Mitochondria-Related Gene BDH1 Implicated in Myopia Risk via Methylation-Regulated Expression: An Integrative Summary-Data Mendelian Randomization Study

Abstract

Purpose

The purpose of this study was to prioritize mitochondria-related causal genes (MRGs) involved in myopia by integrating MRGs-related methylation quantitative trait loci (mQTL), expression quantitative trait loci (eQTL), and protein quantitative trait loci (pQTL) data with myopia genome-wide association study (GWAS) data.

Methods

UK Biobank data were utilized for discovery, and FinnGen data for validation. Summary Data-based Mendelian Randomization (SMR) assessed associations between blood-derived MRG-related QTLs and myopia risk, followed by colocalization analysis to evaluate shared genetic etiology. Multi-omics integration linked methylation, expression, and disease risk.

Results

Our integrative SMR analysis of MRGs prioritized BDH1 as the sole robust candidate. We identified a significant association between increased methylation at a promoter-proximal site (cg02569554) and reduced BDH1 expression (odds ratio [OR] = 0.522, 95% confidence interval [CI] = 0.425–0.641, false discovery rate [FDR] = 8.79 × 10⁻⁹). Although other nominal associations were observed, they did not survive multiple testing correction and are thus considered exploratory.

Conclusions

This study provides integrative genetic evidence that BDH1 may contribute to myopia through methylation-regulated expression. However, the findings remain preliminary due to limited cross-layer consistency and require replication and functional validation.

Translational Relevance

This study prioritizes BDH1 as a mitochondria-related gene that may influence myopia risk via methylation and expression changes, providing a candidate target for mechanistic research and therapeutic intervention.

Introduction

Myopia, commonly known as nearsightedness, is a prevalent refractive error characterized by difficulty seeing distant objects clearly.¹ It affects approximately 30% of the global population,² with higher prevalence rates in East Asian countries, reaching up to 80% to –90% among young adults.³ The clinical features of myopia, including blurred vision for distant objects, eye strain, and headaches, significantly impair quality of life.⁴ More critically, myopia can progress to high myopia, leading to severe ocular complications such as retinal detachment, myopic maculopathy, and glaucoma.⁵ These risks underscore the importance of understanding the etiology of myopia to develop targeted interventions and improve patient outcomes. The etiology of myopia is multifactorial, involving genetic, environmental, and lifestyle factors.⁶ Studies have identified numerous genetic loci associated with myopia,⁷ but the specific causal genes remain largely unclear.

The development of myopia involves axial elongation associated with retinal and choroidal thinning, structural shifts in Bruch’s membrane, and changes in optic disc morphology, resulting in both non-pathological and pathological ocular alterations.⁸ Mitochondria play a central role in cellular energy metabolism in the retina, which is highly dependent on ATP to sustain its functions and physiological processes.⁹ Mitochondrial dysfunction may increase oxidative stress, which in turn damages retinal cells and contributes to various ocular diseases.¹⁰ Mitochondria are also critical in regulating apoptosis,¹¹ maintaining intracellular calcium balance,¹² and participating in redox signaling,¹³ all of which are essential for the structural integrity and functionality of the retina. Any disruption in these processes can have significant implications for eye health.

Accumulating evidence suggests that mitochondrial dysfunction may actively contribute to myopia progression rather than being a secondary consequence. A recent mitochondrial genome-wide association study (GWAS) has identified nine mitochondrial DNA (mtDNA) variants that were significantly associated with high myopia, indicating impaired oxidative phosphorylation and oxidative stress as causal factors.¹⁴ Single-cell RNA sequencing in guinea pigs with lens-induced myopia has shown that dysregulation of mitochondrial respiratory chain genes in cones and Müller glia contributes to myopia progression, whereas restoration of mitochondrial function by electroacupuncture slows disease development.¹⁵ Furthermore, clinical and genetic studies of primary mitochondrial disorders have demonstrated that impaired oxidative phosphorylation, excess production of reactive oxygen species, and apoptosis in energy-demanding retinal cells are central mechanisms of mitochondrial-driven ocular pathology.¹⁶ Together, these findings provide a rationale to systematically investigate mitochondria-related genes (MRGs) in myopia. We therefore hypothesize that genetic variation in MRGs may contribute to myopia development through effects on methylation, gene expression, and protein abundance.

To test this hypothesis, we used Summary Data-based Mendelian Randomization (SMR) analysis.¹⁷ This approach integrates multiple layers of omics data, specifically cis-quantitative trait loci (cis-QTL) for DNA methylation (mQTL), gene expression (eQTL), and protein expression (pQTL), with GWAS data to identify genetic variants that influence molecular traits and link them to myopia. SMR helps distinguish causal relationships from correlations, providing insights into the genetic mechanisms underlying complex diseases.¹⁸ Our workflow began with the collection of MRGs-related blood-based eQTL, mQTL, and pQTL data, which were integrated with myopia GWAS datasets. Colocalization analysis was then conducted to identify shared genetic variants influencing both QTLs and myopia. This design allows us to prioritize candidate MRGs with potential pathogenic or protective roles, thereby providing a basis for mechanistic research and targeted interventions.

Methods

Study Design

The MRGs-related mQTL, eQTL, and pQTL datasets were subjected to SMR analysis to evaluate their associations with myopia GWAS data sourced from public repositories. The results were further examined through colocalization analysis to identify shared causal variants potentially influencing both QTLs and myopia risk. Integration of multi-omics QTL-GWAS analyses was conducted to prioritize candidate genetic variants involved in myopia development. The flowchart of the study is presented in Figure 1.

Figure 1.

Study design. The flowchart illustrates the integration of mitochondria-related gene (MRG) QTL datasets (eQTL, mQTL, and pQTL) with myopia GWAS data from the UK Biobank and FinnGen database. The eQTL dataset includes DNA expression-associated SNPs, the mQTL dataset includes DNA methylation level-associated SNPs, and the pQTL dataset includes protein expression-associated SNPs. These datasets underwent Summary data-based Mendelian Randomization (SMR) analysis. The SMR results were then subjected to colocalization analysis to identify shared causal variants.

Data Sources

The myopia discovery dataset was obtained from the UK Biobank (PheCode 367.1; 1257 cases and 406,530 controls) as part of the Pan-UKB GWAS results released by the Neale Lab¹⁹ and is registered in the GWAS Catalog under accession GCST90435990 (https://www.ebi.ac.uk/gwas/studies/GCST90435990). The validation dataset was sourced from the FinnGen project (release R10; endpoint H7_MYOPIA), comprising 4106 cases and 394,028 controls, as defined by International Classification of Diseases (ICD)-based registry codes (ICD-10 H52.1; ICD-9 3671; and ICD-8 37002). Detailed endpoint information is available through the Risteys portal (https://r10.risteys.finngen.fi/endpoints/H7_MYOPIA; Supplementary Table S1). A total of 1136 MRGs were downloaded from the MitoCarta3.0 database (https://www.broadinstitute.org/mitocarta/mitocarta30-inventory-mammalian-mitochondrial-proteins-and-pathways). Summary data for blood mQTL were derived from a meta-analysis of two European cohorts²⁰: the Brisbane Systems Genetics Study (n = 614) and the Lothian Birth Cohorts (n = 1366). Blood eQTL data were obtained from the eQTLGen consortium,²¹ comprising genetic data on blood gene expression from 31,684 European individuals. Blood pQTL data were sourced from a previous study,²² including 10,708 European individuals.

SMR Analysis

To identify MRGs associated with myopia, we conducted SMR analysis and heterogeneity in dependent instruments (HEIDI) tests using the SMR software tool (SMR version 1.3.1), as previously described.²³ The primary objective was to assess the relationships among MRG methylation, expression, and protein abundance with myopia. The SMR approach integrated summary-level GWAS with mQTL, eQTL, and pQTL data to identify pleiotropic associations that may reflect potential causal relationships. The HEIDI test evaluated whether an observed association is consistent with a single causal variant (pleiotropy) or instead reflects linkage disequilibrium (LD) between distinct causal variants within the same region. Top associated cis-QTLs were selected within a ±1000 kb window around the gene to identify genetic variants. A P value threshold of 5.0 × 10⁻⁸ was applied to ensure robust associations. SNPs with allele frequency differences greater than 0.2 between any 2 datasets, including LD reference samples, QTL summary data, and outcome summary data, were excluded. The maximum allowable proportion of SNPs with allele frequency differences was 0.05 (–diff-freq-prop 0.05). SMR_multi was used to incorporate multiple genetic variants simultaneously, providing a more robust and comprehensive insight into causal relationships.

To investigate regulatory relationships between MRG methylation and transcription, we conducted an integrative analysis combining mQTL and eQTL summary data in BESD format.²⁰ The analysis focused on cis-regions within 2 Mb between methylation sites and gene loci to evaluate whether DNA methylation influences gene expression through shared genetic variants.

In the primary analysis, associations were considered suggestive if they satisfied nominal thresholds of P_SMR < 0.05, P_SMR_multi < 0.05, and P_HEIDI > 0.01. To account for multiple testing across all SMR results, the Benjamini–Hochberg correction was applied. A false discovery rate (FDR) < 0.05 was considered statistically significant.

Colocalization Analysis

To determine if the same genetic variants are responsible for associations observed in both traits, the Bayesian colocalization analysis was used as it accounts for different sources of variation and provides a robust framework compared with other colocalization methods.²⁴ This approach evaluates the probability that a single genetic variant influences both QTL and GWAS traits, thereby providing evidence for shared genetic architecture. Associations meeting nominal thresholds of P_SMR_multi < 0.05, P_SMR < 0.05, and P_HEIDI > 0.01 were subjected to colocalization analysis using the R software package “coloc” to identify shared genetic variants between MRG cis-QTLs and myopia.²⁵ Additionally, SNPs within ±1000 kilobase (kb) of the top cis-QTLs colocalization region windows were analyzed. Five mutually exclusive hypotheses (H0–H4) were tested: no association with either trait (H0), association with only the first trait (H1), association with only the second trait (H2), association with both traits but with different causal variants (H3), and association with both traits sharing the same causal variant (H4). A posterior probability (PPH4) > 0.5 was considered evidence for colocalization.²⁶

Statistical Analysis

All statistical analyses were conducted using R software (version 4.3.0). Manhattan plots and forest plots were generated using the “ggplot2” and “forestplot” packages, respectively. The code for creating SMRLocusPlot and SMREffectPlot was sourced from a previous study.²⁷

Results

Systematic SMR Screening Identifies Nominal Associations Across Multiple Omics Layers

SMR analysis of mQTL-GWAS data from the discovery cohort (Supplementary Table S2) identified 107 methylation loci across 62 MRGs nominally associated with myopia. Among these, 75 loci corresponding to 40 MRGs exhibited colocalization evidence (PPH4 > 0.5; Fig. 2). The BDH1 methylation profile displayed marked positional heterogeneity: in the promoter-proximal region, methylation at cg02569554 (1878 base pair [bp] from transcription start site [TSS]) increased myopia risk (odds ratio [OR] = 1.575, 95% confidence interval [CI] = 1.061–2.339), whereas cg03038685 (1390 bp from TSS) was protective (OR = 0.53, 95% CI = 0.347–0.808). In contrast, gene-body sites (cg04678713 and cg06268398) and a distal site (cg27319530) were consistently protective. These observations indicate complex locus-dependent regulation within BDH1. Although some signals, such as 2 methylation sites in CYP11A1, showed concordant directionality in the validation dataset (Supplementary Table S3), none of the nominal associations identified in our systematic screening survived FDR correction in either cohort.

Figure 2.

SMR analysis of MRG methylation associated with myopia risk. The forest plot demonstrates the results of SMR analysis identifying 107 methylation sites nominally associated with myopia across 62 MRGs (P_SMR < 0.05, P_SMR_multi < 0.05, and P_HEIDI > 0.01). Each row represents a specific gene and its corresponding probe ID, along with the P value from SMR analysis (P_SMR), odds ratio (OR) with 95% confidence intervals (95% CIs), multi-SMR P value, and posterior probability of co-localization (PPH4). The plot highlights sites with significant colocalization evidence (PPH4 > 0.5), indicating shared genetic variants influencing both methylation levels and myopia risk.

SMR analysis of eQTL-GWAS data in the discovery phase (Supplementary Table S4) identified 15 MRGs nominally associated with myopia (Fig. 3). Among these, BDH1 again appeared as a recurrent signal, showing a negative association with myopia risk (OR = 0.359, 95% CI = 0.21–0.616, P_SMR = 0.000), consistent with the promoter-proximal methylation findings indicating potential transcriptional repression. This association was supported by colocalization evidence (PPH4 = 1). A similar protective association was observed for CYP11A1. However, these results were not validated in the FinnGen dataset (Supplementary Table S5) and did not remain significant after FDR correction.

Figure 3.

SMR analysis of MRG expression associated with myopia risk. The forest plot shows the results of eQTL-GWAS SMR analysis identifying 15 MRGs nominally associated with myopia (P_SMR < 0.05, P_SMR_multi < 0.05, and P_HEIDI > 0.01). Each row represents a specific gene, along with P_SMR, OR with 95% CI, multi-SMR P value, and PPH4. ^*PPH4 > 0.5.

SMR analysis of pQTL-GWAS data from the discovery cohort (Supplementary Table S6) identified potential risk associations for NIT2 and PPA2, with colocalization evidence supporting the PPA2 association (Fig. 4). However, these associations were not replicated in the FinnGen_H7_MYOPIA cohort (Supplementary Table S7) and did not survive FDR correction.

Figure 4.

SMR analysis of MRG protein abundance associated with myopia risk. The forest plot shows the SMR analysis results identifying two proteins nominally associated with myopia risk (P_SMR < 0.05, P_SMR_multi < 0.05, and P_HEIDI > 0.01). ^*PPH4 > 0.5.

Integrative Multi-Omics Analysis Highlights a Methylation–Expression Association of BDH1 With Myopia

To explore methylation-mediated transcriptional regulation, we integrated mQTL and eQTL data using SMR with methylation as the exposure and expression as the outcome (Supplementary Table S8). This integrative cross-layer analysis identified BDH1 as the only gene showing convergent methylation (cg02569554) and expression (ENSG00000161267) signals associated with myopia (Fig. 5). The cg02569554 site, located near the TSS, exhibited a strong inverse association with BDH1 expression (OR = 0.522, 95% CI = 0.425–0.641, P_SMR = 5.22 × 10⁻¹⁰, FDR = 8.79 × 10⁻⁹; see the Table). Although the HEIDI test suggested the potential influence of linkage rather than pure pleiotropy (P_HEIDI = 0.025), the locus-level plots provide visual support for this finding. Both cis-mQTL and cis-eQTL analyses showed clear association signals in the BDH1 locus (Figs. 6A, 6B), and the effect-size plots demonstrated that the same genetic variants had opposite effects on methylation and gene expression, as expected for a repressive epigenetic mark (Figs. 6C, 6D). Taken together, these results are consistent with the hypothesis that increased methylation at cg02569554 may suppress BDH1 transcription, thereby potentially contributing to myopia susceptibility.

Figure 5.

Integration of mQTL-GWAS and eQTL-GWAS analyses. The Manhattan plot shows the integration of mQTL-GWAS and eQTL-GWAS data for BDH1. The top plot represents plasma cis-mQTL data, highlighting the methylation site cg02569554 (BDH1). The bottom plot represents plasma cis-eQTL data, highlighting the gene BDH1 (ENSG00000161267). The dashed orange line indicates the P_SMR_multi threshold of 0.05.

Table.

BDH1 Results in the Integration of mQTL and eQTL Analyses

Expo_Probe	Outco_Gene	P_SMR	P_SMR_Multi	P_HEIDI	FDR	OR_SMR	95% CI_SMR
cg02569554	BDH1	5.22 × 10−10	5.22 × 10−10	0.02496161	8.79 × 10−9	0.522	0.425–0.641

Figure 6.

Genetic variants in the BDH1 locus and their impact on methylation, expression, and myopia risk. (A) Locus plot showing the association between genetic variants in the BDH1 locus and BDH1 methylation (cg02569554) in the cis-mQTL analysis. (B) Locus plot demonstrating the association of genetic variants in the BDH1 locus with BDH1 expression in the cis-eQTL analysis. (C) Effect plot illustrating the positive correlation between mQTL effect sizes and GWAS effect sizes for cg02569554 (BDH1), with a multi-SNP-based P value of 0.024. (D) Effect plot showing the negative correlation between eQTL effect sizes and GWAS effect sizes for ENSG00000161267 (BDH1), with a multi-SNP-based P value of 0.004.

Discussion

Our integrative multi-omics analysis prioritizes BDH1 as a novel, epigenetically regulated candidate gene for myopia risk. Our findings suggest a potential regulatory link in which increased methylation at a promoter-proximal site (cg02569554) is significantly associated with reduced BDH1 transcription after FDR correction. To our knowledge, this is the first study to report a potential link between BDH1 regulation and myopia. The location of this CpG site near the BDH1 TSS—a canonical position for transcriptional control—is consistent with a model of methylation-mediated gene silencing, potentially by limiting promoter accessibility or transcription factor recruitment.²⁸ This observation is noteworthy in the context of the biological role of BDH1, which encodes a key mitochondrial enzyme catalyzing the interconversion of β-hydroxybutyrate (BHB) and acetoacetate, thereby maintaining the NAD⁺/NADH redox balance is essential for oxidative phosphorylation and ATP synthesis.²⁹ Notably, BHB itself has been reported to be an epigenetic modifier that influences DNA methylation and histone acetylation, linking ketone body metabolism to chromatin regulation.³⁰ Although not directly assessed in our study, this metabolic–epigenetic crosstalk provides a biologically plausible context for the inverse relationship we observed between BDH1 promoter methylation and its expression.

Because retinal tissue relies heavily on mitochondrial ATP to sustain photoreceptor and retinal pigment epithelial function,³¹ reduced BDH1 expression may compromise ocular energy balance and increase vulnerability to oxidative stress. In glaucoma, BDH1 impacts mitochondrial ATP production through butanoate metabolism,³² suggesting that impaired BDH1 activity can disturb energy homeostasis in ocular tissues. In myopia, similar metabolic vulnerability may be exacerbated by vascular and choroidal alterations characterized by reduced blood flow and tissue hypoxia.³³ Notably, BDH1-dependent ketone body metabolism in retinal endothelial cells can promote pathological angiogenesis in proliferative retinopathy,³⁴ highlighting its role in coupling metabolic and vascular signaling. In contrast, our finding of reduced BDH1 expression associated with myopia may reflect an insufficient metabolic response to hypoxia rather than excessive angiogenic activation, suggesting disease-specific consequences of BDH1 dysregulation. Collectively, these findings suggest that BDH1 promoter hypermethylation may downregulate its expression, disrupt retinal energy and redox balance, and impair metabolic adaptation to hypoxia, thereby contributing to myopia pathogenesis.

This study has several limitations. The most critical is insufficient statistical power. Although we used large-scale GWAS data, the available mQTL, eQTL, and pQTL datasets had comparatively smaller sample sizes, and the genetic effects underlying myopia are likely modest. These factors, combined with stringent multiple testing corrections, limited our ability to detect associations that survive FDR thresholds. Indeed, most associations did not remain significant after FDR correction, with the exception of the BDH1 methylation–expression link, underscoring the exploratory nature of our findings. In addition, the reliance on European cohorts may limit the generalizability of our results to other ethnic groups. The limited number of proteins analyzed in the pQTL datasets restricted our ability to fully capture protein-level implications. Furthermore, the SMR approach, while integrating genetic data, does not account for environmental factors that might influence gene expression or methylation patterns. Last, the limited overlap across mQTL, eQTL, and pQTL results reflects both technical and biological factors. The eQTL and mQTL datasets were generated from large cohorts with broad genomic coverage, whereas the pQTL dataset measured a relatively small number of proteins, limiting statistical power. Differences in sample size, coverage, and measurement platforms further contribute to inconsistencies. Biologically, DNA methylation, mRNA expression, and protein abundance are connected through multi-step regulatory processes and are influenced by factors such as RNA stability, translational control, and post-translational modification. Therefore, concordance across layers is not always expected. In this context, the SMR framework is valuable for identifying genes supported by more than one layer of regulation. In our study, BDH1 was prioritized through an integrative methylation–expression–disease axis, providing biologically plausible support for its role in myopia and a rationale for future functional studies. However, the lack of cross-layer consistency and uncertain reproducibility render the current conclusions preliminary.

In conclusion, this study identifies BDH1 as the only MRG showing a statistically significant methylation–expression association after FDR correction, suggesting that epigenetic regulation of BDH1 may contribute to genetic susceptibility to myopia, although functional validation remains necessary. Other nominal associations were detected but should be regarded as exploratory signals pending further validation. Future studies integrating experimental validation and multiethnic replication are required to clarify the biological relevance of BDH1 and related mitochondrial pathways in refractive development.