Genetics of Circulating Inflammatory Proteins and Iridocyclitis: An Exploratory Mendelian Randomization Study

Abstract

Purpose: This study aimed to explore the potential causal relationship between genetically determined elevated levels of pro-inflammatory cytokines and iridocyclitis, including its subtypes: acute and subacute iridocyclitis (ASIR) and chronic iridocyclitis (CIR).

Methods: A two-sample Mendelian randomization (MR) analysis was conducted using genome-wide association study (GWAS) summary data for inflammatory cytokines (cases, n = 14,824), iridocyclitis (IR; cases, n = 7306 and controls, n = 357,814), and its subtypes (ASIR: cases, n = 6166 and controls, n = 357,814; and CIR: cases, n = 1401 and controls, n = 357,814). The inverse variance‐weighted (IVW) method served as the primary analysis method. Supplementary analytic methods included MR Egger, Weighted median, Weighted mode, and Simple mode methods. Pleiotropy and heterogeneity were evaluated using the Cochran’s Q test, MR Egger intercept test, MR Pleiotropy RESidual Sum and Outlier test (MR-PRESSO), Bayesian colocalization analysis, and Linkage disequilibrium score regression (LDSC) analysis.

Results: Genetically predicted high levels of eotaxin, fibroblast growth factor 23 (FGF23), TNF-related apoptosis-inducing ligand (TRAIL), and Neurotrophin-3 were associated with an increased risk of IR. On the contrary, a high level of interleukin (IL)-2 was associated with a decreased risk of IR. Meanwhile, the IR subgroup analysis demonstrated that high levels of eotaxin and TRAIL were also associated with an increased risk of ASIR. High levels of cystatin D, tumor necrosis factor receptor superfamily member 9 (TNFRSF9), and caspase 8 were associated with an increased risk of CIR. CCL20 and CDCP1 were associated with a decreased risk of CIR. Heterogeneity and pleiotropy tests demonstrated that our findings were stable and reliable.

Conclusions: Inflammatory cytokines are involved in the occurrence of IR and its subtypes. Further studies are warranted to elucidate the precise mechanisms of inflammatory cytokines in IR and its subtypes.

Translational Relevance: The present study highlighted the role of inflammatory cytokines in the development of IR.

Introduction

Uveitis is a group of sight-threatening intraocular inflammatory disorders accounting for 5% to 20% and more than 25% of blindness in developed and developing countries, respectively.^1,2 Uveitis mainly attacks the iris, ciliary body, and choroid, vitreous, and retina.³ Uveitis is characterized as anterior, intermediate, posterior, and panuveitis based on the basis of the anatomical location of ocular inflammation.⁴ Anterior uveitis (AU), the most common uveitis, comprises iritis, anterior cyclitis, and iridocyclitis (IR).⁵ The signs and primary symptoms of IR are blurred vision, eye pain, redness, and photophobia.⁶ IR could present as an acute, subacute, or chronic disease process. 

The pathogenesis of IR remains unclear. Previous studies have demonstrated that genetic susceptibility, infection, immune dysregulation, inflammatory response, and environmental factors that include intestinal microbiota are involved in the occurrence and progression of IR.^7–12 The onset of IR is also associated with a number of systemic autoimmune and autoinflammatory disorders, including ankylosing spondylitis (AS), juvenile idiopathic arthritis (JIA), inflammatory bowel disease (IBD), and Behçet’s disease (BD).¹¹ Furthermore, inflammatory cytokines, such as IL-6, IL-10, IL-17, IL-22, and IL-23, and tumor necrosis factor-alpha (TNF-α) are reported to play pivotal roles in the development of uveitis.^11,13 A previous investigation found significantly higher expression levels of C-C motif chemokine ligand (CCL) 8, CCL13, and CCL20 in the aqueous humor of patients with Human leukocyte antigen B27 (HLA B27) positive acute AU (AAU) compared with aqueous humor from healthy subjects.¹⁴ Moreover, IL-37 mRNA and protein expression levels were higher in peripheral blood mononuclear cells (PBMCs) of patients with AAU compared with PBMC from healthy subjects.¹⁵ However, these findings were primarily from observational studies. Whether these findings indicate a causal link between the abnormal expression of inflammatory cytokines and IR is unclear. Meanwhile, the relationship between numerous inflammatory cytokines and the occurrence of IR remains elusive. 

A Mendelian randomization (MR) study is a causal inference technique that utilizes genetic variants as instrumental variables (IVs) to determine whether an exposure is causally associated with an outcome.¹⁶ An MR study that uses random allele assignment at conception can minimize the effects of potential confounding and reverse causation bias.¹⁶ A recent genome-wide association study (GWAS) discovered several genetic variations associated with inflammatory proteins and yielded the data for the MR study.¹⁷ 

This study investigated the potential causal associations between inflammatory cytokines and the risk of IR and its subtypes using the two-sample MR analysis. 

Methods

Study Design

A two-sample MR analysis was used to investigate the causal relationship between inflammatory cytokines and the risk of IR and its subtypes. The MR study builds upon three principal assumptions: (I) the genetic variants used as IVs in the analysis should be robustly associated with inflammatory cytokines; (II) genetic variants used as IVs for inflammatory cytokines should not be associated with known confounding factors; and (III) genetic variants as IVs should influence the risk of IR only through inflammatory cytokines. Figure 1 depicts the MR study design. 

Figure 1.

 

The overall study design. The MR study builds upon three principal assumptions: (I) the genetic variants used as instrumental variables (IVs) in the analysis should be robustly associated with inflammatory cytokines; (II) genetic variants used as IVs for inflammatory cytokines should not be associated with known confounding factors; and (III) genetic variants as IVs should influence the risk of IR only through inflammatory cytokines.

Data Sources

The GWAS summary data for 91 inflammatory cytokines analyzed in this study was measured using the Olink Target Inflammation panel across 11 cohorts, involving a total of 14,824 participants of European ancestry. The results were then meta-analyzed to provide comprehensive insights.¹⁷ The FinnGen study is a large-scale personalized medicine project involving 500,000 participants from Finnish biobanks. Its primary aim is to generate robust evidence on the influence of genomics on human health. The results of the comprehensive statistical analyses from the FinnGen study have already been published.¹⁸ The diagnosis of IR, acute and subacute iridocyclitis (ASIR), and chronic iridocyclitis (CIR) are based on the International Classification of Diseases, 10th Revision (ICD-10), with the corresponding code H20, H20.0, and H20.1, respectively. Detailed information on the GWAS datasets is provided in Table 1. 

Table 1.

 

Overview of GWAS Data Used in Mendelian Randomization Study

 

Selection of Genetic Instruments for Inflammatory Cytokines

Inflammatory cytokines–related single nucleotide polymorphisms (SNPs) were selected as IVs with a genome-wide significance threshold P value ≤ 5 × 10^–6 and linkage disequilibrium r² ≤ 0.001 within a window of 10,000 kb. Palindrome SNPs were excluded from our MR analysis. The F-statistic was used to assess the strength of the IVs and calculated formulas follows: [(N − K − 1) * R²]/[1 − R²] * K, where: k = number of IVs and R² = proportion of variance by IVs.¹⁹ Individual SNPs with F-statistic values < 10 were considered weak instruments and were excluded. 

MR Analysis

A two-sample MR analysis was performed to investigate the potential causal effects of inflammatory cytokine levels on IR and its subtypes. The principal analysis method was the fixed-effects inverse variance‐weighted (IVW) method.²⁰ The random-effects IVW was used to analyze if there was unexplained statistical heterogeneity.²⁰ In addition, the MR Egger, Weighted median, Weighted mode, and Simple mode methods served as supplementary analysis methods.^21,22 The results are reported statistically as odds ratios (ORs) with 95% confidence intervals (CIs) and P values. 

Sensitivity Analysis

Sensitivity analyses were performed to examine the reliability of the results. Heterogeneity among the IVs was evaluated using Cochran’s Q test and the I² statistic.²³ The MR Egger intercept test was performed to determine potential directional pleiotropy. The MR pleiotropy residual sum and outlier test (MR-PRESSO) was applied to detect outliers and correct for widespread horizontal pleiotropy.²⁴ In addition, the direction of association between inflammatory cytokines and IR and its subtypes was validated using the Steiger directionality test.²⁵ A cross-trait analysis of single variants was conducted using the PhenoScanner online tool to exclude SNPs associated with confounder factors (http://www.phenoscanner.medschl.cam.ac.uk/). The statistical power calculations for the MR analysis were performed using an online tool available at http://cnsgenomics.com/shiny/mRnd/. 

Bayesian Colocalization Analysis

A Bayesian colocalization analysis is performed to assess whether inflammatory cytokines and IRs share the same causal variant.²⁶ We performed a Bayesian colocalization analysis based on posterior probabilities (PPs) for five hypotheses in the current study: (1) PPH₀ = no association between inflammatory cytokines and IR; (2) PPH₁ = association with inflammatory cytokines, not with IR; (3) PPH₂ = association with IR, not with inflammatory cytokines; (4) PPH₃ = association with inflammatory cytokines and IR, two independent SNPs; and (5) PPH₄ = association with inflammatory cytokines and IR, one shared SNP. The SNPs within a 1000 kb window on either side of the lead SNP were used in the colocalization analysis. PPH4 > 0.80 denoted substantial evidence between inflammatory cytokines and IR. 

Linkage Disequilibrium Score Regression

Linkage disequilibrium score regression (LDSC) regression analysis and GWAS summary statistics were executed to evaluate the potential genetic correlation between inflammatory cytokines and IR and its subtypes.²⁷ 

Reverse MR Analysis

A reverse MR study was performed to investigate whether there is a reverse causal effect between IR, including its subtypes: ASIR and CIR and inflammatory cytokines. 

Statistical Analysis

The results of this study were presented in compliance with the Strengthening the Reporting of Observational Studies in Epidemiology Using Mendelian Randomization (STROBE-MR) guidelines (Supplementary Table S1).²⁸ MR analysis was performed using the TwoSampleMR (version 0.5.8) and MR-PRESSO (version 1.0) packages in R Software 4.2.3.^24,29 Bayesian colocalization analysis and LDSC analysis were performed using the “coloc” packages (version 5.2.3) and “ldscr” packages (version 0.1.0), respectively. 

Results

All SNPs selected as IVs are listed in Supplementary Table S2. SNPs associated with confounder factors were excluded from subsequent analysis (Supplementary Table S3). The MR Steiger test revealed that the causal direction of inflammatory cytokines on the risk of IR and its subtypes were not subject to reverse causality (Supplementary Table S4). The F-statistics demonstrated that all IVs were sufficiently strong. Supplementary Table S5 displays the F-statistics value and the proportion of the variance explained (R²). 

Causal Effects of Inflammatory Cytokines on IR

MR IVW analysis revealed that genetically predicted high levels of eotaxin (OR = 1.165, 95% CI = 1.050–1.291, P = 0.0039), Fibroblast growth factor 23 (FGF23; OR = 1.212, 95% CI = 1.014–1.447, P = 0.0344), TNF-related apoptosis-inducing ligand (TRAIL; OR = 1.070, 95% CI = 1.006–1.138, P = 0.0308), and neurotrophin-3 (OR = 1.170, 95% CI = 1.026–1.336, P = 0.0196) were associated with an increased risk of IR. On the contrary, high levels of IL-2 (OR = 0.873, 95% CI = 0.767–0.994, P = 0.0409) were associated with a decreased risk of IR (Fig. 2). 

Figure 2.

 

MR analysis between inflammatory cytokines and IR and its subtypes risk.

Causal Effects of Inflammatory Cytokines on ASIR

The MR IVW results demonstrated that genetically predicted high levels of eotaxin (OR = 1.221, 95% CI = 1.092–1.364, P = 0.0004) and TRAIL (OR = 1.123, 95% CI = 1.050–1.200, P = 0.0007) were associated with an increased risk of ASIR (see Fig. 2). 

Causal Effects of Inflammatory Cytokines on CIR

MR IVW analysis revealed that genetically predicted high levels of cystatin D (OR = 1.300, 95% CI = 1.108–1.524, P = 0.0013), tumor necrosis factor receptor superfamily member 9 (TNFRSF9; OR = 1.285, 95% CI = 1.044–1.581, P = 0.0179), and caspase 8 (OR = 1.388, 95% CI = 1.047–1.842, P = 0.0229) were associated with an increased risk of CIR. On the contrary, a high level of C−C motif chemokine 20 (CCL20; OR = 0.720, 95% CI = 0.545–0.952, P = 0.0211), glial cell line-derived neurotrophic factor (GDNF; OR = 0.675, 95% CI = 0.491–0.929, P = 0.0157), and CUB domain-containing protein 1 (CDCP1; OR = 0.816, 95% CI = 0.688–0.969, P = 0.0200) were associated with a decreased risk of CIR (see Fig. 2). 

However, other genetically predicted inflammatory cytokine levels were not associated with IR, ASIR, and CIR risk (Supplementary Table S6). 

Sensitivity Analysis Results of Inflammatory Cytokines and IR and its Subtypes

There was no evidence of heterogeneity and directional pleiotropy among these inflammatory cytokines associated with IR and its subtype risk (Table 2). MR-PRESSO analysis revealed that the SNP (rs140862652) for TRAIL was an outlier and was removed (see Table 2). All sensitivity analysis results are presented in Supplementary Table S7, Supplementary Table S8, Supplementary Table S9, and Supplementary Table S10. 

Table 2.

 

Sensitivity Analysis of the Causality of Inflammatory Cytokines on Iridocyclitis and its Subtypes

 

Colocalization Analysis of Inflammatory Cytokines and IR and its Subtypes

Eotaxin (PPH₄ = 0.079), FGF23 (PPH₄ = 0.046), IL-2 (PPH₄ = 0.024), neurotrophin-3 (PPH₄ = 0.013), and TRAIL (PPH₄ = 0.011) were found unlikely share a causal variant with IR (Table 3; Fig. 3). Eotaxin (PPH₄ = 0.222) and TRAIL (PPH₄ = 0.011) were unlikely to share a causal variant with ASIR (see Table 3; Fig. 3). Furthermore, there was little support for CCL20 (PPH₄ = 0.272), Caspase 8 (PPH₄ = 0.046), CDCP1 (PPH₄ = 0.054), cystatin D (PPH₄ = 0.112), and TNFRSF9 (PPH₄ = 0.031) sharing a causal variant with CIR (see Table 3; Fig. 3). However, GDNF and CIR had a 0.934 posterior probability of sharing a causal variant, implying that the pleiotropic effects triggered the causal association of GDNF with the risk of CIR (see Table 3; Fig. 3). 

Table 3.

 

Colocalization Analysis of Inflammatory Cytokines With Iridocyclitis and its Subtypes

 

Figure 3.

 

Colocalization analysis between inflammatory cytokines and IR and its subtypes. (1) The X-axis represented the distribution of SNPs in specific DNA regions (A–M). (2) The Y-axis represented the -log10(p) values of SNPs in the GWAS data, with higher values indicating smaller P values, and the lead SNP is at the top. (3) The scatter plot displayed the distribution of SNPs near the lead SNPs from two GWAS datasets.

LDSC Regression Analysis of Inflammatory Cytokines and IR and its Subtypes

The genetic correlation between inflammatory cytokines and IR and its subtypes was estimated using LDSC regression analysis. There was, however, no evidence of statistically significant genetic correlations between inflammatory cytokines and IR and its subtypes (Table 4). 

Table 4.

 

LDSC Regression Analysis of the Genetic Correlation Estimation of Inflammatory Cytokines With Iridocyclitis and its Subtypes

 

The Findings From the reverse MR Analysis

The results demonstrated that genetically predicted IR was not associated with eotaxin (OR = 0.986, 95% CI = 0.956–1.016, P = 0.352), FGF23 (OR = 0.993, 95% CI = 0.964–1.023, P = 0.653), TRAIL (OR = 0.995, 95% CI = 0.965–1.026, P = 0.755), neurotrophin-3 (OR = 0.984, 95% CI = 0.954–1.015, P = 0.303), and IL-2 (OR = 1.010, 95% CI = 0.976–1.046, P = 0.563). Genetically predicted ASIR was not associated with eotaxin (OR = 0.988, 95% CI = 0.961–1.015, P = 0.386) and TRAIL (OR = 0.990, 95% CI = 0.964–1.018, P = 0.487). Genetically predicted CIR was not associated with cystatin D (OR = 0.987, 95% CI = 0.956–1.019, P = 0.426), CCL20 (OR = 1.012, 95% CI = 0.980–1.045, P = 0.478), GDNF (OR = 0.998, 95% CI = 0.967–1.031, P = 0.919), TNFRSF9 (OR = 0.993, 95% CI = 0.958–1.029, P = 0.692), CDCP1 (OR = 1.000, 95% CI = 0.970–1.031, P = 0.997), and caspase 8 (OR = 1.015, 95% CI = 0.982–1.048, P = 0.382). The results were described in the Figure 4. 

Figure 4.

 

The reverse causal effect between IR and its subtypes and inflammatory cytokines.

Discussion

The present study investigated the causal associations of genetically predicted inflammatory cytokines on IR and its subtypes risk using a two-sample MR analysis. The findings revealed that genetically predicted high levels of eotaxin, FGF23, TRAIL, and neurotrophin-3 were associated with an increased risk of IR. Meanwhile, IR subgroup analysis demonstrated that high levels of eotaxin and TRAIL were also associated with an increased risk of ASIR. Genetically predicted high levels of cystatin D, TNFRSF9, and caspase 8 were associated with an increased risk of CIR. However, high levels of IL-2 and inflammatory cytokines (CCL20, GDNF, and CDCP1) were associated with a decreased risk of IR and CIR, respectively. This is the first study to investigate the genetic causal associations between inflammatory cytokines on the risk of IR and its subtypes, making a significant contribution to understanding the genetic basis and the mechanism of IR. 

We discovered that genetically predicted high levels of eotaxin and TRAIL were associated with an increased risk of IR and ASIR. However, FGF23, neurotrophin-3, and IL-2 were associated primarily with IR. Eotaxin (also known as CCL11), a potent eosinophil chemoattractant and member of the CC chemokine family, regulates proinflammatory cytokine production and mediates immune responses, both of which play a significant role in the pathogenesis of inflammatory diseases.^30,31 Eotaxin has been demonstrated to bind to CCR3 receptors and is essential for the recruitment and migration of mast cells, eosinophils, Th2 cells, basophils, neutrophils, and macrophages into inflammatory sites.^32,33 Elevated levels of CCL11 have been revealed in IR associated with inflammatory diseases, including AS, JIA, and IBD.^34–36 TRAIL, a member of the TNF family of proteins, plays a role in regulating apoptosis, inflammation, and immunity.³⁷ Elevated serum TRAIL levels were associated with patients with JIA. A recent MR study showed a link between increased TRAIL levels and AS. FGF23 is a crucial regulator of mineral metabolism and has emerged as a critical regulator of innate and adaptive immune responses.³⁸ Elevated FGF23 levels were associated with patients with AS and IBD.^39,40 Neurotrophin-3 is a secreted growth factor, but its precise role in regulating inflammation and immunity is unknown. For the first time, our study demonstrated a causal relationship between neurotrophin-3 and IR risk. IL-2, a T cell growth factor, is essential for inducing the differentiation of naïve CD4⁺ T cells into regulatory T (T_reg) cells and inhibiting the differentiation of naïve CD4⁺ T cells into Th17 cells.⁴¹ Previous studies demonstrated that Th17 cells are critical in uveitis pathogenesis.¹¹ Patients with uveitis have significantly higher serum and aqueous humor IL-2 levels.⁴² Low-dose IL-2 has shown tremendous promise in treating autoimmune diseases associated with uveitis.⁴² 

Genetically predicted TNFRSF9, caspase 8, cystatin D, CCL20, and CDCP1 were associated with the risk of CIR. TNFRSF9 (CD137), a member of the TNF-receptor superfamily, is expressed not only on the surface of B and T cells but also on the surface of natural killer (NK) cells and dendritic cells (DCs).⁴³ Anti-CD137 mAb potentially inhibited the development of experimental autoimmune uveitis in B10RIII mice.⁴⁴ Caspase 8 is essential in cell apoptosis execution and increases in experimental autoimmune anterior uveitis.⁴⁵ Furthermore, cystatin D (CST5) expression was significantly higher in patients with rheumatoid arthritis, a chronic inflammatory disease.⁴⁶ Our study also discovered that cystatin D was associated with CIR. CCL20 binding to CCR6 promotes the differentiation of naïve CD4⁺ T into Th17 cells while suppressing the differentiation of T_reg cells.⁴⁷ However, our findings on the role of CCL20 in CIR were contradictory. More research is warranted to validate and elucidate the controversial results of CIR. The role of CDCP1 has rarely been studied in immunity and inflammation. CDCP1 (CD318), a CD6 ligand, was demonstrated to suppress T-cell activation.⁴⁸ This finding is consistent with our results that a high CDCP1 level was associated with a lower risk of CIR. 

Colocalization analysis is now widely used in investigations of shared genetic etiology. Shared genetic etiology also demonstrates the potential genetic pleiotropy and genetic confounding of traits.²⁴ Colocalization analysis results revealed a pleiotropic genetic relationship between CDCP1 and CIR. The genetic correlations and potential sample overlap between inflammatory cytokines and IR and its subtypes were evaluated by LDSC regression analysis.⁴⁹ LDSC regression analysis results demonstrated that sample overlap did not affect the relationship between inflammatory cytokines and IR and its subtypes. Overall, the relationships between inflammatory cytokines and IR and its subtypes were causal associations rather than genetic correlations. 

There were some limitations to our investigation. First, our findings may not be generalizable to other populations owing to the GWAS data from the populations of European ancestry. Second, we used a genome-wide significance threshold P value ≤ 5 × 10^–6 rather than a P ≤ 5 × 10^–8 to generate sufficient IVs for MR analysis. However, it was deemed a suitable threshold and used for MR analysis.⁵⁰ Third, undoubtedly, multiple testing correction is vital for statistical analysis. After correcting for multiple testing (0.05/91), genetically predicted high levels of eotaxin was significantly associated with an increased risk of ASIR. However, no significant associations were observed between other inflammatory cytokines and iridocyclitis, including ASIR or chronic IR. It is important to highlight that this study is exploratory MR study, aiming to identify more inflammatory cytokines associated with IR, including its subtypes: ASIR and CIR. We also acknowledged that in large-scale genomic data analyses, the combination of a large sample size and numerous hypothesis tests often requires a more stringent P value threshold, which can reduce the likelihood of detecting significant results. Notably, in some previously published genomic MR studies, P values were reported without correction.^51,52 We did not use the multiple testing correction in this exploratory investigation to identify more inflammatory cytokines linked to IR. Finally, the profiling of inflammatory cytokines in inflammatory cytokines profiling of patients with ASIR differed significantly from that of CIR. This might be explained by distinct clinical disease courses. However, this aspect of our research warrants further exploration. 

Conclusions

We used a two-sample MR to explore the causal effects of inflammatory cytokines on the risk of IR and its subtypes. Inflammatory cytokines (eotaxin, FGF23, TRAIL, neurotrophin-3, and IL-2), (eotaxin and TRAIL), and (TNFRSF9, caspase 8, cystatin D, CCL20, and CDCP1) were associated with the risk of IR, ASIR, and CIR, respectively. Our findings shed light on the role of inflammatory cytokines in the occurrence of IR and its subtypes. Further research is warranted to elucidate the precise mechanisms of inflammatory cytokines in IR and its subtypes. 

Acknowledgments

The authors thank the GWAS Catalog and FinnGen biobank for providing us with the inflammatory cytokines and IR GWAS summary data. We also thank all participants and researchers who contributed to the collection of this data. 

Supported by the National Natural Science Foundation Project (82101108, 81970792, and 82171040), the Medical Science and Technology Project of Henan province of China (Grant No. LHGJ20230471), and the Medical Science and Technology Project of the Health Commission of Henan Province (YXKC2020026). 

Author Contributions: J.W. and H.L. designed the study. H.L., F.L., and F.W. collected and analyzed data. H.L. and F.L. conducted the literature search. H.L. wrote the first draft of the paper. F.W., Z.Q., J.W., and F.L. supervised the study. 

Availability of Data and Materials: The GWAS summary data in the study were downloaded from the FinnGen biobank and the GWAS catalog based on the download link provided in the article. Further information, requests, and inquiries can be directed to the corresponding authors on reasonable request. 

Declarations Ethics Approval and Consent to Participate: All participating studies of GWAS provided informed consent to participate in the original study. The Medical Ethics Committee of The First Affiliated Hospital, and College of Clinical Medicine of Henan University of Science and Technology ruled that no formal ethics approval was required for this study. 

Disclosure: H. Liu, None; F. Li, None; F. Wang, None; Z. Hu, None; L. Du, None; J. Wei, None