Shared Genetic Architecture Between Psychiatric Disorders and Keratoconus: Insights From a Large-Scale Genome-Wide Cross-Trait Analysis

Supplemental Digital Content is Available in the Text.

Abstract

Purpose:

The correlation between keratoconus (KC) and psychiatric disorders has been reported, but whether there is a genetic correlation between the 2 remains unclear.

Methods:

We identified genetic overlaps between KC and 6 psychiatric disorders by using comprehensive summary data from genome-wide association studies. Cross-trait pleiotropic analysis uncovered shared loci and genes, whereas functional annotations and tissue-specific investigations assessed the impact of these pleiotropic genes. Finally, bidirectional Mendelian randomization was applied to explore causal relationships.

Results:

Significant genetic correlation and genetic overlap were observed in 3 of the 6 trait pairs. A multiplicity study using the composite null hypothesis discovered significant possible multiplicity single nucleotide variants across 3 trait pairs, revealing 13 pleiotropic genetic loci, of which 3 were confirmed as colocalized loci with a posterior probability (PP.H4) exceeding 0.75. Significantly, numerous pleiotropic loci were identified across various paired traits, including 11p15.5 and 9q34.3. At the gene level, 137 pleiotropic genes were identified, including PDDC1, PIDD, and PNPLA2. Gene-based analyses indicated that these genes were significantly enriched in the brain and fibroblasts. Pathway analysis highlighted critical roles in Wnt/β-catenin signaling pathway. Mendelian randomization analysis further analyzed the correlation between these diseases.

Conclusions:

This study offers additional evidence of a multifaceted relationship between psychiatric disorders, specifically attention-deficit/hyperactivity disorder, anorexia nervosa, and posttraumatic stress disorder, and KC. At the same time, it may facilitate novel medications and enhanced clinical treatment techniques.

Keratoconus (KC) is linked to detrimental effects such as myopia, irregular astigmatism, corneal scarring, and varying degrees of visual impairment.^1,2The prevalence of KC in the general population is 1.38 per 1000³ although epidemiological studies have revealed significant variability in this prevalence.⁴ A cross-sectional study in Saudi Arabia reported a notably high prevalence of 4.79% for KC in that region.⁵ Epidemiological studies have revealed genetic predisposition, eye rubbing, and UV light exposure as significant risk factors for developing KC.⁴ Individuals with a familial predisposition to KC have up to a 67-fold higher risk of acquiring the condition than the general population.⁶

Several studies have shown a correlation between KC and psychiatric disorders.^7–10 Recently, a large-scale cross-sectional investigation conducted in Israel has revealed an elevated prevalence of KC in individuals with attention-deficit/hyperactivity disorder (ADHD),¹¹ which may be associated with somatization behaviors such as eye rubbing. There exists a significant knowledge gap regarding the potential genetic link and causal relationship between KC and psychiatric disorders.¹² This knowledge gap is particularly critical in refractive surgery decision making. Many patients seeking surgery for myopia or astigmatism may be undiagnosed KC cases or high-risk individuals. If certain psychiatric disorders indicate genetic susceptibility to KC, understanding their genetic correlation would affect preoperative screening and postoperative management. Nonetheless, conventional epidemiological studies may not be sufficient in presenting statistical validity on these subjects.

The High-Definition Likelihood (HDL) approach and the Linkage Disequilibrium Score Regression (LDSC) method,¹³ using pooled data from genome-wide association studies (GWAS), are frequently employed to examine genetic associations among various disorders. Pleiotropic Analysis under Composite Null Hypothesis (PLACO) effectively decreases the false positive rate by considering the influence of covariates, thereby more precisely pinpointing shared genetic etiologies of complex disease signals.¹⁴ The cross-trait analysis employs the correlation among GWAS signals to investigate pleiotropic genetic variant loci across various traits or diseases,^15,16 which may concurrently act as a means for interventions for preventing or treating these conditions.^17,18 Using the most recent GWAS study methodologies, we comprehensively evaluated the genetic correlation, common susceptibility genes, and possible effector connections between KC and psychiatric disorders.

METHODS

GWAS Summary Data Sources

We used a large-scale publicly available GWAS summary statistics for 6 psychiatric disorders and KC. The psychiatric disorder data were obtained from the Psychiatric Genomics Consortium (PGC, https://pgc.unc.edu) and are all based on European populations, including anorexia nervosa (AN, 16,992 cases and 55,525 controls),¹⁹ ADHD (19,099 cases and 34,194 controls),²⁰ bipolar disorder (BIP, 41,917 cases and 371,549 controls),²¹ schizophrenia (SCZ, 11,260 cases and 24,542 controls),²² major depressive disorder (MDD, 30,603 cases and 143,916 controls),²³ and posttraumatic stress disorder (PTS, 23,212 cases and 151,447 controls).²⁴ All studies adhered to identical quality control protocols, employing logistic regression to evaluate relationships between primary characteristics and SNP genotypes in each study while including genetic principal components as covariates in the association analysis. Risk estimates were compiled using fixed-effects inverse variance weighted (IVW) meta-analysis. The GWAS summary data for KC were obtained from the large-scale multiethnic study by Hardcastle et al.²⁵ The discovery cohort comprised 2116 clinically confirmed cases and 24,626 controls, all of European ancestry. In the source study, cases were identified through clinical examination and corneal imaging (eg, Orbscan or Pentacam systems), and controls were defined as individuals without any known corneal disease (Supplemental Table 1, Supplemental Digital Content 1, http://links.lww.com/ICO/B919).

Quality Assurance

This work employed quality control measures to guarantee the accuracy and reliability of the GWAS data. Initially, we omitted SNPs (single nucleotide polymorphisms) located in the MHC (major histocompatibility complex) area, specifically the 25 Mb to 35 Mb segment of chromosome 6, from our analysis.¹⁸ The reason being that this region is omitted from most GWAS studies because of the intricate gene architecture and elevated linkage disequilibrium, which can result in false-positive outcomes. Second, we established a filtering threshold for minor allele frequency (MAF) to mitigate the influence of rare variants. We retained only those with a MAF exceeding 0.01 for all SNPs analyzed, indicating a frequency greater than 1%. This filtering process prioritized common variants, enhancing statistical power and minimizing the level of false positives.

Genome-Wide Susceptibility to Genetic Correlation

We employed the LDSC approach to evaluate the shared genetic architecture between psychiatric disorders and KC.²⁶ The LD scores used in the LDSC analysis were derived from common SNP genotypes of European pedigree samples from the 1000 Genomes Project.²⁷ The standard error was calculated based on the leave-one-out approach in LDSC and subsequently employed to adjust for attenuation bias. The LDSC intercept was used to evaluate probable population overlap between studies.²⁸ Our study highlights the absence of population overlap across psychiatric disorders and KC research, enhancing the findings' reliability. To strengthen the validation of the LDSC results, we implemented the HDL technique, a likelihood-based analytical tool.¹³ HDL facilitates a more efficient use of GWAS summary statistics for estimating genetic correlations. Compared with the LDSC technique, HDL reduces the variance of gene association estimates by approximately 60%, significantly enhancing the precision and robustness of the forecast. The validation of HDL guarantees the reliability of genome-wide genetic overlap analysis.

Hierarchical Examination of Organ Affiliation

To investigate the correlation between KC and psychological illnesses across multiple tissues and organs, we explored the enhancement of SNP heritability for psychosis and KC in specific cells and tissues. Expression profiles from 54 human tissues, including cultured fibroblasts and EBV-transformed lymphocytes, were obtained from the GTEx database,²⁹ which does not include corneal tissue. Therefore, we analyzed all available tissues to comprehensively capture cross-tissue transcriptional patterns that might mediate shared genetic effects between KC and psychiatric disorders.

Identification of Multiplexed Loci and Genes Using PLACO

In this study, we employed the PLACO to systematically present the genetic associations between psychiatric disorders and KC at the SNP level. PLACO is a statistical technique specifically formulated to detect genetic pleiotropy, adept at identifying shared genetic variants across multiple phenotypes.¹⁴ This method enables fast identification of SNPs with significant connections across many diseases. Specifically, we regard SNPs that achieve genome-wide significance levels (P < 5 × 10⁻⁸) as pleiotropic variations. This indicates that these SNPs exhibit robust genetic correlations across several phenotypes, signifying their significance in the progression of these disorders. Identifying these pleiotropic polymorphisms is crucial for uncovering the common genetic foundation linking various psychiatric disorders and KC. Furthermore, to enhance the validation of the biological relevance of these pleiotropic SNPs, we conducted genomic region localization of these risk variants using the Functional Mapping and Annotation Tool (FUMA).³⁰ FUMA enabled the assignment of SNPs to distinct genomic regions (ie, risk loci), thereby offering insights into the potential functions of these variants. Also, Bayesian colocalization analysis was conducted to uncover loci associated with psychiatric disorders and KC.³¹

Gene-Level Exploration Analysis

To investigate the sharing mechanism of the discovered loci, we mapped adjacent genes according to the predominant SNP in each locus. In addition, a multimarker effects analysis of the GWAS data was performed using the method of generalized gene set analysis (MAGMA) to determine the biological functions of these pleiotropic loci. Specifically, MAGMA gene analyses were performed to identify pleiotropic genes by appropriately combining linkage disequilibrium (LD) between markers and by detecting multimarker effects with significance levels adjusted by Bonferroni correction. The MAGMA gene set analysis was conducted to investigate the biological function of the predominant SNPs,³² finally evaluating gene sets from the Molecular Marker Database (MSigDB), which includes aggregated gene sets (c2.all) and Gene Ontology terms (c5.bp, c5.cc, and c5.mf).³³ To prevent false positive outcomes, all examined gene sets underwent Bonferroni correction. A genome-wide tissue-specific enrichment analysis was performed based on 54 GTEx tissues to obtain genome-wide pleiotropic results generated by PLACO. Furthermore, the average expression (log2 transformed) of all identified pleiotropic genes in all 54 GTEx tissues was calculated, and a test for tissue specificity was conducted by examining differentially expressed genes (DEGs) in each tissue (the sign of the t-statistic was precalculated for up-regulated and down-regulated DEGs).

Exploration of Potential Pharmacological Targets in European Populations

The summary data–based Mendelian randomization (SMR) approach integrates summary-level data from GWAS with data from expression quantitative trait loci (eQTL) studies to identify gene expression levels associated with complex traits because of pleiotropy.³⁴ eQTLs denote genetic variants that are linked with gene expression levels; specifically, eQTLs elucidate individual variations in gene expression. eQTL studies identify connections between specific SNPs and gene expression levels, indicating which genetic variants will likely influence gene expression magnitude.

The SMR technique integrates eQTL data with GWAS data to investigate the various processes by which SNPs influence complex characteristics, including diseases. The methods of SMR and HEIDI (Heterogeneity in Dependent Instrument) are used to assess pleiotropic associations between levels of gene expression and complex traits. The primary objective of the SMR analysis is to determine if changes influence the extent of the phenotypic effect of an SNP in gene expression. If SNPs are linked to both gene expression and complex traits, a pleiotropic phenomenon may exist, indicating that the gene could significantly contribute to the genetic foundation of these traits. The HEIDI test further investigates whether this association arises from colocalization, specifically whether the SNP's influence on gene expression originates from the same causal variation as its impact on the complex trait. The HEIDI test results suggest that the association arises via colocalization among various loci instead of a singular pleiotropic effect. In that case, this will enhance our comprehension of the genetic mechanisms governing these complex features.

The SMR technique facilitates the identification of genes linked to KC and psychiatric disorders while also elucidating the regulatory mechanisms connecting genetic variation and phenotype, offering significant insights for target development.

Association Analysis

A bidirectional two-sample Mendelian randomization (MR) analysis was employed to assess the potential causative impact of psychiatric disorders on the risk of KC. In this study, we identified all independently significant SNPs as exposure traits (P < 5 × 10⁻⁸) using the “clumping” function of PLINK 1.9 software, with r² set to 0.001 and a window size of 10,000 KB. The reference panel used for the r² estimate was derived from the 1000 Genomes Project Phase III. The F-statistic (F > 10) was employed to assess the robustness of the instrumental variables (IVs).³⁵ To ascertain the association between these trait pairs, we employed 5 MR approaches for each set of instrumental variables: IVW,³⁶ weighted median method,³⁷ Wald ratio, simple mode, and MR-Egger. Cochran Q statistic was employed to assess the heterogeneity of effect sizes among instrumental factors. Furthermore, a global intercept test for MR-Egger regression was used to identify the extent of horizontal pleiotropy. Figure 1 illustrates the flowchart of this investigation.

FIGURE 1.

Study workflow. Genetic correlation and two-sample Mendelian randomization analyses were performed between KC and psychiatric disorders (PSY). Subsequently, PLACO was applied to conduct multitrait meta-analysis across the trait pairs. Further SNP-level and gene-level analyses were performed to investigate the shared genetic architecture, and colocalization analyses were conducted to identify overlapping loci.

RESULTS

Genetic Correlation Between Psychiatric Disorders and KC

Initially, an evaluation was conducted to ascertain the genetic correlation between 6 psychiatric disorders and KC. The findings from the LDSC and HDL studies are presented in Table 1 (Supplemental Table 2, Supplemental Digital Content 2, http://links.lww.com/ICO/B920). Using the LDSC technique, only ADHD was found to have a genetic association with the KC (rg = 0.1636, P = 4.00 × 10⁻⁴). Substantial genetic associations were identified between ADHD (rg = 0.2429, P = 4.82 × 10⁻⁵), AN (rg = 0.1766, P = 3.77 × 10⁻³), PTS (rg = 0.3209, P = 1.02 × 10⁻²), and KC using the HDL approach. The integrated results will focus on the associations between ADHD, AN, PTS, and KC.

TABLE 1.

Genetic Correlation Between Psychiatric Disorders and KC

Trait Pairs	LDSC		HDL
	rg (SE)	P	rg (SE)	P
KC&AN	0.078 (0.047)	9.49E-02	0.177 (0.061)	3.77E-03
KC&ADHD	0.164 (0.046)	4.00E-04	0.243 (0.06)	4.82E-05
KC&BIP	−0.026 (0.038)	4.91E-01	0.002 (0.059)	9.71E-01
KC&SCZ	−0.018 (0.03)	5.50E-01	−0.002 (0.04)	9.55E-01
KC&MDD	−0.042 (0.071)	5.55E-01	0.009 (0.109)	9.38E-01
KC&PTS	0.083 (0.083)	3.16E-01	0.321 (0.125)	1.02E-02

BIP, bipolar disorder; MDD, major depressive disorder; SCZ, schizophrenia.

Results of Tissue-specific Associations

A subsequent investigation was conducted to see if SNP heritabilities were enriched in specific tissues for 3 psychiatric disorders (ADHD, AN, PTS) and KC, using the S-LDSC approach. S-LDSC was used on pooled GWAS data from several tissue organs to evaluate the significance of genetic enrichment for distinct trait types across different tissues. We acquired a dataset of 54 human tissues from GTEx. After refining the baseline model and all gene sets, we assessed SNP heritability enrichment across various tissues and cell types by analyzing regression coefficient z-scores and their corresponding P-values. Subsequent analysis of tissue-specific risk loci revealed their prevalence in numerous tissues (eg, brain, endocervix, esophagogastroesophageal junction) (Supplemental Figure 1, Supplemental Digital Content 3, http://links.lww.com/ICO/B921; Supplemental Table 3, Supplemental Digital Content 4, http://links.lww.com/ICO/B922).

Identification of Pleiotropic Loci and Genes Associated With Various Psychiatric Disorders and KC

Using the same genetic processes found by the LDSC and HDL approaches, a novel pleiotropic analytic technique PLACO was employed to pinpoint putative pleiotropic loci in KC associated with these 3 psychiatric disorders (Supplemental Figure 2, Supplemental Digital Content 3, http://links.lww.com/ICO/B921). QQ-plots revealed no initial deviation between predicted and observed values, thereby ruling out the possibility of population stratification (Supplemental Figure 3, Supplemental Digital Content 3, http://links.lww.com/ICO/B921). According to the PLACO data, we found 13 pleiotropic genetic risk loci linked to psychiatric disorders and KC using FUMA (P < 5 × 10^–8) (Figure 2; Supplemental Figure 2, Supplemental Digital Content 3, http://links.lww.com/ICO/B921; Supplemental Table 4, Supplemental Digital Content 5, http://links.lww.com/ICO/B923). The ANNOVAR annotation classified 9 out of 13 SNPs (69.2%) as intronic variants. Colocalization analysis identified 3 out of 13 putative pleiotropic loci (23.1%) with a posterior probability (PP.H4) over 0.7, such as 11p15.5 (Table 2; Supplemental Figure 4, Supplemental Digital Content 3, http://links.lww.com/ICO/B921). Significantly, many pleiotropic regions were common among various genome pairs; for instance, the genomic region 11p15.5 was detected in 3 trait pairs, whereas 9q34.3 was found in 2 trait pairs (Supplemental Table 5, Supplemental Digital Content 6, http://links.lww.com/ICO/B924). The MAGMA analysis of gene sets indicated that the pleiotropic loci identified could be involved in several crucial biological pathways, including S100 protein binding and enrichment at pivotal connections in the Wnt/β-catenin signaling pathway, encompassing TCF-dependent signaling, RUNX3 modulation of Wnt signaling, and the assembly of TCF/LEF and CTNNB1 complexes along with their target gene regulation (Fig. 3A; Supplemental Table 6, Supplemental Digital Content 7, http://links.lww.com/ICO/B925). Further tissue-specific analysis indicated that these risk loci were abundant in fibroblasts and the endocervix (Fig. 3B; Supplemental Table 7, Supplemental Digital Content 8, http://links.lww.com/ICO/B926).

FIGURE 2.

Circular diagram presents pleiotropic loci and genes identified by PLACO among 3 trait pairs. Note: Shared loci identified by colocalization analysis are highlighted in orange; shared genes identified by MAGMA analysis are highlighted in blue.

TABLE 2.

Information of 13 Pleiotropic Loci Identified

Trait Pairs	chr	Region	LeadSNPs	P	PP.H4	Nearest Genes
KC-PTS	9:137413954-137487548	9q34.3	rs3132309	6.43E-10	0.031	LIMK2
KC-PTS	11:784775	11p15.5	rs6597978	4.37E-08	0.033	PDDC1, PIDD, PNPLA2
KC-ADHD	3:171595724	3q26.31	rs9868683	7.59E-09	0.126	TMEM212 FNDC3B
KC-ADHD	5:52478715	5q11.2	rs38056	1.01E-11	0.159	ITGA1
KC-ADHD	9:137427625-137436314	9q34.3	rs3118594	4.24E-08	0.094	LIMK2
KC-ADHD	11:721570-825777	11p15.5	rs7479183	1.82E-18	0.909	PDDC1, PIDD, PNPLA2
KC-ADHD	15:67469335-67674711	15q22.33	rs55994097	2.04E-08	0.027	SMAD3
KC-ADHD	18:22681706-22746547	18q11.2	rs4528662	1.64E-08	0.829	ZNF521
KC-AN	3:49017208-50474574	3p21.31	rs4688759	3.50E-08	0.073	ZNF589
KC-AN	11:750849-825777	11p15.5	rs7479183	2.61E-10	0.153	PDDC1, PIDD, PNPLA2
KC-AN	11:47377283-47883337	11p11.2	rs7104036	1.75E-08	0.781	CD244
KC-AN	15:67439629-67707833	15q23	rs62006055	4.97E-10	0.095	SMAD3
KC-AN	20:2049747-2129907	20p13	rs6047046	3.25E-08	0.017	STK35

Lead SNP was the SNP with minimum P values within the corresponding locus.

PP.H4 was the posterior probability of H4 calculated by coloc analysis; the Locus boundary was defined as “chromosome: start–end.”

FIGURE 3.

Bar plot of MAGMA gene-set (A) and tissue-specific (B) analysis for genome-wide pleiotropic results. Note: The red dotted line represents the significance of 0.05 after multiple corrections, and the blue represents the significance of 0.05.

We used different approaches to correlate the detected SNP-level signals with gene-level signals. A total of 137 genes were identified as significant pleiotropic genes shared between various psychiatric disorders and KC, with 14 being unique (Supplemental Table 8, Supplemental Digital Content 9, http://links.lww.com/ICO/B927; Supplemental Figure, Supplemental Digital Content 3, http://links.lww.com/ICO/B921). Supplemental Table 9 (Supplemental Digital Content 10, http://links.lww.com/ICO/B928) provides detailed information on these genes. The MAGMA gene analysis revealed 22 pleiotropic genes that are associated with diverse traits. Among these, PDDC1 and PNPLA2 were recognized as pleiotropic genes for 3 disease pairs, whereas ILMK2 and SMAD3 were identified for 2 disease pairs. eQTL analysis revealed multiple loci associated with relevant genes in brain and eye tissues (eg, PNPLA2). Figure 4 provides a summary of the distribution of pleiotropic genes identified through diverse approaches and across multiple tissues. Notably, certain genes, such as PIDD and PNPLA2, demonstrated significant localization in various tissues when analyzed using different methodologies (Supplemental Table 10, Supplemental Digital Content 11, http://links.lww.com/ICO/B929).

FIGURE 4.

Overview of pleiotropic genes for the psychiatric disorders and KC. The signals represent hits of genes across different trait pairs.

The shared mechanisms between psychiatric disorders and KC may involve particular organs or tissues. Numerous genes, such as CEND1, ITGA1, PIDD, and SMAD3, demonstrated significant differences in expression within brain tissue (Supplemental Figure 6, Supplemental Digital Content 3, http://links.lww.com/ICO/B921; Supplemental Table 11, Supplemental Digital Content 12, http://links.lww.com/ICO/B930). Enrichment analysis further demonstrated that these genes exhibited a significant association with the brain amygdala, whole blood, and thyroid (Supplemental Figure 7, Supplemental Digital Content 3, http://links.lww.com/ICO/B921).

Association Analysis

Mendelian randomization analysis employing the IVW approach indicated no significant causal relationship was detected between psychiatric disorders and KC when KC served as the outcome trait. Inverse Mendelian randomization revealed a strong association (P < 0.05) between KC and psychiatric disorders (AN and PTS). Specifically, as the risk of KC escalated, the risk of AN also climbed (OR = 1.050, 95% CI, 1.022–1.079, P < 0.01). A causal relationship between the onset of KC and the risk of PTS was observed (OR = 1.031, 95% CI, 1.006–1.057, P = 0.016) (Supplemental Figure 8, Supplemental Digital Content 3, http://links.lww.com/ICO/B921; Supplemental Table 12, Supplemental Digital Content 13, http://links.lww.com/ICO/B931).

DISCUSSION

This work, using a comprehensive genetic approach, identified some genetic connections between psychiatric disorders and KC, particularly ADHD, AN, and PTS. Although the MR analysis indicated modest associations between KC and psychiatric disorders such as AN and PTS, the effect sizes were small and should not be interpreted as clinically actionable risk estimates. Instead, these results primarily highlight potential shared genetic susceptibility.

We discovered several genetic risk loci linked to psychiatric disorders and KC, some of which are present in numerous phenotypic pairs (eg, 11p15.5, 9q34.3).^38–41 Prior research had demonstrated that these loci were crucial to the etiology of psychiatric disorders and KC. The gene 11p15.5 (PNPLA2, PIDD) has been documented as related to KC and psychiatric disorders in multiple GWAS studies.^25,42 PNPLA2, which is implicated in lipid metabolism, is a crucial enzyme in the initial step of triglyceride hydrolysis and exhibits significant expression in ocular tissues. The precise pathophysiology of PNPLA2's role in KC remains unclear, with the hypothesis suggesting that it may facilitate the development of KC by destabilizing the corneal architecture via the overexpression of the antisense RNA gene AP006621.⁴³ Although direct evidence linking PNPLA2 to psychiatric disorders remains unverified, research indicates that mutations in PNPLA2 can result in brain lipid metabolism disorders, causing abnormal neutral lipid accumulation, modifying fatty acid composition, and affecting lipid transport in essential areas like the blood–brain barrier and choroid plexus.^44,45 In addition, dysregulation in the production of stress granules may enhance PNPLA2 expression and disrupt cellular stress responses.⁴⁶ The mechanisms above may establish a potential pathological foundation for psychiatric disorders by inducing neuroinflammation and structural alterations in the brain. The PIDD gene is crucial in various biological processes by modulating apoptotic and antiapoptotic pathways through its pivotal function in the PIDDosome complex.⁴⁷ PIDD has been implicated in the pathogenesis of Fuchs endothelial corneal dystrophy by facilitating the apoptosis of corneal endothelial cells.⁴⁸ Furthermore, PIDD1 plays an essential role in the proper gyrification of the developing human neocortex, which is vital for normal cognitive function and behavior.⁴⁹

Our analysis identified that KC has genetic correlations with psychiatric disorders in the 9q34.3 region, specifically including the significant SNP rs3118594 and its adjacent gene ILMK2. The 9q34.3 area was substantially correlated with central corneal thickness (CCT),³⁸ which serves both as a heritable risk factor and as a core component of the pathophysiology of KC. Further analysis revealed that rs3118594 was correlated significantly with CCT in the Latino population, thereby influencing the risk of KC.⁵⁰ Copy Number Variation (CNV) in the 9q34.3 area was linked to psychiatric disorders and neurodevelopmental characteristics, indicating a shared genetic basis between the 2.⁵¹ LIMK2 is crucial for synaptic transmission, plasticity, and memory formation.⁵² Alterations in its signaling are linked to various neurodevelopmental and neurodegenerative disorders, such as autism, intellectual disability, and Alzheimer disease.⁵³ In addition, LIMK2 is instrumental in cell migration by regulating cytoskeletal dynamics and may preserve corneal architecture and functional modulation.⁵⁴ These pieces of evidence collectively that genes in the 9q34.3 area may affect KC and behavioral disorders via many processes.

In addition, our data indicate that several genes previously linked to KC are also associated with psychiatric disorders. SMAD3 (15q22.33) plays a crucial function in KC by modulating the integrity and proper assembly of the corneal extracellular matrix (ECM) through the regulation of the TGF-β signaling pathway.⁵⁵ SMAD3 expression was considerably down-regulated after corneal collagen cross-linking therapy, which enhanced corneal stromal stability and improved the biomechanical capabilities of KC.⁵⁶ Moreover, SMAD3 was found to be significantly linked to regulatory areas of the corneal stroma in a Corneal Remodeling Factor (CRF)-related GWAS investigation, hence, reinforcing its functional relevance in KC.⁵⁷ Notably, SMAD3 exhibits a hypomethylation state after corneal refractive surgery, suggesting its involvement in postoperative repair processes.⁵⁸ These findings indicate that SMAD3 variants could serve as potential biomarkers for preoperative KC risk assessment. A potential association between SMAD3 and psychiatric disorders has been documented, particularly in neurodegenerative diseases linked to TGF-β signaling, where SMAD3 may be implicated in dopaminergic neurodegeneration and the modulation of GABA neurotransmission.⁵⁹ Consequently, SMAD3 may be significant in the shared etiology of KC and psychiatric disorders.

The prevalent genetic architecture identified in our research indicates a shared link between psychiatric disorders and KC. The discovered genes participate in multiple pathways, including the Wnt/β-catenin signaling pathway and S100 protein binding. Prior research had indicated that the Wnt pathway linked to KC is also correlated with psychiatric disorders in this study.⁶⁰ Key components of the Wnt/β-catenin signaling pathway play a role in the pathogenesis of both psychiatric disorders and KC, encompassing TCF-dependent signaling, the modulation of Wnt signaling by RUNX3, the assembly of TCF/LEF and CTNNB1 complexes, and the regulation of their target genes. This pathway is essential for cell proliferation, differentiation, development, and apoptosis, and for maintaining stem cell function and adult tissue homeostasis. Wnt signaling pathways are classified into 2 primary categories: canonical (β-catenin–dependent) and noncanonical (β-catenin–independent) pathways. We identified substantial enrichment of fibroblast pleiotropy for each disease pair. The activation of the Wnt/β-catenin pathway enhances the levels of active β-catenin in corneal fibroblasts, thereby promoting the production of α-smooth muscle actin (α-SMA), collagen 1 (COL1), and fibronectin (FN), indicating its significant role in the fibrosis process.⁶¹ Moreover, conventional Wnt signaling exerts a substantial influence on neurodevelopment. During the initial phases of neurogenesis, Wnt signaling facilitates the self-renewal of neural progenitor cells; conversely, in the intermediate and later stages, it triggers the differentiation of progenitor cells.⁶² Multiple studies indicate a strong correlation between the Wnt/β-catenin signaling system and ADHD,^20,63,64 demonstrating its involvement in ADHD and related symptoms at both genetic and transcriptome levels.⁶⁵ This study identified a shared mechanism between KC and psychiatric disorders via MAGMA gene pathway enrichment analysis, concluding that the Wnt/β-catenin signaling pathway is integral to this process. S100 protein–related pathways have been linked to KC and PTS, with S100A11 overexpression activating the Wnt/β-catenin pathway.⁶⁶ The S100 protein family comprises small calcium-binding proteins that play roles in proliferation, differentiation, apoptosis, Ca²⁺ homeostasis, and energy metabolism through interactions with various target proteins, and in inflammation and migration/invasion.⁶⁷ The expression of S100-A4 protein in the corneal epithelium of patients with KC was found to be up-regulated 1.9-fold compared with normal corneas,⁶⁸ indicating that its increased synthesis may compensate for the reduced number of epithelial cells and corneal thinning caused by degradative enzymes, thereby contributing to the pathogenesis of KC.⁶⁹ PTS is significantly linked to increased blood levels of S-100B,^70,71 predominantly synthesized by astrocytes, serving as a biomarker for cerebral injury while also exhibiting neurotrophic and neuroprotective properties.^72–75

The study still faces some limitations. The data used in this study are at the aggregate level, which restricts our capacity to further stratify the population by characteristics such as sex, age, or other demographics. Second, our analyses were confined to populations of European ancestry, indicating that the results may not be relevant to other populations or ancestries. The small sample size of KC in this investigation may have constrained statistical efficacy, suggesting that we should use caution in interpreting the data. In addition, in line with standard GWAS quality control procedures, we excluded the MHC region because of its complex linkage disequilibrium structure. We recognize this as a limitation, given the established epidemiological associations between KC and atopic diseases. Future studies employing methods capable of handling the genetic complexity of the MHC region may yield further insights.

Nevertheless, despite these limitations, our findings provide important implications for both clinical practice and translational medicine. Within the context of refractive surgery, the association between psychiatric disorders and the risk of KC carries significant implications for surgical decision making and patient management. Psychiatric conditions often precede the clinical onset of KC,⁷⁶ and many patients with KC initially present only with subclinical changes, seeking surgery for myopia or astigmatism without being identified as high risk. Previous studies indicate that AN, often accompanied by nutritional deficiencies, can affect corneal health and postoperative healing.⁷⁷ ADHD has been associated with corneal ulcers, keratitis, and pupillary abnormalities,^78,79 whereas PTS is linked to dry eye disease and neurotrophic keratopathy.^80,81 The above findings suggest that patients with psychiatric disorders may have an increased risk of postoperative complications; however, these conditions do not constitute an absolute contraindication to surgery. On the contrary, integrating genetic susceptibility with clinical assessment of subclinical KC and early signs of psychiatric disorders can provide a more comprehensive preoperative evaluation, enabling surgeons to make better-informed decisions while minimizing risk.

In conclusion, this study not only elucidates the shared genetic basis between KC and psychiatric disorders but also reveals novel therapeutic targets and opportunities to optimize clinical management. The identified pleiotropic genes and pathways—such as PNPLA2, PIDD, and the Wnt/β-catenin signaling pathway—represent strong candidates for drug repurposing and targeted therapy development. Moreover, the observed genetic correlations suggest that patients with KC should be proactively screened for comorbid psychiatric conditions (eg, AN and PTS), and vice versa. Strengthening interdisciplinary collaboration between ophthalmologists and psychiatrists may facilitate earlier detection, more comprehensive risk assessment, and individualized treatment planning. This integrative approach provides a promising framework for future precision medicine research and clinical interventions.

Supplementary Material

cornea-45-1408-s001.xlsx ^{(9.4KB, xlsx)}

cornea-45-1408-s002.xlsx ^{(12KB, xlsx)}

cornea-45-1408-s003.pdf ^{(2MB, pdf)}

cornea-45-1408-s004.xlsx ^{(22.2KB, xlsx)}

cornea-45-1408-s005.xlsx ^{(13.2KB, xlsx)}

cornea-45-1408-s006.xlsx ^{(9.4KB, xlsx)}

cornea-45-1408-s007.xlsx ^{(11KB, xlsx)}

cornea-45-1408-s008.xlsx ^{(10.2KB, xlsx)}

cornea-45-1408-s009.xlsx ^{(16.5KB, xlsx)}

cornea-45-1408-s010.xlsx ^{(107KB, xlsx)}

cornea-45-1408-s011.xlsx ^{(12.7KB, xlsx)}

cornea-45-1408-s012.xlsx ^{(26KB, xlsx)}

cornea-45-1408-s013.xlsx ^{(19.1KB, xlsx)}