Background
Multiple sclerosis (MS) has long been considered a predominantly white matter (WM) disease due to the presence of demyelinated plaques in major WM tracts, observable at a gross anatomical level and using classical histology and conventional magnetic resonance imaging (MRI). Only during the last 15 years has the true extent and clinical impact of cortical grey matter (GM) pathology been documented [
3,
5,
31,
47]. Both neuroimaging and human tissue studies indicate that the extent of GM pathology correlates with disease severity and rate of progression [
6,
8,
22,
51] and is initiated at the earliest stages of MS [
7,
33]. Demyelinated lesions in cortical GM are accompanied by axonal pathology and neuronal and neuropil loss [
36,
37,
43,
47,
60,
62] and the magnitude of this loss is associated with a shorter time to milestones of clinical progression [
29,
37,
51].
There is extensive evidence from numerous studies for significant ongoing inflammation in both the WM and GM of the progressive MS brain, both SPMS and PPMS [
10,
20,
23]. But in contrast to the WM, lesions in the GM are generally not accompanied by significant perivascular infiltrates or evidence of blood-brain barrier leakage [
3,
59]. It is suggested that rather than an abrupt transition from the acute episodes of peripheral immune cell influx that characterise RRMS, it is likely that there is a progressive build-up of sequestered inflammation in the meningeal and WM and GM perivascular spaces that become self-sustaining as MS progresses [
5,
32,
51]. However, we know relatively little about the molecular mechanisms involved in cortical tissue damage and the relationship between inflammation and neurodegeneration. Increased cortical demyelination, neuron and neurite loss is associated with greater meningeal inflammation [
10,
29,
31,
33,
36,
37,
54], including the presence of aberrant tertiary lymphoid-like structures in a significant proportion of cases [
29]. Moreover, a strong association between meningeal inflammation and severity of pathology has also been shown in the spinal cord ([
2,
11] and in brain biopsies from patients with a recent diagnosis of MS [
33]. Such lymphoid-like aggregates of immune cells are suggested to drive chronic inflammation in target organs in many other inflammatory or autoimmune conditions [
40,
46,
55] by accelerating and/or maintaining the disease process. However, the molecular mechanisms by which they drive chronic disease are not well established.
Further investigation of the nature of leptomeningeal infiltrates in MS has shown that, in addition to B-lymphocyte aggregates, they contain both CD4
+ and CD8
+ T-lymphocytes [
37], many of which express IFN-γ [
53], and myeloid cells expressing TNF [
24]. Gene expression for TNF and IFN-γ is increased in SPMS cases exhibiting meningeal lymphoid-like structures, together with increased protein levels in the cerebrospinal fluid (CSF) [
35]. TNF and IFN-γ have been shown to act synergistically to increase apoptosis in human oligodendrocytes in culture [
48] and to upregulate TNFR1 expression, thereby rendering oligodendrocytes responsive to TNF [
1]. These data imply a direct role for meningeal immune infiltrates in releasing relevant inflammatory mediators that may diffuse through the adjacent cerebral cortex to, directly and/or indirectly, mediate demyelination and neurodegeneration [
37].
Whereas soluble TNF (sTNF) signals predominantly via TNFR1 to promote pro-inflammatory reactions, transmembrane TNF (tmTNF) signals via both TNFR1 and TNFR2 to activate protective and homeostatic functions. Aberrant TNF production plays a role in the pathogenetic mechanisms of many autoimmune and chronic inflammatory conditions, including rheumatoid arthritis, Crohn’s disease, psoriasis, systemic lupus erythematosus, type II diabetes and atherosclerosis [
28,
49]. Chronic overexpression of TNF by astrocytes in mice causes CNS inflammation, oligodendrocyte apoptosis, demyelination and neurological dysfunction [
30], even in the absence of mature lymphocytes. Selectively blocking the effects of soluble TNF by treating EAE mice with a dominant negative TNF monomer improved clinical outcome by reducing the production of pro-inflammatory cytokines and chemokines, whilst leaving tmTNF free to signal via TNFR2 to promote repair and neuroprotection [
4,
9,
57]. Thus, there is increasing evidence that TNF plays a major role in the pathogenesis of MS via TNFR1 signalling.
In order to identify some of the signalling pathways that may be involved in the increase in cortical pathology in response to an increased inflammatory milieu in the subarachnoid space, we have carried out a gene expression profiling study on subpial cortical GM lesions and nearby normal appearing GM from the motor cortex of MS cases with substantial meningeal infiltration, compared with cases with only mild meningeal inflammation and non-neurological controls. In addition to highlighting a number of the key molecular mechanisms of cortical injury in progressive MS, our results suggest that the degree of meningeal inflammation affects the balance between TNFR1 pro-cell death and TNFR1/TNFR2 pro-cell survival signalling, which then determines the severity of the pathology.
Discussion
Chronic inflammation that is sequestered within the CNS is suggested to be one of the main drivers of the accumulation of neurological deficit during the progressive stages of MS [
5,
13] and is manifest in the cerebral cortical grey matter as a build-up of meningeal immune cell infiltrates [
29,
36], subpial demyelination [
3,
31,
36] and neuronal and axonal damage [
8,
25,
37,
43,
47,
62]. We show here, using gene expression profiling of demyelinated and normal appearing grey matter from the motor cortex of secondary progressive MS cases with high and low levels of meningeal immune cell infiltrates, that at the RNA transcript level, there is a change in the balance of TNF signalling pathways from TNFR1/2-activated NFκB-dependent cell survival towards TNFR1 activated RIPK3 dependent necroptotic cell death with increasing levels and organisation of meningeal infiltration. Such an altered balance, varying in its extent between different MS cases, may help explain the heterogeneity seen in the degree of cortical pathology and its contribution to disease progression.
Previous studies of gene expression changes in MS cortical grey matter have studied MS cases that have not been stratified in any way and have considered relatively small numbers [
19,
58]. When the present data from the study of 20 SPMS cases are considered without any prior stratification, the results are largely in agreement with previous studies and identify changes to individual genes and gene networks involved in activated microglial function, inflammatory processes, oxidative stress [
20] immunoglobulin synthesis [
58], neuronal damage and mitochondrial dysfunction [
19]. The finding of Ig-related genes as one of the group of genes with the highest level of upregulation is consistent with a previous finding [
58] and may be explained not only by the presence of contaminating RNA from the adjacent meninges containing B cells and plasma cells but also by the increased frequency of perivascular B cell infiltrates previously identified in cortical grey matter from MS cases with increased meningeal inflammation [
38]. Previous studies have also highlighted the presence of neuronal and axonal degeneration in the NAGM [
19] and have suggested their role in the accumulation of irreversible disability in progressive MS. Our finding of significant decreases in gene expression for multiple neuronal and synaptic proteins confirms and extends this to show that similar changes occur in both GMLs and NAGM, suggesting that the mechanisms leading to neuronal degeneration may be largely independent of demyelination.
Although previous studies have illuminated a number of downstream pathogenetic mechanisms involved in cortical damage [
18‐
20,
26], they have not been able to identify possible initial inflammatory and/or neurodegenerative triggers for the subsequent cascade of pathological events [
5,
12]. In light of our previous findings suggesting a link between meningeal inflammation and increased demyelinating and neurodegenerative pathology, we have specifically sought to understand the molecular mechanisms by which increasing cortical pathology occurs by stratifying well-characterised MS cases into those with high levels of meningeal infiltrates with lymphoid-like tissue formation and those with lower levels of diffuse meningeal infiltrates. The presence of lymphoid-like structures in the meninges of a substantial proportion of cases with secondary progressive MS (SPMS), that associates with more extensive subpial cortical damage, early disease onset and rapid clinical progression [
29,
37,
51], suggests that diffusion of cytotoxic and myelinotoxic factors from the inflamed meninges across the compromised glia limitans might have a major role in causing injury in the adjacent cortical GM. Furthermore, increased levels of gene and protein expression for TNF and IFN-γ are found when increased meningeal infiltration is seen [
35]. Therefore, it is not unexpected that there would be an upregulation of TNF signalling pathways in the MS cortical grey matter.
Substantial evidence exists for a role for TNF in the pathogenesis of MS [
28,
52], both in the relapsing-remitting stage that is driven by the peripheral immune response, and also in the progressive stage when a compartmentalised inflammatory response may predominate. Our finding that an unbiased pathway analysis of cortical grey matter tissue revealed changes to multiple pathways and processes involving TNF/TNFR interaction extends this evidence to include a role in the pathogenesis of cortical injury. The finding of significant changes to TNF/TNFR signalling pathways at the bulk transcript level when comparing all MS cases to controls, which then were shown to vary according to the level of meningeal inflammation, provides further confirmation of a major role for TNF in the genesis of cortical pathology in MS. Local delivery of TNF into the CNS by intraventricular injection exacerbates EAE clinical disease [
56] and local transgenic production in the CNS by astrocytes results in an inflammatory demyelinating pathology [
30,
50]. The different responses to binding of sTNF and tmTNF to TNFR1 and TNFR2 in the inflamed CNS have recently been elucidated in EAE mice [
4,
34,
57,
64] and demonstrate that, whereas binding of sTNF to TNFR1 results in the production of pro-inflammatory cytokines and increased pathology, binding of tmTNF to TNFR2 promotes remyelination and neuroprotection. TNFR1-mediated signalling in the absence of TNFR2 expression results in exacerbated chronic EAE disease [
34]. The change in the balance of TNFR1- vs TNFR2-mediated signalling pathways at the gene expression level in the MS cortical GM in response to the degree of meningeal inflammation supports these animal studies. However, it needs to be emphasised that our data suggests that there is a change in the balance of TNFR2 to TNFR1 signalling, not a complete shift. Although we have separated the MS cases into those with organised lymphoid-like structures in the meninges (i.e. high levels of inflammation) and those without (lower levels of inflammation), in reality it is a biological continuum. Therefore, follicle-negative cases also have a variable level of cortical demyelination, neurodegeneration and diffuse meningeal infiltrates, but at a lower level than the follicle-positive cases.
Although the levels of TNFR1 and TNFR2 protein were not significantly different between F+ SPMS and F− SPMS cases, our further analysis of the downstream pathways suggests that the increased pro-inflammatory reactions in the MS meninges, indicated by increased levels in patient CSF [
35], direct soluble TNF/TNFR1 interaction towards RIPK1/RIPK3/MLKL-mediated necroptosis, rather than caspase 8-dependent apoptosis or IKK/NFkB-dependent cell survival. This is in agreement with a recent study in which RIPK3-mediated necroptosis was demonstrated in oligodendrocytes in the MS brain under caspase 8-deficient conditions [
42]. Similar to this study, we found that reduced CASP8 and c-FLIP gene expression was accompanied by increased expression of genes, such as CYLD and RNF11, involved in de-ubiquitination of RIPK1, that in combination with increased RIPK3 expression would direct TNF signalling towards necroptosis [
14,
41,
44]. Our data adds to this finding by suggesting that this signalling/cell fate may be directed by soluble mediators originating from the inflamed meninges and may also be involved in neuronal damage in MS, although this will require confirmation via single-cell RNA analysis and further protein localisation studies. However, in keeping with this idea, the predominant localisation of TNFR1 in the MS cortical grey matter was in neurons and oligodendrocytes and a recent study of the topography of demyelination and neurodegeneration in MS highlighted the association of oxidative injury to neurons with an increased inflammatory process in the meninges [
25]. The shift away from apoptosis agrees with a number of studies showing that only rarely could apoptotic neurons be observed in the MS cortical layers [
37,
47]. However, definitive evidence for this change in the balance of cell death pathways in the MS cortex must await specific and reliable histological and molecular markers of necroptosis [
44], which are currently not available. It is also not possible to say whether such changes in TNF signalling leading to neuronal cell death are specific to MS or may occur in other chronic CNS inflammatory conditions characterised by meningeal inflammation [
20]. Tuberculous meningitis (TBM) is probably the most studied of the non-MS conditions that can give rise to a more long-term meningeal inflammation. Whilst elevated TNF levels are seen in the CSF of some TBM patients, there is no consistent pattern and the levels are not related to severity or clinical course of disease [
39,
45]. No detailed molecular study of TNF signalling in brain tissue from TBM has been carried out and elevated meningeal inflammation is present for a much shorter time than is the case in MS. However, it is possible that there are effects of chronic TNF cytotoxicity in the TBM brain. Although it has been shown that subpial cortical demyelination is not a feature of TBM [
20,
37], it is not known whether there is a significant loss of cortical neurons similar to that seen in MS. To date, it has not been possible to procure suitable frozen tissue samples from non-MS chronic inflammatory CNS conditions that have a similar disease duration.
When the level of inflammation was lower in the MS meninges, TNF/TNFR1 and TNF/TNFR2 signalling appeared to be directed more towards IKK/NFkB-mediated cell survival. The upregulation of CASP8, TRADD, FADD and RIPK1 in the presence of increased c-FLIP expression, as seen in the F− SPMS cortex, would be expected to inhibit both necroptotic and apoptotic signalling and lead to IKK complex formation and NFκB activation and translocation [
14]. Our finding of TNFR2 expression predominantly in cortical astrocytes and microglia, in particular in the most external cortical layers, and the concomitant evidence of significant upregulation of TNFR2 gene expression only in F− SPMS but not in F+ SPMS cases, suggests that this mechanism may protect both oligodendrocytes and neurons via indirect pathways. Selective stimulation of human-TNFR2 on astrocytes in culture has been shown to lead to leukaemia inhibitory factor secretion, which promotes oligodendrocyte survival and differentiation [
21] and stimulation of TNFR2 on mouse microglia leads to the upregulation of anti-inflammatory cytokines [
61]. Thus, the lack of activation of TNFR2-mediated pathways in the MS cortex in the presence of increased meningeal inflammation would have a deleterious effect not only by exacerbating cell loss/damage but also by inhibiting tissue repair. Again, it needs to be noted that the degree in shift in the balance of these competing TNF signalling pathways is variable from case to case. Single-cell RNA sequencing approaches will be needed to verify the cellular compartments in which these changes in TNF signalling are taking place.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.