Identifying the Involvement of Gut Microbiota in Retinal Vein Occlusion by Mendelian Randomization and Genetic Correlation Analysis

Abstract

Purpose: Previous researches have suggested an important association between gut microbiota (GM) and vascular pathologies such as atherosclerosis. This study aimed to explore the association between 196 GM taxa and retinal vein occlusion (RVO).

Methods: This study used Mendelian randomization (MR), linkage disequilibrium score regression (LDSC), and polygenic overlap analysis. Genome-wide association study (GWAS) data associated with 196 GM taxa was obtained from the MiBioGen consortium, involving a large number of European-ancestry participants. GWAS data of RVO was obtained from the FinnGen consortium and another study that also involved European-ancestry participants. Inverse-variance weighted was used as the primary approach for MR estimation. Moreover, LDSC and polygenic overlap analyses were performed to evaluate the genetic correlation between GM taxa and RVO.

Results: The MR results identified the association of six GM taxa, including class Bacilli, order Lactobacillales, family Streptococcaceae, genus Clostridium innocuum group, genus Family XIII AD3011 group, and genus Subdoligranulum with the development of RVO. In addition, the polygenic overlap analysis supported the genetic association between GM and RVO.

Conclusions: Our findings confirmed the association between six GM taxa and the development of RVO, thereby highlighting the effects of GM on retinal vascular health.

Translational Relevance: The results may provide the rationale for developing GM-based strategies for preventing the onset of RVO.

Introduction

Retinal vein occlusion (RVO) is a relatively common retinal vascular disorder and an important cause of visual impairment and blindness globally. Mainly, RVO has two subtypes, including central RVO and branch RVO, depending on the obstruction location of retinal vein.¹ Worldwide, more than 25 million people have been or are being affected by RVO,² with approximately 15 new cases per 100,000 population per year.³ Reported risk factors of RVO includes old age, hyperlipidemia, diabetes, and hypertension.^4–7 Currently, available treatment for RVO mainly includes antivascular endothelial growth factor agents and corticosteroids.⁸ However, the pathogenesis of RVO is still not fully understood, with the result that there is no effective preventative strategy for RVO. 

Gut microbiota (GM) has been suggested to be implicated in vascular pathologies such as atherosclerosis⁹ and eye disorders such as age-related macular degeneration (AMD).¹⁰ Notably, the interplay of GM in retinal vascular diseases such as RVO has been reported.^11,12 However, comprehensive evaluation of the association between the vast majority of GM taxa and RVO is still undone. 

By using exposure-associated genetic variants as instruments, Mendelian randomization (MR) can examine and determine the association between exposures and outcomes and overcome the bias of observational studies.^13,14 This approach has been used to explore risk factors of ocular diseases such as glaucoma,^15,16 AMD,¹⁰ and diabetic retinopathy.¹⁷ Linkage disequilibrium score regression (LDSC) analysis¹⁸ is an approach designed for quantifying the genetic correlation between two traits by using genome-wide summary statistics. LDSC analysis can provide information about the shared genetic etiology between traits. However, LDSC analysis only reports overall genetic correlation, failing to capture mixtures of effect in different directions. Polygenic overlap analysis¹⁹ can report the fraction of causal genetic variants shared by two traits over the total number of causal genetic variants across these two traits. 

In this study, we performed MR, LDSC, and polygenic overlap analysis to uncover potential association between 196 GM taxa and development of RVO to provide novel insights into our understanding of the pathogenesis of RVO. As a result, six GM taxa were identified to be associated with development of RVO and shared genetic basis with RVO. 

Methods

Study Design

The theoretical basis of MR analysis^20,21 are presented in Figure 1. In this study, we applied MR approaches to estimate the association between 196 GM taxa and development of RVO. This study was in compliance with the guidelines of Strengthening the Reporting of Observational Studies in Epidemiology using Mendelian randomization. 

Figure 1.

 

The theoretical basis and three basic assumptions of MR analysis.

Data Acquisition

The genome-wide association studies (GWAS) data of GM taxa was obtained from MiBioGen consortium.²² In MiBioGen consortium, 211 GM taxa (with 15 unknown taxa in them) were categorized into five levels (including phylum, class, order, family, and genus). To maintain the accuracy of the results, the 15 unknown GM taxa were removed in analysis. The GWAS data of RVO were obtained from FinnGen (775 cases and 308,633 controls, used for discovery)²³ and UK Biobank Consortium (387,189 European participants, used for validation).²⁴ The RVO cases were defined by H34.8 in the International Classification of Disease-10, including central and branch RVO cases. All these studies involved solely European-ancestry participants. GWAS data used in this study was publicly available and approved by the ethics committees of original institutions, following the tenets of the Declaration of Helsinki. 

Selection and Filtration of Instrumental Genetic Variables

The single-nucleotide polymorphisms (SNPs), which are used as instrumental variables, were selected and filtrated through the following criteria: (1) genome-wide P for exposure (GM taxa) < 1.0 × 10^–5; (2) without linkage disequilibrium (r² < 0.01 and window size 10,000 kb); (3) effect allele frequency > 0.01; (4) palindromic SNPs were removed; and (5) F-statistic > 10.^25,26 

MR Analysis

The validity and accuracy of MR results rely on the absence of heterogeneity and pleiotropy.²⁷ Therefore we used the random-effect inverse-variance weighted (IVW) method²⁸ as the primary method, and weighted-median and weighted-mode methods were used to improve the robustness of IVW-derived results as what they are designed for.^29,30 To evaluate the heterogeneity and detect pleiotropy, Cochran's Q test and MR-Egger intercept test were performed.³¹ Moreover, the MR Pleiotropy RESidual Sum and Outlier (MR-PRESSO)³² test was performed to identified outlier SNPs. The cutoff of sensitivity tests was set as 0.05. Moreover, we performed Benjamini-Hochberg procedure to control false discovery rate (FDR) and minimize the false-positive results. The cutoff of FDR was set as 20% (0.2). 

Genetic Correlation and Polygenic Overlap Analyses Between GM Taxa and RVO

For a pair of traits, LDSC analysis¹⁸ can quantify the genetic correlation for the two traits using GWAS summary statistics. However, LDSC analysis only reports overall genetic correlation, failing to capture mixtures of effect in different directions. Therefore polygenic overlap analysis was performed to further assess the genetic correlation between GM taxa and RVO. Polygenic overlap analysis can report the fraction of causal genetic variants shared by two traits over the total number of causal genetic variants across these two traits. MiXeR¹⁹ is a statistical tool, quantifying polygenic overlap between traits by using summary GWAS data. MiXeR enables a more comprehensive quantification of genetic correlation than the LDSC method and provides more information about the genetic relationships between traits. 

Statistical Analysis

R (version 4.0.1) and the TwoSampleMR³³ R packages were used for analyses. Results with IVW-P < 0.05 was considered significant. 

Results

Discovery Stage Results of MR Analysis

The list of 196 GM taxa is in Supplementary Table S1. In the discovery stage, 23 GM taxa were found to be associated with risk of RVO (Fig. 2, Supplementary Table S2). The 23 GM taxa significantly associated with RVO risk were phylum Actinobacteria, class Bacilli, class Coriobacteriia, order Coriobacteriales, order Lactobacillales, order NB1n, family Coriobacteriaceae, family Streptococcaceae, genus Clostridium innocuum group, genus Eubacterium rectale group, genus Eubacterium ventriosum group, genus Butyrivibrio, genus Desulfovibrio, genus Erysipelotrichaceae UCG003, genus Escherichia Shigella, genus Family XIII AD3011 group, genus Holdemanella, genus Howardella, genus Lachnospiraceae UCG001, genus Ruminiclostridium9, genus Ruminococcaceae UCG011, genus Streptococcus, genus Subdoligranulum. Results of weighted median and weighted mode methods supported the robustness of IVW-derived results (Supplementary Table S3). In addition, Cochran's Q test, MR-Egger intercept test, and MR-PRESSO global test all suggested no apparent heterogeneity or pleiotropy in these 23 results (P > 0.05, Supplementary Tables S2 and S4). In this discovery stage, class Bacilli, order Lactobacillales, order NB1n, genus Family XIII AD3011 group, and genus Ruminiclostridium9 passed the FDR correction (Supplementary Table S8). 

Figure 2.

 

MR results of discovery stage. (A) P values of IVW, heterogeneity test and pleiotropy test of GM taxa, and risk of RVO in discovery stage; (B) Forest plot of the significant MR results in discovery stage. CI, confidence interval; OR, odds ratio.

Validation Stage Results of MR Analysis

In the validation stage, 25 GM taxa were identified to be associated with risk of RVO (Fig. 3, Supplementary Table S5). The 25 GM taxa significantly associated with RVO risk were phylum Euryarchaeota, class Bacilli, class Betaproteobacteria, class Erysipelotrichia, order Bacillales, order Bifidobacteriales, order Clostridiales, order Erysipelotrichales, order Lactobacillales, family Bifidobacteriaceae, family Clostridialesvadin BB60 group, family Erysipelotrichaceae, family Porphyromonadaceae, family Streptococcaceae, genus Clostridium innocuum group, genus Ruminococcusgauvreauii group, genus Alistipes, genus Dialister, genus Family XIII AD3011 group, genus Intestinibacter, genus Lachnospiraceae ND3007 group, genus Oxalobacter, genus Roseburia, genus Ruminococcaceae UCG004, genus Subdoligranulum. Results of weighted-median and weighted-mode methods supported the robustness of IVW-derived results (Supplementary Table S6). In addition, Cochran's Q test, MR-Egger intercept test, and MR-PRESSO global test all suggested no apparent heterogeneity or pleiotropy in these results (P > 0.05, Supplementary Tables S5 and S7). Notably, the effect of six GM taxa (class Bacilli, order Lactobacillales, family Streptococcaceae, genus Clostridium innocuum group, genus Family XIII AD3011 group, and genus Subdoligranulum) on the risk of RVO was validated in this stage (Table 1). In this stage, only genus Intestinibacter passed the FDR correction (Supplementary Table S9). 

Figure 3.

 

MR results of validation stage. (A) P values of IVW, heterogeneity test and pleiotropy test of GM taxa and risk of RVO in validation stage; (B) Forest plot of the significant MR results in validation stage. CI, confidence interval; OR, odds ratio.

Table 1.

 

Significant MR Results for GM Taxa and Risk of RVO in Both Discovery and Validation Stages

 

Genetic Correlation and Polygenic Overlap Analyses for the Six GM Taxa and RVO

Genetic correlation (LDSC) analysis indicated no significant genetic correlation between these six GM taxa and RVO (Table 2). Polygenic overlap (MiXeR) analysis results are presented as a Venn diagram of unique and shared polygenic components across traits. For class Bacilli and RVO, MiXeR estimates that 36.3 K variants causally influence class Bacilli and 0.6 K influence RVO. Among these variants, 0.5 K are shared between them, which have a high genetic correlation (Fig. 4A). For order Lactobacillales and RVO, MiXeR estimates that 63.0 K variants causally influence order Lactobacillales and 0.7 K influence RVO and 0.4 K are shared between them (Fig. 4B). For family Streptococcaceae and RVO, MiXeR estimates that 80.3 K variants causally influence family Streptococcaceae and 0.7 K influence RVO and 0.5 K are shared between them (Fig. 4C). For genus Clostridium innocuum group and RVO, MiXeR estimates that 2.9 K variants causally influence genus Clostridium innocuum group and 0.8 K influence RVO and 0.4 K are shared between them (Fig. 5A). For genus Family XIII AD3011 group and RVO, MiXeR estimates that 18.0 K variants causally influence genus Family XIII AD3011 group and 0.7 K influence RVO and 0.4 K are shared between them (Fig. 5B). For genus Subdoligranulum and RVO, MiXeR estimates that 1.8 K variants causally influence genus Subdoligranulum and 0.7 K influence RVO and 0.3 K are shared between them (Fig. 5C). 

Table 2.

 

Genetic Correlation Between GM Taxa and RVO Calculated by LDSC Analysis

 

Figure 4.

 

Venn diagrams of unique and shared polygenic components at the causal level and conditional Q–Q plots of observed versus expected-log10 P values in the primary trait as a function of significance of association with a secondary trait at the level of P ≤ 0.1 (orange lines), P ≤ 0.01 (green lines), P ≤ 0.001 (red lines), showing polygenic overlap between class Bacilli (A), order Lactobacillales (B), family Streptococcaceae (C) and RVO.

Figure 5.

 

Venn diagrams of unique and shared polygenic components at the causal level and conditional Q–Q plots of observed versus expected-log10 P values in the primary trait as a function of significance of association with a secondary trait at the level of P ≤ 0.1 (orange lines), P ≤ 0.01 (green lines), P ≤ 0.001 (red lines), showing polygenic overlap between genus Clostridium innocuum group (A), genus Family XIII AD3011 group (B), genus Subdoligranulum (C) and RVO.

Discussion

Considering the potential health burden related to RVO, the preventive strategies of this disease present critical unmet medical needs. It is urgently needed to identify biomarkers for early diagnosis or drug targets for prevention. Despite recent advances, the relationship between GM and RVO has not been determined. In addition, considering the variety of GM, it is still needed to clarify specific taxa of GM that significantly contribute to the development of RVO, to understand the pathogenesis of RVO in a more accurate and precise way. As far as we know it, this is the first study to comprehensively explore the association between GM taxa and RVO. In this study, we identified 6 GM taxa (class Bacilli, order Lactobacillales, family Streptococcaceae, genus Clostridium innocuum group, genus Family XIII AD3011 group, and genus Subdoligranulum) positively associated with the incidence of RVO. Although only class Bacilli and genus Family XIII AD3011 group passed FDR correction in discovery stage, the other 4 GM taxa were still worthy of further investigation. Notably, because of the relative accessibility of GM taxa, the findings in this study may open up new opportunities for the development of strategies for modifying the risk of RVO. 

Mounting evidence from GWAS studies suggests an abundance of shared genetic influences between multiple human traits. Genetic correlation analysis and polygenic overlap analysis may deepen our understanding of cross-trait genetic architectures. Here in this study, we performed genetic correlation and polygenic overlap analyses by LDSC¹⁸ and MiXeR¹⁹ for the six RVO-associated GM taxa identified in this study and RVO. Despite no significant genetic correlation, these six GM taxa and RVO shared causal variants, more or less, thereby highlighting the importance of them in the pathogenesis of RVO and providing potential targets for further exploring the pathophysiology of RVO. 

GM of an individual, which is originally obtained from the mother, will experience considerable changes after being exposed to the external environment. Healthy homeostasis of GM has been associated with development and normal function of the central nervous system³⁴ through a gut-brain axis. Similarly, the gut-eye axis^35–37 represents the potential of GM (its presence and activity) to influence ocular health. Inflammation induced by microbial dysbiosis has been implicated in the progression of ocular diseases such as AMD.³⁸ Moreover, a gut-vascular axis has also been proposed to explain the influence of GM on vascular health.⁹ 

GM imbalance has been associated with cardiovascular diseases and considered as a risk factor for pathogenesis of cardiovascular diseases.³⁹ GM alterations modify the metabolic homeostasis, causing structural and functional alterations in situ and exacerbating the cardiovascular pathological scenarios.⁴⁰ Experimental evidence has demonstrated how GM alterations and their metabolites led to the initiation or worsening of vascular pathology, describing a crosstalk between the gut and blood vessels.⁴¹ GM release bioactive molecules that can be transported into the systemic circulation and act in peripheral districts. GM-associated metabolites include short-chain fatty acids, amines, phenols, ammonia, thiols, and indoles, produced through the process defined as saccharolytic or proteolytic pathway.⁴² Some of these metabolites can negatively interfere with cardiovascular districts via their potential toxicity.⁴³ Shift in the GM (microbial dysbiosis) could lead to the corresponding alterations in metabolic products, especially short-chain fatty acids,^44,45 thereby leading to systemic inflammation⁴⁶ and onset of various disorders.^37,47 GM alterations can influence the permeability of gut and lead other metabolites such as trimethylamine-N-oxide, peptidoglycans, hydrogen sulfide, nitric oxide, carbon monoxide, and methane to cross the intestinal epithelium, affecting the vascular district.^48–50 

Aging has been considered as the main risk factor for vascular homeostasis defects and the development of vascular diseases. Notably, GM has been found to be associated with aging and aging-related diseases.^51,52 Age-related alterations of GM can lead to chronic inflammation and metabolic dysfunction, which in turn affect aging and increase the risk of age-related diseases.⁵¹ The relationship between GM taxa and accelerated aging was hypothesized to be mediated by spermidine, a natural polyamine, which is crucial for cell development, proliferation, and tissue regeneration, with anti-inflammatory and antioxidant properties.⁵³ The decline of endogenous spermidine produced by GM has been observed and stimulating the production of spermidine by GM has been considered as a promising anti-aging strategy.^51,53 In addition, Mou et al.⁵² suggested that GM contributes to aging and aging-related diseases through the induction of systemic chronic inflammation. Collectively, these six GM taxa identified in this study might promote the development of RVO via a gut-vascular axis.⁹ 

There were some limitations. First, only European-ancestry individuals were included in this study, which hinders the direct application of the findings into other populations. Second, the results only suggested the associations between GM taxa and higher risk of RVO, without investigating the underlying mechanisms. We did propose some potential mechanisms, which still require to be confirmed by further researches. Third, the association between GM and RVO subtypes such as central and branch, ischemic and non-ischemic, was not investigated, which limited the significance of this study and required to be discussed in future studies. Fourth, systemic diseases such as hypertension or medication taking will also impact the development of RVO, which is not sufficiently discussed during the analysis in this study. Last, the results are not validated in local cohorts, which limited the applicability of the findings in this study. 

In conclusion, this study provides the evidence that six GM taxa (class Bacilli, order Lactobacillales, family Streptococcaceae, genus Clostridium innocuum group, genus Family XIII AD3011 group, and genus Subdoligranulum) shared a genetic basis with RVO and are associated with the development of RVO, which highlights the implication of GM in the pathophysiology of RVO. The findings may contribute to the development of new strategies for preventing the onset of RVO. 

Acknowledgments

Supported by The Funding for Scientific Research Projects from Wuhan Municipal Health Commission (WX23Z33). 

Disclosure: S. Lei, None; Y. Liu, None