Острая ТРЕМ2 ингибиция истощает MAFB-высокую микроглию и препятствует воссоединению.

Myelin is the multilayered insulating sheath that surrounds the axons of many neurons to enhance the speed and efficiency of electrical signal conduction along the axons. Myelin loss, or demyelination, is a pathological process underlying a wide range of neurological conditions affecting both the central and peripheral nervous systems, including multiple sclerosis (MS), stroke, traumatic brain injury, and spinal cord injury ( 1 , 2 ). Furthermore, white matter loss is a prominent feature of neurodegenerative diseases such as Alzheimer’s disease (AD) and is a hallmark of normal aging, correlating with cognitive decline ( 3 ). Remyelination, the process of regenerating lost myelin, plays a critical role in functional recovery. This process involves the recruitment and proliferation of oligodendrocyte progenitor cells (OPCs), followed by their differentiation into myelinating oligodendrocytes and the regeneration of myelin sheaths ( 4 – 6 ).

Microglia are key players in remyelination, acting at multiple stages ( 7 – 9 ). These cells are highly efficient at removing myelin debris, which inhibit OPCs maturation and remyelination ( 10 ). This debris clearance process is slower with age, contributing to the age-dependent decline in remyelination capacity. Microglia also contribute to OPCs proliferation and recruitment, as well as their subsequent pruning and differentiation ( 11 ). They secrete factors, such as CCL2, which promote OPCs migration to demyelinated sites ( 12 ). Furthermore, microglia express factors like Transglutaminase 2, which is essential for OPCs differentiation ( 13 ). Other microglia-derived factors, such as Insulin-like Growth Factor 1 (IGF-1), also promote OPCs differentiation ( 14 ). Finally, microglia may support the ensheathment of axons by newly formed oligodendrocytes, stabilizing both the axon and the myelinating cell.

To explore how microglia influence demyelination and remyelination processes in the central nervous system (CNS), researchers have widely utilized the Cuprizone (CPZ) model ( 15 , 16 ). CPZ is a copper-chelating agent that causes mitochondrial dysfunction of oligodendrocytes, causing their death and subsequent demyelination in specific regions of the CNS, such as the corpus callosum. Mice are typically fed a diet supplemented with CPZ for 4 to 6 wk. As demyelination progresses, microglia act as the first responders, shifting from a resting state to an activated state called “disease-associated microglia” (DAM) to clear damaged myelin ( 17 , 18 ). Once CPZ is removed from the diet, the repair process initiates and is usually completed within 2 to 4 wk. During this time, OPCs differentiate into mature oligodendrocytes, producing myelin, leading to the repair of the damaged myelin sheaths. Simultaneously, microglia express a transcription factor called MAFB ( 17 ). MAFB is known to play a key role in the final stages of microglial development from yolk sac progenitors, supporting the acquisition and maintenance of microglial identity ( 19 ). As such, it may be crucial for restoring microglial homeostasis.

Triggering Receptor Expressed on Myeloid cells 2 (TREM2) is a cell surface receptor expressed on microglia ( 20 ). It plays a critical role in maintaining their metabolic fitness and sustaining essential functions, such as phagocytosis of apoptotic neurons, amyloid plaques, and myelin debris. TREM2 also supports microglial proliferation and chemokine production. Homozygous variants in the TREM2 gene that prevent its expression or function lead to severe white matter dystrophy, resulting in an early-onset dementia known as Nasu–Hakola disease ( 21 ). This highlights a potential role for TREM2 in supporting myelination. In mice, activation of TREM2 through intraperitoneal administration of TREM2 agonistic monoclonal antibodies (mAb) increased clearance of myelin debris from demyelinating lesions caused by CPZ ( 22 ) or by intracranial injection of lysolecithin ( 23 ). This activation promoted new myelin formation and preserved axonal integrity. Conversely, mice with a constitutive deletion of the Trem2 gene treated with CPZ exhibited defective clearance of myelin debris and impaired remyelination ( 17 , 24 – 26 ). However, the impact of acute TREM2 blockade on microglial responses during demyelination and remyelination remains unexplored.

To address this, we employed an anti-TREM2 blocking mAb in both the demyelination and remyelination phases of the CPZ model and performed global transcriptomic analyses at the single-cell level. Our findings reveal that anti-TREM2 depletes microglia expressing high levels of the transcription factor MAFB (MAFB hi ) and impairs remyelination. Supporting the importance of MAFB hi microglia in myelin repair, similar cells were identified in humans in both active and chronically active MS lesions. These results suggest a critical role of MAFB hi microglia in remyelination and highlight the necessity of TREM2 for maintaining this microglial population.

Results

TREM2 Blockade by the Anti-TREM2 mAb 178 Impairs Remyelination.

To characterize the role of TREM2 blockade in the CNS during de- and remyelination, 8-wk-old male wild-type (WT) mice were divided into 5 experimental groups ( Fig. 1 A ). A) Untreated: fed regular chow for 5 wk. B) CPZ + control (Ctrl) IgG2a: fed 0.2% CPZ diet for 5 wk to induce demyelination and injected intraperitoneally (i.p.) with Ctrl IgG2a during demyelination. C) CPZ + anti-TREM2: fed 0.2% CPZ diet and injected i.p. with anti-TREM2 during demyelination. D) Remyelination + Ctrl IgG2a: fed 0.2% CPZ diet initially, then returned to regular chow for 2 wk to allow remyelination, and injected i.p. with Ctrl IgG2a during remyelination. E) Remyelination + anti-TREM2: fed 0.2% CPZ diet initially, then returned to regular chow for 2 wk, and injected i.p. with anti-TREM2 during remyelination. This design allowed for the assessment of demyelination and remyelination with and without TREM2 blockade, alongside an untreated group for comparison. All blocking experiments were conducted with the anti-TREM2 mAb 178, which has been previously shown to block TREM2 function in vitro and in vivo ( 27 – 31 ). This anti-TREM2 antibody reached concentrations of 1 to 4 ng/mg in the brain tissue both after five weekly administrations during the demyelination and after two weekly administrations during the remyelination ( Fig. 1 B and C ).

Blockade of TREM2 affects remyelination and the transcriptomes of brain cells. ( A ) Schematic diagram of the experimental strategy. In the untreated group, 2-mo-old male B6 mice were fed with regular chow for 5 wk. In the demyelination groups, mice were fed with CPZ and were injected either with Ctrl IgG2a (60 mg/kg) or anti-TREM2 (αTREM2—60 mg/kg) once a week for 5 consecutive weeks. In the remyelination groups, mice were fed with CPZ for 5 wk and then were injected with Ctrl IgG2a or αTREM2 for 2 consecutive weeks after CPZ withdrawal. After the mice were killed, their brains were collected for TEM or snRNA-seq. ( B and C ) ELISA quantification of αTREM2 concentrations in the plasma ( B ) and brain lysates ( C ) 1 wk after the final antibody injection in either the demyelination or remyelination group. n = 7 for plasma, n = 6 for brain tissue. The solid circles represent the demyelination samples ( n = 3 or 4). The empty circles represent the remyelination samples ( n = 3 in each group). Original magnification, 3,000×. (Scale bar, 2 μm.) The scale bar in the zoomed-in images is 0.5 μm. ( D ) Representative TEM images of the corpus callosum in each condition. The arrowheads in blue indicate myelinated axons, the arrowheads in orange indicate nonmyelinated axons, and the arrowheads in red indicate the myelin abnormalities, such as myelin outfolding and unraveling ( 32 ). ( E ) Quantification of the percentage of myelinated axons vs. total axons. Untreated group, n = 3 brains; demyelination + Ctrl IgG2a group, n = 5 brains; demyelination + αTREM2, n = 6 brains; remyelination + Ctrl IgG2a group, n = 7 brains; and remyelination + TREM2 group, n = 7 brains. Five images were analyzed per mouse. Each dot represents the ratio of myelinated axons per image. Data are presented as mean ± SEM. P values by one-way ANOVA with Tukey’s multiple comparisons test. ( F ) UMAP plot of 95,255 nuclei showing 7 distinguished cell-type identities determined by expression of specific markers. n = 2 mice in untreated; n = 3 in demyelination + αTREM2; n = 3 in demyelination + Ctrl IgG2a; n = 3 in remyelination + αTREM2; and n = 3 in remyelination + Ctrl IgG2a. ( G ) A dotplot shows the feature genes for each cell type. ( H and I ) Number of DEGs in each cell type by comparisons between αTREM2 vs. Ctrl IgG2 in demyelination ( H ) or remyelination ( I ). Log2(fold change) > 0.5, adjusted P < 0.05, nonparametric Wilcoxon rank-sum test, Bonferroni correction.

To investigate the impact of blocking TREM2 on myelin damage and repair, we used transmission electron microscopy (TEM) to assess the proportion of myelinated axons relative to total axons. We also measured myelin sheath thickness of myelinated axons using the G-ratio, calculated as the ratio of the inner axonal diameter to the outer diameter of the axon, including its myelin sheath. We further quantified the number of myelin debris within a defined area of the corpus callosum. After 5 wk of CPZ treatment, all treated groups, including those receiving anti-TREM2 or Ctrl IgG2a, showed a significant reduction in myelination compared to the untreated group, as demonstrated by decreased proportions of myelinated axon, increased G-ratios, and accumulation of myelin debris ( Fig. 1 D and E and SI Appendix , Fig. S1 A–C ). After a 2-wk recovery period without CPZ, both the anti-TREM2 and Ctrl IgG2a groups exhibited significant increases in myelinated axons and reductions in myelin debris, although the myelin sheaths remained thinner. Blocking TREM2 specifically reduced the ratio of myelinated axons, while it did not significantly affect the reduction in myelin debris or the thickness of the myelin sheaths during remyelination ( Fig. 1 D and E and SI Appendix , Fig. S1 A – C ). These findings suggest that TREM2 is required in myelin repair, particularly in promoting recovery of myelinated axons.

Anti-TREM2 Influences the Transcriptomes of Brain Cells Primarily during Remyelination.

We sought to determine the transcriptional changes in brain cells underlying the effects of anti-TREM2. Tissues from the cerebral cortex and corpus callosum were collected at the end of each treatment for single nucleus RNA-sequencing (snRNA-seq) analysis. In total, 95,255 nuclei from all five groups were analyzed after quality control ( SI Appendix , Fig. S2 A–C ). Unsupervised clustering yielded 16 populations ( Dataset S1 ), annotated as excitatory or inhibitory neurons, oligodendrocytes, astrocytes, microglia, OPCs, vascular cells, and fibroblasts, based on cluster marker genes and known cell type markers ( 17 ) ( Fig. 1 F and G and Dataset S1 ). The analysis of differentially expressed genes (DEGs) between anti-TREM2 and Ctrl IgG2a across various brain cell types revealed a significant increase in DEGs during the remyelination phase, while changes during demyelination were less pronounced ( Fig. 1 H and I and Dataset S1 ). These findings indicate that the anti-TREM2 treatment primarily influences remyelination. Moreover, anti-TREM2 primarily affected the frequency of microglial nuclei during remyelination ( SI Appendix , Fig. S2 D ), leading to approximately a 25% reduction compared to the Ctrl IgG2a group. Thus, blocking TREM2 primarily affects microglia, with secondary effects on the transcriptome of other brain cell types.

Anti-TREM2 Ablates the Mafb hi Microglia during Remyelination.

Since TREM2 is expressed in microglia, we initially focused on understanding the impact of anti-TREM2 on microglia. Subclustering of microglia from all five experimental groups identified 8 subsets with distinct cluster markers ( Fig. 2 A and B and Dataset S2 ) ( 17 ). Cluster 2 represented homeostatic microglia (HM) enriched with genes such as Tmem119 and P2ry12 ( Fig. 2 B , SI Appendix , Fig. S3 A and B , and Dataset S2 ). Clusters 3 and 4 showed gene expression profiles characteristic of disease-associated microglia (DAM) ( 33 ), marked by the expression of genes such as Lpl , Csf1 , and Cst7 ( Fig. 2 B ). A progression was observed between the two DAM clusters. Cluster 4 (DAM1) expressed higher levels of homeostatic markers, including P2ry12 and Tmem119 , and showed lower activity in chemokine and cytokine-related pathways ( SI Appendix , Fig. S3 A and C ). Cluster 3 (DAM2) exhibited a more pronounced activation state, with reduced homeostatic gene expression and increased chemokine and cytokine activity ( SI Appendix , Fig. S3 C and Dataset S2 ). Cluster 0, identified as transition microglia (TM), displayed intermediate characteristics between HM and DAM1/DAM2. This cluster was enriched for genes involved in cholesterol metabolism and transfer, consistent with a transitional state that bridges HM and DAM ( Fig. 2 B and SI Appendix , Fig. S3 A ). Cluster 1 ( Mafb hi ) represented a subset of microglia characterized by high expression of the Mafb gene, combining some features of DAM and HM ( Fig. 2 B , SI Appendix , Fig. S3 D , and Dataset S2 ). These cells were in a partially activated state, expressing higher levels of Apoe , Lyz2 , B2m , H2-D1 , and H2-K1 compared to HM ( SI Appendix , Fig. S3 E ). Pearson correlation analysis showed a close relationship of Mafb hi with HM, corroborating the limited activation state of Mafb hi ; in contrast, clusters 3 and 4 reflected more activated states ( SI Appendix , Fig. S3 F ). Additional microglial populations included cluster 5, enriched for interferon (IFN)-stimulated genes such as Irf7 and Oasl2 (IFN-R microglia); cluster 6, enriched for the MHC class II genes H2-Eb1 and H2-Aa (MHCII microglia); and cluster 7, defined by high expression of proliferation genes Top2a and Mki67 (Proliferating microglia—PM) ( Fig. 2 A and B ).

TREM2 blockade ablates the Mafb hi microglia during remyelination. ( A ) UMAP plot of reclustered microglia identifying eight subclusters. n = 7,700 microglial nuclei. HM, homeostatic microglia; Mafb hi , microglia with high levels of Mafb transcript; TM, transition microglia; DAM, disease-associated microglia; IFN-R, interferon responsive microglia; MHCII, MHCII + microglia; PM, proliferating microglia. ( B ) A dotplot shows the feature genes for each subcluster. ( C ) The Monocle prediction of the microglia transition trajectory with Seurat’s cluster information in ( A ) mapped alongside pseudotime tree. The top 500 genes with the highest variability in expression were used to construct the pseudotime tree. ( D ) Microglia on the tree are colored by pseudotime. ( E ) The dotplot shows the transcription factors activities and gene motif enrichment for each subcluster. ( F ) DNA logos show the binding motif of the transcription factors. The images are created by JASPAR. ( G ) UMAP plot of microglia subclusters split by condition. n = 502 nuclei in untreated; n = 2,300 nuclei in demyelination + Ctrl IgG2a; n = 2,310 nuclei in demyelination + αTREM2; n = 1,418 nuclei in remyelination + Ctrl IgG2a; and n = 1,170 nuclei in remyelination + αTREM2. ( H ) Volcano plot showing the DEGs in αTREM2 vs. Ctrl IgG2a in remyelination. Log2(fold change) > 0.5, adjusted P < 0.05, nonparametric Wilcoxon rank-sum test, Bonferroni correction. ( I ) Pathways and their genes down-regulated in αTREM2 group in remyelination. P value by the hypergeometric test.

Cell trajectory analysis using Monocle2 revealed a gradual shift in microglial states ( 34 ), starting with less activated states represented by the HM or Mafb hi cluster, which were predominant at the initial time point. This trajectory culminated in a fully activated state corresponding to DAM subpopulations. Along the activation pathway, other activated states, such as MHCII and IFN-R, branched off, highlighting the diverse functional responses of microglia during activation ( Fig. 2 C and D ). These findings were corroborated by feature gene expression patterns ( SI Appendix , Fig. S3 G ). To identify regulons, i.e., transcription factors driving specific microglial signatures, we applied the Single-Cell rEgulatory Network Inference and Clustering (SCENIC) algorithm to our microglial snRNA-seq data ( Dataset S3 ) ( 35 ). Among the 133 significant regulons with high activity scores, MITF, JUN, and BHLHE40 were prominent in positively regulating DAM1, DAM2, and PM populations ( Fig. 2 E and F ). MAFB exhibited significant activity in regulating the Mafb hi subset. SPI1 showed heightened activity in controlling microglia in DAM2 and Mafb hi subsets. Furthermore, IRF7 and RUNX3 displayed strong activity in regulating IFN-R and PM populations, respectively ( Fig. 2 E and F ).

We then analyzed the microglia clusters separately for each treatment group. To achieve this, we generated individual UMAPs representing all the cell clusters in each treatment ( Fig. 2 G ), and pie charts to illustrate the contribution of each treatment group to each cluster ( SI Appendix , Fig. S3 H ). Demyelination was associated with a marked reduction of HM paired with a clear expansion of TM, DAM1, DAM2, and PM ( Fig. 2 G and SI Appendix , Fig. S3 H ). The Mafb hi cluster was also moderately increased. These changes were similar in anti-TREM2 and Ctrl IgG2a treated groups. CPZ withdrawal and remyelination led to a decline in the TM, DAM1/2, PM, MHCII, and IFN-R populations, with an expansion of the Mafb hi subset, likely indicating a shift toward reduced activation and return to homeostasis ( Fig. 2 G and SI Appendix , Fig. S3 H ). Remarkably, anti-TREM2 markedly reduced the Mafb hi subset compared to Ctrl IgG2a, while increasing the HM and PM populations ( Fig. 2 G and SI Appendix , Fig. S3 H ). Analysis of DEGs in the demyelination model confirmed minimal impact of the anti-TREM2 treatment, as evidenced by the limited number of DEGs between anti-TREM2 vs. Ctrl IgG2a groups ( SI Appendix , Fig. S3 I ). Conversely the anti-TREM2 treatment during remyelination downregulated genes indicating microglial activation ( Csf1 , Irf8 , Mafb , Tgfbr2 , and Plcg2 ), endocytosis ( Prkca , Ptpn1 , Wasl , and Ehd4 ), and the Tyrobp causal network ( C1qc , Maf , and Nrros ) ( Fig. 2 H and I ). We conclude that the primary effect of anti-TREM2 is to ablate the Mafb hi microglia and their protective functions during remyelination. Interestingly, analysis of Trem2 mRNA expression in microglia revealed a marked increase in Trem2 + microglia during demyelination, which sharply decreased during remyelination ( SI Appendix, Fig. S3 J and K ). The lower efficacy of anti-TREM2 in demyelination vs. remyelination may result from its inability to effectively target all Trem2 + microglia during demyelination.

TREM2 Blockade Impairs the Generation of Mature Oligodendrocytes during Remyelination.

We examined whether TREM2 blockade has cell extrinsic effects on oligodendrocytes transcriptome. Subclustering of oligodendrocyte lineage cells from all five experimental groups identified into 6 populations that were defined based on published markers as mature oligodendrocytes (MOLs, clusters 0, 3), myelin-forming oligodendrocytes or newly formed oligodendrocytes/differentiation-committed oligodendrocytes (MFOL/NFOL/COP, cluster 4), OPCs (cluster 2), proliferating OPCs (POPCs, cluster 5), and demyelination-associated oligodendrocytes (DOLs, cluster 1) ( Fig. 3 A and B and Dataset S4 ) ( 17 , 36 ). Notably, DOLs exhibited a characteristic gene expression profile with markers of activation, senescence, antioxidative response, and extracellular matrix remodeling such as Serpina3n , Apoe , Abca1 , C4b , H2-D1 , Cdkn1a , Bax , and Col5a3 ) ( SI Appendix , Fig. S4 A and Dataset S4 ).

TREM2 blockade impairs remyelination and astrocyte activation. ( A ) UMAP plot of reclustered oligodendrocytes and OPC identifying six subclusters. n = 11,656 nuclei. ( B ) A dotplot shows the feature genes for each subcluster. ( C ) UMAP plot of oligodendrocytes and OPC subclusters split by condition. n = 2,559 nuclei in untreated; n = 2,069 nuclei in demyelination + Ctrl IgG2a; n = 2,136 nuclei in demyelination + αTREM2; n = 2,348 nuclei in remyelination + Ctrl IgG2a; and n = 2,544 nuclei in remyelination + αTREM2. ( D and E ) Pathways and their genes down-regulated in oligodendrocytes ( D ) or OPC ( E ) in the remyelination + αTREM2 group. ( F ) UMAP plot of reclustered astrocytes identifying seven subclusters. n = 4,220 nuclei. ( G ) A dotplot exhibits the feature genes for each subcluster. ( H ) Pearson correlation analysis of astrocyte subclusters. ( I ) UMAP plot of astrocyte subclusters split by condition. n = 1,137 nuclei in untreated; n = 759 nuclei in demyelination + Ctrl IgG2a; n = 831 nuclei in demyelination + αTREM2; n = 625 nuclei in remyelination + Ctrl IgG2a; and n = 868 nuclei in remyelination + αTREM2. ( J ) Pathways and their genes down-regulated in the remyelination + αTREM2 group. P value by the hypergeometric test.

We subsequently created distinct UMAPs to visualize the cell clusters for each treatment ( Fig. 3 C ) and used pie charts to depict the proportion of each treatment group within each cluster ( SI Appendix , Fig. S4 B ). CPZ treatment led to a reduction of MOLs (clusters 0, 3) and MFOL/NFOL/COP (cluster 4) along with an expansion of DOLs (cluster 1). CPZ withdrawal and subsequent remyelination led to the reconstitution of MOLs, MFOLs, and MFOL/NFOL/COP, along with a reduction in DOLs. TREM2 blockade led to a slight increase in the abundance of DOLs, especially during remyelination ( Fig. 3 C and SI Appendix , Fig. S4 B ). DEGs analysis consistently showed that anti-TREM2 reduced the expression of genes associated with oligodendrocyte differentiation, with a moderate effect during demyelination and a stronger effect during remyelination ( Fig. 3 D and SI Appendix , Fig. S4 C and D ). Anti-TREM2 also affected OPCs, especially during remyelination. It led to a reduction in levels of key transcription factors, such as Sox8 and Olig1 , which are crucial for the differentiation and maturation of oligodendrocytes ( Fig. 3 E and SI Appendix , Fig. S4 E and F ). Thus, the ablation of Mafb hi microglia by anti-TREM2 indirectly disrupts the coordinated gene expression programs in OPCs and oligodendrocytes, which are essential for effective remyelination. This disruption ultimately leads to incomplete remyelination.

TREM2 Blockade Hinders Astrocyte Activation.

Subclustering of astrocytes identified 8 distinct populations ( Fig. 3 F and Dataset S5 ). Cluster 0 expressed panastrocyte genes; cluster 2 was enriched in genes related to extracellular matrix production, such as Agt and Itih3 ( Fig. 3 G ). Cluster 1 showed markers of metabolic activation, such as mTOR pathway genes ( Slc7a11 and Slc7a5 ) ( Fig. 3 G and SI Appendix , Fig. S5 A ). Cluster 3 displayed a signature similar to the homeostatic cluster 0, while cluster 4 was closer to the activated cluster 1 ( Fig. 3 G and SI Appendix , Fig. S5 B ). Pearson correlation analysis corroborated similarities of cluster 3 with 0 and cluster 4 with 1 ( Fig. 3 H ). Cluster 5 expressed the synaptogenic gene Thbs4 and the neurogenesis-related gene Igfbp5 ; cluster 6 showed high expression of Gfap and Vim ; cluster 7 expressed IFN-stimulated genes ( Fig. 3 G , SI Appendix , Fig. S5 C and D , and Dataset S5 ).

To demonstrate the impact of each treatment, we created distinct UMAPs to visualize the cell clusters for each group ( Fig. 3 I ) and used pie charts to show the proportion of each treatment group within each cluster ( SI Appendix , Fig. S5 E ). Astrocytes in the untreated group were primarily represented by clusters 0 and 2. Demyelination led to the expansion of clusters 1, 7, and to a minor extent clusters 5 and 6; conversely cluster 0 was reduced. These changes were modestly affected by anti-TREM2 ( Fig. 3 I and SI Appendix , Fig. S5 E ). During remyelination, cluster 4 expanded in the Ctrl IgG2a group, whereas cluster 3 was more prominent in the anti-TREM2 group ( Dataset S5 ). In contrast, clusters 2 and 7 were both reduced during remyelination, regardless of the treatment with anti-TREM2 or Ctrl IgG2a ( Fig. 3 I and SI Appendix , Fig. S5 E ). DEGs analysis revealed that anti-TREM2 mainly influenced the expression of astrocyte activation markers during remyelination, whereas its impact during demyelination was minimal ( Fig. 3 J and SI Appendix , Fig. S5 F and G ). Overall, these findings show that anti-TREM2-induced ablation of Mafb hi microglia during remyelination indirectly reduces astrocyte activation.

TREM2 Blockade Is Associated with Markers of Vascular Dysfunction and Synapse Disruption during Remyelination.

Our initial snRNA-seq data suggested that anti-TREM2 induces differential gene expression in vascular cells compared to Ctrl IgG2a treatment ( Fig. 1 I and Dataset S1 ). Thus, we performed a more detailed analysis of vascular-mesenchymal cell clusters ( Dataset S6 ). These included endothelial cells (Endothelial 1 and Endothelial 2), pericytes, smooth muscle cells (SMCs), and fibroblasts (fibroblasts 1-3) ( SI Appendix , Fig. S6 A and B ). Compared to Endothelial 1, Endothelial 2 exhibited elevated expression of stress-response genes along with signatures associated with apoptosis, autophagy, cytoskeleton organization, and inflammation ( SI Appendix , Fig. S6 C and D and Dataset S6 ). An analysis of subcluster frequencies revealed that anti-TREM2 led to a decrease in pericytes, while the Endothelial 2 subcluster expanded during remyelination ( SI Appendix , Fig. S6 E and F ). These findings suggest the presence of vascular dysfunction in the anti-TREM2 group during remyelination.

Analysis of snRNA-seq data from neurons identified a number of inhibitory and excitatory neuronal clusters ( Dataset S6 ). No significant difference was detected in the proportion of neuronal nuclei with or without TREM2 blockade during either demyelination or remyelination ( SI Appendix , Figs. S2 D and S7 A – C ). Additionally, DEGs analysis of neuronal subclusters revealed minimal effects of anti-TREM2 during demyelination ( SI Appendix , Fig. S7 D and E and Dataset S1 ). However, during remyelination, anti-TREM2 was associated with the downregulation of genes related to synapse organization and neuron projection development in excitatory neurons, as well as genes involved in neuron migration and homeostasis in inhibitory neurons ( SI Appendix , Fig. S7 F – I and Dataset S1 ). These results indicate that the impairment of remyelination caused by anti-TREM2 leads to malfunctioning neurons.

TREM2 Blockade Impacts Multiple Ligand–Receptor Pathways across Various Brain Cell Types.

We used NicheNet analysis of DEGs between anti-TREM2 and Ctrl IgG2a treatments in remyelination to identify ligand–receptor pairs across various brain cell types affected by TREM2 blockade. The top ligands of microglial receptors impacted by anti-TREM2 included: Csf1 , derived from glia, which interacts with microglial Csf1r to promote survival and proliferation; Gas6 , which stimulates phagocytosis in microglia and other glial cells via Mertk , Axl and Tyro3 ; Apoe , which impacts lipid metabolism through binding to Lrp1 , Sorl1 and Vldlr ; and Vcam , which mediates cell adhesion via integrins Itgb1/2 ( SI Appendix , Fig. S8 A and B ). The most significant ligands of oligodendrocyte receptors affected by anti-TREM2 included: microglial Apoe and Gas6 ; F11r (encoding JAM-A) which promotes cell adhesion through Jam3 ; and Prosaposin , which binds to Gpr37 ( SI Appendix , Fig. S8 C and D ). Ligands engaging astrocyte receptors included microglial Apoe , Gas6 , as well as Fgf13 , which interacts with Fgf receptors 1-3 ( SI Appendix , Fig. S8 E and F ). Overall, NicheNet analysis corroborated that TREM2 blockade during remyelination weakens survival, phagocytic, metabolic, and adhesion pathways essential for the glial response to myelin damage.

MAFB hi Microglia Are Present in Acute and Acute/Chronic MS Lesions.

We investigated whether Mafb hi microglia could be identified in human demyelination. For this purpose, we compared the mouse Mafb hi module with microglial signatures in human MS. A reanalysis of the human MS snRNA-seq dataset from Schirmer et al. ( 37 ) showed that microglia coexpressed TREM2 and MAFB ( SI Appendix , Fig. S9 A – D ). In this dataset, MAFB was also weakly expressed in inhibitory neurons. Reclustering of microglia revealed three distinct clusters spanning three different MS conditions ( Fig. 4 A and B and Dataset S7 ). Within these clusters, cluster 1 stood out due to its pronounced MAFB expression. This cluster was selectively enriched within acute/chronic active MS lesions ( Fig. 4 C and D and Dataset S7 ). Reanalysis of another snRNA-seq dataset from Absinta et al. ( 38 ) corroborated coexpression of TREM2 and MAFB in microglia ( SI Appendix , Fig. S9 E – H ). Conversely, MAFB was not detectable in neurons. Reclustering of microglia revealed three microglia clusters across five different MS conditions ( Fig. 4 E and F and Dataset S7 ). Among these, cluster 0 exhibited elevated MAFB expression ( Fig. 4 G and H ) alongside the concurrent expression of microglial markers of activation, including APOE , GPNMB , BHLHE40 , and C1QC ( Dataset S7 ). Importantly, cluster 0 primarily consisted of microglia from the edges of chronic active lesions, which are characterized by ongoing inflammatory activity. MAFB was not detected in microglia from the core of chronic lesions, where the damage to myelin is more extensive and less active, with less inflammation compared to the edges of the lesions ( Fig. 4 E and F and Dataset S7 ).

MAFB + microglia are enriched in active and chronic active lesions of MS. ( A ) UMAP plot from a snRNA-seq dataset of MS and control brain specimens (ref. 37 ), showing identities of reclustered microglia as determined by expression of specific markers. ( B ) UMAP plot of microglia subclusters split by pathology. ( C ) UMAP plots showing expression pattern of MAFB in microglia subclusters. ( D ) Violin plots of microglial expression of MAFB in each type of MS lesion. P value by the Wilcoxon test. ( E ) UMAP plot from snRNA-seq dataset of MS and control brain specimens (ref. 38 ), showing identities of reclustered microglia based on expression of specific markers. ( F ) UMAP plot of microglia subclusters split by pathology. ( G ) UMAP plots showing expression pattern of MAFB in microglia subclusters. ( H ) Violin plot of microglial expression of MAFB in each pathology. P value by the Wilcoxon test. ( I ) Neuropathological assessment by immunohistochemistry or immunofluorescence of the expression of MAFB and TREM2 in the normal white matter of postmortem healthy donors and in active, chronic active, and chronic inactive lesions of MS cases. The arrows in black or white indicate the MAFB + cells. Scale bars in LFB and MHC class II immunohistochemistry panels are 100 μm; Scale bars in MAFB immunohistochemistry panels are 25 μm; Scale bars in TREM2 and MAFB immunofluorescence panels are 10 μm. n = 4 in control white matter; n = 10 in active lesion; n = 10 in chronic active lesion; n = 10 chronic inactive lesion. ( J ) Semiquantitative analysis of the number of TREM2 + MAFB + microglia in the white matter across different pathologies. Data are presented as mean ± SEM. P values by one-way ANOVA with Tukey’s multiple comparisons test. ( K and L ) Immunohistochemistry of MAFB + macrophages within perivascular inflammatory cuffs ( K ) and meningeal infiltrates ( L ).

We further validated the MAFB expression pattern in MS patients vs. controls at the protein level. MS postmortem brain specimens were selected based on the presence of at least one active, one chronic active, and one inactive lesion ( Dataset S8 ). The presence and activity of these lesions in the MS cases and control was assessed by immunohistochemistry for MHC class II, combined with myelin staining with Luxol fast blue ( Fig. 4 I ). Active lesions encompassed newly inflamed areas characterized by a sharp border, and the presence of myeloid cells within the lesion which indicated ongoing inflammation and active demyelination. Chronic active lesions consisted of an inactive center with ongoing demyelination at the edges which contained myeloid cells and represented a slowly expanding form of damage. Inactive lesions consisted of areas of demyelination that lacked active inflammation and had no myeloid cells present, representing past damage. Assessment of the expression levels of TREM2 and MAFB, revealed enhanced expression of MAFB in TREM2 + microglia within both active and chronic active lesions but not inactive lesions ( Fig. 4 I and J ). Additionally, immunohistochemistry showed MAFB expression in the macrophage nuclei within perivascular inflammatory cuffs and meningeal infiltrates in MS patients ( Fig. 4 K and L ). We conclude that the MAFB expression is associated with microglial and macrophage response to ongoing demyelination in MS.

Discussion

During demyelination, microglia become activated to clear myelin debris and support the regeneration of myelin by oligodendrocytes. During disease recovery, microglia regain their homeostatic signature ( 18 , 39 ). Microglia activation is a graded response: three well-known activation states have been previously identified at the transcriptional state: transition, DAM1, and DAM2 ( 33 ). These likely represent a progression from a homeostatic state to a highly activated and protective state that is required for clearance of tissue damage and tissue repair. Our study identifies another activation state characterized by high expression of MAFB. This transcription factor is known to restrict macrophage proliferation and induce their differentiation ( 40 , 41 ). Additionally, it promotes the final stages of microglial development from yolk sac progenitors, facilitating the acquisition and maintenance of microglia identity ( 19 ). Since MAFB-expressing microglia become dominant when CPZ treatment ends and remyelination begins, MAFB expression may indicate a dynamic shift in microglial function toward a homeostatic program. This shift coincides with the simultaneous processes of reduced demyelination and active remyelination. This finding suggests a crucial role of MAFB hi microglia in completing the repair process. Future studies using a conditional Mafb knockout mouse model will be crucial for understanding how MAFB influences the tissue repair functions of microglia.

TREM2 sustains activation of microglia during multiple neurodegenerative disease, including AD and demyelination disorders. It achieves this in part by stimulating the mTOR signaling pathway, which is necessary for protein synthesis and energy production ( 42 ). In our study, we found that acutely blocking TREM2 with mAb 178 during remyelination eliminated Mafb hi microglia, prematurely halting their repair functions and leading to incomplete remyelination. It is still unclear whether TREM2 signaling directly influences MAFB by changing its activity or expression. Alternatively, TREM2’s role may not be specific to MAFB hi microglia but may instead support the metabolic needs of all activated microglia, including both DAM and MAFB hi . In addition to directly ablating Mafb hi microglia, anti-TREM2 treatment also disrupted OPCs differentiation into mature, myelin-forming oligodendrocytes. Furthermore, it reduced astrocyte activation and impaired neuronal synapse formation. The NicheNet analysis unveiled a reduced network of interactions among microglia, oligodendrocytes, and astrocytes during remyelination following TREM2 blocking, resulting in weakened survival, phagocytic, metabolic, and adhesion pathways essential for remyelination. Thus, our data indicate that blocking microglial TREM2 and depleting Mafb hi microglia is linked to secondary disruptions in oligodendrocyte, neuronal, and astrocyte interactions essential for remyelination.

Notably, deletion of the Trem2 gene substantially blocked microglial activation and differentiation into DAM during both demyelination and remyelination ( 17 ). In contrast, acute blockade of TREM2 with mAb 178 failed to prevent the formation of DAM during demyelination. The limited efficacy of the antibody may result from its inability to target all expanding TREM2 + microglia during demyelination and/or to block the heightened signaling driven by abundant TREM2 ligands, such as myelin debris, in the demyelinated environment. Moreover, soluble TREM2, generated by the shedding of TREM2’s extracellular domain ( 43 ), may act as a sink, limiting the antibody’s availability at its target site. The antibody’s limited ability to cross the blood–brain barrier and achieve sufficient concentrations may have hindered its effectiveness in fully blocking microglial activation. Surprisingly, the anti-TREM2 agonist antibodies hPara.09 and mAb5 were found to inhibit microglial debris clearance and myelin recovery following lysolecithin-induced demyelination ( 44 ). A comparison with the anti-TREM2 mAb 178 is needed to determine whether this discrepancy is due to differences in recognized epitopes or binding affinities.

Consistent with our identification of a Mafb hi signature in a mouse model of demyelination and remyelination, we found high expression of MAFB in human microglia associated with MS lesions. This transcriptional signature was particularly prominent in microglial clusters linked to acute and chronic active MS lesions, which are characterized by ongoing demyelination and inflammatory infiltrates. Immunofluorescence analyses further confirmed the presence of MAFB + cells, many of which coexpressed TREM2, in areas of active lesions and chronic active lesions exhibiting ongoing inflammation and demyelination. In contrast, MAFB + cells were absent in inactive lesions, which consisted of demyelinated areas lacking active inflammation and indicative of past damage. MAFB was also expressed in macrophages within perivascular spaces and meninges of patients with active MS. Whether MAFB + cells are bona fide microglia, resident macrophages, or infiltrating macrophages, the expression of MAFB signifies ongoing inflammation and suggests the activation of mechanisms to modulate inflammatory responses to injury.

In conclusion, this study highlights the relevance of MAFB hi microglia in the response to demyelination, their dependency on TREM2, and their potential impact in promoting remyelination in human MS. Although agonistic anti-TREM2 antibodies are currently being tested in preclinical studies and trials for AD ( 45 – 49 ), some of these therapeutics may also be suitable to facilitate myelin repair in demyelinating diseases.

Materials and Methods

The detailed experimental procedures, including antibody generation, CPZ model, TEM, isolation of nuclei from frozen brain tissue, snRNA-seq with data processing and analysis, reanalysis of public MS datasets, examination of postmortem tissue samples by immunofluorescence and immunohistochemistry, as well as statistical methods are thoroughly described in SI Appendix , Materials and Methods .

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Dataset S03 (XLSX)

Dataset S04 (XLSX)

Dataset S05 (XLSX)

Dataset S06 (XLSX)

Dataset S07 (XLSX)

Dataset S08 (XLSX)