Shared genetic architecture of psychiatric disorders and ocular diseases: Evidence from genome-wide analyses

Abstract

Objective

Psychiatric disorders are frequently comorbid with ocular diseases, yet the shared genetic basis and potential causal links remain unclear. This study aimed to systematically investigate the shared genetic architecture, bidirectional causality, and implicated biological pathways between common psychiatric disorders and ocular diseases.

Methods

We analyzed four psychiatric disorders and eight ocular diseases using an integrative genome-wide analytical framework. We used linkage disequilibrium score regression (LDSC) to estimate genome-wide genetic correlations and local genetic correlation analysis (LAVA) to identify region-specific shared signals. Bidirectional Mendelian randomization (MR) was performed to assess causal relationships. Shared loci were identified using conditional/conjunctive false discovery rate (cond/conjFDR) and evaluated by colocalization. Variants were then annotated and mapped to genes using FUMA, followed by Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment and protein–protein interaction (PPI) network analyses to prioritize pathways and hub genes.

Results

LDSC revealed significant genetic correlations between anxiety disorder and conjunctivitis, and between major depressive disorder (MDD) and several ocular diseases, including cataract, conjunctivitis, and keratitis. LAVA further confirmed shared, directionally consistent local genetic effects between MDD and these ocular phenotypes across multiple genomic regions. Bidirectional MR suggested that MDD increases the risk of cataract, conjunctivitis, and keratitis, with evidence for bidirectional causality between cataract and MDD. The conjFDR analysis (conjFDR < 0.05) identified approximately 70 independent shared loci (lead SNPs). Functional annotation prioritized representative SNPs with potential functional and regulatory evidence, including rs1029871 and rs678 for bipolar disorder–cataract, rs1042602 for MDD–cataract, and rs12185233 and rs17651549 shared between schizophrenia and both cataract and conjunctivitis. Enrichment analyses of mapped genes suggested that pathways related to hyaluronic acid and amino acid metabolism, cholinergic signaling, and immune–inflammatory processes may contribute to the comorbidity between psychiatric and ocular diseases.

Conclusion

This genome-wide study reveals genetic links between psychiatric and ocular diseases and provides evidence for bidirectional causality in selected phenotypes. Metabolism–inflammation–related pathways may contribute to comorbidity along the eye–brain axis, offering leads for future mechanistic validation.

1. Introduction

Psychiatric disorders are highly prevalent worldwide and pose a major public health challenge, affecting both health and quality of life. Common psychiatric disorders include major depressive disorder (MDD), bipolar disorder (BD), anxiety disorders (ANX), and schizophrenia (SCZ). From 1990–2019, global disability-adjusted life years (DALYs) due to psychiatric disorders increased from 3.1 % to 4.9 %(Collaborators, 2022). According to the World Health Organization (WHO), by 2019, approximately 970 million people worldwide were affected by psychiatric disorders, with anxiety and depression being the most common. In recent years, large-scale genome-wide association studies (GWAS) and multi-omics approaches have substantially advanced our understanding of the genetic underpinnings of psychiatric disorders, identifying numerous risk loci related to neurotransmitter systems, immune–inflammatory responses, energy metabolism, and neurodevelopment(Warren et al., 2024, Sæther et al., 2023, Zou et al., 2025, Schork et al., 2019). Nevertheless, how these genetic findings interact across different disease systems to shape clinical phenotypes and comorbidity patterns remains an important unresolved question.

Although anatomically distinct, the eye and brain are highly interconnected in terms of embryonic development, anatomical structure, and physiological function, collectively referred to as the "eye–brain axis" (Nguyen et al., 2021). As an extension of the central nervous system, the retina shares molecular regulatory mechanisms with the brain, particularly in neurons, glial cells, and the vascular system (London et al., 2013). In recent years, clinical epidemiological studies have revealed significant comorbidity between psychiatric and ocular diseases. For example, patients with cataracts have a significantly increased risk of MDD, while those undergoing cataract surgery show a markedly lower risk compared to those who do not (Chen et al., 2020). Patients with SCZ, MDD, and BD also face higher risks of developing glaucoma (Liu et al., 2020). Similarly, patients with glaucoma have a 1.5-fold higher risk of MDD and a 2.6-fold higher risk of ANX compared to the general population (Zhang et al., 2024b). Moreover, studies have shown that diabetic retinopathy (DR) is associated with an increased risk of SCZ and BD (Zhang et al., 2025, Chen et al., 2024a), while its comorbidity with MDD and ANX is considered one of the most common eye–brain comorbidity patterns (Dong et al., 2025). A cohort study conducted in a Chinese population revealed that, after adjusting for comorbidities, dry eye syndrome (DES) showed higher prevalence rates in patients with MDD, BD, and neurosis (Liang et al., 2020). However, whether these clinical associations are underpinned by shared genetic factors and biological mechanisms remains unclear, and systematic mechanistic studies are still lacking.

With the rapid development of GWAS, numerous genetic loci associated with psychiatric and ocular diseases have been identified. However, most studies have focused on individual disease domains, with few attempts to integrate findings across different systems or disease types. To address this gap, the present study examines four common psychiatric disorders (ANX, MDD, BD, and SCZ) alongside eight major ocular diseases, including cataract, DR, age-related macular degeneration (AMD), keratitis, conjunctivitis, DES, myopia, and glaucoma. We employed large-scale genome-wide data and multiple statistical genetics methods to systematically evaluate genetic correlations and potential shared biological pathways (Fig. 1). Moreover, this study advances an interdisciplinary research paradigm and provides additional empirical support for the eye–brain axis hypothesis. Investigating the molecular and genetic pathways shared between psychiatric and ocular diseases could reveal novel opportunities for early diagnosis and intervention. Such insights may help translate mechanistic understanding into practical strategies for clinical management.

Fig. 1.

Study overview. AMD, age-related macular degeneration; DR, diabetic retinopathy; ANX, anxiety disorder; BD, bipolar disorder; MDD, major depressive disorder; SCZ, schizophrenia; LDSC, linkage disequilibrium score regression; LAVA, Local Analysis of [co]Variant Association; MR, Mendelian randomization; cond/conjFDR, conditional/conjunctive false discovery rate; CADD, Combined Annotation Dependent Depletion; RDB, RegulomeDB.

2. Materials and methods

2.1. GWAS data

We included four common psychiatric disorders and eight common ocular diseases in our analysis. Among the psychiatric disorders, data for BD and SCZ were obtained from the Psychiatric Genomics Consortium (PGC3) (https://pgc.unc.edu/), while ANX and MDD data were sourced from GWAS studies (Li et al., 2024, Als et al., 2023). All samples were of European ancestry. The ocular disease data for AMD, conjunctivitis, DR, keratitis, and myopia were obtained from the FinnGen R11 release, and the remaining data were extracted from the GWAS Catalog (https://www.ebi.ac.uk/gwas/). All samples were of European ancestry, except for the cataract dataset, in which 77 % of the participants were of European descent. Detailed sample information is provided in Table 1.

Table 1.

GWAS summary statistics data.

Phenotypes	PMID/FinnGen ID	Years	Population	Sample size
				Cases	Control
ANX	37945807	2023	European	74973	400243
BD	39843750	2024	European	156643	2799462
MDD	37464041	2023	European	170756	329443
SCZ	35396580	2022	European	53386	77258
AMD	finngen_R11_H7_AMD	2023	European	11023	432955
Cataract	34127677	2021	77.5 % European	67844	438399
Conjunctivitis	finngen_R11_H7_CONJUNCTIVITIS	2023	European	36160	413573
DR	finngen_R11_DM_RETINOPATHY_EXMORE	2023	European	12681	51410
DES	34737426	2021	European	1417	454931
Glaucoma	34737426	2021	European	1190	455158
Keratitis	finngen_R11_H7_KERATITIS	2023	European	14741	429396
Myopia	finngen_R11_H7_MYOPIA	2023	European	4732	432955

ANX, anxiety disorder; BD, bipolar disorder; MDD, major depressive disorder; SCZ, schizophrenia; AMD, age-related macular degeneration; DR, diabetic retinopathy; DES, dry eye syndrome.

2.2. Genetic correlation analysis

We used linkage disequilibrium score regression (LDSC) to estimate the genetic correlation between psychiatric and ocular diseases (Bulik-Sullivan et al., 2015). GWAS summary statistics were first standardized by removing low-frequency minor alleles, ambiguous strand SNPs, and non-SNP variants. Based on the LDSC regression model, genetic covariance and correlation were estimated by assessing the association between SNP effect sizes and linkage disequilibrium (LD) scores (Ni et al., 2018). The European subset of the 1000 Genomes Project was used as the LD reference panel (Auton et al., 2015). To account for multiple testing, false discovery rate (FDR) correction was applied to the results.

Based on the initial LDSC results, we performed Local Analysis of Variant Association (LAVA) on phenotype pairs that remained genome-wide significant (FDR-adjusted p < 0.05). LAVA focuses on regional genetic correlations, enabling a finer dissection of local genetic contributions to overall genetic effects (Werme et al., 2022). The genome was divided into 2500 independent blocks based on European LD patterns, and local correlations were estimated within each block.

2.3. Bidirectional Mendelian randomization (MR) analysis

MR is a statistical approach that leverages genetic variants as instrumental variables to infer causal relationships between exposures and diseases, thereby effectively mitigating confounding inherent in traditional observational studies(Emdin et al., 2017). To investigate the causal relationships between common psychiatric and ocular diseases, we conducted a bidirectional MR analysis. This study was based on the three core assumptions of MR: relevance, independence, and exclusion restriction (Davies et al., 2018).

Single nucleotide polymorphisms (SNPs) reaching genome-wide significance were selected as instrumental variables, using thresholds of p < 5 × 10⁻⁶ and p < 5 × 10⁻⁸. SNPs in linkage disequilibrium (r² ≥ 0.001) within a 10,000-kb window were excluded. Additionally, to minimize weak instrument bias, the F-statistic for each instrument was calculated using the formula F = R²(N − k − 1) / [k(1 − R²)], where SNPs with F < 10 were considered weak instruments and excluded (Supplementary Table S1) (Burgess and Thompson, 2011).

To ensure robustness and accuracy, multiple MR methods were employed, including MR-Egger, weighted median, inverse variance weighted (IVW), simple mode, and weighted mode (Bowden et al., 2016b, Bowden et al., 2016a, Hartwig et al., 2017). To ensure the robustness of the results, we conducted multiple sensitivity analyses. Heterogeneity was assessed by calculating the Q statistic for both MR-Egger and IVW methods; a p-value less than 0.05 indicated significant heterogeneity, in which case a random-effects IVW model was applied for subsequent analyses. Furthermore, the MR-Egger intercept test was used to detect potential pleiotropy of the instrumental variables; a significant intercept (p < 0.05) suggested possible bias (Burgess et al., 2017). The MR-PRESSO method was utilized to detect and remove potential outliers (Verbanck et al., 2018). All analyses were conducted using the R software (version 4.3.2) with the "TwoSampleMR" and "MRPRESSO" packages.

2.4. cond/conjFDR and colocalization analysis

We applied the conditional false discovery rate (condFDR) and conjunctive false discovery rate (conjFDR) methods to identify shared genetic loci between psychiatric and ocular diseases (Andreassen et al., 2013). The condFDR is a Bayesian model-free method for GWAS summary statistics that enhances the discovery of phenotype-associated SNPs by leveraging overlapping SNP associations. The conjFDR combines condFDR values from multiple phenotypes to identify loci jointly associated with all relevant traits (Smeland et al., 2020). Genetic loci with conjFDR < 0.05 were considered shared loci and were carried forward for downstream analyses.

Due to the complex and unique LD structure of the major histocompatibility complex (MHC) region, we excluded the extended MHC region (chromosome 6: 25,119,106–33,854,733), the chromosome 8p23.1 region (chromosome 8: 7,200,000–12,500,000), and the MAPT region (chromosome 17: 40,000,000–47,000,000) prior to fitting condFDR/conjFDR models (Schwartzman and Lin, 2011). These exclusion regions were defined according to the human genome build 19 to avoid bias.

We performed Bayesian colocalization analysis (coloc) using the R package "coloc" (R version 4.3.2). This approach estimates posterior probabilities to assess whether two phenotypes share the same causal variant within a specific genomic region (Giambartolomei et al., 2014). A locus was considered colocalized when the posterior probability of hypothesis 4 (PP.H4), which indicates a shared causal variant, was greater than 0.75.

2.5. Locus definition and functional annotation

We performed functional mapping and gene annotation using FUMA (https://fuma.ctglab.nl/) to define genetic loci for shared SNPs identified by condFDR/conjFDR and coloc(Watanabe et al., 2017). For each phenotype pair, SNPs reaching conjFDR < 0.05 were used as the input set for LD-based clumping and locus definition in FUMA. Among these, candidate SNPs with LD r² < 0.6 were considered independent significant SNPs, and independent SNPs with LD r² < 0.1 were designated as lead SNPs. SNPs with LD r² ≥ 0.6 defined the boundaries of a lead SNP’s locus. Loci within 250 kb of each other were merged into a single locus. All LD calculations referenced genotype data from European populations in the 1000 Genomes Project. Additionally, we determined the effect direction of shared loci by comparing the Z-scores in the GWAS data.

We prioritized significant SNPs located in protein-coding exons and performed functional annotation using ANNOVAR. Moreover, functional analyses of these SNPs were performed using CADD scores, RegulomeDB scores, and chromatin state information from the FUMA platform. Specifically, CADD scores indicate the potential deleteriousness of SNPs to protein function, while RegulomeDB scores predict whether SNPs reside in putative regulatory regions that may influence transcriptional regulation (Kircher et al., 2014, Boyle et al., 2012).

2.6. Gene mapping and functional analysis

To identify genes mapped by significant SNPs, we employed the SNP2GENE function in FUMA, mapping SNPs according to the following criteria. (A) Positional mapping: SNPs were mapped to protein-coding genes within a 10-kb window. (B) eQTL mapping: cis-eQTL SNPs were linked to gene expression, focusing on eQTL data from the EyeGEx and PsychENCODE databases. (C) 3D chromatin interaction mapping: genomic regions were mapped based on physical interactions inferred from the three-dimensional chromatin structure.

We performed functional analysis of the mapped genes. First, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were performed using the R package clusterProfiler, to examine biological processes (BP), molecular functions (MF), cellular components (CC), and enriched pathways associated with these genes. Protein-protein interaction (PPI) networks were constructed using the STRING database (https://string-db.org), and hub proteins were identified via Cytoscape(Szklarczyk et al., 2019).

3. Results

3.1. Genome-wide and local genetic correlation analysis

LDSC analysis revealed significant genetic correlations between ANX and conjunctivitis (genetic correlation [rg] = 0.1839, p = 0.0004) as well as keratitis (rg = 0.1995, p = 0.0288). BD showed a negative genetic correlation with DR (rg = −0.1455, p = 0.0154), and a positive correlation with myopia (rg = 0.0997, p = 0.0227). MDD exhibited varying degrees of genetic correlation with AMD, cataract, conjunctivitis, and keratitis (rg: 0.0803–0.3246; p: 0.0389–1.91 ×10⁻⁸). In addition, SCZ showed a genetic correlation with conjunctivitis (rg = 0.0765, p = 0.024) (Fig. 2A). After FDR correction, the genetic correlations between ANX and conjunctivitis, as well as between MDD and cataract, conjunctivitis, and keratitis, remained statistically significant (Supplementary Table 2).

Fig. 2.

Genetic correlation analysis between psychiatric disorders and ocular diseases. (A) Genome-wide genetic correlations estimated using LDSC regression. Red indicates positive correlation and blue indicates negative correlation; darker colors represent stronger correlations. Significance levels are indicated as ***p < 0.0005, **p < 0.005, *p < 0.05. (B–E) Local genetic correlations between significant phenotype pairs identified by LAVA analysis. Light red and light blue dots indicate nominal significance (p < 0.05), while dark red dots indicate associations surviving Bonferroni correction (p < 0.05). MDD, major depressive disorder; BD, bipolar disorder; SCZ, schizophrenia; ANX, anxiety disorder; AMD, age-related macular degeneration; DR, diabetic retinopathy.

For phenotype pairs that remained significant after FDR correction, we further performed LAVA analysis to explore local genetic architecture. Bivariate LAVA analysis revealed that, after Bonferroni correction, MDD and cataract shared significant and directionally consistent local genetic effects in one genomic region, MDD and conjunctivitis in four regions, and MDD and keratitis in one region (Fig. 2B–E, Supplementary Table 3).

3.2. Bidirectional Mendelian randomization analysis

The bidirectional MR analysis revealed significant causal associations between MDD and multiple ocular diseases. MDD showed positive causal effects on cataract (IVW OR = 1.10, 95 % CI: 1.03–1.18), conjunctivitis (IVW OR = 1.16, 95 % CI: 1.07–1.27), glaucoma (IVW OR = 1.58, 95 % CI: 1.04–2.39), and keratitis (IVW OR = 1.22, 95 % CI: 1.08–1.37). After FDR correction, the associations with cataract, conjunctivitis, and keratitis remained statistically significant. In the reverse MR analysis, cataract (IVW OR = 1.07, 95 % CI: 1.02–1.11) and glaucoma (IVW OR = 1.02, 95 % CI: 1.01–1.03) also exhibited positive causal effects on MDD. The effect of cataract on MDD remained significant after FDR correction.

Additionally, we identified a bidirectional causal relationship between BD and DR, which remained significant after FDR correction. Notably, the causal effects were negative in both directions (Fig. 3, Supplementary Table 4). Given the potential influence of selection bias, phenotype heterogeneity, and correlated horizontal pleiotropy, this pattern should be interpreted cautiously and warrants validation in independent datasets. All MR findings were supported by at least one sensitivity test. Although some IVs exhibited heterogeneity, the overall instrument strength was adequate, with all F-statistics ≥ 20 (Supplementary Table 5).

Fig. 3.

Bidirectional MR results between psychiatric disorders and ocular diseases using the IVW method. *p < 0.05. MR, Mendelian randomization; IVW, inverse-variance weighted; ANX, anxiety disorder; BD, bipolar disorder; MDD, major depressive disorder; SCZ, schizophrenia; AMD, age-related macular degeneration; DR, diabetic retinopathy.

3.3. Identification and validation of shared genetic loci

The conjFDR approach was employed to identify shared genetic loci between psychiatric and ocular diseases. At a conjFDR threshold < 0.05, multiple shared loci were identified between MDD and various ocular diseases. Specifically, MDD shared 2, 3, 3, 1, 1, 4, and 5 genomic loci with AMD, cataract, conjunctivitis, DR, DES, glaucoma, and keratitis, respectively (Fig. 4). Furthermore, ANX shared one locus each with AMD and myopia. BD shared 7, 3, 1, 2, and 1 loci with cataract, conjunctivitis, DES, glaucoma, and keratitis, respectively. Similar to MDD, SCZ shared 4, 13, 12, 1, 1, 3, and 1 loci with AMD, cataract, conjunctivitis, DR, DES, glaucoma, and keratitis, respectively (Supplementary Table 6, Supplementary Figures 1–3).

Fig. 4.

Identification of shared genetic loci between MDD and ocular diseases. Manhattan plots display phenotype-pair–specific shared loci between MDD and each of the following ocular diseases: AMD, cataract, conjunctivitis, DR, DES, glaucoma, keratitis, and myopia. The red dashed line represents the significance threshold of conjFDR < 0.05. Lead SNPs within the shared loci are highlighted in green. MDD, major depressive disorder; AMD, age-related macular degeneration; DR, diabetic retinopathy.

To further verify whether the shared genetic loci were driven by the same causal variants, we conducted colocalization analysis. The results indicated that the PP.H4 values at loci shared between MDD and ocular diseases were all greater than 0.75. In contrast, the four loci associated with BD and cataract, along with one locus linked to BD and conjunctivitis, exhibited PP.H4 values below 0.75. Similarly, six loci related to SCZ and cataract and two loci related to SCZ and conjunctivitis also showed PP.H4 values under 0.75. Additionally, the single locus shared between SCZ and AMD failed to produce an estimable colocalization probability (Supplementary Table 6).

3.4. Functional annotation of candidate SNPs

Detailed information on genome-wide significant and independent SNP loci associated with psychiatric disorders and ocular diseases can be found in Supplementary Table 7. Among the shared genetic loci identified by condFDR/conjFDR and supported by colocalization analysis, three SNPs (CADD > 20) and four SNPs (RegulomeDB score ≤ 2) associated with BD and cataract suggest potential functional effects and regulatory activity, respectively. Notably, rs1029871 and rs678, located near NEK4 and ITIH1, meet both criteria, indicating they may mediate common mechanisms underlying both disorders (Table 2). The SNP rs1042602 (CADD > 20), associated with MDD and cataract, is linked to genes including TYR. For SCZ and ocular diseases, rs12185233 and rs17651549, shared between SCZ and both cataract and conjunctivitis, also show dual evidence of functionality (Supplementary Table 8). Some SNPs, such as rs12637632, have been previously reported to be associated with both BD and cataract, further supporting the robustness of the observed genetic overlap (Supplementary Table 9).

Table 2.

Functional annotation of shared SNP loci between psychiatric disorders and ocular diseases.

Trait1	Trait2	rsID	Chr:BP	effect_allele	other_allele	Lead_SNP	Locus	nearestGene	exonic_function	CADD	RDB
BD	Cataract	rs68136184	1:150199039	C	CTCCTCT	rs10888575	1	ANP32E	nonframeshift substitution	12.79	NA
BD	Cataract	rs4434138	3:52556890	G	A	rs12637632	2	STAB1	nonsynonymous SNV	15.49	5
BD	Cataract	rs11177	3:52721305	A	G	rs12637632	2	GNL3	nonsynonymous SNV	21.5	NA
BD	Cataract	rs2289247	3:52727257	A	G	rs12637632	2	GNL3	nonsynonymous SNV	13.84	NA
BD	Cataract	rs6617	3:52740182	G	C	rs12637632	2	SPCS1	nonsynonymous SNV	10.3	1 f
BD	Cataract	rs1029871	3:52797634	C	G	rs12637632	2	NEK4	nonsynonymous SNV	24.2	1 f
BD	Cataract	rs678	3:52820981	T	A	rs12637632	2	ITIH1	nonsynonymous SNV	25.1	1 f
BD	Cataract	rs1042779	3:52821011	G	A	rs12637632	2	ITIH1	nonsynonymous SNV	10.36	1 f
BD	Cataract	rs68094128	3:52821992	C	CT	rs12637632	2	ITIH1	frameshift substitution	2.894	NA
BD	Conjunctivitis	rs369636	5:140230370	A	C	rs59479	2	PCDHA gene cluster	synonymous SNV	9.649	5
BD	Conjunctivitis	rs369639	5:140230371	A	C	rs59479	2	PCDHA gene cluster	synonymous SNV	8.73	5
BD	Conjunctivitis	rs251364	5:140237387	T	G	rs59479	2	PCDHA gene cluster	frameshift substitution	6.733	NA
BD	Conjunctivitis	rs630162	5:140237548	A	G	rs59479	2	PCDHA gene cluster	synonymous SNV	2.123	NA
BD	Conjunctivitis	rs11321479	5:140242451	GC	G	rs59479	2	PCDHA gene cluster	synonymous SNV	24.8	NA
BD	Conjunctivitis	rs251369	5:140242479	T	A	rs59479	2	PCDHA gene cluster	synonymous SNV	9.661	5
BD	Conjunctivitis	rs155807	5:140347053	T	C	rs59479	2	PCDHA gene cluster	synonymous SNV	18.39	NA
BD	Gaucoma	rs6558407	8:144995494	T	C	rs11786083	1	PLEC	nonsynonymous SNV	18.85	5
BD	Gaucoma	rs7833924	8:144996029	G	A	rs11786083	1	PLEC	nonsynonymous SNV	15.99	4
BD	Gaucoma	rs7002002	8:144997927	A	G	rs11786083	1	PLEC	nonsynonymous SNV	15.22	4
BD	Gaucoma	rs55895668	8:145001031	C	T	rs11786083	1	PLEC	nonsynonymous SNV	16.13	4
BD	Gaucoma	rs11136334	8:145001588	T	C	rs11786083	1	PLEC	nonsynonymous SNV	17.29	2b
BD	Gaucoma	rs11136336	8:145007187	A	G	rs11786083	1	PLEC	nonsynonymous SNV	17.08	2b
BD	Gaucoma	rs11136343	8:145058986	G	A	rs11786083	1	PARP10	nonsynonymous SNV	0.078	NA
BD	Gaucoma	rs11136344	8:145059425	C	T	rs11786083	1	PARP10	nonsynonymous SNV	4.67	NA
MDD	Cataract	rs1042602	11:88911696	C	A	rs1042602	1	TYR	nonsynonymous SNV	24.2	7
MDD	Conjunctivitis	rs9891146	17:65988049	T	C	rs56404791	2	C17orf58	nonsynonymous SNV	5.191	4
MDD	Keratitis	rs11604671	11:113268059	G	A	rs11214589	2	ANKK1	nonsynonymous SNV	4.484	5
MDD	Keratitis	rs2734849	11:113270160	A	G	rs11214589	2	ANKK1	nonsynonymous SNV	0.001	3a
SCZ	AMD	rs17803620	15:89804043	T	C	rs6496568	3	FANCI	nonsynonymous SNV	23	7
SCZ	AMD	rs2283432	15:89836228	C	G	rs6496568	3	FANCI	nonsynonymous SNV	22.2	6
SCZ	Cataract	rs4434138	3:52556890	G	A	rs67409736	1	STAB1	nonsynonymous SNV	15.49	5
SCZ	Cataract	rs13303	3:52558008	T	C	rs67409736	1	STAB1	nonsynonymous SNV	9.423	1d
SCZ	Cataract	rs66782572	3:52567617	A	G	rs67409736	1	NT5DC2	nonsynonymous SNV	10.12	4
SCZ	Cataract	rs3617	3:52833805	A	C	rs67409736	1	ITIH3	nonsynonymous SNV	7.773	NA
SCZ	Cataract	rs3774729	3:63982082	A	G	rs6798742	2	ATXN7	nonsynonymous SNV	13.2	5
SCZ	Cataract	rs12949256	17:43507297	T	C	rs1635298	5	ARHGAP27	nonsynonymous SNV	14.99	4
SCZ	Cataract	rs16940674	17:43910507	T	C	rs1635298	5	CRHR1	nonsynonymous SNV	17.55	1 f
SCZ	Cataract	rs16940681	17:43912159	C	G	rs1635298	5	CRHR1	nonsynonymous SNV	5.04	4
SCZ	Cataract	rs62621252	17:43922942	C	T	rs1635298	5	MAPT-AS1:SPPL2C	nonsynonymous SNV	6.315	5
SCZ	Cataract	rs62054815	17:43923266	A	G	rs1635298	5	MAPT-AS1:SPPL2C	nonsynonymous SNV	0.018	5
SCZ	Cataract	rs12185233	17:43923654	C	G	rs1635298	5	MAPT-AS1:SPPL2C	nonsynonymous SNV	23.5	1 f
SCZ	Cataract	rs12185268	17:43923683	G	A	rs1635298	5	MAPT-AS1:SPPL2C	nonsynonymous SNV	8.654	1 f
SCZ	Cataract	rs12185235	17:43923703	T	C	rs1635298	5	MAPT-AS1:SPPL2C	nonsynonymous SNV	5.04	1 f
SCZ	Cataract	rs12373139	17:43924130	A	G	rs1635298	5	MAPT-AS1:SPPL2C	nonsynonymous SNV	4.046	1 f
SCZ	Cataract	rs12373142	17:43924200	G	C	rs1635298	5	MAPT-AS1:SPPL2C	nonsynonymous SNV	7.462	1 f
SCZ	Cataract	rs754512	17:44055647	T	A	rs1635298	5	MAPT	stopgain	4.834	1d
SCZ	Cataract	rs63750417	17:44060775	T	C	rs1635298	5	MAPT	nonsynonymous SNV	11.85	5
SCZ	Cataract	rs62063786	17:44061023	A	G	rs1635298	5	MAPT	nonsynonymous SNV	7.864	5
SCZ	Cataract	rs62063787	17:44061036	C	T	rs1635298	5	MAPT	nonsynonymous SNV	0.172	5
SCZ	Cataract	rs17651549	17:44061278	T	C	rs1635298	5	MAPT	nonsynonymous SNV	26.8	1 f
SCZ	Cataract	rs10445337	17:44067400	C	T	rs1635298	5	MAPT	nonsynonymous SNV	19.15	1 f
SCZ	Cataract	rs62063857	17:44076665	G	A	rs1635298	5	MAPT:STH	nonsynonymous SNV	1.506	7
SCZ	Cataract	rs34579536	17:44108906	G	A	rs1635298	5	KANSL1	nonsynonymous SNV	14.36	3a
SCZ	Cataract	rs34043286	17:44117119	G	A	rs1635298	5	KANSL1	nonsynonymous SNV	21	4
SCZ	Cataract	rs553586616	17:44248837	C	T	rs1635298	5	KANSL1	nonsynonymous SNV	21.8	NA
SCZ	Cataract	rs543087095	17:44249199	G	T	rs1635298	5	KANSL1	nonsynonymous SNV	26.4	NA
SCZ	Cataract	rs566760142	17:44408004	C	A	rs1635298	5	ARL17B:LRRC37A	nonsynonymous SNV	15.22	NA
SCZ	Cataract	rs199913382	17:44625866	C	A	rs1635298	5	LRRC37A2:ARL17A	nonsynonymous SNV	19.03	7
SCZ	Conjunctivitis	rs251353	5:140228164	C	A	rs155359	6	PCDHA gene cluster	nonsynonymous SNV	15.12	NA
SCZ	Conjunctivitis	rs251355	5:140229368	G	C	rs155359	6	PCDHA gene cluster	nonsynonymous SNV	15.05	2b
SCZ	Conjunctivitis	rs251362	5:140236950	C	G	rs155359	6	PCDHA gene cluster	nonsynonymous SNV	17	4
SCZ	Conjunctivitis	rs246074	5:140307969	C	G	rs155359	6	PCDHA gene cluster	nonsynonymous SNV	16.72	1 f
SCZ	Conjunctivitis	rs1193851	11:65386206	G	C	rs1205259	8	PCNXL3	nonsynonymous SNV	22.4	2b
SCZ	Conjunctivitis	rs706792	12:50467644	T	G	rs7308692	9	ASIC1	nonsynonymous SNV	19	NA
SCZ	Conjunctivitis	rs7302981	12:50537815	A	G	rs7308692	9	RP4–605O3.4:CERS5	nonsynonymous SNV	16.3	6
SCZ	Conjunctivitis	rs17751061	19:19413092	T	C	rs10401714	10	SUGP1	nonsynonymous SNV	33	5
SCZ	Gaucoma	rs3774729	3:63982082	A	G	rs9879045	1	ATXN7	nonsynonymous SNV	13.2	5
SCZ	Myopia	rs16969968	15:78882925	A	G	rs11633958	1	CHRNA5:RP11–650L12.2	nonsynonymous SNV	14.97	5

Trait1: Psychiatric disorder; Trait2: Ocular disease; rsID: dbSNP identifier; Chr:BP: Chromosome and base pair position; effect_allele/other_allele: Effect and non-effect alleles; Lead_SNP: Index SNP at the locus; Locus: Locus number; nearestGene: Closest gene to SNP; exonic_function: Exonic functional consequence; CADD: Combined Annotation Dependent Depletion score; RDB: RegulomeDB score.

3.5. Functional analysis of mapped genes

We mapped the shared SNPs between BD and cataract, BD and conjunctivitis, and BD and glaucoma to 28, 28, and 12 protein-coding genes, respectively. The shared SNPs between MDD and ocular diseases were mapped to between 1 and 16 protein-coding genes. For SCZ and ocular diseases, between 4 and 150 protein-coding genes were mapped, with the SCZ–conjunctivitis pair including up to 150 potential genes (Supplementary Table 10). Notably, several genes overlapped among SNPs shared by BD or SCZ and cataract or conjunctivitis, suggesting a substantial shared genetic basis and potentially convergent pathological pathways between these psychiatric and ocular diseases.

Subsequently, we performed GO and KEGG pathway enrichment analyses, as well as PPI network analysis, for all mapped genes. GO enrichment results revealed several key biological processes potentially shared between psychiatric and ocular diseases, including hyaluronic acid metabolism (ITIH family genes), cell adhesion (PCDH family genes), cholinergic signaling (CHRNA genes), and immune regulatory pathways. These findings support the hypothesis of a molecular link along the eye–brain axis involving metabolic–inflammatory processes (Supplementary Table 11).

KEGG pathway enrichment analysis further revealed that psychiatric and ocular diseases may be interconnected through a series of shared molecular mechanisms, including amino acid metabolism (ALAS1, GLYCTK), cholinergic signaling (CHRNA), ABC transporter pathways (ABCA6, ABCA10), and lysosomal function (WDR7). These results underscore the pivotal role of metabolic dysregulation and neural signaling in the comorbidity between psychiatric and ocular diseases (Supplementary Table 12).

Given the extensive overlap in candidate pathogenic genes and pathways shared between BD or SCZ and cataract or conjunctivitis, we further conducted PPI network analyses of these gene sets. The PPI network constructed for BD and cataract comprised 27 nodes and 23 edges. Using the CytoHubba and MCODE plugins, five hub genes were identified: ITIH3, GLT8D1, NEK4, GNL3, and ITIH1. The BD and conjunctivitis PPI network contained 28 nodes and 58 edges, from which nine key genes within the PCDHA2–11 cluster (excluding PCDHA5) were extracted. The SCZ and cataract PPI network included 59 nodes and 68 edges, with ten hub genes identified: MAPT, ARHGAP27, CRHR1, LRRC37A2, ARL17A, PLEKHM1, SPPL2C, KANSL1, LRRC37A, and ARL17B. The SCZ and conjunctivitis PPI network was the most complex, comprising 141 nodes and 171 edges. Seven key genes were identified within the PCDHA2–10 cluster, excluding PCDHA5 and PCDHA9 (Supplementary Figures 4–7).

4. Discussion

The eye and brain share substantial homology in embryonic origin, anatomical organization, and functional regulation, forming a closely interacting “eye–brain axis” (Nguyen et al., 2021). Recent epidemiological studies have reported significant comorbidity between common ocular diseases such as glaucoma and cataract and psychiatric disorders such as MDD and SCZ (Chen et al., 2020, Liu et al., 2020, Zhang et al., 2024b). However, the underlying genetic and molecular mechanisms contributing to this comorbidity remain poorly understood. In this study, we systematically investigated the genetic correlations and potential molecular mechanisms linking psychiatric and ocular diseases by integrating genome-wide and local genetic correlation analyses, bidirectional MR, cross-trait identification and validation of shared genetic loci, functional annotation, and gene-level functional analyses.

At the genetic-correlation level, we identified robust genetic correlations between ANX and conjunctivitis, and between MDD and cataract, conjunctivitis, and keratitis. Local analysis using LAVA further validated shared, directionally consistent local genetic effects between MDD and these ocular phenotypes across multiple genomic regions. This pattern indicates that the comorbidity is unlikely to be driven by a single locus and is more consistent with a shared polygenic background spanning multiple genomic segments.

MR results supported positive causal effects of MDD on cataract, conjunctivitis, keratitis, and glaucoma, whereas reverse-direction MR also suggested that cataract and glaucoma may exert causal effects on MDD. These findings align with prior observational studies reporting increased risk of MDD among patients with inflammatory ocular diseases, cataract, and glaucoma (Chen et al., 2020, Zhang et al., 2024b, Vakros et al., 2021). Several factors may underlie this association. First, patients with MDD exhibit chronic systemic inflammation, and elevated inflammatory markers also play a critical role in ocular surface inflammation (Gałecki and Talarowska, 2018, Bielory and Friedlaender, 2008, Chigbu et al., 2024). Second, MDD is often accompanied by appetite loss, malnutrition, and metabolic dysregulation, all of which can impair lens metabolism and increase the risk of cataract(Zhang et al., 2024d). Previous studies have identified oxidative stress as a key pathogenic factor in cataract formation, and similarly, oxidative stress is pivotal in the pathophysiology of depression (Spector, 1995, Bhatt et al., 2020). This shared biological basis suggests potential bidirectional and reinforcing mechanisms between the two conditions. Additionally, the use of antidepressants may reduce tear secretion and elevate intraocular pressure, thereby increasing the risk of inflammatory ocular diseases and glaucoma (Le et al., 2024).

To investigate the potential shared genetic basis between psychiatric and ocular diseases, we applied cond/conjFDR in combination with colocalization analysis to identify and validate shared loci. The results revealed varying numbers of shared loci across different disease pairs. Among these loci, 54.3 % exhibited concordant allele effect directions, suggesting that they may influence psychiatric and ocular disease risk through similar biological mechanisms. Conversely, 45.7 % showed opposite effect directions, indicating directionally divergent effects of the same variant across phenotypes. Similarly, we also observed bidirectional negative causal effects in MR between BD and DR that remained significant after FDR correction. Such a pattern may reflect selection bias, heterogeneity in phenotype definitions, population structure, or correlated horizontal pleiotropy, in addition to any true biological relationship. Therefore, these discordant or negative signals should be viewed as hypothesis-generating and require replication in independent datasets and validation through functional studies before drawing mechanistic conclusions.

After annotating significant shared SNPs located in protein-coding exons, we found that several loci exhibited deleterious effects on protein function and regulatory effects on gene expression across different diseases. Our study found that rs1029871 and rs678 may exert both deleterious and regulatory effects in the development of BD and cataract. These loci and their corresponding genes, NEK4 and ITIH1, have been previously associated with BD and SCZ, although no direct evidence has yet linked NEK4 or ITIH1 to cataract (Scott et al., 2009, Zhang et al., 2024a). NEK4 overexpression on the surface of glutamatergic neurons has been shown to promote dendritic branching, axonal elongation, and oxidative stress in cultured primary neurons, suggesting its important role in regulating neuronal morphology and function (Zhang et al., 2024a). As a member of the NEK gene family, NEK4 also contributes to mitochondrial respiration and dynamics, mitochondrial DNA maintenance, stress responses, and cell death, indicating that it may also play a role in the pathogenesis of cataract (Basei et al., 2024, Babizhayev, 2011, Babizhayev and Yegorov, 2016). The inter-α-trypsin inhibitor (ITI) family consists of protease inhibitors highly expressed in the liver and known to stabilize the extracellular matrix by binding to hyaluronic acid (Enghild et al., 1991, Chen et al., 1994). ITIH1, a heavy chain member of this family, has been associated with inflammatory responses, tumorigenesis, and psychiatric disorders (Zhang et al., 2024c). Although direct evidence linking ITIH1 to cataract is lacking, its role in inflammation regulation and extracellular matrix homeostasis may be relevant to cataract pathophysiology and warrants further investigation.

Our study also identified rs17651549, located in the MAPT gene, as potentially involved in the comorbidity between SCZ and cataract, along with other members of the MAPT gene region. Microtubule-associated protein tau (MAPT), located on chromosome 17q21.31 and composed of 15 exons, encodes the tau protein, which plays a critical role in stabilizing neuronal microtubules. It has been widely implicated in dementia and various neurodegenerative disorders (Strang et al., 2019, Labbé et al., 2016, Caffrey and Wade-Martins, 2007). Recent studies have also implicated MAPT in the pathogenesis of psychiatric disorders, particularly SCZ, and it has been proposed as a potential therapeutic target (Dang et al., 2025, Salenius et al., 2024). Additionally, we identified rs1042602 as a potentially relevant locus for the comorbidity between MDD and cataract. This SNP is located at chromosome 11q14.3 within the tyrosinase (TYR) gene. TYR encodes tyrosinase, a rate-limiting enzyme in melanin synthesis, expressed in retinal pigment epithelial cells and cutaneous melanocytes (Olivares and Solano, 2009). It catalyzes the conversion of tyrosine to DOPA and subsequently to dopaquinone, and has been associated with an increased risk of several ocular diseases, including cataract (Paviani et al., 2020, Boutin et al., 2020, Chen et al., 2024b). While no direct evidence links TYR to the onset of MDD, the tyrosine metabolism pathway is closely related to dopamine synthesis, dysfunction of which is a key mechanism in MDD pathophysiology (Daubner et al., 2011, Belujon and Grace, 2017). Therefore, TYR may exert an indirect effect on the comorbidity between MDD and cataract by modulating polygenic networks involved in neurotransmitter synthesis and metabolism.

Enrichment analyses of protein-coding genes further confirmed that inflammation and metabolism play central roles in the mechanisms underlying the comorbidity between ocular and psychiatric disorders. In particular, repeated enrichment of the ITIH gene family in the comorbidity between BD, SCZ, and cataract suggests that hyaluronic acid metabolism may serve as a key molecular hub underlying the eye–brain axis (Zhang et al., 2024c). Furthermore, abnormalities in amino acid metabolism, especially disruptions in glycine and serine pathways mediated by ALAS1 and GLYCTK, may represent common mechanisms underlying these diseases (Sumiyoshi et al., 2004, Yang and Svensson, 2008).

In the comorbidity between BD or SCZ and conjunctivitis, the enrichment of the PCDHA gene family suggests that impaired cell adhesion could lead to damage to the conjunctival epithelial barrier (Shen and Zhang, 2024). Meanwhile, alterations in CHRNA3 and CHRNA5 may affect cholinergic signaling, thereby influencing synaptic plasticity and contributing to the pathological features of SCZ (Wu et al., 2022). The enrichment of NEK7, a key regulator of the NLRP3 inflammasome, and WDR7, a gene involved in lysosomal pathways, suggests a synergistic role of neuroimmune activation and ocular tissue degeneration in driving comorbidity. Finally, by analyzing PPI networks, we identified key genes involved in the comorbidity between BD, SCZ, and cataract or conjunctivitis, thereby deepening our understanding of the shared mechanisms and offering new directions for future research, while providing prioritized candidates for downstream validation.

Several factors may affect the interpretation and generalizability of our findings. First, the GWAS datasets were derived from European ancestry, which may limit their generalizability to other populations. Some ocular phenotypes, including conjunctivitis and keratitis, were not stratified by subtype, which may mask disease-specific mechanisms. LDSC and LAVA depend on adequate heritability and sample size, so estimates for low-powered traits should be interpreted cautiously. While we employed a range of MR approaches alongside sensitivity analyses, the potential influence of horizontal pleiotropy and weak instruments cannot be entirely ruled out. In addition, functional annotation should be interpreted with caution. Gene and pathway assignments can be affected by local LD structure, heterogeneity in phenotype definitions, and the SNP to gene mapping strategy. Therefore, these results are best viewed as hypothesis generating. Replication in independent datasets and validation in functional experiments are needed before making mechanistic or translational claims.

Despite these limitations, our study provides a comprehensive and integrative analysis of the genetic architecture underlying the comorbidity between psychiatric and ocular diseases. By leveraging large-scale genome-wide datasets and multiple complementary analytical approaches, we offer novel insights into the molecular mechanisms of the eye–brain axis. Future studies incorporating multi-ancestry cohorts, refined clinical phenotyping, and functional experiments will be essential to validate and extend these findings, ultimately advancing the development of targeted interventions and personalized treatment strategies.

5. Conclusion

Our study characterizes shared genetic architecture along the eye–brain axis, identifying genetic correlations and evidence of bidirectional causality between MDD and cataract, together with shared risk loci linking SCZ, BD, and multiple ocular phenotypes. Functional enrichment and PPI analyses nominate inflammation, metabolic dysregulation, and neural signaling pathways as convergent biological themes underlying these associations. Overall, these findings provide a genetic framework for eye–brain comorbidity and generate testable hypotheses for replication and downstream functional validation.

Declaration of generative AI and AI-assisted technologies in the manuscript preparation process

During the preparation of this work, the authors used ChatGPT (OpenAI) to assist with language polishing and clarity improvement. After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the content of the published article.

Institutional review board statement

Not applicable.

Informed consent statement

Not applicable.

Funding

This work was supported by the Shandong Province Medicine and Health Science and Technology Development Programme Project (Grant No. 202203090255), the Natural Science Foundation of Shandong Province (Grant No. ZR2023MH068), and the Brain Science and Brain-like Intelligence Technology – National Science and Technology Major Project (Project Grant No. 2021ZD0200600; Subproject No. 2021ZD0200602).

CRediT authorship contribution statement

Cong Zhou: Visualization, Formal analysis. Jun-jie Luan: Writing – review & editing, Visualization, Resources, Project administration. Xiang-wen Wang: Validation, Formal analysis. Wei-jian Han: Resources, Data curation. Ping Sun: Supervision, Resources, Project administration, Funding acquisition. Xiao-jia Ma: Software, Data curation. Ting-ting Wang: Software. Zhao-yang Zhang: Validation, Resources, Methodology, Conceptualization. Huan-xi Liu: Writing – review & editing, Writing – original draft, Visualization, Validation, Methodology, Formal analysis.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary material

Supplementary material

Supplementary material