Shared mechanisms across the major psychiatric and neurodegenerative diseases

The study investigates whether major psychiatric disorders that typically manifest in early or mid-adulthood (major depressive disorder, bipolar disorder, schizophrenia, anxiety disorders, post-traumatic stress disorder, problematic alcohol use, neuroticism, insomnia) share genetic and molecular mechanisms with neurodegenerative diseases that typically manifest in late life (Alzheimer’s disease, Lewy body dementia, Parkinson’s disease, amyotrophic lateral sclerosis, frontotemporal dementia). Epidemiologic evidence shows individuals with psychiatric conditions have up to four-fold increased risk of later neurodegeneration, and about 65% of people with neurodegenerative diseases experience psychiatric symptoms during illness. Prior work also suggests shared genetic risk (e.g., between schizophrenia and Parkinson’s disease). The authors hypothesize a shared genetic and molecular basis across these psychiatric and neurodegenerative diseases, which could inform early interventions and shared therapeutic targets, potentially mitigating later-life neurodegeneration.

The paper contextualizes prior findings: elevated dementia risk following psychiatric disorders; high prevalence of neuropsychiatric symptoms in neurodegenerative diseases; and evidence of shared genetic architecture between schizophrenia and Parkinson’s disease. It notes established roles of abnormal protein accumulation and interactions in neurodegeneration, emerging evidence for synaptic transmission and mitochondrial dysfunction in disease mechanisms, and limited correlations between brain mRNA and protein abundances necessitating direct proteomic analyses. Previous genetic correlation studies within psychiatric disorders and among brain structures are referenced, motivating a broader cross-domain analysis of psychiatric, neurodegenerative, and brain structural traits.

The study analyzed 25 GWAS spanning psychiatric disorders, neurodegenerative diseases, and brain structure phenotypes, restricting to European-ancestry summary statistics. Pairwise genetic correlations were estimated using LD score regression, robust to sample overlap and LD structure. To identify gene products mediating disease risk via brain protein abundance, the authors integrated GWAS with deeply profiled human brain proteomes (n = 722; predominantly frontal cortex) using a PWAS framework (FUSION) after stringent proteomic QC, normalization, covariate regression, and surrogate variable analysis; 9363 proteins remained post-QC and 2909 with significant cis-SNP heritability were modeled with 500 kb windows using multiple predictive models (BLUP, LASSO, elastic net, BSLMM), selecting best-performing weights. Cis-mediated causal inference combined: (i) PWAS associations (FDR < 0.05); (ii) SMR to test mediation by protein, with HEIDI to exclude LD-driven associations (retain HEIDI p > 0.05); and/or (iii) COLOC colocalization with PP4 > 0.5. Proteins meeting these criteria were termed cis causal proteins. For trans effects, among independent genome-wide significant GWAS SNPs, trans-pQTLs were mapped (p < 5×10^-8, outside 500 kb of target gene), followed by trans-SMR/HEIDI with Bonferroni correction to declare trans causal proteins. Parallel transcriptomic analyses (TWAS, SMR/HEIDI, COLOC) were conducted using human brain transcriptomes (n ≈ 858–888 across descriptions; mainly dPFC) after normalization and covariate adjustment, yielding causal mRNAs and comparison with causal proteins. Functional characterization included: (a) brain region expression profiling of shared causal genes using Allen Brain Atlas microarray Z-scores across amygdala, nucleus accumbens, anterior cingulate, frontal cortex, hippocampus, locus coeruleus, substantia nigra, and temporal lobe; (b) cell-type specificity using human single-cell RNA-seq from dPFC of cognitively normal donors; (c) protein–protein interaction mapping using BioGRID curated physical PPIs, with bootstrap testing for enrichment of cross-group PPIs; and (d) gene set enrichment analysis (GSEA; GO BP/MF/CC, Reactome, KEGG, WikiPathways) on shared and interacting proteins, plus targeted enrichment using curated synaptic proteome and MitoCarta3.0 for mitochondrial proteins. Quality control included removal of proteins with extensive missingness, batch and technical covariate regression, outlier sample removal via PCA, and genetic QC (call rate, MAF, HWE, relatedness, ancestry PCs). Statistical significance used FDR correction where appropriate.

• LD score regression revealed multiple significant genetic correlations within psychiatric disorders and within brain structural traits, and positive correlations among AD, PD, and LBD. Importantly, there were significant cross-domain correlations: among MDD, BD, PTSD, AUD, anxiety symptoms; and between AD and anxiety symptoms (FDR < 0.05). • Integrative proteogenomic analyses identified 651 cis-regulated proteins consistent with causality/pleiotropy across the studied traits (PWAS + SMR/HEIDI and/or COLOC). Trait-level counts included, for example, MDD: 98 proteins (96 cis, 2 trans); BD: 70; schizophrenia: 158; insomnia: 29; AD (two GWAS): 18 and 4; PD: 20; several brain structural traits with varying counts. • Within-group sharing of causal proteins was substantial: psychiatric traits shared on average 45% of their causal proteins (range 31–75%); neurodegenerative traits shared 21% on average (0–50%); brain structure traits shared 30% on average (0–50%). Specific multi-trait shared proteins within groups were noted (e.g., CTNND1, CNNM2 across multiple psychiatric traits; shared loci between AD and PD). • Cross-domain sharing: 13 causal proteins were shared between psychiatric and neurodegenerative diseases (30% of neurodegenerative causal proteins; 13/42), spanning pairs, triads, and quartets of traits. There was a positive relationship between genetic correlation magnitude and percent shared causal proteins across psychiatric–neurodegenerative pairs (Spearman rho = 0.39; p = 0.01). • PPIs were enriched across domains: 12 physical PPIs connected 19 psychiatric and 30 neurodegenerative causal proteins (with 6 shared proteins), representing a 2.6-fold enrichment over chance (bootstrap p = 0.003). Overall, 118 psychiatric and neurodegenerative causal proteins formed an interacting network. • Enrichment analyses implicated synaptic transmission pathways (SNARE complex, vesicle-mediated transport, synaptic vesicle recycling, neurotransmitter secretion), immune/myeloid leukocyte activation, and mitochondrial components and processes. Among the 118 interacting proteins, 102 were synaptic (2.2-fold enrichment; p < 9.5 × 10^-7) and 24 were mitochondrial (3.7-fold enrichment; p < 2.3 × 10^-9), enriched across mitochondrial matrix and inner membrane components. • Brain region and cell-type expression analyses showed heterogeneous regional expression among the 13 cross-domain shared proteins; 10 of 13 had cell-type-enriched expression with notable enrichment in excitatory neurons (e.g., CCDC6, DOC2A, SPATA2, STX1B), inhibitory neurons, and oligodendrocytes. • Transcript-level integration found 615 causal mRNAs for psychiatric traits and 61 for neurodegenerative diseases, with 24 shared mRNAs (37.5% of neurodegenerative causal mRNAs). There were 171 PPIs among 145 psychiatric and neurodegenerative causal mRNAs (2.0-fold over chance; p = 0.008). Approximately 39.6% of causal mRNAs were also causal proteins, and 32.3% of causal proteins were causal mRNAs. ADAM10 and CCDC6 emerged as shared at both mRNA and protein levels.

The findings support a shared genetic and molecular architecture between major psychiatric disorders and neurodegenerative diseases, addressing the central hypothesis. Significant cross-domain genetic correlations and identification of 13 shared causal proteins, together with a 2.6-fold excess of cross-group PPIs, highlight common etiologic pathways. Functional enrichment in synaptic transmission (including SNARE complex and SNAP receptor activity), immune/myeloid activation, and mitochondrial processes indicates convergent mechanisms potentially acting early in disease courses. The link between degree of genetic correlation and shared causal proteins further suggests that genetic overlap manifests at the molecular network level. These shared pathways provide a rationale for developing therapeutics targeting synaptic vesicle cycling, neurotransmitter release machinery, immune modulation, and mitochondrial function that may benefit both early-life psychiatric conditions and late-life neurodegenerative diseases and their neuropsychiatric symptoms. The partial concordance between causal mRNAs and proteins aligns with known modest mRNA–protein correlations in brain tissue, underscoring the value of proteomic integration for target discovery.

This work demonstrates shared genetic susceptibility and molecular pathophysiology between major psychiatric and neurodegenerative diseases. The study identifies 651 causal brain proteins across traits, including 13 proteins shared between psychiatric and neurodegenerative disorders, and an interacting network of 118 cross-domain causal proteins enriched for synaptic, immune, and mitochondrial functions. These results nominate mechanistic pathways and specific brain proteins as promising therapeutic targets with potential cross-disorder efficacy. Future research should extend GWAS power for underpowered neurodegenerative traits (e.g., LBD, FTD), increase the depth and breadth of human brain proteomes (including single-cell proteomics), expand analyses to diverse ancestries to enhance generalizability, and perform mechanistic studies of shared proteins (e.g., ADAM10, MAPT-related pathways) in relevant model systems, aiming to inform early interventions that may mitigate later neurodegeneration.

Key limitations include limited statistical power for traits with smaller GWAS (e.g., LBD, FTD), potentially leading to underdetection of causal proteins. MAPT did not reach significance for AD in PWAS (nominal p = 0.027, FDR = 0.52), likely reflecting power constraints. The proteomic analyses could not fully account for cell-type heterogeneity due to the lack of human single-cell proteomic data. The number of deeply profiled human brain proteomes, while the largest to date (n = 722), still constrains discovery; larger proteomic cohorts should increase yield. Analyses were limited to European ancestry, which may restrict generalizability to other populations. Modest correspondence between mRNA and protein levels in brain implies that transcript-only analyses may miss protein-level effects, although this was mitigated by proteomic integration.