Introduction
Multiple sclerosis (MS) is the most common chronic neurodegenerative and neuroinflammatory disease in young adults [
8]. At disease onset, the majority of MS patients present with a relapsing–remitting disease course (RRMS), and despite availability of many disease-modifying therapies most RRMS patients will eventually develop secondary progressive MS (SPMS) [
20,
28,
36]. The transition into SPMS is characterized by gradual worsening of neurological disability without periods of remission and currently no efficient therapeutic options exist for the majority of patients [
16]. The main cause of disease progression in SPMS is neurodegeneration, which encompasses tissue damage in the white matter (WM) [
29] and accumulating pathology in the grey matter (GM), including demyelination and loss of neurons and synapses in the cerebral cortex [
25,
31,
41]. Multiple studies have shown that the degree of cortical pathology provides a better correlate for progression of clinical disability than the number of WM lesions [
5,
6,
15]. The lack of therapies for SPMS can, therefore, be largely attributed to our incomplete understanding of the mechanisms behind cortical neurodegeneration. This, in turn, is severely hindered by absence of suitable animal models. However, a number of studies have now indicated that chronic, compartmentalized inflammation of the nearby leptomeninges is likely to drive many aspects of cortical pathology [
23,
31].
Meningeal inflammation in SPMS is characterized by accumulation of immune cells, including B, T and myeloid cells, either diffusely present or in aggregates resembling tertiary lymphoid follicles [
3,
48]. The degree of inflammation and the presence of follicles both associate with the severity of cortical pathology, possibly by production and subsequent diffusion of pro-inflammatory cytokines into the cortex [
32,
46]. Indeed, in a recently developed animal model for chronic meningeal inflammation, chronic overexpression of two of these cytokines, TNFα and IFNγ, in the leptomeninges of rats was recently shown to induce robust meningeal inflammation, cortical demyelination and neuronal loss as seen in SPMS [
24].
Microglia are the brain-resident immune cells, and as such, they continuously scan their environment for structural damage or invading pathogens using their highly motile processes [
38,
50]. In addition, microglia are crucial for development and maintenance of neuronal networks by facilitating synaptic plasticity [
55] and directly interacting with neuronal cell bodies to monitor and protect neuronal function [
9]. Accordingly, upon brain insults, microglia actively adapt their shape and function to restore brain homeostasis [
7,
10,
49,
50]. Despite this, multiple lines of evidence point towards an active contribution of microglia to neurodegeneration in many chronic neuroinflammatory and neurodegenerative diseases [
4,
11]. In both the previously mentioned in vivo model for chronic MS-like meningeal inflammation and in SPMS donors, meningeal inflammation strongly associates with activation of cortical microglia [
24,
30]. However, it remains to be investigated whether cortical microglia augment the pro-inflammatory signal coming from the meninges and thereby contribute to neurodegeneration in SPMS cortex.
In the SPMS cortex, activated microglia have been found in close proximity to apical dendrites, neurites and neuronal soma, but whether these microglia contribute or try to salvage ongoing neurodegeneration is currently unknown [
41]. To study the link between meningeal inflammation, microglial behavior and neurodegeneration in SPMS, we used post-mortem brain tissue from SPMS donors and a recently developed in vivo model for chronic MS-like meningeal inflammation [
24]. We identified two subgroups of MS cortical areas with distinct patterns of microglial morphology, density, and protein expression. Interestingly, these two subgroups of MS cortex also showed separate levels of local meningeal inflammation and differed in the extent of neuronal damage. Highly similar microglial phenotypes and associations with neuronal damage were observed in vivo, where these phenotypes existed at different time points after the onset of progressive MS-like meningeal inflammation. Taken together, our data suggest that in progressive MS, chronic meningeal inflammation induces phenotypic changes in cortical microglia that differentially associate with cortical neurodegeneration in a time-dependent manner.
Discussion
Currently, therapeutic options for progressive MS patients are limited [
16]. This is partly attributed to our lack of understanding of the pathological mechanisms driving the disease, which in turn is a consequence of a dearth of suitable animal models for progressive MS [
33]. In this study, using both post-mortem tissue from progressive MS cases and a recently developed animal model for progressive MS-related cortical pathology [
24], we aimed to uncover mechanisms that drive neurodegeneration in progressive MS cortex and further establish the use of this new animal model to study underlying pathological processes.
Using extensive morphological analyses and quantification of several well-known markers for microglial activation, we could separate our cohort of progressive MS tissue into three distinct clusters. Cortices in two of these clusters (termed MS1 and MS2) contained a microglia population that significantly differed from controls. Microglia in MS1 cortical areas were characterized by a high cellular density and elevated expression of both HLA class II and CD68; whereas MS2 microglia were defined by a hyper-ramified morphology and low P2Y12 expression. Both MS-specific microglial phenotypes were linked to increased meningeal inflammation, but only MS2 microglia associated with increased presence of B cells in the adjacent meninges. Remarkably, we detected cortical microglia that were very similar to both MS1 and MS2 in an animal model for chronic MS-like meningeal inflammation (CMI). Cortical microglia found 1 month after the induction of meningeal inflammation resembled microglia in MS1 areas, while microglia at 2 months after induction shared many features with MS2 microglia. We further show that MS1 microglia spatially associate with neuronal cell bodies, whereas significant neuronal loss was restricted to the MS2 cortex. The spatial association between microglia and neurons was related to the removal of pre-synapses from the neuronal soma, and accompanied by increased pre-synapse phagocytosis by MS1 microglia. Again, neuronal changes in CMI animals reflected what we observed in progressive MS tissue, with cortical microglia 1 month after induction of meningeal inflammation closely apposing neuronal somata, and neuronal loss in the upper cortical layers being most prominent after 2 months. Furthermore, we observed a removal of pre-synapses from the soma of neurons that were in close contact with microglia in CMI animals and found evidence for pre-synapse phagocytosis by microglia at both time points.
Given the striking similarities between the two MS-specific clusters and the cortex of CMI animals after 1 and 2 months, it is tempting to speculate that the changes observed in MS1 cortex are caused by early stage meningeal inflammation and those observed in MS2 cortex by more chronic meningeal inflammation. As B cells are most prominent in the meninges of both MS2 cortex and 2-month CMI animals, this would suggest that at first T cells and myeloid cells populate the MS meninges, with the number of meningeal B cells rising over time. This also fits with the finding that the number of meningeal B cells and levels of B-cell-related cytokines in the cerebrospinal fluid associate with disease progression [
30,
32,
45]. Furthermore, the extent of meningeal inflammation in MS donors, and more specifically the number of meningeal B cells, has been strongly linked to the presence of tertiary lymphoid-like follicles [
3,
30]. Unfortunately, we could identify too few of these lymphocyte clusters in the meninges of our tissue cohort to test whether they indeed are found predominantly in the MS2 cluster, as one would expect. Likewise, it would be interesting to explore whether these follicles also occur in CMI animals, which could open up studies examining their development and, eventually, used to find therapeutics that could block their formation.
Similar to findings under different neuroinflammatory conditions [
7,
13], we show that the juxtaposition of microglia to neuronal somata leads to displacement of pre-synapses from the cell soma. At the same time, we observed increased pre-synapse phagocytosis by microglia. Interestingly, both displacement and phagocytosis of pre-synapses was predominantly seen in MS1 cortex. In contrast, significant neuronal loss was restricted to cortical layers 2 and 3 of MS2 cases, which is in line with a recent study showing a specific loss of CUX2
+ neurons in the same layers, which spatially associated with inflamed meninges [
47]. Given the differences in neuronal pathology between MS1 and MS2 areas, we speculate that pre-synaptic displacement and stripping by MS1 microglia could represent a protective response to neuroinflammation and prevent substantial neuronal loss, as has previously been observed during LPS-induced neuroinflammation [
7]. In contrast, MS2 microglia might have lost their ability to protect neurons from degenerating and perhaps even actively contribute to neuronal damage. Alternatively, it might be that pre-synaptic removal and the more pro-inflammatory profile (decreased P2Y12 and increased HLA class II and CD68 expression) of microglia in MS1 cortex is damaging to neurons but to such an extent that this not yet leads to an observable drop in neuronal density. The latter explanation seems to be supported by our recent finding of necroptotic neuronal death in non-immunized 1-month CMI rats. [
43]. However, the number of neurons undergoing necroptosis at 1 month after injection is very small (< 5 per mm
2), which might be due to the presence of neuroprotective MS1-like microglia. Hence, future research is needed to investigate whether these dichotomous microglial phenotypes are neuroprotective/neurodegenerative and whether they are mainly driven by differences in the neuroinflammatory environment, e.g. more meningeal B cells in MS2 cases, or the result of microglial exhaustion over time.
Surprisingly, we did not detect an overall decrease in pre-synaptic density in SPMS cortex, despite evidence for increased pre-synapse phagocytosis by microglia. Although loss of synapses has been extensively reported in hippocampus [
14,
35], thalamus [
54] and spinal cord of MS donors [
42], evidence for a loss of synapses in the cortex is more ambiguous. Reduced spine density on apical dendrites, indicating post-synaptic loss, has been shown in both myelinated and demyelinated cortex of MS donors [
25]. Whereas other studies, using similar methods to ours, only found a significant loss of Synaptophysin in leukocortical demyelinated lesions but not in normal appearing MS cortex [
53] and no correlation with cortical atrophy [
44]. Furthermore, the lack of decreased pre-synaptic density in our cohort might be explained by the low rate of pre-synapse removal (< 0.5% of pre-synapses within microglial lysosomes) in progressive MS cortex, which might allow for compensatory synaptogenesis. Alternatively, concomitant cortical thinning in these patients might obscure a reduction in synapse density despite an overall loss of synapses. Similarly, we also did not observe a significant difference in parameters of disease progression between the MS clusters, which one might expect given the differences in meningeal inflammation and neuronal loss and their strong correlation with disease severity [
23,
31]. However, this could be a limitation of our experimental setup, in which we only analyzed one cortical region in each patient. And given the strong spatial association between meningeal inflammation, microglial activation and neuronal damage [
31,
47], pathology in this cortical region might not be predictive for the extent of tissue damage in the remaining cortex.
Microglia have been extensively studied in the context of age-related neurodegenerative diseases and white matter lesions in MS (for review see [
4,
11]), however, their role in MS-related cortical pathology remains relatively obscure. Studies in amyotrophic lateral sclerosis (ALS), Alzheimer´s disease (AD) and white matter MS lesions and their respective animal models have identified what appears to be a shared microglial response to (neuronal) damage, which involves retraction of their processes and significant changes in protein and/or mRNA expression. This phenotype has been termed either MGnD (microglia neurodegenerative phenotype) [
27] or DAM (disease-associated microglia) [
26]. Whether these microglial phenotypes are beneficial or detrimental remains a matter of debate and is likely time and disease-dependent. Here, we show that cortical microglia in MS respond rather differently than in ALS and AD and acquire an almost hyper-ramified morphology throughout the cortex, especially in the MS2 cortex. Similarly, we found that microglial activation markers, HLA and CD68 [
34,
52], are not altered in MS2 microglia. Lastly, the expression of the homeostatic marker TMEM119 did not significantly differ from controls in all clusters unlike in MGnD/DAM microglia, which lose TMEM119 expression [
26,
27]. Interestingly, hyper-ramified microglia have previously been observed in animal models of ischemic stroke [
37] and chronic stress [
21,
22,
51], which might be mediated by specific neuron-microglia interactions. For instance, increasing neuronal activity with neurotransmitter agonists can induce hyper-ramification of microglia via ATP signaling [
12,
18,
39], while depleting CX3CR1 in microglia, which binds to neuronal fractalkine, yielded mice resistant to stress-induced microglial hyper-ramification [
21]. As we could replicate the microglial alterations seen in both MS1 and MS2 cortex by experimentally inducing chronic meningeal inflammation in vivo, we propose that these alterations are caused by MS-related meningeal inflammation directly, through cytokine signaling for instance, and indirectly, via altered neuronal function. Since CMI is induced by chronic overexpression of both TNFα and IFNγ in meninges of the sagittal sulcus, and TNFα and IFNγ are both extensively produced in inflamed MS meninges [
19], we assume that these cytokines are involved in driving the phenotypic changes in microglia. Future research should elucidate whether the difference between MS1 and MS2 microglia is caused by (1) a difference in duration of exposure to TNFα and IFNγ, (2) involvement of other (B-cell derived) inflammatory factors, and/or (3) altered neuron-microglia signaling.
In this study, the results obtained from what we coined the CMI animal model, corroborate the initial experiments using this animal model [
24], including persistent meningeal inflammation, cortical demyelination and neuronal loss. Although our data indicate that MOG immunization prior to lentiviral injection does not significantly alter the extent of meningeal inflammation, we did find that the difference between 1- and 2-month microglia was exacerbated in MOG-immunized rats, which is why we decided to focus on these animals. It would be interesting to explore if this difference is caused by increased cortical demyelination in MOG-immunized animals [
24], or by the presence of MOG-primed lymphocytes in the meninges.
Taken together, we convincingly show that the experimental model of chronic meningeal inflammation closely mimics meningeal inflammation-induced cortical pathology in progressive MS patients. Although this model strongly emphasizes the role of TNFα and IFNγ in driving cortical pathology, and thereby likely over-simplifies the complex pathological processes in MS meninges and cortex, we are convinced that this model will be a valuable addition to our toolbox for studying progressive MS.
In conclusion, we have uncovered two distinct MS-specific microglial phenotypes in the cortex of progressive MS donors that are driven by local meningeal inflammation and differentially associate with neuronal damage. Results obtained in a novel experimental model for chronic MS-like meningeal inflammation suggest that these phenotypes may occur sequentially and that microglia lose their protective properties over time, leading to neuronal loss. Hence, timely targeting of the processes contributing to microglial activation in the progressive MS cortex provides an interesting therapeutic strategy to combat progressive MS.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.