New insights into genetic comorbidity mechanisms: type 2 diabetes and primary open-angle glaucoma

Abstract

Aims

To investigate the shared genetic mechanisms between type 2 diabetes (T2D) and primary open-angle glaucoma (POAG). Using large-scale genome-wide association study (GWAS) data, we performed single nucleotide polymorphism (SNP) level analysis to detect pleiotropic variants and loci, paired eQTL mapping analysis and gene-level analysis to identify candidate pleiotropic genes. In addition, Mendelian randomisation (MR) analysis was performed to assess causal associations.

Materials and methods

We used POAG GWAS data from Finngen (9565 cases and 430 250 controls) and T2D GWAS data from 55 555 European ancestry samples. We used Linkage Disequilibrium SCore (LDSC) regression to assess the genetic association between T2D and POAG and further used PLeiotropic Analysis under the COmposite null hypothesis (PLACO) to identify shared genetic variants between paired traits. Finally, we further used MR analysis to explore the causal association between T2D and POAG at the genetic level.

Results

The LDSC results and MR analysis revealed that the T2D effect was significantly higher than that of the POAG (OR=1.09, 95% CI 1.03 to 1.14, p=1.50×10⁻³). The PLACO property analysis determined that the T2D sum POAG shared 178 individual SNPs, separate localisation of 79 individual causes. The five most popular choices are based on the effectiveness of CCND2, SVEP1, ST6GAL1, TCF7L2 and HMGA2. expression quantitative trait loci mapping further revealed 36 genes with regulatory roles in optic nerve-related brain tissues. Functional enrichment analyses indicated that these pleiotropic genes are involved in neurodevelopmental, neuroprotective and metabolic pathways, with tissue-specific enrichment observed in neural, pancreatic, adipose and retinal tissues. It is possible to present the main comorbid mechanisms of T2D and POAG.

Conclusions

Our study provides new insights into the aetiology and pathogenesis of T2D and POAG at the genetic level.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Type 2 diabetes (T2D) has been linked to primary open-angle glaucoma (POAG) in observational studies, but their genetic relationship and causality remain uncertain.

WHAT THIS STUDY ADDS

• This study identifies a shared genetic basis and confirms a causal effect of T2D on POAG. It highlights 79 pleiotropic genes involved in neurodevelopmental and metabolic pathways.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• The findings offer genetic insight into T2D–POAG comorbidity and suggest new targets for early diagnosis and personalised treatment.

Introduction

Glaucoma is a group of diseases characterised by progressive degeneration of retinal ganglion cells and is one of the leading causes of permanent vision loss.^{1 2} The latest Global Burden of Disease study³ revealed that as of 2020, the prevalence of glaucoma in people over 50 years of age is 2.04%, which has become the second greatest threat to human vision impairment. Primary open-angle glaucoma (POAG) is the main phenotype of glaucoma, accounting for about 70% of total glaucoma cases.^{1 4 5} Its pathophysiology is characterised by an open drainage angle but increased resistance to atrial fluid drainage.² Notably, POAG-induced vision loss is slow, and the body receives neurologic compensation for the area of visual loss during the course of the disease.^{6 7} These features reduce early detection of the disease and delay optimal treatment, ultimately decreasing the patient’s quality of life.⁸ To date, no treatment has been able to restore glaucoma-induced visual impairment, making early detection and prevention an effective way to reduce the burden of the disease.⁴ However, the aetiology, pathogenesis and risk factors of POAG are not fully understood, which makes early prevention of the disease difficult.⁹

Type 2 diabetes (T2D) is one of the most common chronic metabolic diseases worldwide, accounting for nearly 90% of the estimated 537 million cases of diabetes worldwide.^{10 11} More seriously, the trend of T2D is increasing at a younger age, with the prevalence of the disease under the age of 40 rising year by year, posing a new threat to global public health.¹² Due to its widespread prevalence worldwide and high comorbidity of glaucoma, T2D is now recognised as a systemic risk factor for glaucoma prevention.^{13 14} However, most significant associations between T2D and POAG originate from observational studies,15,17 and the causal association between the two is currently inconclusive. Of note, two recent Mendelian randomisation (MR) studies have attempted to demonstrate a genetic causal association between T2D and POAG based on genome-wide association study (GWAS) data, but there was significant variability in the results: the causal association between T2D and POAG was marginally significant in populations of European ancestry but not in other populations.^{9 18} Therefore, it is important to further search for specific genomic variants or loci that explain genome-wide genetic correlations and to delve deeper into the common genetic aetiology and mechanisms between these two diseases.¹⁹ Utilising the correlation of GWAS signals to investigate pleiotropic genetic variants or loci between multiple traits and to identify potential targets for genetic intervention in T2D and POAG can contribute to the prevention or treatment of these two diseases.^{20 21}

In this genome-wide pleiotropic association study, utilising large-scale GWAS pooled data, we performed single nucleotide polymorphism (SNP) level analyses to detect pleiotropic variants and loci, followed by paired colocalisation analyses to identify colocalised sites and gene-level analyses to identify candidate pleiotropic genes for unravelling the potential shared genetic aetiology and mechanisms for T2D and POAG. Afterwards, pathway enrichment analysis was used to explain the underlying biological mechanisms in greater depth. In addition, MR analyses were performed to assess causality and partially characterise different types of pleiotropy. Our work provides new insights into the aetiology and pathogenesis of T2D and POAG at the genetic level and contributes to the early prevention of both diseases.

Material and methods

Data sources

We got the GWAS summary data from publicly accessible data of European origin to verify that the data samples were of the same ethnic origin and that the sample sizes were more than 20 000 to assure statistical power. In addition, considering that sample overlap when performing two-sample MR can bias the results, we performed MR analyses using POAG GWAS data from Finngen (N_case=9565; N_{control =} 430 250).²² The GWAS data for type 2 diabetes came from the published large meta-GWAS study that included a sample of 55 555 European ancestry.²³ Online supplemental data 1 contains more extensive details about the aforementioned datasets.

Study design and quality control

Figure 1 illustrates the study overview. The data shown above were detailed in the initial article for inclusion criteria and data quality checks. In addition, we performed additional quality control on the above GWAS data, including removal of SNPs without rsID and uniform alignment with the hg19 human reference genome.

Figure 1. Overall study design of genome-wide cross-trait analysis. GWAS summary statistics of primary open-angle glaucoma (POAG) and type 2 diabetes (T2D) were retrieved from publicly available GWAS. Global correlation analysis and two-way two-sample Mendelian randomisation (MR) analysis between POAG and T2D were conducted, followed by multi-trait meta-analysis of POAG and T2D GWASs using PLACO. Based on the results, SNP-level analysis and gene-level analysis were further implemented to investigate the shared genetics underlying POAG and T2D. A schematic illustration was used in the figure to outline the analytical workflow. eQTL, expression quantitative trait loci; FUMA, Functional Mapping and Annotation; GWAS, genome-wide association study; LDSC, Linkage Disequilibrium Score; MAGMA, Multi-marker Analysis of GenoMic Annotation; PLACO, PLeiotropic Analysis under the COmposite null hypothesis; SNP, single nucleotide polymorphism.

Genetic correlation analysis

We employed cross-trait Linkage Disequilibrium SCore (LDSC) regression to assess the genetic correlation between T2D and POAG.²⁴ Due to the sample overlap, we configured LDSC to have no intercept limit. For the LDSC analysis, we used well-imputed HapMap3 variants along with precomputed LDSC derived from the 1000 Genomes Project Phase 3, specifically for individuals of European ancestry.

Cross-trait meta-analysis using PLACO

Genetic correlation refers to the genome-wide average sharing of genetic effects between traits. To identify shared genetic variation between pairwise traits, we conducted a cross-trait GWAS meta-analysis using the recently developed PLeiotropic Analysis under the COmposite null hypothesis (PLACO).²⁵ This analysis included a multiplicity of effects evaluation based on the p<0.05 criterion within the LDSC frameworks. PLACO can explain the potential correlation between two traits to determine the pleiotropic SNPs for a given variant. The PLACO test is based on the hypothesis H₀: β_trait1×β_trait2=0 vs H₁: β_trait1×β_trait2≠0, with the test statistic T_PLACO=Z_trait1×Z_trait2. For each pair of traits, trait1 and trait2 were denoted as T2D and POAG, respectively. These are: Z_trait1 and Z_trait2 as the observed Z-scores of SNPs from related GWAS summary data, respectively. β_trait1 and β_trait2 indicate the effect sizes for T2D and POAG, respectively. To account for potential sample overlap, we decorrelated the Z-scores using the correlation matrix calculated from GWAS summary data. Significant pleiotropic variations were defined as SNPs with P_PLACO<5× 10⁻⁸.

SNP-level analysis using Functional Mapping and Annotation

We further used the online platform Functional Mapping and Annotation (FUMA, https://fuma.ctglab.nl/) to annotate the significant pleiotropic SNPs identified by PLACO, aiming to identify pleiotropic loci.²⁶ FUMA annotation was conducted using default settings, employing the 1000 Genomes Project Phase 3 for individuals of European ancestry as the reference panel. Independent significant SNPs were defined when p<5×10⁻⁸ and r²<0.6 between SNPs. Lead SNPs, a subset of significant SNPs, are defined if they are independent from each other at r²<0.1. The genomic risk loci were discovered by combining the linkage disequilibrium (LD) blocks of independent significant SNPs that are closest to each other (250 kb). Additionally, we performed functional evaluations using Annotate Variation categories, Combined Annotation-Dependent Depletion (CADD) scores and RegulomeDB (RDB) scores.²⁷ CADD scores assess and quantify the potential deleteriousness of single nucleotide variants, with higher scores indicating a greater likelihood of harm. RDB scores serve as a categorical measure derived from expression quantitative trait loci (eQTL) and chromatin markers, ranging from 1a to 7; lower scores suggest variants with stronger evidence of regulatory function.

Gene-level analysis using Multi-marker Analysis of GenoMic Annotation

Gene-based association analysis is generally more powerful in identifying genetic risk variations compared with SNP-level analysis, as genes are more closely related to biological functions than individual SNPs. This approach enhances our understanding of the biological processes underlying cross-trait characteristics.

To identify candidate pleiotropic genes, we conducted a gene-level Multi-marker Analysis of GenoMic Annotation (MAGMA) based on the results from PLACO for each pair of traits.²⁸ Utilising NCBI build 37.3, we incorporated MAGMA Gene IDs and the locations of 19 427 protein-coding genes. Through this genetic association analysis, we successfully identified significant pleiotropic genes for each pair of traits. In the MAGMA analysis of the PLACO data, the significant gene was proclaimed at both the locus-specific Bonferroni-corrected p<0.05 /N (N: total number of genes tested for each pair of traits).

To investigate the functional consequences of shared genetic loci between POAG and T2D, we performed eQTL mapping using the SNP2GENE module in FUMA. To enhance the biological interpretability of pleiotropic gene signals, we prioritised the integration of eQTL datasets derived from optic nerve-related tissues.²⁹ Specifically, we used the brain cortex and brain hypothalamus-specific eQTL data from the GTEx V.8 dataset, given its relevance to optic nerve pathways and neurodegenerative processes. This tissue-relevant eQTL mapping strategy enables the evaluation of whether identified pleiotropic genes exhibit trait-associated regulatory effects in disease-relevant contexts.

Cell type-specific analyses

To identify tissue-specific or cell type-specific enrichment of SNP heritability associated with POAG–T2D comorbidity, we performed stratified Linkage Disequilibrium SCore regression.³⁰ This approach evaluates whether the heritability of complex traits is disproportionately enriched in functional annotations derived from gene expression profiles across different tissues. Specifically, we used tissue-specific gene expression data from the GTEx project and the Franke lab dataset to test whether the heritability of POAG and T2D is enriched in genes that are highly expressed in specific tissues.

Pathway enrichment analysis and gene set analysis

To better understand the underlying biological mechanisms of T2D and POAG or their comorbidity, we performed the gene set enrichment analysis (GSEA) and Cytoscape on the results of MAGMA using the Metascape online tool, using gene sets derived from Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes pathway database, Reactome Gene Sets, Canonical Pathways, CORUM and WikiPathways.^{31 32}

In addition to pathway enrichment analysis conducted using Metascape, we performed gene set analysis with MAGMA to evaluate the potential association between T2D and POAG–associated genetic variants and gene sets from MSigDB,²⁸ specifically those within the GO database.²⁶ Unlike traditional enrichment tools that rely solely on input gene lists, MAGMA leverages full GWAS summary statistics, accounts for LD and implements gene-level association models. This complementary approach enables a more comprehensive and statistically robust interpretation of the polygenic architecture underlying primary POAG and T2D.

MR analysis

Building on the findings from LDSC analyses, we further investigated the causal association between T2D and POAG at the genetic level using two-sample MR analysis. MR analysis employs genetic variants as instrumental variables to estimate the causal association between an exposure and an outcome.³³

For the selection of genetic instruments, we established a stringent threshold of (p<5×10⁻⁸). We used the 1000 Genomes Project Phase 3 (European population) as the reference panel and performed LD clumping to identify independent SNPs with (r²<0.001) within a 10 000 kb window. The inverse variance-weighted (IVW) method served as our primary MR analysis approach, allowing us to obtain a high-powered estimate by regressing the outcome effect coefficients against the exposure effect coefficients.³⁴

To assess the robustness of the IVW estimates against horizontal pleiotropy, we conducted sensitivity analyses using MR-Egger regression and the weighted median approach.^{35 36} MR-Egger regression facilitated the evaluation of horizontal pleiotropy by examining the intercepts of the genetic instruments. Additionally, we performed the MR Steiger directionality test to assess whether the observed genetic instruments explain more variance in the exposure (T2D) than in the outcome (POAG), thereby confirming the inferred causal direction.

All statistical analyses were performed using the ‘TwoSampleMR’ in R (V.4.3.0).

Code availability

We used publicly available software for analysis in this study. Here, we list the URLs for the software: LDSC (https://github.com/bulik/ldsc); FUMA (https://fuma.ctglab.nl/); PLACO (https://github.com/RayDebashree/PLACO); GWAS Catalogue (https://www.ebi.ac.uk/gwas/home); Two-Sample MR (https://mrcieu.github.io/TwoSampleMR/articles/introduction.html).

Results

Genetic correlation between POAG and T2D

In figure 2a, we visualised the cross-trait GWAS results between POAG and T2D. Given the limitations of traditional epidemiological studies in establishing a causal association between T2D and POAG, we explored their association at the genetic level using large-scale GWAS summary statistics. Based on LDSC analysis, we observed a significant genetic correlation between T2D and POAG (p=0.014) (figure 2b and online supplemental data 2).

Figure 2. Results of genetic association between POAG and T2D and results of two-sample Mendelian randomisation. (a) Manhattan plot of pleiotropic analysis results for POAG and T2D. (b) LDSC genetic association results between POAG and T2D. (c) Forest plots of two sample MR results. LDSC, Linkage Disequilibrium Score; MR, Mendelian randomisation; POAG, primary open-angle glaucoma; T2D, type 2 diabetes.

The results of MR analysis

Based on the results of LDSC, we further used two-sample MR to explore the causal association between POAG and T2D. To avoid the bias caused by overlapping samples, we chose the GWAS data from the finn population of POAG for the analysis, and based on the inverse variance weighted as our main analytical method, we found that T2D would increase the risk of POAG (OR=1.09, 95% CI 1.03 to 1.14, p=1.50×10⁻³) (figure 2c and online supplemental data 3). In addition, the results of our sensitivity analysis show that the direction of the effect is consistent with that of IVW, further demonstrating the robustness of our results (MR Egger: OR=1.09; weighted median: OR=1.06) (figure 2c and online supplemental data 3). In addition, the MR Steiger directionality test indicated that reverse causality is unlikely, with a statistically significant result (p value <0.001), supporting the causal direction from T2D to POAG.

Annotation of pleiotropic SNPs using FUMA

In our cross-trait meta-analysis, we identified 178 pleiotropic SNPs shared between T2D and POAG, localised to 79 genes, respectively (online supplemental data 4). The five most significant candidate pleiotropic genes were CCND2, SVEP1, ST6GAL1, TCF7L2 and HMGA2 (online supplemental data 5). Through eQTL mapping based on SNP2GENE in FUMA, we identified 36 genes mapped in optic nerve-relevant brain tissues. These genes are predominantly located on chromosomes 9, 11 and 12. Notably, a cluster of eQTL-mapped genes was observed on chromosome 11, including ABCC8, APIP, SLC39A13, PSMC3, RAPSN, FAM180B, C1QTNF4, MTCH2, AGBL2 and OR5B3, suggesting potential regulatory hotspots contributing to the genetic overlap between POAG and T2D (online supplemental data 6).

Tissue expression analysis based on MAGMA and GTEx transcriptomic data across 53 tissue types revealed that pleiotropic genes between POAG and T2D were predominantly enriched in Brain_Cerebellum, Brain_Cerebellar_Hemisphere and Thyroid tissues (figure 3a). Further cell type-specific expression analysis demonstrated significant enrichment in Neural Stem Cells, Adipose_Subcutaneous, Thyroid, Pancreas, and Lung (online supplemental data 7), suggesting a diverse tissue involvement underlying the shared genetic architecture.

Figure 3. Tissue enrichment results and pathway enrichment results for pleiotropic genes. (a) MAGMA Tissue Expression Analysis. (b) Bar graph of enriched terms across pleiotropic genes. (c) Gene ontology (GO) enrichment analysis of MAGMA-prioritised genes. Each dot represents a significantly enriched GO term. Dot size indicates the number of associated genes; the x-axis shows –log10(P-value). MAGMA, Multi-marker Analysis of GenoMic Annotation.

These findings highlight the potential role of neurodevelopmental and neurodegenerative processes (as implicated by neural stem cell and brain tissue enrichment), alongside metabolic and endocrine regulation (as indicated by adipose, pancreas and thyroid enrichment), in mediating the comorbidity between POAG and T2D. Collectively, these results underscore the relevance of both nervous system and metabolic tissues in the pleiotropic mechanisms linking POAG and T2D.

Pathway enrichment results

The pathway enrichment results revealed that pleiotropic genes were significantly enriched in pathways related to Human T-cell leukaemia virus 1 infection, insulin secretion, negative regulation of small molecule metabolic processes, regulation of epithelial cell migration and regulation of binding (figure 3b). Notably, pleiotropic genes were also enriched in the retinal morphogenesis in lens-based eyes pathway, suggesting potential involvement in developmental processes critical for retinal integrity. Retinal morphogenesis is highly dependent on intercellular communication and microenvironmental homeostasis; disruptions in these processes may underlie retinal vulnerability in both T2D and POAG. These findings imply that both systemic metabolic dysfunction and local retinal regulatory disturbances could contribute to shared pathological mechanisms.

Further GSEA using MAGMA showed significant enrichment of pleiotropic genes in negative regulation of neuron death, regulation of insulin secretion, insulin secretion and endoplasmic reticulum calcium ion homeostasis (figure 3c). These results strengthen the hypothesis that pleiotropic genes influence neuroprotection, insulin regulation and calcium homeostasis in the endoplasmic reticulum—biological processes that are intimately linked to the pathophysiology of both T2D and POAG.

Discussion

Based on GWAS data related to T2D and POAG, we applied the comprehensive pleiotropic genomic analysis to explore genetic associations and pleiotropic candidate genes between T2D and POAG. The LDSC analysis revealed a significant positive genetic correlation between T2D and POAG. To further explore whether there was a causal association between T2D and POAG at the genetic level, two-sample MR was employed to explore the causal association between these two diseases. Traditional epidemiologic studies are unable to provide a high level of strength of evidence because of confounding factors and limited sample size. In addition, the results of previous traditional epidemiologic studies on T2D versus POAG were contradictory.³⁷ In a cohort study³⁸ from the Netherlands that followed 3837 participants who did not have POAG at baseline for 6.5 years, POAG occurred in only 87 patients, and the relative risk of T2D for POAG (RR=0.65 (0.25–1.64)) was not statistically significant after adjusting for confounders. A subsequent comparative cross-sectional study from Nepal (n=189)³⁹ attempted to explore the statistical association between T2D and POAG, and still no statistically significant results were observed (p=0.757). Interestingly, a recent large prospective cohort study (n=11 260)⁴⁰ utilising multivariate logistic regression observed that untreated T2D was associated with a higher risk of POAG (OR=1.50 (1.06–2.13)). This suggests that sample size may be one of the most significant factors influencing the results of traditional epidemiology, with a lack of sample size limiting its ability to detect causal associations. Based on this, we used GWAS data with the largest sample to date as well as MR to provide a better strength of evidence.

In our cross-trait meta-analysis, we identified 178 pleiotropic SNPs shared between T2D and POAG, localised to 79 genes, respectively. The five most significant candidate pleiotropic genes were CCND2, SVEP1, ST6GAL1, TCF7L2 and HMGA2. CCND2 is the gene encoding the cell cycle protein D2, which promotes the step in the cell cycle known as the G1-S transition and is a key controller of cell growth and division (proliferation) rates in vivo.⁴¹ Notably, in mouse experiments, cell cycle protein D2 exerts an extremely significant effect on pancreatic islet β-cell division and proliferation, with a threefold reduction in β-cell mass in cell cycle protein D2-/-mice compared with normal mice.^{42 43} This suggests that the CCND2 gene may influence T2D risk by controlling islet β-cell proliferation. In addition, recent reviews have summarised the effects of CCND2 on human neurological development, particularly glial cell proliferation,^{44 45} and neuroinflammation and glial cells are key factors in the pathogenesis of POAG.⁴⁶ Therefore, CCND2 may play an important role in mediating the genetic comorbidity between T2D and POAG.

SVEP1 is a large extracellular chimeric protein with protein interaction and adhesion functions.⁴⁷ Previous MR studies have shown that increased SVEP1 protein significantly increases the risk of developing T2D (p=0.0004).⁴⁸ In addition, human genetic or proteomic studies have identified an association between SVEP1 with glaucoma.⁴⁹ ST6GAL1 adds α2–6-linked sialic acid to N-glycosylated proteins and is overexpressed in a variety of human malignancies.⁵⁰ Numerous studies have shown that ST6GAL1 variants are significantly associated with the risk of T2D.51,53 However, the association between ST6GAL1 and POAG has not been reported. It is noteworthy that ST6GAL1 regulates angiogenesis through the regulation of integrins and PECAM,⁵⁴ and the most obvious factors in the development of optic neuropathy in POAG include impaired regulation of ocular blood flow.⁴⁶ This implies that ST6GAL1 is likely to affect POAG by regulating hemodynamics. TCF7L2, the most potent motif for T2D risk, directly regulates the expression of genes involved in lipid and glucose metabolism and functions as a central transcriptional regulatory molecule in the metabolic programme of adipocytes.⁵⁵ Numerous animal studies^{56 57} and population epidemiologic studies⁵⁸ have confirmed the association of variants and low expression of TCF7L2 with a high risk of T2D. However, there is still limited understanding of the association between the TCF7L2 gene and POAG. Our study localised the TCF7L2 gene as the shared genetic basis of T2D-POAG, which we assume is mainly due to the crucial role of TCF7L2 in the Wnt signalling pathway. On the one hand, the Wnt signalling pathway must be realised through the cell content-dependent expression of TCF7L2 and its numerous isoform variants.⁵⁵ On the other hand, the Wnt pathway alters intraocular pressure by regulating adhesion junctions and cell contacts in human trabecular meshwork cells, which in turn affects the pathogenesis of POAG.⁵⁹ To date, the genetic association between HMGA2 and T2D has been confirmed in GWAS studies of Chinese Mongolians and African Americans,60,62 which may be related to the link between the HMGA2 gene and pancreatic islet dysfunction.⁶³ Downstream of the T2D-POAG genetic association, the important role of HMGA2 in eye development may explain the pathogenesis of POAG.⁶⁴ Moreover, HMGA2 has been reported as a candidate susceptibility gene for POAG in a previous GWAS study of the Japanese population.⁶⁵

Our pathway and gene set enrichment analyses provide mechanistic insights into the shared genetic basis of POAG and T2D. Pleiotropic genes were enriched in pathways related to insulin secretion, small molecule metabolism, epithelial cell migration and retinal morphogenesis—biological processes critical to both metabolic regulation and ocular development. Notably, the enrichment of genes involved in retinal morphogenesis in lens-based eyes suggests that dysregulation of retinal development and microenvironmental homeostasis may contribute to the susceptibility of retinal tissues in both diseases.

Further enrichment in gene sets associated with negative regulation of neuron death, regulation of insulin secretion and endoplasmic reticulum calcium ion homeostasis supports the hypothesis that pleiotropic genes exert effects through neuroprotective and metabolic pathways. These pathways are central to maintaining neuronal integrity and systemic metabolic balance, implying that disruptions in shared cellular regulatory mechanisms—particularly in the retina, pancreas and nervous system—may underlie the comorbidity of POAG and T2D. Together, these findings underscore the importance of integrated metabolic and neurobiological processes in the pathogenesis of these complex diseases.

Several limitations of our study must be recognised. First, we focused only on data with European ancestry; therefore, large GWAS datasets for T2D and POAG in other ethnic populations are needed. Second, our analysis explored the association of traits only from a genetic perspective; further consideration of the influence of environmental and psychological factors is needed.

Conclusion

In this study, we conducted a comprehensive genome-wide cross-trait analysis using the largest available GWAS dataset for POAG to date, aiming to elucidate its genetic architecture and causal association with T2D. Our findings demonstrate a significant shared genetic basis between POAG and T2D, with MR analyses supporting a causal effect of T2D on POAG, but not vice versa. Through PLACO, we identified 178 pleiotropic SNPs mapped to 79 genes, with CCND2, SVEP1, ST6GAL1, TCF7L2 and HMGA2 emerging as top candidate pleiotropic genes. Furthermore, eQTL mapping revealed 36 genes with regulatory relevance in optic nerve-related brain tissues.

Collectively, our integrative analyses suggest that these pleiotropic genes are predominantly involved in neurodevelopmental, neuroprotective and metabolic pathways. Tissue-specific and cell-type-specific expression enrichment in neural, pancreatic, adipose and retinal tissues further underscores the role of both neuronal integrity and metabolic regulation in mediating the comorbidity between POAG and T2D. These findings provide novel insights into the shared genetic mechanisms and potential biological pathways linking these two complex diseases. Incorporating pleiotropic loci into genetic risk models may enhance early identification of high-risk individuals, while tissue-specific regulatory markers could inform precision diagnostics and therapeutic strategies for POAG in the context of metabolic disorders such as T2D.

Moreover, in the context of complex diseases like POAG and T2D, AI (artificial intelligence)-driven approaches may facilitate the interpretation of large-scale genomic datasets, enhance the identification of pleiotropic mechanisms and enable the development of personalised risk prediction models, all while maintaining compliance with evolving data governance standards.⁶⁶ Integrating these technologies with pleiotropy-informed genetic insights may accelerate the translation of our findings into clinical applications.

Data availability statement

Data are available in a public, open access repository. All data relevant to the study are included in the article or uploaded as supplementary information.