both fluid and immune cells from the CSF and is connected to the deep cervical lymph nodes [66]. In experimental autoimmune encephalomyelitis (EAE) mice models, ablation of meningeal lymphatic vasculature reduced the inflammatory response [68]. Since still a lot about ‘CNS lymphatics’ and its role in neuroinflammation is to be revealed, whether relevant differences exist in immunosurveillance between brain and spinal cord is unknown.
Blood spinal cord barrier
Animal studies show that the blood-spinal cord barrier (BSCB) is not just a continuation of the blood-brain barrier (BBB). Each has its own morphology and features, making the BSCB an independent physiological entity [69]. Proteins that make up the tight junctions differ between the BSCB and BBB, i.e. decreased expression of zonula occludens-1 and occludin in BSCB compared to BBB [69]. Varying experiments suggest that in normal conditions the BSCB is more permeable than the BBB, also for cytokines [69]. Regional differences seem to exist regarding the passage of interferon-gamma (IFN\(\gamma\); see ‘Immunological’ below) and tumor necrosis factor-alpha (TNF\(\alpha\)) from the periphery into nervous tissue. TNF\(\alpha\) is an immunomodulatory cytokine produced by a variety of cells, which depending on receptor binding exerts a pro-inflammatory (TNFR1; causing secondary neuronal and axonal damage in MS) or anti-inflammatory / protective effect (TNFR2; remyelinaton and neuroprotection in MS) [70]. The permeability for IFN\(\gamma\) and TNF\(\alpha\) is greatest in the cervical spinal cord, followed by the lumbar and thoracic regions [71]. This possibly contributes to why, within the cord, there is a predilection for lesion formation in the upper cervical cord. Though current evidence is mainly from animal studies, differences in BBB/BSCB permeability might have a role in the regional differences in lesion pathogenesis and may also be relevant for DMT drug distribution, since some DMTs have been shown to penetrate the CNS with possibly central activity as well (e.g. cladribine, dimethyl fumarate, S1P inhibitors)[72].
Immunological
Are there immunological factors that influence localization of MS pathology? To start with the innate immune system within the different CNS regions, a recent transcriptomic study shows that, in human CNS tissue samples (brain, cerebellum, spinal cord) from MS patients and controls, there is a strong regional glial subtype diversity when this evaluated in controls. However, these differences are less evident in MS tissue samples. The authors suggest that the pattern of transcriptomic changes in MS are shared across CNS regions and converge to specific pathways, foremost those regulating cellular stress and immune activation [73].
In the adaptive immune system, there is evidence from EAE experiments that the balance between CD4+ T cell subsets and their cytokine profiles influence spatial localization of inflammatory activity. The balance between Th1 and Th17 cells is thought to determine whether there is more brain or spinal cord inflammation, foremost mediated by the cytokines IFN\(\gamma\) and IL-17 [74]. There is some early evidence that this is also the case in MS in humans [75]. Collectively, these studies suggest that IFN\(\gamma\) can play both a protective and a pathogenic role in CNS autoimmunity in part by the differential modulation of chemokine production in the brain versus spinal cord.
Corroborating the theory of immunological actors from the adaptive immune system influencing the degree of involvement of the different CNS regions, is the fact that there seems to be immune system-related genetic loci influencing predisposition of lesion localization [76]. Different studies were conducted evaluating whether different HLA alleles influence lesion localization in MS patients. HLA-DRB1*1501 has been most often associated with more extensive spinal cord pathology [77–79]. HLA-DRB1 encodes for the beta chain of the major histocompatibility complex class II, whose ligands are the CD4 receptors of helper T cells. Currently, to our knowledge, there are no other known single-nucleotide polymorphisms that have been associated with increased spinal cord involvement relative to the brain in MS. This would be of interest (i) to better understand the immunological pathways involved in localization of MS pathology and (ii) to find predictors of cord involvement and subsequently identify subgroups which would benefit from routine spinal cord MRI monitoring.
Remyelination and repair
Finally, the degree of repair capabilities is possibly different between brain and cord. In a post-mortem study in tissue blocks from 120 RRMS/SPMS patients, a much larger proportion of lesions in the brain showed complete remyelination compared to the spinal cord [80]. In contrast, in the studies by Bramow et al[81] and Luchetti et al[82] (both SPMS/PPMS specimens), the proportion remyelinated plaques was comparable between brain and spinal cord. In all three studies, spinal cord lesions were considerably less often mixed active/inactive (smouldering lesions) and more often inactive.
Remyelination is mediated by surviving mature oligodendrocytes within a lesion, de novo oligodendrocytes (from oligodendrocyte progenitor cells) [83] or Schwann cells. The latter are only seen in spinal cord lesions [84, 85]. A few small studies show there being less remyelination-capable oligodendrocytes in cord lesions compared to brain lesions, but the authors argue that this may be related to the age of the lesions or a possible vulnerability of cord lesions to repeated bouts of demyelination [86, 87]. There is an increasing focus on the development of remyelination therapies. Given the current sparse knowledge on spinal cord remyelination and the significant impact of cord lesions on disability, a better understanding of remyelination in the cord and how it differs from remyelination in the brain is a potential important target for future research.
2.3 Spinal cord pathology in MS
Given the drivers of regional differences discussed above, we review what the current literature teaches us about spinal cord MS pathology (demyelination, inflammation and axonal loss) and where it is different from brain pathology. Methods are needed for in vivo quantification of disease, not only for clinical use, but also for research to be able to study MS pathology in the early phase of the disease. Here, MRI is the main instrument which offers a variety of contrasts and techniques. For an in-depth summary on the application of advanced MRI techniques for the cord and their limitations, we refer to the reviews by Combes et al [8] and Moccia et al [88]. Figure 2.2 provides an overview of changes in quantitative MRI parameters in the spinal cord for lesioned and normal-appearing tissue.
Demyelination
The most prominent pathological hallmark of MS is demyelination with relative preservation of axons throughout the CNS [89]. But, how can we measure it? Pathological studies can accurately determine and classify different MS pathophysiological processes like demyelination using stains specific for various myelin degradation products and immunohistochemistry. However, such studies are inherently limited by a biased sample; Post-mortem samples are mostly from patients with a longer disease duration, generally in the progressive phase of the disease. Also, pathological studies provide only a snapshot in time and do give limited longitudinal information.
To assess demyelination in vivo, conventionally T2 contrast is used, but has a low specificity as it is influenced by other changes in intra-/extracellular water. Additionally, it is well established that there are also pathological changes in the normal-appearing white (NAWM) and gray matter (NAGM). Advanced techniques including myelin water imaging (MWI), magnetization transfer (MTR) and diffusion-tensor imaging (DTI), provide means to quantify demyelination in the normal-appearing spinal cord (NASC).
Lastly, electrophysiological tools like motor (MEP) and somatosensory evoked potentials (SSEP) can in theory provide a more direct measure of spinal cord neuronal function than structural imaging, where increased latencies can be caused by demyelination (and also axonal loss), but are not specific. Whether latencies are of longitudinal value in MS is not studied.
Demyelination in the spinal cord
The traditional pathological hallmark of MS are sharply demarcated focal inflammatory lesions with primarily demyelination, which can involve the WM as well as GM [6]. Additionally, there are also diffuse changes within the MS NASC, including demyelination, e.g. diffuse subpial demyelination in the cord [90]. Radiological studies have taught us a lot on topographical patterns of focal lesions. However, for diffuse changes this remains challenging and we still mainly rely on observations from pathological samples. Findings from these studies are discussed below.
Spatial organization
At the time of the first clinical episode approximately half of patients have cord lesions indicative of focal demyelination on MRI [91, 92]. In early RRMS patients, most focal lesions are found in the upper cervical cord (~40%), followed by the lower cervical cord (~20%) and lower thoracic cord (~25%), with the least lesions found in the upper thoracic cord (~15%) [50]. Different theories exist explaining this predilection of focal lesion for the upper cervical cord. Possibly this has to do with the regional difference in blood-spinal cord barrier permeability and differences in venous drainage. Another theory is based on regional differences in myelin protein composition. However, in a study by Trotter [93], differences in composition between the cervical and thoracic cord seem limited, but no distinction was made between the upper and lower cervical cord. Finally, the larger amount of mechanical stress (due to movement) in the upper cervical cord compared to the lower cervical/thoracic cord could also contribute to this predilection [63].
Regarding the organization of lesions among the cord tracts, MS cord lesions are predominantly found in dorsal and lateral columns and less in the anterior and central portions of the spinal cord. Lesions in the lateral, anterior and central part occur relatively more often in progressive subtypes [94]. In histopathological samples, the GM is more extensively involved than the WM in the spinal cord [95, 96]. Lesions in the spinal cord grey matter were more frequently inactive and active inflammatory lesions were more prevalent in the white matter [64]. Recent radiological studies found that the grey matter is equally to more extensively involved in progressive subtypes, while in relapsing-remitting MS the lesion volume was higher for white than for grey matter (normalized for WM and GM volume, respectively)[55, 94].
With radiological as well as pathological evidence for periventricular and cortical lesion gradients in the brain [55, 97], it is theorized that a CSF-mediated process, secondary to meningeal intrathecal inflammation, may play a role [98]. In a 7T MRI study, spinal cord lesions were mostly localized in the outer- (subpial) and innermost layers (surrounding the central canal), supporting this theory [55]. Reduction in MTR also seems to be greatest nearer to the subpial surface and near to the central canal in regions [99]. In contrast, a recent histopathological study found relative sparing of the subpial zone for demyelination in MS spinal cords. The authors argue that this discrepancy with the findings in radiological studies is possibly related to the greater sensitivity of pathological assessment over MRI or that the subpial signal changes on MRI are partially due to pathological process other than demyelination, e.g. reactive gliosis [64].
In addition to demyelination within plaques, diffuse partial demyelination can be found in the NASC in MS, which is more common in PPMS [49, 100] and which is also found in in vivo MWI [101, 102]. Decreases in MTR are found in lesioned spinal cord as well as NASC [99], however, these MTR changes are not specific for demyelination since it is also influenced by axonal loss, inflammation and other changes in water content [103].
Temporal characteristics
Not much is known on the course of spinal focal lesions. In a study where brain and cord MRI were preformed monthly during a year, gadolinium-enhancement of cord lesions (in particular the thoracic cord) was less common than for brain lesions. Additionally, when there was enhancement of cord lesions, this never lasted >1 month for cord lesions, whereas longer enhancement was common for brain lesions [104]. This highlights the regional differences in barrier function.
Studies applying advanced MRI techniques show that N-acetyl aspartate levels (NAA; spectroscopy) [105] and radial diffusivity (RD; diffusion-tensor imaging) are associated with recovery after a spinal relapse [106, 107]. NAA is produced by mitochondria. It is theorized that the initial decrease in NAA corresponds to decreased mitochondrial activity in the presence of inflammatory mediators. With the subsequent increase in NAA reflecting the enhanced mitochondrial activity necessary for maintaining axonal conduction [105]. The mechanism behind the increase in RD is that with myelin damage, there is increased diffusion perpendicular to the fiber direction, which is associated with demyelination in pathological studies [108] and with greater disability in vivo [109–112]. MWI has shown good correlation with histological myelin staining in brain as well as spinal cord [113–116]. The myelin water fraction (MWF) in vivo is decreased in MS lesions and is seen to partly recover with remyelination [113, 117–119]. In PPMS, a MWI study found a 10% decrease of myelin in the NAWM of the cervical cord over 2 years, while the myelin water fraction in controls remained stable, which provides another argument for the presence of diffuse demyelination in progressive disease [101].
To our knowledge, it has not been studied whether metabolites or diffusion-measures evolve differently in cord compared to the brain in lesions/normal-appearing tissue. Potentially, this could provide clues in identifying regional differences in repair mechanisms.
Inflammation
Inflammation is the force driving demyelination and axonal loss. Immunohistochemistry in pathological samples can provide a highly detailed view of the inflammatory process and involved immune cells. Here too, it only provides a static picture with overrepresentation of progressive subtype MS patients. Gaining insight into the inflammatory process in vivo is more challenging: With conventional and most advanced MRI techniques we only visualize indirect effects of inflammation (e.g. edema using T2 contrast/relaxation and in MTR), increased metabolic activity (lactate in MRS) or glutamate production by activated leucocytes, microglia and macrophages (using MRS). More recently, there has been increased interest in positron emission tomography (PET) using radioligands specifically targeted at microglia and macrophages. This taught us that activated macrophages and microglia are present in the NAWM of the brain and this correlated with disability accumulation over time [120, 121]. The application of PET imaging in the spinal cord is more challenging than for the brain, making it a less useful tool to investigate microglial activity in the cord.
It is also possible to determine biochemical markers for neuroinflammation like e.g. immunoglobulins, oligoclonal bands, lymphocyte subsets in CSF and/or serum. However, knowledge on these markers in relation to spinal cord pathology is limited.
Inflammation in the spinal cord
New active lesions are hypercellular with loss of myelin and dense infiltration of monocytes/microglia. T cells are found perivascularly and throughout the lesion. There are, however, much less T cells within lesions than there are macrophages/microglia [6], while in the spinal meninges, T cell infiltration is prominent [90]. Meningeal inflammation and lymphoid-like structures within the meninges are found not only in the brain, but also in the spinal meninges [122]. Androdias and colleagues hypothesized that possibly activated meningeal T cells, through cytokines like IFN\(\gamma\), cause an increase of parenchymal microglia recruitment and activity [90]. However, Reali et al found that more severe spinal cord pathology in post-mortem samples from SPMS patients was correlated with the density of B cell, but not T cell, in spinal meningeal infiltrates [122]. Given that progressive subtypes are associated with more extensive spinal cord involvement and the poor response to disease-modifying therapies in these patients, this raises the question if it is possible that the limited effect of therapy in the progressive subtypes might be due to the intrathecal persistence of pathogenic B and T cell clones driving parenchymal inflammation and which are poorly accessible for peripherally delivered therapies [90]. While complement and immunoglobulin depositions can be found in lesions in brain and spinal cord, whether there is in fact antibody-mediated demyelination (direct, or indirect via the activation of autoreactive T cells) or as a result of complement-based mechanisms (independent or dependent of antibodies), is still uncertain [123–125].
When lesions are mixed active/inactive (smouldering), they are demyelinated and have a hypocellular center with a border rim of activated macrophages/microglia. These lesions are mainly found in the progressive disease stage and like mentioned earlier are rarely found in the spinal cord [80, 82, 125]. The macrophages/microglia in the rim are iron-laden. These paramagnetic rims can be detected with susceptibility-weighted MRI, with good sensitivity and specificity in histopathological-MRI correlation studies for the brain [126]. Due to technical constraints and the limited presence of smouldering lesions in the cord, paramagnetic rim lesions have not yet been visualized in the spinal cord.
Finally, when there are only few T cells and macrophages/microglia left with almost complete depletion of oligodendrocytes, a lesion is considered inactive.
In regard to biochemical biomarkers, Oechtering et al [127] recently showed that intrathecal IgM synthesis in early MS is associated with clinical spinal cord syndromes at first presentation and the number of spinal T2/T1-Gd+ lesions, while there was no association with cerebral lesions in that study. The authors hypothesized that there is a higher concentration of IgM locally in the spinal canal compared to brain CSF, which is based on the observations that leptomeningeally produced proteins have higher concentrations in lumbar compared to ventricular CSF [128].
In the normal-appearing GM and WM of the cord, different studies found more CD3+ T-cells and CD68+ macrophages/microglia in MS patients compared to controls [78, 90], which, in turn, is possibly also associated with the degree of axonal loss [90]. HLA-DRB1*15 positive patients show significantly increased T cell inflammation in both normal appearing WM and GM compared to HLA-DRB1*15 negative patients, also when controlled for the presence and activity of plaques [78]. In a radiological study, carriage of HLA-DRB1*15 in MS patients was associated with only more spinal cord lesions and not brain lesions [77].
Axonal loss
Axonal loss can occur within focal plaques secondary to inflammatory demyelination, but also outside focal lesions. Loss of axons outside lesions can be attributed to Wallerian degeneration along the tracts secondary to focal plaques. However, the role of Wallerian degeneration seems limited, since no close relationship between lesion load and axonal loss exists [129, 130]. Therefore, in regard to axonal loss, there seems to be an important role of diffuse inflammation [90]. This chronic self-sustaining compartmentalized inflammation (mainly microglia and astrocytes) produces neurotoxic inflammatory mediators (reactive oxygen species, cytokines, chemokines) and can lead to axonal injury [131].
In histopathology, axonal loss can be determined by using an axon specific stain and then counting axons per surface area on microscopy. In vivo, on conventional MRI, increased T1 relaxation (so called black holes) somewhat correlates with axonal loss, but is also influenced by demyelination and other changes in intra/extracellular water. Additionally, you can determine brain and spinal cord volume/atrophy on conventional MRI. Here, interestingly, the atrophy rate at follow-up is higher in the spinal cords than in the brains of MS patients [132]. However, the correlation of spinal cord volume loss and axonal density seems to be limited [133], since it is only an indirect measurement and is also influenced by demyelination and inflammation [134]. Moreover, brain or spinal cord volume loss entails more than disease-related atrophy, like volume loss related to aging-related changes, lifestyle, comorbidities and physiological fluctuations (hydration state, time of day). Furthermore, there are technical factors that are of influence, like inter-/intrascanner variability, angle of the neck, motion and what segmentation technique is applied [40, 135]. Lastly, one has to take into account pseudo-atrophy, a phenomenon where there is a temporary increase in volume due to inflammation, which then decreases over time or with treatment, causing a relative volume loss. Taken together, this makes it challenging to draw conclusions about axonal loss in MS from spinal atrophy MRI studies.
Some imaging makers using advanced MR techniques have been proposed for axonal loss: NAA, which is detected almost exclusively in neurons, has shown a correlation with spinal axonal loss within lesions and the normal-appearing cord [136], but this relation has only be studied to a limited extent. In mouse models, spinal axial diffusivity (AD; diffusion tensor imaging) decreases with axonal loss [137, 138]. However, in histopathological studies of human MS patients, the AD measure does not seem to discriminate in the degree of axonal injury [108]. It is hypothesized that AD decreases only in acute axonal damage and then increases due to gliosis and the increased extracellular space in the chronic phase [108]. Also, the neurite dispersion index (a metric in ‘neurite orientation dispersion and density imaging’) estimates the fraction of axons and dendrites within the neural tissue and is thought to be primarily sensitive to axonal loss, though, it is influenced by demyelination too [139]. A histopathological correlation has been shown within lesions [139], but not yet for the normal-appearing WM/GM.
Neurofilament light chain (Nfl), as a cytoskeletal component of neurons, gets released in the extracellular space after neuronal damage and is an increasingly studied biomarker in MS [140]. It has been found to correlate with brain as well as spinal cord atrophy [141], but whether it directly correlates with spinal axonal loss is unknown.
In summary, a specific in vivo marker for axonal loss is still lacking, as changes in current markers are generally not specific for axonal loss but also affected by other pathological processes (demyelination, inflammation, gliosis) in MS.
Axonal loss in the spinal cord
From post-mortem studies it is known that there is extensive axonal loss (~30-60%) in the MS spinal cord compared to controls without MS [129, 134, 142], on all spinal cord levels in the WM as well as GM [133, 143]. The degree of spinal axonal loss is correlated with disease duration [133]. Besides the neuroaxonal injury within plaques, there is diffuse axonal loss in the normal-appearing brain and spinal cord tissue [133], for which different mechanisms have been proposed: Firstly, there is Wallerian degeneration along tracts secondary to focal plaques, which plays a limited role given that plaque load in the spinal cord does not correlate with axonal density [129, 133]. Also, there is the pronounced inflammation and microglial activation outside plaques which markedly exceeds that seen in classical Wallerian degeneration. The diffuse axonal loss in the NASC is thought to be related to this diffuse inflammation, with different studies showing correlations between diffuse axonal loss and the presence of microglia [90, 122].
Spinal cord atrophy MRI studies show that cord atrophy is already detectable in early disease [144, 145], is independent of lesions [146] and is associated with disability [147–149] as well as being predictive for future disability accrual [150]. Mainly cord GM atrophy seems to be associated with a progressive phenotype [150, 151]. Cord atrophy is more pronounced in the cervical compared to the thoracic and lumbar cord and occurs in the WM as well as GM [149]. The atrophy rate of the cord is higher than for the brain in MS with faster tissue loss in progressive subtypes [152]. There is evidence that there is a cranio-caudal pattern of atrophy with tissue loss at C1/C2 in early RRMS, progressing in caudal direction with loss in the lower cervical segment in progressive MS. It is theorized that this due a higher myelin content and white matter fiber density of the upper cord levels [153]. However, to what extent axonal loss contributes to atrophy detected on MRI remains uncertain.
2.4 Treatment & spinal cord pathology
Spinal cord outcome measures have scarcely been used in MS clinical trials, causing knowledge on the effect of DMTs on spinal pathology to be limited. The number/volume of new spinal cord lesions has, to our knowledge, only been an outcome measure in a single trial: Laquinimod was compared against placebo in PPMS patients and no treatment effect on formation of new T2 cervical lesions was found [131]. Also, from real-world observational studies no specific data exists on the impact of DMTs on the formation of spinal cord lesions. Spinal cord MRI is included in some observational studies with “no evidence of disease activity” (NEDA) as outcome, but these do not specifically report the number of new spinal cord lesions during follow-up [154–156]. Given that approximately 10% of patients show asymptomatic spinal MRI activity in absence of new brain lesions [33, 157], there are arguments that spinal cord MRI should become part of future MS drug trials. For mouse models, there is evidence for at least some DMTs including fingolimod, fumaric esters and glatiramer acetate, that they also exert an effect on inflammation in the cord [158–161].
A few trials evaluated the effect of DMTs on cord atrophy [162]. Various small trials found no beneficial effect of interferons and glatiramer acetate on cord atrophy in RRMS and PPMS [163–167]. In the large INFORM trial (336 patients in treatment arm) fingolimod too did not show a favourable effect on cord atrophy after 3 years in PPMS [168]. The aforementioned laquinimod-trial, in addition to not reducing lesion formation in the cord, found no evidence for reduction of the cord atrophy rate in PPMS [131].
2.5 Conclusion
The spinal cord is a separate entity in MS pathophysiology with relevant differences to the brain, not just anatomically (volume, white/grey matter organization, vascularization), but also in barrier functions (differences in function and composition of the blood-spinal cord barrier compared to blood-brain barrier) and possibly in repair mechanisms. Also, immunological phenotypes seem to influence localization of inflammatory activity within the CNS. Altogether, this forms an underpinning for the observation that brain and cord pathology progress independently in MS patients in in vivo radiological studies.
These findings argue that brain and cord should be evaluated separately regarding treatment. As currently knowledge on the effect of DMTs on MS cord pathology is scarce, it should be a topic for further study, with it holding important chances in improving treatment strategies in MS patients with more spinal cord involvement.
With the toolbox being increasingly filled with improved advanced MRI techniques and immunological/biochemical markers, new possibilities exist in growing our understanding of differences between the spinal- and brain disease process in MS, how this is affected by treatment and how remyelination/repair varies between CNS locations.