Cortical microglia acquire a different morphological appearance in progressive MS
To assess the involvement of microglia in cortical pathology in progressive MS patients, we first quantified microglial number and morphological complexity in cortical layers 1, 3 and 5/6 using confocal microscopy on IBA1+ stained sections of MS brains and non-neurological controls (Fig. 1a, b). While the density of IBA1+ microglia in MS cortex was variable and not significantly different from controls (Fig. 1c), morphological analysis of a large number of individually traced microglia revealed a more ramified microglia morphology in MS cortex, as indicated by the area under the curve (AUC) of the Sholl analysis  (Fig. 1d, e). This morphological change was most evident in neuronal layer 3 (Fig. 1f) and was further corroborated on a subset of MS and control cases stained for TMEM119, a microglia-specific protein (Online Resource 1a–c). Both microglial density and morphology were positively associated throughout the different cortical layers (Online Resource 1d). Further analyzes of microglial morphology revealed an increase in the peak of the AUC, which displays the maximum number of intersections in the Sholl analysis, a larger reach of the branches (wingspan), and increased number and length of branches and number of junctions in MS microglia located in neuronal layer 3 (Fig. 1f and Online Resource 1e). Lastly, we observed a trend towards a larger microglial soma in all analyzed cortical layers of progressive MS brains (Fig. 1f). Absence of IBA1+ cells with an amoeboid morphology and co-expression of P2Y12 with virtually all non-vessel associated IBA1+ cells indicates limited infiltration of peripheral myeloid cells in MS cortex (Online Resource 1f).
Heterogeneous microglial marker expression in the progressive MS cortex
To further characterize cortical microglia, we quantified the expression of homeostatic microglia markers P2Y12 (Fig. 2a) and TMEM119 (Fig. 2b), and two commonly used activation markers, HLA class II (Fig. 2c) and CD68 (Fig. 2d), in IBA1+ microglia in cortical layer 3. We found that P2Y12 expression was significantly lower in MS cortical microglia compared to controls (Fig. 2b), while we observed a trend towards increased microglial HLA class II expression (Fig. 2f). TMEM119 and CD68 were not differently expressed (Fig. 2h) although expression of all four markers was highly variable within the MS group.
Cortical microglia morphology, density, and protein expression reveal two distinct MS subgroups
To explore if the variation in both morphology and protein expression of cortical microglia within the analyzed MS cortical areas could be explained by the existence of subgroups with different microglial phenotypes, we used principle component analysis (PCA) and included parameters of microglia morphology, protein expression and cell density. PCA visualization showed the presence of three different subgroups separating MS and control cases based on their microglial phenotype (Fig. 2i). Then, using K-means clustering, MS cases were assigned to three distinct clusters which we named MS0, MS1, and MS2 (Fig. 2i). The three donors in the MS0 subgroup had a microglia phenotype that showed close resemblance to those in control (Fig. 2j). MS1 was mainly defined by low P2Y12, high HLA class II, high CD68 and a markedly increased microglia density, whereas MS2 was characterized by low P2Y12 expression and increased morphological complexity as quantified by Sholl analysis (AUC), higher number of junctions and branches and increased total branch length (Fig. 2j and Online Resource 2a). We did not find significant differences in age of death, disease duration, progressive MS duration and time in wheelchair between the MS subgroups (Online Resource 2b). For the remainder of the study, we focused on MS-specific subgroups MS1 and MS2, as the MS0 group was limited (n = 3) and contained microglia that were similar to those in controls.
Increased meningeal T and B cells in progressive MS brains
Next, we explored whether meningeal inflammation is associated with the different progressive MS subgroups as several studies have already described meningeal inflammation as an important driver of cortical pathology in progressive MS [23, 31, 38]. We found that CD19+ B cells and CD3+ T cells were increased in the meninges of progressive MS donors, while the levels of meningeal IBA1+ myeloid cells remained unchanged (Fig. 3a, b and Online Resource 3a). Interestingly, the increase in meningeal CD19+ B cells was most notable in the MS2 group (Fig. 3c), and correlated with increased morphological complexity of cortical microglia. To further dissect the involvement of meningeal T cells, we analyzed the presence of CD4+ and CD8+ T cells (Fig. 3d). As expected, both CD4+ and CD8+ T-cell numbers were increased in a large subset of MS cases (Fig. 3e, and Online Resource 3b), and similar CD4/CD8 T-cell ratios shows that they were equally induced in the meninges overlying MS1 and MS2 cortex. (Fig. 3f).
Experimental chronic meningeal inflammation time-dependently induces MS1- and MS2-like microglia
To investigate whether meningeal inflammation could drive the microglial changes we observed in progressive MS cortex, we made use of a novel animal model of chronic experimental meningeal inflammation (CMI; Online Resource 4a) which has recently been shown to replicate important features of cortical pathology in progressive MS patients . In this model, a subclinical autoimmune response to myelin was induced and lentiviral vectors carrying the TNFα and IFNγ genes were injected in the sagittal sulcus below the meningeal dura mater layer. As expected, we observed an increase in meningeal CD3+ T cells, IBA1+ myeloid cells and CD79a+ B cells in the sagittal sulcus of CMI rats 1 and 2 months after lentiviral injection. Especially CD79a+ B cells were most strongly increased in numbers after 2 months (Fig. 4a–c and Online Resource 4b). Similar to the previous study, we did not observe significant differences in meningeal immune cell infiltration between CMI rats, which underwent subclinical MOG immunization prior to lentiviral injection; and IFA control rats, which were only injected with IFA prior to lentiviral injections (Online Resource 4c).
Next, we compared microglial density, morphology and protein expression in the cortex surrounding the sagittal sulcus in CMI rats with appropriate controls. Here, we found that after 1 month, CMI animals displayed an increased microglia density (Fig. 4d) and a less ramified morphology (Fig. 4e–i). In contrast, after 2 months, microglial density was almost back to control levels (Fig. 4b–d) whereas morphological complexity of microglia was similar to or greater than in controls (Fig. 4e–f). We observed little to no changes in microglial soma size in both 1 and 2 month CMI animals (Fig S4d). Importantly, virtually all non-vessel-associated IBA1+ cells in the cortex of CMI rats were positive for P2Y12, indicating very little or absent infiltration of peripheral myeloid cells (Online Resource 4e). Quantification of P2Y12 and HLA class II expression in cortical layer 3 microglia of CMI rats (Fig. 4j–m) revealed a clear reduction in P2Y12 expression after both 1 and 2 months (Fig. 4j–i), whereas expression of HLA class II was highly upregulated in CMI 1 month microglia but almost back to levels seen in control animals after 2 months (Fig. 4m). We found similar changes in cortical microglial phenotype in IFA control animals after 1 and 2 months, though most changes were less pronounced than in the MOG-immunized CMI animals (Online Resource 4f).
Taken together, cortical microglia in the CMI model at 1 month differ substantially from those at 2 months after lentiviral injection. Interestingly, microglia seen after 1 month carry many features of those seen in MS donors previously allocated to the MS1 subgroup, including a high microglial density, a less ramified morphology, low P2Y12 and high HLA class II expression. In addition, characteristics of CMI rats 2 months after injection, i.e. higher number of meningeal B cells, lower microglial density and HLA class II expression, but an increased morphological complexity, closely resembles the MS2 subgroup.
MS1 and MS2 subgroups differentially associate with cortical neurodegeneration
We next questioned whether the two MS clusters differentially associated with local tissue damage and for this purpose, we first quantified demyelination in MOG-stained sections. As expected, we observed that the majority of progressive MS tissues sections had one or more cortical lesions, mostly of the subpial type, which were completely absent from controls (Fig. 5a). However, we did not detect any differences between MS1 and MS2 cortical areas in either lesion load (as % of total cortex; Fig. 5b) or MOG+ area in the different cortical layers (Fig. 5c). We subsequently compared neuronal density per cortical layer using HuC/D immunolabeling (Fig. 5d). As it was recently reported that neurons in layer 2 and 3 of the cortex are especially prone to degenerate in progressive MS cortex , we also decided to include cortical layers 2 and 4 in this analysis. In line with a previous study , we observe a lower neuronal density in the upper cortical layers of the majority of progressive MS cases, but this only reached statistical significance in cortical layers 2 and 3 of MS2 cortex (Fig. 5e).
Several studies have described close contact between microglia and neurons in neuroinflammatory conditions [13, 49], leading to removal of pre-synaptic input from the neuronal soma. Here, using microglial IBA1 and neuronal HuC/D co-labeling, we show that in the MS1 cortex a higher percentage of neuronal somata was directly associated with one or more microglial cell bodies (Fig. 5f–g). Similarly, we also observed a higher percentage of microglia that were contacting neurons in both MS1 and MS2 (Online Resource 5a). Next, we found an almost complete loss of Synaptophysin+ (Syn) pre-synaptic input on the part of the neuronal soma that was occupied by a microglial cell body (Fig. 5h). In addition, we show an overall reduction of pre-synaptic input on neuronal soma in direct contact with one or more microglia in MS cortex (Fig. 5i).
Next, we assessed whether cortical microglia in MS1 and MS2 areas were engaged in phagocytosis of pre-synapses by quantifying the presence of Syn+ structures in LAMP1+ lysosomes in microglia (Fig. 5j). Similar to what happens in the thalamus of MS donors , we found a significant increase in pre-synaptic phagocytosis specifically in MS1 cortical microglia (Fig. 5j–k), which is in line with an overall increased phagocytic capacity of MS1 microglia as indicated by increased CD68 expression (Fig. 2) and corroborated here by increased LAMP1 expression (Online Resource 5b). Remarkably, we did not observe a decrease in the total pre-synaptic density in layer 3 of either MS1 or MS2 cortex (Fig. 5l), which might be explained by the fact that the percentage of all pre-synapses that were located in microglial lysosomes was only around 0.3% (Online Resource 5c). A small fraction of pre-synapses were located in lysosomes of non-microglial cells, but there was no difference between control, MS1 and MS2 cortex (Online Resource 5d).
CMI induces time-dependent cortical neurodegeneration
To assess associations with cortical tissue damage in CMI animals, we first measured demyelination in the cortical layers extending from the sagittal sulcus in MOG-stained sections (Fig. 6a). As shown previously, CMI induces cortical demyelination most strongly after 2 months and in the layers closest to the sulcus . Due to the large variability of MOG+ area in all groups, we did not detect any significant difference in layers 1 or 2 (Fig. 6b). Moreover, neuronal density in layer 1 was decreased at both time points, whereas we only found neuronal loss in layer 2 after 2 months (Fig. 6c, d).
In line with what we observed in MS1 cortical areas, the number of neuronal somata directly associated with microglia cell bodies was significantly increased in layer 3 of CMI 1 month rats, whereas after 2 months this was almost back to control levels (Fig. 6e–f). At both time points, these contacts resulted in removal of pre-synaptic input from the neuronal soma (Fig. 6g–h). Phagocytosis of vGAT+ pre-synapses by microglia was increased after 1 and 2 months, albeit with a large variation between animals (Fig. 6i–j). Again, we did not observe a total loss of vGAT+ pre-synapses in cortical layer 3 of CMI rats (Fig. 6k), possibly due to the small percentage of pre-synapses found in microglial lysosomes (Online Resource 6a). Furthermore, pre-synapse phagocytosis by non-microglia cells was not significantly different between groups (Online Resource 6b).