Integrative analysis reveals the autoantibodyome neuroimmune signature of neurodegeneration

Summary

Neurodegenerative disorders, including Alzheimer’s disease (AD), Parkinson’s disease (PD), and multiple sclerosis (MS), are complex conditions driven by systemic dysregulation that transcends the central nervous system. An integrative systems immunology framework was applied to characterize the neuroimmune “autoantibodyome” across neurodegeneration through an individual participant data meta-analysis of five protein microarray datasets, comprising 596 samples from patients with AD, PD, or MS and healthy controls. We mapped differentially reactive autoantibodies stratified by their targets, unveiling shared features among diseases, such as blood-brain barrier impairment and amplified pro-inflammatory activation, alongside disease-specific perturbations in neuroimmune processes, including short-term memory (AD), skeletal muscle contraction (PD), and pain perception (MS). We identified convergent dysregulation of various autoantibodies targeting diverse synaptic transmission pathways, including gamma-aminobutyric acid (GABA)ergic and glutamatergic signaling. These results indicate the potential of the autoantibodyome to interact with and report on central alterations, suggesting that neurodegeneration may be better understood as a systemic dyshomeostasis.

Graphical abstract

Highlights

• •Hallmark neurodegenerative clinical manifestations are reflected in autoantibodies
• •Blood-brain barrier impairment and immune activation are shared features
• •Autoantibodies target an interconnected network of synaptic molecules
• •GABA, glutamate, serotonin, and dopamine pathways are targeted by autoantibodies

Immunology; Neuroscience

Introduction

Neurodegenerative diseases are characterized by progressive neuronal dysfunction and death, accumulation of host proteins, and prominent immune dysregulation.^1,2 Beyond their personal and familial impact, these diseases pose a significant burden on healthcare systems and economies.³ Alzheimer’s disease (AD), Parkinson’s disease (PD), and multiple sclerosis (MS) are the neurodegenerative diseases with the highest prevalence and disability-adjusted life years (DALYs) worldwide, collectively estimated to have affected over 73 million people in 2023.⁴ By 2050, when more than 2 billion people will be over the age of 60, their prevalence is expected to triple.⁵ As age is the primary risk factor for neurodegenerative diseases, this growing population becomes increasingly vulnerable to conditions that devastate memory, cognition, and autonomy.

AD, the leading cause of dementia, predominantly affects individuals over 65 years of age and is characterized by a progressive decline in memory, comprehension, and reasoning.^6,7 Its pathology involves amyloid-β plaque deposition and hyperphosphorylated tau protein tangles, with mild cognitive impairment (MCI) often representing a prodromal stage of AD.⁸ PD is the second most common neurodegenerative disorder, and like AD, its prevalence increases markedly with age, but it differs in its clinical profile. The manifestations encompass motor and non-motor symptoms, such as tremor, rigidity, bradykinesia, sleep disturbances, and depression, arising from α-synuclein (SNCA) aggregation, loss of substantia nigra neurons, and dopaminergic denervation.^9,10 Early-stage PD (earlyPD) corresponds to Hoehn and Yahr stages I–II, where symptoms are present but balance remains preserved.¹¹ In contrast, MS primarily affects younger adults, especially women between 20 and 50 years of age.¹² Its pathogenesis is driven by immune-mediated inflammation leading to demyelination and neurodegeneration, producing symptoms such as cognitive dysfunction, fatigue, optic neuritis, and depression.^13,14 The most common clinical course at onset is relapsing-remitting MS (RRMS), characterized by acute episodes of neurological deficit followed by remission periods, which can later evolve into secondary-progressive MS (SPMS), marked by an increasing accumulation of disability.^15,16

In these three neurodegenerative conditions, the immune system plays an important role in disease development. While inflammation is one of the primary immune mechanisms associated with the development of these diseases, the adaptive response has also been shown to contribute to neurodegeneration.¹⁷ Autoimmunity is proposed as a mechanism for AD and PD pathophysiology and is a defining characteristic of MS.^18,19 Although AD, PD, and MS differ in etiology and clinical manifestations, they converge on a systemic axis of neuroimmune dysregulation and progressive central nervous system (CNS) decline.²⁰ Within this context, autoantibodies provide a unique vantage point,²¹ capturing the multifaceted interplay between immune surveillance and neurodegeneration.

Autoantibodies, immune molecules directed against the self,^22,23 along with the broader “autoantibodyome,”²¹ here understood as the repertoire of such antibodies within an organism,²⁴ have been independently investigated in the context of neurodegenerative diseases.^25,26 Recent advances in high-throughput proteomic technologies now provide unprecedented access to autoantibody seromics.²⁷ In particular, the ProtoArray Human Protein Microarray platform²⁸ used in this study encompasses over 9,000 full-length human proteins, providing a comprehensive and unbiased view of systemic autoantibody patterns. Differential reactivity analysis²⁹ enables the identification of differentially reactive autoantibodies (DRAs) in neurodegenerative diseases, which have levels significantly altered compared to controls, enabling the exploration of previously inaccessible dimensions of the neuroimmune interface. While previous studies have mostly examined the autoantibodyome as a source of diagnostic biomarkers for individual neurodegenerative diseases,^{25,30,31,32,33,34} a systematic analysis of the neuroimmunological systems and processes targeted—including a comprehensive characterization of autoantibody specificities across diseases and stages and their cross-disease convergence—has not yet been undertaken.

Here, we implemented an integrative systems immunology framework,^35,36,37,38 combining large-scale IgG autoantibodyome profiling, network-based inference, and cross-cohort harmonization³⁹ to systematically map neuroimmune reactivity across AD, PD, and MS. By resolving both disease-specific and convergent autoantibody signatures,^22,23 we uncovered peripheral immune fingerprints that mirror central neurodegenerative processes while consistently implicating synaptic and neurotransmission-related pathways. This systemic perspective provides a rationale to explain why therapeutic strategies directed at single neurotransmitters or synaptic components have thus far failed.⁴⁰ Indeed, neurodegeneration may reflect not an isolated defect but a network-wide disruption,⁴¹ affected by multiple autoantibody interactions.^22,23 Accordingly, the autoantibodyome emerges as both a functional readout of neuroimmune network perturbations and a reservoir of accessible biomarkers for patient stratification and the design of novel therapeutic strategies for neurodegenerative diseases.

Results

Neuroimmune autoantibody signatures indicate systemic crosstalk in neurodegenerative diseases

We first mapped DRAs across neurodegenerative subgroups compared with healthy controls, following the workflow summarized in Figure 1. As seen in the bar plot, this analysis revealed both increased and decreased DRAs, indicating that neurodegenerative pathophysiology involves bidirectional alterations of circulating autoantibodies. Notably, earlyPD exhibited the largest number of DRAs, even surpassing established PD, suggesting that systemic immune dysregulation may peak before symptoms worsen and thus potentially trigger the pathophysiological processes underlying disease initiation and manifestation. Conversely, patients with MCI displayed fewer DRAs than those with AD, indicating that contrasting autoantibody patterns occur with disease progression in AD or PD.

Figure 1.

Integrative systems immunology workflow for neuroimmune autoantibodyome analysis

Overview of the three main stages of analysis: (1) Autoantibodyome data acquisition, including the Preferred Reporting Item for Systematic reviews and Meta-Analyses (PRISMA) workflow for dataset selection from the public repository GEO, serum autoantibody quantification using high-density human protein microarrays (>9,000 proteins), and patient stratification into major disease groups (Alzheimer’s disease [AD], Parkinson’s disease [PD], multiple sclerosis [MS]), subgroups (AD and mild cognitive impairment [MCI], PD and early-stage PD [earlyPD], MS, relapsing-remitting MS [RRMS], and secondary progressive MS [SPMS]), and healthy controls. (2) Data processing and individual participant data meta-analysis, comprising background correction, normalization, duplicate aggregation, identifier conversion, and batch effect correction, resulting in a merged dataset for downstream analyses. Differential reactivity analysis identified differentially reactive autoantibodies (DRAs) at an adjusted p value < 0.05. On the right, bar plots show the count of DRAs per major disease group and subgroup, stratified by fold-change direction (upregulated in red, downregulated in blue). (3) Neuroimmune, synaptic, and neurotransmission mapping, integrating The Human Protein Atlas, SynGO, and CellChat to characterize DRAs targeting proteins expressed in the immune and/or nervous systems, synaptic structures and processes, and neurotransmitter ligand-receptor pairs. Functional enrichment highlighted biological processes related to the neuroimmune axis and synaptic signaling, indicating systems-level interactions between circulating autoantibodies, neurobiology, and immune regulation. See Tables S5 for patient metadata and Table S8 for all DRAs. Data underlying this figure are derived from data curation (Table S6) and integrated GEO datasets (GSE62283, GSE39087, GSE74763, GSE137422, and GSE95718). Created with BioRender.com.

Classification of targets showed that more DRAs mapped to immune-related antigens than to nervous-related antigens, although a substantial fraction of proteins overlapped between both systems, as shown in Figure 2A by the distribution of immune, nervous, and shared immune-nervous targets. While participants with earlyPD and AD presented the highest numbers of immune and/or nervous DRAs, the proportion of targets across systems was similar among all subgroups. This pleiotropic targeting, which may contribute simultaneously to immune regulation and neuronal function, reflects a systemic disruption of neuroimmune homeostasis. Network-level visualization in Figure 2B further highlighted this complexity, revealing shared and disease-specific antigenic hubs across AD, PD, and MS subgroups. While many autoantigens were uniquely associated with individual disease states, several of the most significant DRAs, such as CD247, SOCS2, and RGS13, were common to multiple conditions, suggesting convergent autoimmune dysregulation.

Figure 2.

Neuroimmune autoantibody signatures indicate systemic crosstalk in neurodegenerative diseases

(A) Count of differentially reactive autoantibodies (DRAs) classified by system: immune antigens (dark green), nervous antigens (yellow-green), or both immune and nervous (lime green), across disease subgroups.

(B) Network of autoantibody targets showing DRAs mapped to immune and/or nervous system proteins (indicated by shapes) across disease subgroups (indicated by colors). The top three targets (by p value) of each subgroup are labeled. Shared and subgroup-specific hubs illustrate both common and unique neuroimmune alterations.

(C) Venn diagrams showing the overlap of neuroimmune DRAs across the three major groups (AD, PD, and MS), stratified by regulation direction. Group-specific and intersecting DRAs are highlighted and further enriched.

(D) Top enriched biological processes (BPs) of the intersecting neuroimmune DRAs across the major groups. The top 20 BPs (ranked by p value) reflect common processes potentially dysregulated in the three conditions. See also Figure S1 for the top 35 enriched BPs of the group-specific DRA sets and Table S1 for the functional enrichment of common or unique neuroimmune DRA sets. Data underlying this figure are derived from integrated GEO datasets (GSE62283, GSE39087, GSE74763, GSE137422, and GSE95718).

Next, we characterized shared and disease-specific neuroimmune autoantibody signatures across the major neurodegenerative disorders (AD, PD, and MS), as summarized in Figure 2C, which depicts the overlap of up- and downregulated DRAs among the three diseases. 17 autoantibodies were found consistently upregulated, while 25 were consistently downregulated across all three diseases, pointing to shared neuroimmune processes despite distinct clinical phenotypes. Functional enrichment of these shared DRAs, with the 20 most significant biological processes (BPs) represented in Figure 2D and all enriched BPs of shared and unique sets in Table S1, revealed two complementary signatures. Downregulated DRAs were linked to blood-brain barrier (BBB) integrity (e.g., GJB6, MBP), sensory perception, and chemokine production processes, suggesting weakened barrier function and dyshomeostasis of sensory and immunoregulatory functions. Conversely, upregulated DRAs enriched processes related to leukocyte activation, NF-κB signaling (e.g., IL18RAP, S100A8), and other immune effector processes, indicating a heightened inflammatory and immune activation signature across participants with neurodegenerative diseases.

Disease-specific functional enrichment analyses, with the 35 most significant BPs represented in Figure S1 and all enriched BPs of shared and unique sets in Table S1, revealed distinct neuroimmune profiles. In participants with AD, we found downregulation of lymphocyte differentiation and IL-8 production, along with upregulation of the ERBB signaling pathway, which is related to neuregulins and the development of neurodegenerative diseases.^42,43 In participants with PD, there was downregulation of synaptic processes such as vesicle fusion and actin filament organization, alongside upregulation of immune cell activation. In participants with MS, we observed downregulation of DNA repair pathways and upregulation of synaptic and sensory development processes. Interestingly, some enriched processes were shared between diseases, even though the DRAs were unique to each condition and regulated in opposite directions; for example, “vesicle fusion to plasma membrane” was downregulated in participants with PD but upregulated in participants with MS, suggesting both common biological axes and disease-specific mechanistic divergence.

We alternatively characterized the shared and disease-specific neuroimmune BPs obtained from the enrichment of all DRAs, rather than the selective enrichment of neuroimmune DRAs, across the three major groups, with the most significant BPs represented in Figure S2 and all enriched processes by subgroup in Table S2. Shared BPs included leukocyte migration and proliferation, response to type I interferon, humoral immune response, natural killer (NK) cell activation, and neuron projection extension. Similar to the enrichment of group-specific neuroimmune DRAs (Figure S1), the group-specific neuroimmune BPs enriched in participants with AD and PD were associated with leukocyte migration and activation processes, with the addition of synaptic vesicle processes in participants with PD and MS. Therefore, the convergence of the neuroimmune signatures was observed through both the selection of neuroimmune DRAs and neuroimmune BPs.

Altogether, these results shed light on dysregulation of the neuroimmune axis, as reflected through autoantibodies: BBB impairment, amplified immune and pro-inflammatory activation, and altered synaptic function. This shared imbalance, together with the disease-specific neuroimmune characterization, suggests that autoantibodies may serve as systemic reporters of central and peripheral disruption in participants with AD, PD, and MS.

Neurodegenerative disease hallmarks emerge from autoantibody repertoires

Moving beyond the identification of common and group-wise specific neuroimmune DRAs, the analysis was expanded to encompass all DRAs within each subgroup, investigating their associated BPs. This approach allowed us to evaluate the immune, nervous, and neuroimmune BPs, as well as their convergence or divergence across subgroups. As shown in the clustering of enriched BPs in Figure 3A and the distribution of immune, nervous, and shared immune-nervous BPs in Figure 3B, all subgroups displayed significant enrichment of BPs associated with immune and/or nervous system functions, underscoring a common systemic architecture involving the neuroimmune axis.

Figure 3.

Neurodegenerative hallmarks emerge from neuroimmune biological processes enriched by autoantibody repertoires

(A) Uniform manifold approximation and projection (UMAP) of significantly enriched Gene Ontology (GO) biological processes (BPs) by disease group, clustered based on term frequency-inverse document frequency (TF-IDF) of gene sets. Node size indicates significance (–log₁₀p-value), and node color indicates system classification (immune, nervous, or both).

(B) Count of significantly enriched BPs per group, categorized as immune (dark green), nervous (yellow-green), or both immune and nervous (lime green). Classification is detailed in Table S9, and enrichment results are provided in Table S1.

(C) Network of top significantly enriched (p value < 0.01) immune and/or nervous system BPs across disease groups. Circles represent BPs, with size proportional to degree of connectivity; shapes indicate system classification; node color indicates disease group. The top immune BP and top nervous BP (by p value) are labeled for each group.

(D) Top immune BPs (by p value), unique or shared across subgroups, include processes related to lymphocyte and natural killer cell activation, as well as T and B cell differentiation.

(E) Top nervous BPs (by p value), unique or shared across subgroups, include unique disease-characteristic processes such as “short-term memory” in AD, “skeletal muscle contraction” in PD, and “sensory perception of pain” in MS, as well as shared processes related to synaptic vesicle dynamics and neurotransmitter secretion. See also Figure S2 for shared and unique neuroimmune BPs of the major groups and Table S2 for the functional enrichment by subgroup. Data underlying this figure are derived from integrated GEO datasets (GSE62283, GSE39087, GSE74763, GSE137422, and GSE95718).

We sought to determine which of these processes were unique to the specific neurodegenerative disease subgroups, with Figure 3C representing the network of top enriched BPs across subgroups. Several BPs were uniquely enriched in specific subgroups, such as “inflammatory response to wounding” and “short-term memory” in AD, “lymphocyte chemotaxis” and “synaptic vesicle cycle” in earlyPD, and “regulation of inflammatory response” and “axonogenesis” in RRMS (see Table S2 for all enriched BPs by subgroup). Remarkably, some of these unique processes are hallmarks of the clinical manifestations of the specific neurodegenerative diseases, such as “short-term memory” for AD, “skeletal muscle contraction” for PD, and “sensory perception of pain” for MS. This indicates that circulating autoantibodies can reflect characteristic and specific disease processes, highlighting their potential role as biomarkers and overall indicators of pathogenesis and dysregulation.

The most significant unique or shared immune-related BPs (Figure 3D) pointed to convergent mechanisms of immune cell activation and differentiation, particularly involving T and B cells, NK cells, and granulocytes, suggesting a shared immunopathological axis across neurodegeneration. In parallel, the most significant unique or shared nervous-related BPs (Figure 3E) highlighted processes such as “neurotransmitter secretion,” “synaptic vesicle exocytosis,” and “neuron apoptotic process,” indicating that synaptic processes and neurotransmission are conserved neurofunctional disruptions reflected by circulating autoantibodies, which we further explored.

We additionally performed these analyses with the stricter adjusted p value < 0.01 DRAs and found similar results, with the list of DRAs, enrichment results by subgroup, and semantic similarity matrix available in Table S3. While the total number of DRAs was reduced to about half across subgroups (Figure S3A), there were 661 immune and/or nervous BPs enriched in total, which is slightly larger than the 638 immune and/or nervous terms enriched with the adjusted p value < 0.05 DRAs. Overall, 292 terms were found in both, indicating that about half of the terms were shared. However, considering the redundancy of Gene Ontology (GO) terms, semantic similarity analysis between the neuroimmune enriched terms for the two thresholds was 0.914, indicating a very high overall similarity. The top unique and shared immune or nervous BPs (Figures S3B and S3C) also indicated that the immune processes were represented by lymphocyte and leukocyte activation, as well as T cell differentiation, while nervous BPs related to neuron morphogenesis and synaptic vesicle cycle processes.

Synapsis-associated autoantibody signatures in neurodegenerative diseases

The synapse is a key locus of dysfunction in neurodegenerative diseases.^40,44 Our aim was to determine whether DRAs targeted synapse-related proteins and how these antibodies could relate to synaptic processes and cellular components (CCs). Indeed, across all subgroups and conditions, we found DRAs targeting synapse proteins and significantly enriched synapse-associated GO terms across BPs and CCs, indicating that the autoantibodyome may capture alterations at the core of neuronal communication, as represented in Figure 4A, with the distribution of DRAs against synaptic targets and enriched terms.

Figure 4.

Synapsis-associated autoantibody signatures in neurodegenerative diseases

(A) Count of synapse-associated differentially reactive autoantibodies (DRAs), Gene Ontology (GO) biological processes (BPs), and GO cellular components (CCs), stratified by disease subgroup.

(B) Synapse-associated GO terms significantly enriched in each subgroup, including pre- and post-synaptic BPs and CCs. Symbol size denotes the overlap of DRAs and term-associated genes, while color intensity indicates statistical significance (-log₁₀p-value).

(C) Network of the top 15 enriched synapse-related BPs and CCs (ranked by p value) and their corresponding DRA targets. Circles represent GO terms, with size proportional to the degree of connectivity; triangles represent DRAs, with orientation indicating directionality (up- or downregulated) and color indicating disease subgroup. See Table S4 for the functional enrichment for SynGO 2024. Data underlying this figure are derived from integrated GEO datasets (GSE62283, GSE39087, GSE74763, GSE137422, and GSE95718).

Although all disease groups converged on synaptic reactivity, distinct subgroups were enriched for specific synaptic compartments, as represented in Figure 4B, showing the BPs and CCs enriched by subgroup, with complete enrichment results and p values available in Table S4. The three major groups enriched processes such as “modulation of chemical synaptic transmission” and “regulation of synaptic vesicle priming,” although these differed between subgroups. AD enriched processes related to the modulation of synaptic transmission, regulation of vesicle processes, and post-synaptic components; MCI enriched few pre-synaptic processes. PD enriched regulatory and cytoskeletal processes as well as post-synaptic components; earlyPD enriched neurotransmitter uptake, neurotransmitter receptor insertion, and vesicle exocytosis and endocytosis processes, along with both pre- and post-synaptic components. MS enriched few synaptic terms; RRMS mainly enriched pre-synaptic processes and components; SPMS had the highest number of enriched synaptic terms, encompassing both pre- and post-synapse processes, neurotransmitter uptake, synapse structure, and membrane potential, among others. This comprehensive profile in SPMS suggests a stage-dependent escalation, indicating a potential widespread collapse of synaptic machinery at this stage.

A focused network analysis of the top 15 enriched synapse-related BPs and CCs in Figure 4C showed that the related DRAs were mostly upregulated in AD, while in earlyPD these were mostly downregulated. A notable example is the SNCA protein, which was associated with “synaptic vesicle endocytosis” and targeted by downregulated autoantibodies. Given SNCA’s known accumulation and role in PD pathology,⁴⁵ its reduced autoantibody levels could reflect decreased levels of normal SNCA due to misfolding and accumulation of abnormal molecules in the CNS.

In addition to global synaptic processes, autoantibodies targeting molecules related to neurotransmitters were identified, including receptors (e.g., GABRA3, GRIA2) and transporters (e.g., SLC1A6, SLC6A6). These findings suggest that autoantibodies could play a role in synaptic communication by interacting with proteins involved in neurotransmitter release, uptake, and receptor signaling, which could provide a mechanistic link between immune dysregulation and neuronal dysfunction in AD, PD, and MS. Thus, we sought to characterize the autoantibodyome landscape within the context of neurotransmission.

Neurotransmitter pathways are targeted by neurodegeneration autoantibodies

Having identified DRAs directed to synaptic targets, we next examined how the autoantibodies associated with neurodegenerative diseases could affect neurotransmission pathways. To this end, we mapped DRAs targeting neurotransmitter ligands and receptors, summarizing the potentially affected ligand-receptor pairs and their source pathways in Figures 5A–5C (disease-representative subgroups) and Figure S4 (subgroups with fewer features), where the potentially affected ligand-receptor pairs and their source pathways are shown. Across the three major groups (AD, PD, and MS), we identified convergent targeting of the GABA pathway (upregulated DRAs against the transporter SLC6A6), the glutamate pathway (downregulated DRAs against the receptor GRIA2), and serotonin and dopamine pathways. These neurotransmitter systems are broadly altered in neurodegeneration,⁴⁶ and our results indicate that this disruption is reflected in the circulating autoantibodyome.

Figure 5.

Neurotransmitter pathways are targeted by neurodegeneration autoantibodies

(A–C) Chord diagrams representing the relationships between neurotransmitter ligands (bottom) and receptors (top) targeted by differentially reactive autoantibodies (DRAs). Elements targeted by upregulated DRAs are shown in red, elements targeted by downregulated DRAs are shown in blue, and non-targeted elements are shown in black. The representative subgroups with the largest count of features from each disease are shown: AD (A, AD subgroup), PD (B, earlyPD subgroup), and MS (C, SPMS subgroup). Ribbon colors indicate neurotransmitter pathways.

(D) Proposed events leading to the emergence of autoantibodies against neurotransmission molecules in neurodegenerative diseases. Created with BioRender.com. See Figure S4 for subgroups with fewer features and Table S10 for the list of neurotransmission ligand-receptor interactions targeted by DRAs. Data underlying this figure are derived from integrated GEO datasets (GSE62283, GSE39087, GSE74763, GSE137422, and GSE95718).

In the AD subgroup (Figure 5A), additional DRAs include upregulated reactivity against the gamma-aminobutyric acid (GABA) receptor GABRA3 and the glycine receptor GLRA2 and downregulated reactivity against the thyrotropin-releasing hormone (TRH), the calcitonin-related protein RAMP1, and the serotonin pathway enzyme TPH1. Glycine, TRH, and calcitonin pathways have previously been investigated in AD for their neuroprotective effects and as possible therapeutic interventions.^47,48,49 In contrast, participants with MCI did not present DRAs directed against neurotransmitter ligands or receptors, suggesting that these alterations occur with disease progression.

In the earlyPD subgroup (Figure 5B), TRH and RAMP1 are also targeted but by upregulated DRAs. Additional targeted ligands included transporters of the GABA (downregulated DRAs against SLC6A13) and glutamate (upregulated against SLC1A6 and SLC1A7) pathways and the glycine pathway enzyme SHMT1 (downregulated DRAs). EarlyPD further shows upregulated DRAs targeting dopamine and serotonin pathways (DDC, HTR5A, HTR1E) and downregulated DRAs against MERTK, a receptor related to pro-melanin concentrating hormone (PMCH). In the PD subgroup (Figure S4A), neurotransmission-related DRAs are fewer and largely restricted to shared targets in the GABA (upregulated against SLC6A6), glutamate (downregulated against GRIA2), and TRH (downregulated against TRH) pathways.

In the MS subgroup (Figure S4B), PMCH is the only neurotransmission-related target targeted by downregulated DRAs. The RRMS subgroup (Figure S4C) shows DRAs involving the PMCH pathway (upregulated DRAs against MERTK) and downregulated DRAs against the serotonin receptor HTR1E. The SPMS subgroup (Figure 5C) features DRA targets restricted to the GABA (upregulated against SLC6A6 and downregulated against GABRA3) and glutamate (upregulated against SLC1A6 and downregulated against GRIA2) pathways. Overall, MS stages recapitulate the GABAergic and glutamatergic signatures observed in AD and PD but with a narrower set of neurotransmission-related targets.

PMCH and its neuropeptide product, melanin-concentrating hormone (MCH), have been associated with the regulation of sleep and feeding behavior, cognition, and stress, with potential applications as an antidepressant.⁵⁰ The MCH system has been shown to modulate the activity of dopamine, GABA, and glutamate neurons and has been investigated in the contexts of AD and PD,^51,52 but not in MS, to the best of our knowledge. Therefore, our findings highlight PMCH as a previously unexplored yet potentially interesting therapeutic target for further investigation in MS.

Discussion

Our study of the circulating autoantibodyome across AD, PD, and MS is based on the re-analysis of publicly available microarray datasets, to which we applied an integrative systems immunology framework to autoantibodies, as previously implemented by our group^22,23 but not employed in the original studies. This approach enabled a broad characterization of autoantibodies targeting neuroimmunological networks in neurodegenerative diseases, revealing a complex neuroimmune landscape in which peripheral immune signatures mirror CNS dysfunction. Our findings indicate that these conditions are associated with bidirectional dysregulation of the autoantibody repertoire, exemplified by both the amplification of pro-inflammatory responses and the reduction of potentially regulatory or homeostatic functions, such as impairment of the BBB. This duality suggests that neurodegeneration possibly reflects not only pathogenic gain-of-function but also the erosion of protective and regulatory mechanisms.

The characterized DRAs are defined as autoantibodies that exhibit significantly increased or decreased signals compared to controls, reflecting the levels of autoantibodies binding to self-antigens. Although predominantly studied for their role in autoimmune pathogenesis, our findings align with the perspective that a systemic network of autoantibodies is a natural component of the immune system, which becomes dysregulated in the context of disease.^21,22,34,53 In this regard, autoantibodies hold potential as easily accessible blood biomarkers that are quick to measure and affordable, extending beyond autoimmune diseases,⁵⁴ which has been skillfully studied by the original authors and other groups.^55,56

On the functional aspect of autoantibodies, the implications of their binding are diverse. Beyond neutralizing soluble components, upon receptor interaction, autoantibodies can induce receptor internalization and have agonistic, antagonistic, or allosteric effects.^24,57,58 Thus, neither their functional nor their pathophysiological roles can be inferred solely from abundance or binding. For instance, autoantibodies against SNCA have recently been shown to exert a neuroprotective effect by inhibiting SNCA aggregation,⁵⁹ but they also correlate with the occurrence of familial PD.⁶⁰ Notably, autoantibodies have been found to both prevent and induce neurodegeneration,⁶¹ and these opposing effects reinforce the need for functional assays to distinguish protective from pathogenic autoantibodies and their mechanisms of action.

Recent findings have demonstrated in vitro that autoantibodies can indeed affect synapses,^62,63,64,65 supporting the notion that, on a larger scale, the autoantibodyome may not only be active at neuroimmune interfaces but also interact with molecular components of the synaptic machinery, potentially shaping circuit-level communication and plasticity. In this work, we found convergence of autoantibodies targeting synaptic and neurotransmission-related processes. Although dysfunction of GABAergic, glutamatergic, serotonergic, and dopaminergic pathways is a well-established hallmark of neurodegenerative disorders,⁴⁶ our analysis indicates that these same systems are also targeted by the autoantibodies. This intersection underscores the concept of neurodegeneration as a network disorder rather than a defect confined to discrete neurotransmitters or proteins,^40,41 providing a possible explanation for the limited efficacy of therapeutic strategies aimed at single molecular targets.⁴⁰

In addition to the canonical neurotransmitter systems, we identified autoantibodies against less-explored pathways, such as PMCH, TRH, and calcitonin. These neuroendocrine pathways have been linked to CNS effects such as neuroprotection, mood regulation, and cognition^49,50,51,52 but also to systemic effects such as food intake, lipid metabolism, and locomotor activity.^48,66 The emergence of these signals indicates the potential of the autoantibodyome to illuminate less-studied pathways that converge with neurodegenerative symptomatology, warranting future investigation into their role in pathophysiology and potential therapeutic intervention.

A central question concerns whether these circulating autoantibodies can access the CNS to modulate their synaptic targets. While the CNS was once considered an immune-privileged site, accumulating evidence shows that neuroinflammation and BBB dysfunction are pervasive in neurodegeneration^20,67 (Figure 5D). In MS, intrathecal synthesis of pathogenic autoantibodies and the migration of B cells into the CNS are well-established hallmarks, supported by recent mechanistic studies demonstrating CNS-directed autoantibody signatures and BBB alterations characteristic of active disease.^68,69 In PD, increasing evidence indicates that BBB disruption is an early and progressive event, enabling peripheral immune molecules, including autoantibodies, to access the brain parenchyma, in line with recent high-resolution imaging and molecular studies of vascular permeability.^70,71 Such permeability would allow systemic autoantibodies to infiltrate the CNS, bind to neuronal epitopes, and exacerbate inflammatory cascades.⁷² The growing approval of monoclonal antibodies for the treatment of AD supports that immunoglobulins can indeed penetrate CNS compartments, interact with CNS targets, and provide clinical benefit.^{73,74,75,76,77} Thus, it is conceivable that autoantibodies can also actively interact with the pathological development process.

Future perspectives

As this is a high-throughput, hypothesis-generating study, future work should prioritize experimental validation of key DRA targets in independent cohorts and mechanistic models. This should include the accurate measurement of autoantibody levels and functional readouts of synaptic and neurotransmission-related processes in the presence of autoantibodies to evaluate their pathophysiological effects. Characterizing BBB interaction and the trafficking of immune components through it may help delineate the conditions under which circulating autoantibodies gain access to CNS compartments and interact with their targets. Longitudinal and stage-stratified studies in AD, PD, and MS could further establish the temporal relationship between these autoantibodies and neurodegenerative progression, probing their potential as biomarkers. These insights may then support potential therapeutic approaches, such as B cell blockade, to restore neuroimmune homeostasis in neurodegenerative diseases.

In conclusion, our integrative systems immunology analysis indicates that the circulating autoantibodyome captures hallmark BPs and key perturbations in neurotransmitter pathways in neurodegenerative disease. The observed bidirectional neuroimmune interactions, involving autoantibodies directed against nervous system components, support a view of the immune and nervous systems as a coupled network rather than discrete compartments. These findings are consistent with a model in which the autoantibody repertoire can both report synaptic pathology and modulate neuronal communication, thereby linking immune dysregulation to neuronal dysfunction. In addition to their potential as readily accessible blood-based biomarkers for CNS conditions, the identified autoantibody signatures highlight neuroendocrine and neurotransmission-related pathways as candidate targets for therapeutic intervention.

Limitations of the study

This study has several limitations. First, the analysis relies exclusively on publicly available protein microarray datasets. Although this design enabled large-scale comparisons and explicit adjustment for age, sex, and platform version, it also imposed important constraints: one platform lacked MS subtype classification, reducing disease-stage resolution; all datasets were restricted to approximately 9,000 protein targets and to IgG autoantibodies, limiting the fraction of the neuroimmunological autoantibodyome that could be interrogated; and subgroup sample sizes were imbalanced, which may bias detection toward better-powered groups. Additionally, the consolidation of subgroups into major disease groups (AD, PD, MS) may introduce biological heterogeneity, and other relevant confounders, such as treatment status, comorbidities, and environmental exposures, could not be systematically controlled due to incomplete metadata.

Second, our discovery-driven integrative systems immunology approach identifies autoantibody signatures converging on synaptic and neurotransmitter pathways and is consistent with a model in which BBB dysfunction could facilitate CNS access of circulating autoantibodies. However, we did not generate direct evidence of autoantibody penetration into the CNS or of physical interaction with the proposed targets; these inferences require experimental validation in cellular and in vivo models. Finally, the functional consequences of the identified autoantibody-target interactions remain unresolved. Altered binding does not necessarily imply pathogenic, neutral, or protective effects, and key parameters such as antibody specificity, affinity, and isotype, as well as effector functions, were not assessed. Mechanistic studies will therefore be essential to establish causality and determine the therapeutic relevance of the neuroimmune autoantibody signatures described here.

Resource availability

Lead contact

Requests for further information and resources should be directed to and will be fulfilled by the lead contact, Otavio Cabral-Marques (otavio.cmarques@usp.br).

Materials availability

This study did not generate new, unique reagents.

Data and code availability

• •This paper analyzes existing, publicly available protein microarray datasets accessible on the GEO database under the following accession numbers: GSE62283, GSE39087, GSE74763, GSE137422, and GSE95718.
• •All original code used in this study has been deposited at Zenodo and is publicly available at https://doi.org/10.5281/zenodo.17833682 as of the date of publication.
• •Analysis results and full tables are provided in the supplemental information.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

We thank the São Paulo Research Foundation (grants 2023/14417-2 to J.N.U., 2024/08016-8 to A.L.N., 2025/07090-2 to F.Y.N.d.V., 2024/22162-7 to R.G.N., 2025/02768-0 to D.G.d.A.F., 2023/13356-0 to D.L.M.F., 2019/14526-0 and 2020/04667-3 to G.C.-M., 2018/149332 to H.I.N., 2020/16246-2 to I.S.F., and 2018/18886-9 to O.C.-M.) for financial support. We also thank the National Council for Scientific and Technological Development (CNPq) (grants 140013/2025-3 to A.S.A., 130027/2023-5 to Y.L.G.C., 102430/2022-5 to L.F.S., and 309482/2022-4 to O.C.-M.) and the Coordination for the Improvement of Higher Education Personnel (CAPES) (CAPES/PROEX grants 88887.917898/2023-00 to J.N.U., 88887.801068/2023-00 to A.L.N., 88887.699840/2022-00 to F.Y.N.d.V., 88887.082794/2024-00 to R.S.S., and 88887.196113/2025-00 to I.S.F.) for financial support. J.N.U. was further supported by Charité, the Medizinische Hochschule Brandenburg Theodor Fontane, and the German Academic Exchange Service (DAAD; Ref. No. 91898528). G.M. was supported by grants from the German Federal Ministry of Education and Research (BMBF) and the German Research Foundation (DFG: EXPAND-PD project #CA2816/1) and through the BIH Center for Regenerative Therapies (BCRT) and the Berlin-Brandenburg School for Regenerative Therapies (BSRT: GSC203), respectively, and in part by the European Union’s Horizon 2020 Research and Innovation Program and the grant agreements no. 733006 (PACE), 779293 (HIPGEN), 754995 (EU-TRAIN), and 101095635 (PROTO). We thank DeMarshall and Nagele et al. for the rich data made publicly available, which was utilized in this study.

Author contributions

J.N.U.: Methodology, Software, Formal Analysis, Data Curation, Visualization, Writing - Original Draft, Writing - Review & Editing. A.L.N., F.Y.N.d.V., Y.L.G.C., A.S.A., R.G.N., D.G.d.A.F., R.S.S.: Software, Formal Analysis, Visualization, Writing - Review & Editing. L.F.S., G.C-M., T.A.K., N.O.C., G.M., A.H.C.M., R.J.S.D., H.I.N., G.R., I.S.F., H.D.D.: Writing - Review & Editing. D.L.M.F.: Methodology, Software. O.C-M.: Conceptualization, Writing - Original Draft, Writing - Review & Editing, Supervision, Project Administration, Resources, Funding Acquisition.

Declaration of interests

The authors declare no competing interests.

Declaration of generative AI and AI-assisted technologies in the writing process

During the preparation of this work, the author(s) used ChatGPT and EditGPT in order to improve the readability and language of the manuscript. After using this tool/service, the author(s) reviewed and edited the content as needed and take(s) full responsibility for the content of the published article.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Deposited data
Human autoantibody Invitrogen ProtoArray V5.0 dataset	DeMarshall et al.30	GEO: GSE62283
Human autoantibody Invitrogen ProtoArray V5.0 dataset	Nagele et al.34	GEO: GSE39087
Human autoantibody Invitrogen ProtoArray V5.0 dataset	DeMarshall et al.32	GEO: GSE74763
Human autoantibody Invitrogen ProtoArray V5.1 dataset	DeMarshall et al.31	GEO: GSE137422
Human autoantibody Invitrogen ProtoArray V5.1 dataset	DeMarshall et al.33	GEO: GSE95718
Software and algorithms
R (version 4.4.1)	CRAN	https://cran.r-project.org/bin/windows/base/old/4.4.1/
PAA	Turewicz et al.78	https://doi.org/10.1093/bioinformatics/btw037
PAWER	Fishman et al.29	https://doi.org/10.1186/s12859-020-03722-z
limma	Ritchie et al.79	https://doi.org/10.1093/nar/gkv007
sva (ComBat)	Leek et al.39	https://doi.org/10.1093/bioinformatics/bts034
cyclic LOESS	Ballman et al.80	https://doi.org/10.1093/bioinformatics/bth327
Human Protein Atlas	Uhlén et al.81	https://www.proteinatlas.org
EMBL-EBI Ontology Lookup Service	EMBL-EBI	https://www.ebi.ac.uk/ols4/
Rols	Gatto, L.	https://doi.org/10.18129/B9.bioc.rols
Uberon ontology	Mungall et al.82	https://doi.org/10.1186/gb-2012-13-1-r5
Gene Ontology	The Gene Ontology Consortium83,84	https://doi.org/10.1093/genetics/iyad031
SynGO (release 1.2, 20231201)	Koopmans et al.85	https://syngoportal.org
clusterProfiler	Yu et al.86; Xu et al.87	https://doi.org/10.1038/s41596-024-01020-z
Appyter for Enrichment Analysis Visualization	Clarck et al.88	https://doi.org/10.1016/j.patter.2021.100213
Enrichr	Chen et al.89	https://maayanlab.cloud/Enrichr/
GOSemSim	Yu et al.90,91	https://doi.org/10.1007/978-1-0716-0301-7_11
CellChat	Jin et al.92,93	https://doi.org/10.1038/s41596-024-01045-4
Codes and pipelines	This paper	https://doi.org/10.5281/zenodo.17833682

Experimental model and study participant details

This study is a secondary analysis of publicly available human serum microarray datasets profiling autoantibody reactivity in neurodegenerative diseases and healthy controls (n = 223). We analyzed ProtoArray Human Protein Microarray data from five GEO series: GSE62283,³⁰ GSE39087,³⁴ GSE74763,³² GSE137422,³¹ and GSE95718.³³ Patient samples were classified by condition into seven subgroups: Alzheimer’s disease (AD, n = 45), mild cognitive impairment (MCI, n = 71), Parkinson’s disease (PD, n = 55), early-stage PD (earlyPD, n = 121), multiple sclerosis (MS, n = 30), relapsing-remitting MS (RRMS, n = 31), and secondary progressive MS (SPMS, n = 20). The subgroups were consolidated into three major disease groups: AD (AD and MCI, n = 116), PD (PD and earlyPD, n = 176), and MS (MS, RRMS, and SPMS, n = 81). Participant information (age, sex, diagnosis) is available in Table S5. The influence of age and sex was considered and adjusted for in the differential reactivity analysis, to focus on the impact of disease on the differentially expressed autoantibodies. No further patient information was available. All ethical approvals and consents were obtained in the original studies and are referenced therein. No new human recruitment or sample collection was performed in the present study.

Method details

Data curation and acquisition

We conducted a systematic search of the Gene Expression Omnibus (GEO) database⁹⁴ (https://www.ncbi.nlm.nih.gov/gds) to identify publicly available high-throughput human autoantibody array datasets. The following query was used: ((“autoantibodies”[MeSH Terms] OR autoantibody[All Fields]) AND “Homo sapiens”[porgn] AND “gse”[Filter] AND “Protein profiling by protein array”[Filter]), yielding 69 results as of April 19, 2023 (all search results and dataset information are available in Table S6). The following four inclusion criteria were applied to filter the datasets: (1) neurological conditions, (2) >4,000 autoantibody targets, (3) availability of sex information, and (4) availability of age information. After manual curation, five datasets were retained for downstream analysis: GSE62283,³⁰ GSE39087,³⁴ GSE74763,³² GSE137422,³¹ and GSE95718.³³ Raw ProtoArray data and metadata were retrieved from GEO for the five datasets.

Processing of autoantibody arrays

Duplicate samples, non-healthy controls, and samples lacking age or sex information, as well as those with conflicting metadata across datasets, were excluded. After filtering, n = 596 samples remained. Due to the low technical variation observed among samples processed on the same platform version, raw.gpr files were jointly processed by platform version using R (version 4.4.1), RStudio, and the following packages: PAA,⁷⁸ PAWER,²⁹ limma,⁷⁹ and sva.³⁹ Preprocessing steps included background correction using the normexp method,⁷⁹ normalization via cyclic LOESS,⁸⁰ conversion of target identifiers to gene symbols, and aggregation of duplicate features by median reactivity.

Dataset integration

Batch effects across datasets were corrected using the ComBat method³⁹ (Figure S5). Principal component analysis (PCA) was performed using the singular value decomposition⁹⁵ method, with autoantibody targets as variables. Batch-corrected data from the two platform versions were combined under an individual participant data meta-analysis (IPD-MA or mega-analysis⁹⁶) approach. The final processed expression matrix and corresponding sample metadata are available in Tables S7 and S5, respectively.

Differential reactivity analysis

Differential autoantibody reactivity analysis comparing each subgroup with controls was performed using the limma package,⁷⁹ adjusting for age and sex as covariates. Multiple testing correction was applied using the Benjamini-Hochberg method.⁹⁷ Differential reactive autoantibodies (DRAs) were defined as those with an adjusted p-value <0.05, unless otherwise indicated, and are listed in Table S8. Additional confirmatory analyses were performed with DRAs selected by the adjusted p-value <0.01 threshold, listed in Table S3. The DRAs of the major disease groups (AD, PD, MS) were defined as the autoantibodies exhibiting significant differential reactivity in any of their subgroups.

Neuroimmune and synaptic target profiling of differential autoantibodies

Autoantibody targets associated with the immune or nervous systems were identified using the Human Protein Atlas⁸¹ (https://www.proteinatlas.org). Immune-related targets (n = 5339) were defined as those classified under the immune cell category, immune cell lineage, bone marrow, or lymphoid tissues, which exhibited at least a 4-fold higher normalized expression level compared to other tissues (enriched) or relative to the mean across all other tissues (enhanced). Nervous system–related targets (n = 3226) were defined as proteins classified under the brain category, brain, or choroid plexus tissues that were enriched or enhanced in these.

Biological processes (BPs) were classified as nervous system- or immune system-related based on their association with the Uberon⁸² ontology terms “nervous system” (UBERON:0001016) or “immune system” (UBERON:0002405), including all descendant terms. Ontology information was retrieved from the EMBL-EBI Ontology Lookup Service⁹⁸ (https://www.ebi.ac.uk/ols4/) using the rols R package. The complete annotation lists of nervous and immune system targets and BPs are provided in Table S9.

Synaptic targets and Gene Ontology (GO)^83,84 terms related to synapse were obtained from the synapse knowledge base SynGO,⁸⁵ dataset release 1.2 (“20231201″), available at (https://syngoportal.org/).

Functional enrichment analysis

Functional enrichment analyses were performed separately for common and condition-specific neuroimmune DRA targets, categorized by major disease group and direction of fold change. Overrepresentation tests for Gene Ontology Biological Processes (GO BPs) were conducted using the ClusterProfiler R package.^86,87 For each target set, shared (intersection), AD-specific, PD-specific, and MS-specific, significantly enriched GO BPs (p-value <0.05) were increasingly ranked by p-value and the top terms were selected.

For the analysis of neuroimmune biological processes, functional enrichment was performed using all the DRA targets, stratified by disease group and fold change direction. The overrepresentation test for GO BPs was conducted using the ClusterProfiler package. Significantly enriched terms (p-value <0.05) were classified by system and clustered based on the term frequency-inverse document frequency (TF-IDF) of gene sets, followed by application of the Leiden community detection algorithm, using the Enrichment Analysis Visualization Appyter.⁸⁸ The most significant unique and shared neuroimmune BPs were ranked and selected by p-value. Complete enrichment results are provided in Table S1 for the common and unique neuroimmune DRA sets and Table S2 for all DRAs by subgroup.

Synapse-related functional enrichment analysis was performed using targets associated with the nervous or immune systems, stratified by disease group and fold change direction. The analysis was conducted with the SynGO 2024 database via the enrichment analysis tool Enrichr⁸⁹ (https://maayanlab.cloud/Enrichr/). Enriched synaptic terms were filtered using a significance threshold of p-value <0.05. Complete enrichment results for SynGO 2024 are provided in Table S4.

Semantic similarity analysis of functional enrichment results

The semantic similarity analysis of the sets of enriched terms for adjusted p-value <0.01 DRAs and adjusted p-value <0.05 DRAs was performed with the GOSemSim package,^90,91 which accounts for the GO acyclic graph structure to calculate similarity between terms. The Wang⁹⁹ method and score combination method rcmax were applied, and the similarity matrix is available in Table S3.

DRA-associated neurotransmission ligand-receptor interaction mapping

Ligand–receptor interactions related to neurotransmitters potentially affected by DRAs were identified using the CellChat database^92,93 via the CellChat R package. The neurotransmitter-related signaling pathways were extracted from the database as those consistently annotated as such. Filtered interaction results are presented in Table S10.

Quantification and statistical analysis

All statistical analyses were conducted in R 4.4.1 and RStudio using packages and resources listed in the key resources table. Each serum sample hybridized on a ProtoArray constitutes one biological replicate (n), and no technical replicates were available in the datasets.

Array processing, normalization, batch effect correction by dataset were applied as described in method details. Data quality and correction performance were evaluated by PCA and distributional diagnostics (Figure S5). Differential reactivity analysis was performed with limma,⁷⁹ with age and sex added as covariates in the linear model. Effect sizes (log2-fold changes) and moderated t-statistics were computed per autoantibody. Multiple testing was corrected with the Benjamini–Hochberg method to control the false discovery rate and obtain the adjusted p-values. Significant DRAs were defined as those with an adjusted p-value <0.05. For enrichment analyses, terms with p-value <0.05 were reported, with directionality considered when applicable.

Published: January 23, 2026

Supplemental information

Table S4. Functional enrichment results for SynGO 2024, filtered by p value < 0.05, related to

Figure 4