Вспыхнувшая Микроглия, как Макропгейдж в Центральной Нервиционной системе Продромальной болезни Паркинсона.

Introduction

Inflammation with activation of both the adaptive and innate immune system is critical in mediating tissue repair 1 . While some degree of inflammation is observed in virtually all central nervous system (CNS) diseases representing responses to tissue injury, primary T cell mediated autoimmunity such as Multiple Sclerosis (MS) can be a primary cause of CNS inflammation. Parkinson’s Disease (PD) is a neurodegenerative disorder, predominantly in older populations, characterized by the progressive loss of dopaminergic neurons and the accumulation of α-synuclein aggregates in the brain. Both genetic and pathologic evidence support the notion that CNS inflammation contributes to PD pathogenesis 2 – 6 . Specifically, the presence of TNFα in microglia and the recent discovery of autoreactive α-synuclein T cells in early PD 7 raises the hypothesis that autoreactive T cells drive early CNS inflammation in prodromal PD, as is seen in relapsing-remitting MS, with later, secondary neurodegeneration manifesting as classic PD 8 – 13 .

We recently identified myeloid cells in the cerebrospinal fluid (CSF) expressing microglia-associated cell surface proteins, such as APOE, C1Q, and TREM2, that are presumably derived from a monocyte lineage. While first observed in higher numbers in the CSF of patients with HIV 14 , they are also observed in the CSF of healthy subjects. These CSF microglia-like macrophages share gene expression patterns with border-associated macrophages located along the brain vasculature at the interface of the brain parenchyma 15 , 16 . Surprisingly, the frequency of this population of microglia-like macrophages was markedly decreased in the CSF of patients with autoimmune, relapsing-remitting MS as compared to age-matched healthy controls 17 . With highly effective B cell depletion therapy, this population re-emerged in the CSF of MS patients, with associated increases in myeloid TNFα expression in both CSF and blood, suggesting an important role in autoimmune disease pathogenesis 17 .

One early clinical manifestation of α-synucleinopathy is Rapid Eye Movement Sleep Behavior Disorder (RBD), a parasomnia that can appear years or even decades before the onset of typical symptoms of PD or Dementia with Lewy Bodies (DLB) 18 . Thus, RBD serves as an early indicator of future α-synuclein-related neurodegeneration. Additionally, there is strong epidemiologic data indicating a link between autoimmune disorders, such as inflammatory bowel disease and PD, adding further support for a prominent role for the immune system in mediating the inception and progression PD pathology 19 , 20 . Thus, examining whether there is CNS inflammation in RBD as a manifestation of prodromal synucleinopathy and determining the pathways mediating inflammation can provide a rationale for clinical experiments to prevent the disease and to provide more convincing evidence that an early autoimmune process proceeds a secondary neurodegenerative form of disease.

Here, we performed single-cell RNAseq analysis of CSF and blood from 84 individuals, including 34 with RBD, 18 with PD without RBD, 15 PD and RBD, 15 age-matched healthy controls, and 29 patients with recent onset MS. We unexpectedly identified a lymphocytic pleocytosis in the CSF of patients with RBD, that was not present in patients with longer standing PD, similar to what is observed in patients with autoimmune MS. However, in marked contrast to MS, single-cell RNAseq in prodromal PD revealed an increase in the frequency and cell numbers of CSF microglia-like macrophages expressing IL6-JAK-STAT3 and TNFα signaling pathways. This inflammatory signature was shared with that of brain microglia of PD patients and CNS-dural sinus myeloid cells and border-associated macrophages in PD model mice expressing human truncated α-synuclein in dopaminergic neurons. These findings reveal a myeloid-mediated TNFα inflammatory process in the CNS during the prodromal stage of PD, suggesting a newly described pathologic process in disease etiology.

Results

Single-cell immune atlas of PBMCs and CSF in prodromal and early stage of PD

To profile the immunological changes in prodromal and across PD stages, we generated a single-cell RNAseq immune atlas from paired peripheral blood mononuclear cells (PBMCs) and CSF cells. A total of 84 participants were enrolled, of whom 73 contributed blood and CSF samples suitable for analysis and categorized into four groups: RBD (n=36), PD without RBD (described as PD; n=15), PD with RBD (described as PD-RBD; n=18), and age-matched healthy controls (HC; n=15) ( Table S1 , 2 ). Clinical and demographic information was collected, and PBMCs and CSF samples were obtained from all participants. The demographic and clinical characteristics of our study population were similar to other cohorts ( Table S1 ). The mean ages of the subjects in the four groups ranged from 64 to 89 years; 77% were males. Of the RBD group, 19 (53%) were hyposmic, as defined by an UPSIT score below the 15 th percentile for age and gender 21 , 7 (19%) evidenced reduced dopamine transporter (DaT) binding in the striatum, and 25 (69%) showed Synuclein Aggregating Activity (SAA) in their CSF. In all, 26 (72%) of the RBD subjects met MDS Prodromal PD Likelihood Ratio criteria for prodromal PD 22 , and were thus defined as a “high-risk” population. Of the subjects who entered with a diagnosis of RBD, six were diagnosed with either PD or Lewy Body Dementia (LBD) during the course of the study.

We first examined the CSF for evidence of inflammation by quantitative analysis of approximately ~30 mL of CSF from each subject. Unexpectedly, subjects with RBD showed an increase of CSF cells compared to age-matched HC subjects (3.3 + 2.3 vs 1.4 ± 0.7 cells/uL; p = 0.045, ANOVA followed by Tukey-Kramer post-hoc test; Fig. 1A ), indicative of inflammation in the CNS of prodromal PD patients. We then performed droplet-based single-cell RNAseq with VDJ sequencing using the 10x Genomics Chromium platform. After stringent quality control, 523,521 cells from PBMCs (73 donors) and 259,560 cells from CSF (72 donors) were retained ( Fig. 1B – F , Fig. S1A ). The average number of genes and percentage of mitochondrial genes detected per cell across all samples were 1,833 genes and 2.67% in PBMCs and 1,929 genes and 1.83% in CSF, respectively, reflecting high-quality single-cell transcriptomic data across both sample types. Based on the clusters defined by unsupervised clustering, we identified major immune cell types: B cells, myeloid cells, T cells, and NK cells in both PBMCs and CSF. Subclusters within these groups were further annotated based on known marker genes, resulting in 24 clusters in PBMCs and 20 clusters in CSF ( Fig. 1D – G , Fig. S1B , C ). The cell cluster annotations were validated using CellTypist 23 ( Fig. S2E , F ).

In PBMCs, the proportions of immune cells were as follows: B cells 8.09%, myeloid cells 20.8%, T cells 59.6%, and NK cells 11.0% ( Fig. 1D ). In contrast, CSF showed distinctly different proportions consistent with our previous findings 24 : B cells 0.53%, myeloid cells 23.8%, T cells 73.0%, and NK cells 2.59% ( Fig. 1F ). Compared to PBMCs, CSF had a higher proportion of T cells and myeloid cells, and a lower proportion of B cells and NK cells. At the subcluster level, specific differences were observed. For example, populations such as CD4 T naive cells (15.9% in PBMC, 5.62% in CSF), CD16 + monocytes (3.56% in PBMC, 0.09% in CSF), and CD16 + NK cells (10.0% in PBMC, 0.34% in CSF) were less abundant in CSF, while clusters such as cDC2 (1.06% in PBMC, 3.98% in CSF) and CD4 T central memory (Tcm) cells (12.3% in PBMC, 41.6% in CSF) were enriched. As we previously reported 14 , microglia-like macrophages in the CSF (termed CSF Mac), which expressed C1QC and TREM2 specifically, were the most abundant myeloid cell population in CSF and were uniquely observed in this compartment. These findings suggest that, while the CSF immune environment largely consists of circulating immune cells similar to those in PBMCs, it exhibits significant differences in cell composition. The CSF likely represents a unique immune microenvironment linked to the CNS and CNS borders. Overall, we successfully constructed a single-cell immune cell atlas of PBMCs and CSF in RBD and PD.

Pronounced immunological changes in prodromal PD

To identify PD-specific immune alterations, we analyzed our single-cell immune atlas to assess both quantitative changes (shifts in cell frequencies and absolute cell numbers) and qualitative changes (alterations in gene expression). To evaluate changes in cell frequencies, we used a Bayesian model 25 to compare RBD, PD, and PD-RBD groups against age-matched healthy controls ( Fig. 2A – C ). RBD exhibits heterogeneity in features such as hyposmia and CSF SAA status, both of which are known to be associated with PD onset 26 . Using these factors, we further stratified RBD individuals based on predicted PD risk 22 . This analysis revealed a significant increase in the CSF Mac population in the RBD high-risk group ( Fig. 2B , C ). Additionally, the absolute numbers of CSF Mac, CD4 + Tnaive, Tcm, and Tem cells in CSF were increased as compared to healthy controls ( Fig. 2D ) while there was a decrease in the relative proportion of CD4 + Tcm cells and CD8 + T effector memory/terminally differentiated effector memory (em/emra) cells in the RBD group (FDR < 0.05 and < 0.1 respectively). As previously described 27 , we observed an increase in CD4 + T naive cells in the peripheral blood of prodromal PD (FDR < 0.05).

We next examined gene expression changes within each cluster ( Fig. S2A ). Compared to PBMCs, we observed more pronounced gene expression changes in CSF, especially in CD4 + T cells, CD14 + monocytes, and the CSF Mac population. Notably, when stratifying RBD patients by MDS Prodromal PD probability score, the RBD high-probability group exhibited the highest number of differentially expressed genes (DEGs). These findings suggest that both cell frequency and gene expression changes are most pronounced in the RBD high-probability group, with relatively smaller differences observed in PD and PD-RBD. Furthermore, in CSF, CD4 + T cells and CSF Mac demonstrated coordinated changes in both cell numbers and gene expression.

To interpret these gene expression changes, we calculated scores for each cell population across conditions using Hallmark gene sets from the Molecular Signatures Database (MSigDB) ( Fig. 2E , S2B – F , Table S3 , 4 ). In PBMCs, oxidative phosphorylation was broadly upregulated in T cells from RBD and PD-RBD patients, suggesting that peripheral circulating T cells maintain a homeostatic state, consistent with the observed increase in naive CD4 + T cells. In CSF, the most pronounced gene program changes were detected in myeloid populations, particularly in CSF Mac. In the RBD group, significant enrichment of pathways such as mitotic spindle (Mean change = 0.640, padj=4.33 × 10 −302 ), inflammatory response (Mean change = 0.689, padj=2.45 × 10 −189 ), IL6-JAK-STAT3 signaling (Mean change = 0.447, padj=3.97 × 10 −166 ), and TNFα signaling via NF-κB (Mean change = 0.807, padj=2.84 × 10 −96 ), were observed. These findings suggest heightened cellular activity and immune responses in the CSF Mac population in prodromal PD.

Inflammatory signature in MS versus RBD CSFs

A major paradox in the field of autoimmunity revolves around the TNFα pathway, presumably relating to the different genetic architectures of human autoimmune diseases 28 , 29 . Specifically, while anti-TNFα therapy is an effective treatment for Inflammatory Bowel Disease (IBD) and rheumatoid arthritis (RA), it paradoxically induces exacerbations in patients with MS 30 , 31 . Moreover, three independent studies have shown a highly significant decrease in the incidence of PD in IBD patients on anti-TNFα therapy with an increased concordance of IBD and PD 19 , 20 . Based on these observations, we hypothesized that the CSF Mac population in prodromal PD would be characterized by elevated TNFα signaling and exhibit a distinct gene expression profile compared to MS. We analyzed single-cell dataset of CSF cells from 29 MS patients and 7 age-matched healthy individuals. As recently reported in a different MS dataset 17 , we observed a marked reduction in the CSF Mac population that was confirmed with a recent meta-analysis of other MS datasets ( Fig. 3C ) 16 , 17 , 32 . Gene set enrichment analysis revealed a pattern opposite to that observed in RBD, with strong enrichment of gene sets in T cells, B cells, and dendritic cells, while enrichment in CSF macrophages was minimal ( Fig. 3E , Table S5 ). Notably, TNFα signaling via NF- κ B showed a negative change in MS CSF Mac (Mean change=−0.716, padj=4.75 × 10 −5 ). Furthermore, we applied (1) a detailed analysis pipeline specialized for circulating CD4 + T cells 33 , 34 and (2) TCR-based estimation of T cell activity 35 . Both approaches consistently suggested that T cells are not activated in PD or RBD ( Supplementary Note ). These findings suggest that the regulatory mechanisms governing CSF Mac population differ between MS and PD, particularly in the context of TNFα signaling pathway.

Inflammatory signature in CSF Macrophages

We conducted a detailed analysis of CSF Mac, which exhibited notable quantitative and qualitative changes. As previously described, CSF Mac population shares transcriptomic similarity with microglia and expresses macrophage markers, such as C1QC , APOE , and TREM2, at higher levels compared to other myeloid populations ( Fig. 3A – C ) 14 , 16 , 17 . Previous studies also suggested this population resembles border-associated macrophages 15 , 16 . To further analyze the tissue specificity of this CSF Mac population and its similarity with microglia, we integrated myeloid populations from both PBMCs and CSF with publicly available datasets of microglia 36 and cross-tissue myeloid cells 23 , comparing their expression profiles ( Fig. S3 ). While the analysis as expected revealed that CSF Mac population was transcriptionally close to microglia and tissue-resident macrophages, distinct differences were also observed. For instance, SPP1 , P2RY12 , and TMEM119 were more highly expressed in microglia than in CSF Mac ( Fig. S3 ). These findings highlight both the similarities and the divergence in expression patterns between CSF Mac and microglia. We next compared the gene expression of CSF Mac between controls and RBD. The analysis revealed upregulation of HLA class II molecules and their regulator CIITA , as well as increased expression of HIF1A , mitochondrial genes, AP1 family genes, JAK2 , TLR2 , and TNFRSF1A and TNFRSF1B in RBD ( Fig. 3D , E , Table S6 ). These changes were consistent with hallmark gene set analyses, which showed elevated scores for pathways such as inflammatory response and IL6-JAK-STAT3 signaling in RBD patients ( Fig. 3F ). Both RNA seq and flow cytometry analysis revealed an increase in integrin alpha 4 subunit (ITGA4), a key molecule for CNS homing, in CD16 + monocytes from the blood of RBD low-risk individuals, suggesting that monocyte migration to the CNS is more active during the prodromal phase of PD ( Fig. 3G ). This finding suggests that the activation of myeloid cells in peripheral blood is reflected later than in the CSF.

We further hypothesized that CSF Mac might be functionally linked to microglial changes occurring in the brains of PD patients. To test this hypothesis, we utilized publicly available single-nucleus RNAseq data profiling multiple brain regions from 100 donors, including 75 PD cases 37 . The analysis revealed that genes upregulated in CSF Mac in RBD were also significantly upregulated in the primary motor cortex and prefrontal cortex of PD brains ( Fig. S4A , B ). These results suggest that CSF Mac are not only activated during the prodromal phase of PD but also reflect microglial changes occurring within the PD brain. In addition, we attempted to examine myeloid populations in the CNS border region. As dural tissue was not available, we examined dural myeloid cells from PD model mice expressing aggregation-prone truncated human α-synuclein under the control of the tyrosine hydroxylase promoter 38 (material and methods). The dural macrophage expressed CSF Mac marker genes such as Trem2 , Apoe , and C1qc ( Fig. S4C ). We found that TNFα related genes and MHC class II genes were upregulated in PD model mice, consistent with the changes observed in CSF Mac in prodromal PD. These data suggest that synchronized gene expression changes occur in brain parenchymal microglia, CNS-broader macrophage, and CSF Mac in prodromal and clinical stages of PD.

Enhanced cell-cell communication in the immune landscape of RBD

Next, we examined cell-cell interactions in the CSF using CellPhoneDB 39 . The ligand—receptor interaction prediction suggested that myeloid cells, including CSF Mac and cDC2, would exhibit robust ligand—receptor signals and that there would be more interactions between lymphocytes and myeloid cells than among lymphocytes alone ( Fig. 4A ). Focusing on the RBD group, which observed more pronounced immunological changes, we investigated cell-cell interaction pairs with substantial shifts in signaling dynamics. We found that ligands associated with signaling were moderately elevated (FDR<0.2) in lymphocyte subsets such as CD4 Tcm, NK CD16, and memory B cells, whereas receptor signals were enhanced (FDR<0.2) in myeloid subsets including cDC2 and CSF Mac ( Fig. 4B ). Notably, the strongest elevation in receptor mediated signals occurred among myeloid cells themselves, suggesting the presence of autonomous feedback activation. To explore these analyses further, we examined autonomous ligand—receptor pairs in CSF Mac and identified several upregulated pairs in RBD, including C3 – C3AR1 , CD55 – ADGRE5 , and LGALS9 – HAVCR2 (Tim3), which are associated with immunoregulatory functions ( Fig. 4C ). Additionally, we detected neurodegeneration-related signals mediated by amyloid precursor protein ( APP – CD74 , APP – SORL1 ) and prion protein ( PRNP – ADGRG6 ). It is also known that aggregated α-synuclein can induce the activation of myeloid cells 40 . These ligand—receptor pairs may modulate the activation state of CSF macrophages and influence the CNS microenvironment. In particular, CSF Mac may help process and clear disease-associated proteins such as α-synuclein, while simultaneously suppressing excessive inflammation via Tim3 and complement regulation as reported in microglia 41 ( Fig. 4D ). Collectively, these dynamic and comprehensive changes in cell-cell interactions appear to play a critical role in shaping the immune landscape during the prodromal stage of PD.

Genetic prioritization of cell populations highlighted myeloid population in CSF

To determine whether integrating known genetic information with our immune cell atlas could aid in prioritizing relevant cell populations in PD, we analyzed both polygenic and rare variant components. First, we used scDRS 42 to integrate GWAS data associated with PD 4 , RBD 43 , and LBD 44 traits, evaluating the enrichment of polygenic signals in individual cell types ( Fig. S5A , B ). The analysis revealed the strongest polygenic signal in CSF Mac for LBD. Note that LBD 44 shares genetic risks with both PD and Alzheimer’s disease, suggesting it may not be a pure synucleinopathy. For PD, significant associations were observed in cDC2 and pDC in blood and pDC in CSF. No significant associations were detected for RBD, although polygenic signals appeared to accumulate in myeloid populations within the CSF.

We further examined the expression of PARK genes, which have been identified as causative variants in familial PD, within the CSF ( Fig. S6 ). Although some genes, such as PARK7 and CHCHD2 , displayed ubiquitous expression, the overall trend indicated higher expression within myeloid cells. When focusing on CSF Mac, we observed that multiple genes, including SNCA (encoding α-synuclein), PRKN , LRRK2 , PINK1 , and VPS35 were upregulated in RBD and PD. In contrast, PARK7 and CHCHD2 showed higher expression in controls, with reduced levels in RBD and PD, suggesting a loss-of-function role of these genes in PD. Together, these results suggest that, beyond polygenic signals, familial Parkinson’s disease variants also converge in CSF myeloid cell populations, especially in CSF Mac.

Discussion

While emerging evidence suggests an underlying inflammatory etiology for PD as demonstrated by the presence of α-synuclein autoreactive T cells in the circulation of a proportion of patients with manifest and prodromal PD, pathological studies reveal only a minimal presence of CD8 + T cells in the substantia nigra of PD patients early in disease pathogenesis 45 . To elucidate early mechanisms of disease pathogenesis in PD, we performed single-cell RNAseq analysis of paired CSF and blood from 84 subjects with different stages of PD and compared them to 15 age-matched healthy subjects and 29 patients with MS. We found a previously unrecognized pleocytosis that is most pronounced in patients with RBD, a prodromal synucleinopathy. In prodromal PD CSF, single-cell RNAseq analysis revealed increased frequencies of a population of microglia like macrophages, termed CSF Mac, exhibiting inflammatory responses characterized by increased expression of IL6-JAK-STAT3 and TNFRSF1A and TNFRSF1B signaling pathways, with a prominent lack of T cell activation. This is the opposite of what is observed in MS CSF where these is a loss of CSF Mac and increases in activated T cells. Single cell analysis of peripheral blood monocytes in in early prodromal PD revealed increased expression of ITGA4 in circulating CD16 monocytes that was confirmed by flow cytometry consistent with monocyte migration into the CNS. Thus, these findings reveal a myeloid mediated inflammatory process in the CNS of patients with prodromal PD, suggesting a novel pathologic mechanism in disease etiology of PD.

We compared immune populations with single-cell resolution in the CSF of subjects with new-onset MS and RBD. In contrast to MS CSF, where we observed a decreased frequency of CSF Mac, this macrophage population expressing microglia-like features was increased in the CSF of patients with prodromal PD, with a pronounced TNF/IL-6/JAK-STAT signature. Inhibition of the JAK/STAT pathway has been implicated in exerting protective effects in PD 46 – 48 . In addition, treatment of MS with B cell depletion is 98% effective in preventing new inflammatory exacerbations; concurrently, the CSF Mac population becomes exponentially expanded in the CSF with increased secretion of TNFα by circulating monocytes 17 . Taken together, these data suggest fundamental differences in the activity of TNFα pathways between prodromal PD and MS. Moreover, these experimental results are consistent with large-scale epidemiologic investigations of patients with IBD, demonstrating that anti-TNFα can substantially decrease the incidence of new-onset PD 19 .

Our observations comparing PD and MS are in line with a fundamental observation of human inflammatory diseases. Specifically, the therapeutic efficacy found with blocking an inflammatory pathway in one autoimmune disease can lead to flare-ups in other autoimmune diseases. Two prominent examples include anti-IL-17 blockade that has strong efficacy in treating psoriasis but has the risk of new onset or exacerbation of IBD 49 . Another example is anti-TNFα blockade, an effective treatment in IBD and rheumatoid arthritis that can cause exacerbations in patients with MS 31 . These paradoxical observations are related to the different genetic architectures of autoimmune diseases where a risk haplotype in one disease can be protective in another 28 . Thus, we hypothesized that comparing the TNFα pathways among different CNS inflammatory diseases may give insight into disease pathogenesis. However, only a randomized control investigation of patients with prodromal PD can provide definitive evidence for a role of TNFα in the pathogenesis of PD.

Compared to the significant changes in the CSF myeloid populations, relatively minor changes in T cell populations were observed in prodromal PD. Specifically, our analysis replicated previous studies 27 showing increases in circulating and CSF naïve CD4 + cells in prodromal PD and PD with reductions in the proportion of CD4 + Tcm cells and CD8 + Temra cells in CSF. Additionally, recent studies by Sulzer and co-workers demonstrated that activated α-synuclein reactive T cells in mouse models of PD cause inflammation in the gut tissue with the absence of CNS inflammation. These results are in contrast to T and B cell populations in patients with early relapsing/remitting MS, which display markedly inflammatory signals in CSF T cells and in parenchymal lesions which are relatively absent in prodromal PD. The prominent activation of CSF myeloid cells in the absence of T cell activation in prodromal PD suggest a fundamentally different mechanism of autonomous neuroinflammation mediated predominantly by the innate as opposed to the adaptive immune system. In this regard, synchronized myeloid cell activations in brain parenchymal microglia, CNS-broader macrophage are observed that are in equilibrium with human CSF Mac.

In summary, these observations in mouse models and direct examination of humans with prodromal PD leads us to hypothesize that the disease begins with activation of α-synuclein reactive T cells in the gut, perhaps related to the known microbiome dysbiosis in PD patients, leading to aggregation of α-synuclein in the intestines. It is known that α-synuclein can propagate from the gut, via the vagus nerve, to the brain 50 , where it has been shown that α-synuclein aggregates can activate CNS border-associated macrophages 40 and dural myeloid cells, which are synchronized with the CSF Mac. In this model, α-synuclein–reactive T cells may act locally in the dura sinuses as suggested by mouse model and not directly in the brain parenchyma, consistent with our observations in human that in total contribute to the degeneration of CNS dopamine-producing neurons. Together, these data suggest a novel mechanism of myeloid cells as opposed to T cell-mediated autoimmune disease.

Supplementary Material