Genetic correlation between circulating cytokines and risk of three ophthalmic diseases: a bidirectional two-sample Mendelian randomization study

Abstract

Purpose

Pathogenesis and the associated risk factors of cataracts, glaucoma, and age-related macular degeneration (AMD) remain unclear. We aimed to investigate causal relationships between circulating cytokine levels and the development of these diseases.

Patients and methods

Genetic instrumental variables for circulating cytokines were derived from a genome-wide association study of 8293 European participants. Summary-level data for AMD, glaucoma, and senile cataract were obtained from the FinnGen database. The inverse variance weighted (IVW) was the main Mendelian randomization (MR) analysis method. The Cochran’s Q, MR-Egger regression, and MR pleiotropy residual sum and outlier test were used for sensitivity analysis.

Results

Based on the IVW method, MR analysis demonstrated five circulating cytokines suggestively associated with AMD (SCGF-β, 1.099 [95%CI, 1.037–1.166], P = 0.002; SCF, 1.155 [95%CI, 1.015–1.315], P = 0.029; MCP-1, 1.103 [95%CI, 1.012–1.202], P = 0.026; IL-10, 1.102 [95%CI, 1.012–1.200], P = 0.025; eotaxin, 1.086 [95%CI, 1.002–1.176], P = 0.044), five suggestively linked with glaucoma (MCP-1, 0.945 [95%CI, 0.894–0.999], P = 0.047; IL1ra, 0.886 [95%CI, 0.809–0.969], P = 0.008; IL-1β, 0.866 [95%CI, 0.762–0.983], P = 0.027; IL-9, 0.908 [95%CI, 0.841–0.980], P = 0.014; IL2ra, 1.065 [95%CI, 1.004–1.130], P = 0.035), and four suggestively associated with senile cataract (TRAIL, 1.043 [95%CI, 1.009–1.077], P = 0.011; IL-16, 1.032 [95%CI, 1.001–1.064], P = 0.046; IL1ra, 0.942 [95%CI, 0.887–0.999], P = 0.047; FGF-basic, 1.144 [95%CI, 1.052–1.244], P = 0.002). Furthermore, sensitivity analysis results supported the above associations.

Conclusion

This study highlights the involvement of several circulating cytokines in the development ophthalmic diseases and holds potential as viable pharmacological targets for these diseases.

Mendelian randomization, circulating cytokines, ophthalmic diseases, causality

Introduction

Cataracts, glaucoma, and age-related macular degeneration (AMD) are the three leading causes of visual impairment and blindness in older adults [1]. In 2020, approximately 33.6 million adults aged ≥ 50 years were blind globally [2], and the leading causes of vision impairment in these populations were cataracts, followed by glaucoma and AMD [3]. In older people, visual impairment and blindness can lead to a reduced quality of life [4] and serious complications, such as depression [5], cognitive impairment [6], and even death [7].

Cytokines are a class of small-molecule soluble proteins secreted by various immune or non-immune cells [8], and play an important role in regulating cell growth and differentiation [9]. Numerous studies have demonstrated a significant correlation between cytokines and the risk of AMD [10], glaucoma [11], and cataracts [12]. The findings suggest preventive measures or interventions targeting inflammatory pathways may be viable strategies to reduce the risk of developing these ophthalmic diseases. However, the sample sizes in these studies mentioned above were relatively small, and results could be affected by various factors, such as unmeasured confounding, reverse causation, and biases.

Mendelian randomization (MR) analysis uses single nucleotide polymorphisms (SNPs) as genetic instrumental variables (IVs) for exposure factors to determine their potential causal relationship with outcomes [13]. Notably, genetic IVs are unmodifiable [14], allowing for the analysis of unbiased estimates of the effect of exposure on outcomes and alleviating concerns about reverse causality [15, 16]. Recently, a meta-analysis of genome-wide association studies (GWASs) assessed the genetic basis of 41 circulating cytokines [17]; these findings opened up new avenues for exploring potential associations between these factors and prevalent ophthalmic conditions. Therefore, we aimed to assess potential causal associations between these 41 circulating cytokines and the risk of developing the aforementioned ophthalmic diseases using a bidirectional two-sample MR design.

Results

MR analysis of confounders associated with AMD, glaucoma, and senile cataract

The effects of various confounders on the presence of AMD, glaucoma, and senile cataract were analyzed using two-sample MR. The analysis showed that BMI, hypertension, and diabetes were relevant to the presence of AMD (BMI, 1.118 [95%CI, 1.005–1.244], P = 0.041), glaucoma (hypertension: 2.619 [95%CI, 1.384–4.958], P = 0.003; diabetes: 5.986 [95%CI, 1.810–19.791], P = 0.003), and senile cataract (BMI: 1.106 [95%CI, 1.041–1.175], P = 0.001; hypertension: 1.837 [95%CI, 1.199–2.816], P = 0.005; diabetes: 3.853 [95%CI, 1.595–9.305], P = 0.003) (Table 1).

Table 1

MR results of confounders for AMD, glaucoma, and senile cataract.

Exposure	Outcome	Method	nSNP	Beta	SE	P-value	OR (95%CI)
IOP	AMD	IVWmre	12	0.007	0.010	0.474	1.007 (0.988–1.027)
	Glaucoma	IVWmre	6	0.010	0.007	0.139	1.010 (0.997–1.024)
	Senile Cataract	IVWmre	12	0.000	0.004	0.982	1.000 (0.992–1.008)
Alcohol	AMD	IVW	4	−0.287	0.396	0.469	0.751 (0.346–1.631)
	Glaucoma	IVWmre	4	−0.011	0.499	0.982	0.989 (0.372–2.632)
	Senile Cataract	IVW	4	−0.282	0.193	0.144	0.754 (0.516–1.101)
BMI	AMD	IVWmre	474	0.112	0.055	0.041	1.118 (1.005–1.244)
	Glaucoma	IVWmre	473	0.014	0.045	0.756	1.014 (0.928–1.108)
	Senile Cataract	IVWmre	471	0.101	0.031	0.001	1.106 (1.041–1.175)
Smoking	AMD	IVWmre	33	0.857	0.480	0.074	2.356 (0.920–6.037)
	Glaucoma	IVWmre	32	−0.015	0.327	0.963	0.985 (0.519–1.871)
	Senile Cataract	IVW	33	0.118	0.217	0.586	1.126 (0.735–1.724)
Diabetes	AMD	IVWmre	57	0.384	0.811	0.636	1.468 (0.300–7.198)
	Glaucoma	IVWmre	60	1.789	0.610	0.003	5.986 (1.810–19.791)
	Senile Cataract	IVWmre	58	1.349	0.450	0.003	3.853 (1.595–9.305)
Hypertension	AMD	IVW	63	0.879	0.471	0.062	2.409 (0.956–6.069)
	Glaucoma	IVW	64	0.963	0.326	0.003	2.619 (1.384–4.958)
	Senile Cataract	IVW	63	0.608	0.218	0.005	1.837 (1.199–2.816)

Abbreviations: IVW_mre, inverse variance weighted (multiplicative random effects); IOP, Intraocular pressure; BMI, Body mass index; AMD, age-related macular degeneration; SNPs, single nucleotide polymorphisms; SE, standard error; OR, odds ratio; CI, confidence interval.

MR analysis of 41 circulating cytokines associated with AMD, glaucoma, and senile cataract

We screened 27 circulating cytokines with a total of 71 SNPs when the screening threshold was (r² < 0.001, kb = 10 000, P < 5 × 10⁻⁰⁸), and only 9 circulating cytokines with a number of SNPs > 3 (Supplementary Material, Table S1). After relaxing the threshold to P < 5 × 10⁻⁰⁶, we extracted 452 SNPs associated with 41 circulating cytokines (Supplementary Material, Table S2). After removing confounder-related SNPs and outlier SNPs (Supplementary Material, Table S3), a total of 1050 SNPs for three ophthalmic diseases were identified after harmonizing alleles between circulating cytokines and three ophthalmic diseases, and they were strongly associated with 41 circulating cytokines, with an F-statistic range of IV of 11–790, indicating the robustness of IVs (Supplementary Material, Table S4; AMD: 350 SNPs; glaucoma: 352 SNPs; senile cataract: 348 SNPs).

Based on the IVW method, volcano plot revealed that six circulating cytokines were significantly associated with AMD, five with glaucoma, and four with senile cataract (Fig. 1). Detailed MR results are shown in Supplementary Material, Table S5. Reverse MR results revealed significant associations between AMD and the cytokine β-NGF (0.932 [95% CI, 0.872–0.997], P = 0.040) and between senile cataract and the cytokine eotaxin (1.123 [95% CI, 1.006–1.253], P = 0.038) (Supplementary Material, Table S6). After removing reverse-differential circulating cytokines, we finally identified five circulating cytokines suggestively associated with AMD (SCGF-β, 1.099 [95%CI, 1.037–1.166], P = 0.002; SCF, 1.155 [95%CI, 1.015–1.315], P = 0.029; MCP-1, 1.103 [95%CI, 1.012–1.202], P = 0.026; IL-10, 1.102 [95%CI, 1.012–1.200], P = 0.025; eotaxin, 1.086 [95%CI, 1.002–1.176], P = 0.044), five suggestively linked with glaucoma (MCP-1, 0.945 [95%CI, 0.894–0.999], P = 0.047; IL1ra, 0.886 [95%CI, 0.809–0.969], P = 0.008; IL-1β, 0.866 [95%CI, 0.762–0.983], P = 0.027; IL-9, 0.908 [95%CI, 0.841–0.980], P = 0.014; IL2ra, 1.065 [95%CI, 1.004–1.130], P = 0.035), and four suggestively associated with senile cataract (TRAIL, 1.043 [95%CI, 1.009–1.077], P = 0.011; IL-16, 1.032 [95%CI, 1.001–1.064], P = 0.046; IL1ra, 0.942 [95%CI, 0.887–0.999], P = 0.047; FGF-basic, 1.144 [95%CI, 1.052–1.244], P = 0.002) (Fig. 2). The analysis results of the four MR Methods are presented in Supplementary Material, Table S7. Cochran’s IVW Q test did not find evidence of heterogeneity for 14 circulating cytokines (all P-values for Cochran’s Q test were > 0.05, Supplementary Material, Table S8), and MR-Egger regression analyses did not detect potential directional pleiotropy across SNPs (intercept P-values > 0.05, Supplementary Material, Table S9). In addition, MR-PRESSO results indicated that no SNPs had pleiotropy. The scatter plot analysis of individual SNPs effects showed that as the effect of individual SNPs on circulating cytokine levels increased, the promotional effect of individual SNPs on ophthalmic diseases also increased, and vice versa. This is consistent with the results of the MR analysis (Supplementary Material, Fig. S1). MR leave-one-out tests showed that the effect values were stable after excluding a single SNP from the 14 circulating cytokines (all points lie on the side of the 0 scale) (Supplementary Material, Fig. S2). However, we did not find suggestive associations of 14 circulating cytokines with the risk of the three ophthalmic diseases in the UK Biobank, and detailed results are shown in Supplementary Material, Table S7.

Figure 1

Volcano plot of MR results for the association between 41 circulating cytokines and the risk of AMD, glaucoma, and senile cataract. Abbreviations: AMD, age-related macular degeneration; OR, odds ratio; β-NGF, beta-nerve growth factor; TRAIL, TNF-related apoptosis inducing ligand; TNF-β, tumor necrosis factor beta; SCGF-β, stem cell growth factor beta; SCF, stem cell factor; IL-16, interleukin-16; MCP-1, monocyte chemoattractant protein-1; IL-10, interleukin-10; IL1ra, interleukin-1-receptor antagonist; IL-1-β, interleukin-1-beta; IL2ra, interleukin-2 receptor antagonist; FGF-basic, fibroblast growth factor basic.

Figure 2

Forest plot showing Mendelian randomization results of the causal relationship of circulating cytokines with the risk of AMD, glaucoma, and senile cataract. Abbreviations: AMD, age-related macular degeneration; SNPs, single nucleotide polymorphisms; OR, odds ratio; CI, confidence interval; SCGF-β, stem cell growth factor beta; SCF, stem cell factor; MCP-1, monocyte chemoattractant protein-1; IL-10, interleukin-10; β-NGF, beta-nerve growth factor; MIG, monokine induced by gamma interferon; IL-12p70, interleukin-12p70; IL1ra, interleukin-1-receptor antagonist; IL-1-β, interleukin-1-beta; IL2ra, interleukin-2 receptor antagonist; TRAIL, TNF-related apoptosis-inducing ligand; IP10, interferon gamma-induced protein 10; IL-9, interleukin-9; IL-16, interleukin-16; FGF-basic, fibroblast growth factor basic.

MR analysis of 41 circulating cytokines associated with dry AMD, wet AMD, PACG, and PACG

We further analyzed the correlation of 41 circulating cytokines with dry AMD, wet AMD, PACG, and PACG, and showed that four circulating cytokines showed suggestive association with dry AMD (β-NGF, 1.162 [95%CI, 1.032–1.307], P = 0.013; SCGF-β, 1.127 [95%CI, 1.052–1.208], P = 0.001; MIG, 1.087 [95%CI, 1.002–1.180], P = 0.045; MCP-1, 1.127 [95%CI, 1.022–1.243], P = 0.017), two suggestively associated with wet AMD (IL-12p70, 1.129 [95%CI, 1.012–1.259], P = 0.030; IL-10, 1.137 [95%CI, 1.009–1.281], P = 0.036), five suggestively associated with POAG (TRAIL, 1.067 [95%CI, 1.009–1.129], P = 0.023; IP-10, 1.143 [95%CI, 1.034–1.263], P = 0.009; IL1ra, 0.866 [95%CI, 0.771–0.973], P = 0.016; IL-1-β, 0.833 [95%CI, 0.723–0.961], P = 0.012, IL-9, 0.887 [95%CI, 0.789–0.997], P = 0.044), and one suggestively associated with PACG (eotaxin, 1.247 [95%CI, 1.008–1.544], P = 0.042) (Supplementary Material, Fig. S3 & Fig. 2). Detailed results of the four MR methods are shown in Supplementary Material, Table S7. The scatter plot analysis of the causal effects of individual SNPs for 12 circulating cytokines were consistent with the results of the MR analysis (Supplementary Material, Fig. S4). MR leave-one-out sensitivity analysis showed that removing a specific SNP of 12 cytokine-associated SNPs did not change the results (Supplementary Material, Fig. S5).

Colocalization analysis of regions of shared genetic variation between circulating cytokines and ophthalmic diseases

The results of colocalization analysis indicated the presence of shared genetic variants between SCGF-β and AMD (PPH₄ = 0.964, rs187503377, nearest_genes: NT5DC3). No region of shared genetic variation existed between the other circulating cytokines and ophthalmic diseases, suggesting that the suggestive associations between circulating cytokines and ophthalmic diseases are entirely cytokine-determined (Table 2 & Supplementary Material, Figs S6 and S7).

Table 2

Overview of MR analysis for circulating cytokines with suggestive causal relationships with AMD, dry AMD, wet AMD, glaucoma, POAG, PACG, and senile cataract.

Exposure	Outcome	Method	Significance threshold: P < 5e-06						Colocalization
			nSNP	P-value	OR (95%CI)	PH	PP	PG	nSNP	PPH4
SCGF-β	AMD	IVW	14	0.002	1.099 (1.037,1.166)	0.429	0.921	0.386	534	0.964
SCF		IVW	8	0.029	1.155 (1.015,1.315)	0.777	0.308	0.796	422	0.010
MCP1		IVW	14	0.026	1.103 (1.012,1.202)	0.352	0.159	0.331	309	0.007
IL-10		IVW	9	0.025	1.102 (1.012,1.200)	0.863	0.589	0.829	344	0.013
Eotaxin		IVW	15	0.044	1.086 (1.002,1.176)	0.403	0.694	0.454	273	0.021
β-NGF	Dry AMD	IVW	7	0.013	1.162 (1.032,1.307)	0.708	0.713	0.743	379	0.013
SCGF-β		IVW	14	0.001	1.127 (1.052,1.208)	0.733	0.720	0.727	534	0.656
MIG		IVW	14	0.045	1.087 (1.002,1.180)	0.510	0.418	0.543	308	0.010
MCP1		IVW	14	0.017	1.127 (1.022,1.243)	0.859	0.158	0.860	309	0.011
IL-12p70	Wet AMD	IVW	10	0.030	1.129 (1.012,1.259)	0.177	0.622	0.324	341	0.029
IL-10		IVW	10	0.036	1.137 (1.009,1.281)	0.297	0.852	0.434	344	0.029
MCP1	Glaucoma	IVW	14	0.047	0.945 (0.894,0.999)	0.962	0.621	0.969	309	0.010
IL1ra		IVW	6	0.008	0.886 (0.809,0.969)	0.216	0.319	0.277	416	0.372
IL-1-β		IVW	4	0.027	0.866 (0.762,0.983)	0.135	0.457	0.395	440	0.343
IL-9		IVW	6	0.014	0.908 (0.841,0.980)	0.784	0.439	0.828	597	0.014
IL2ra		IVW	6	0.035	1.065 (1.004,1.130)	0.257	0.650	0.399	445	0.051
TRAIL	POAG	IVW	14	0.023	1.067 (1.009,1.129)	0.548	0.707	0.549	264	0.114
IP10		IVW	9	0.009	1.143 (1.034,1.263)	0.293	0.423	0.327	307	0.027
IL1ra		IVW	6	0.016	0.866 (0.771,0.973)	0.661	0.641	0.686	-	-
IL-1-β		IVW	4	0.012	0.833 (0.723,0.961)	0.467	0.727	0.550	-	-
IL-9		IVW	6	0.044	0.887 (0.789,0.997)	0.432	0.448	0.534	-	-
Eotaxin	PACG	IVW	15	0.042	1.247 (1.008,1.544)	0.329	0.428	0.378	273	0.121
TRAIL	Senile Cataract	IVW	14	0.011	1.043 (1.009,1.077)	0.093	0.352	0.114	264	0.185
IL-16		IVW	10	0.046	1.032 (1.001,1.064)	0.441	0.150	0.324	276	0.009
IL1ra		IVW	6	0.047	0.942 (0.887,0.999)	0.297	0.238	0.327	416	0.013
FGF-basic		IVW	5	0.002	1.144 (1.052,1.244)	0.528	0.215	0.614	379	0.009

P_H, P-value for heterogeneity test; P_P, P-value for pleiotropy test; P_G, P-value for global test (MR-PRESSSO). Abbreviations: AMD, age-related macular degeneration; POAG, primary open-angle glaucoma; PACG, primary angle closure glaucoma; IVW, inverse variance weighted; SNPs, single nucleotide polymorphisms; OR, odds ratio; CI, confidence interval; PPH₄, posterior probability (PP) of H₄ (PPH₄).

Joint analysis of SNP-gene mapping of circulating cytokines with suggestive associations with three ophthalmic diseases and differential genes from public databases

The MR results demonstrated that MCP-1 was associated with both AMD and glaucoma and IL1ra with glaucoma and senile cataract. We analyzed the genes corresponding to the SNPs using the ENSEMBL online VEP tool (asia.ensembl.org). Detailed information is shown in Supplementary Material, Table S10. Differential genes in the GEO datasets GSE221042, GSE27276, and GSE213546 were analyzed by GEO2R (Supplementary Material, Tables S11–S13), and co-analyzed with SNP mapping genes to identify key genes. The results showed that the gene ACKR2 was associated with AMD, genes RDM1, and MACROD2 with senile cataract, whereas no genes associated with glaucoma were identified (Fig. 3).

Figure 3

Venn diagram of intersecting genes of SNP mapped genes with differential genes from the GEO database. Abbreviations: AMD, age-related macular degeneration; GEO, gene expression omnibus.

Relevant drugs for differential circulating cytokines and predictions for key genes

Of the 12 circulating cytokines showing suggestive causal associations with AMD, glaucoma, and senile cataract, six (MCP-1, IL-10, eotaxin, IL1ra, IL-1-β, and FGF-basic) had records of past or current clinical drug development programs, and four drugs (canakinumab, pentosan polysulfate, foreskin keratinocyte (neonatal) and sucralfate) were determined to have pharmacologic effects on two circulating cytokines (IL-1-β and FGF-basic) (Supplementary Material, Table S14). Indications and related conditions for these drugs include cryopyrin-associated periodic syndromes, chronic interstitial cystitis, prostatitis, and peptic ulcer. Conversely, other remaining circulating cytokines, for which our MR analysis highlighted evidence of an association, have limited observational epidemiologic support.

Discussion

In this study, we conducted an MR analysis to explore the causal relationship between circulating cytokines and three ophthalmic diseases. The results revealed suggestive evidence that circulating cytokines SCGF-β, SCF, MCP-1, IL-10, and eotaxin were associated with a higher AMD risk. Circulating cytokines MCP-1, IL1ra, IL-1-β, and IL-9 were associated with a lower glaucoma risk, and IL2ra was associated with higher glaucoma risk. Additionally, circulating cytokines TRAIL, IL-16, and FGF-basic were associated with a higher senile cataract risk, and IL1ra was associated with lower senile cataract risk. Most circulating cytokines are related to inflammation, which is a major mechanism in ophthalmic disease development. Drugs targeting circulating cytokines β-NGF, FGF-basic, and IL1ra have been approved for treating ocular surface diseases [18], and inhibitors targeting eotaxin are also currently in clinical trials for treating AMD [19]. Other circulating cytokines, such as IL-16 and IL-1β, have been extensively studied in three ophthalmic diseases [20, 21], and may be potential drug targets for treating these ophthalmic diseases in the future.

Circulating cytokines MCP-1 and IL-1ra are present concurrently in two ophthalmic diseases, suggesting that a common therapeutic target may exist for both diseases. MCP-1 plays a pivotal role in the inflammatory process by enhancing other inflammatory factors’ expression and facilitating inflammatory cell migration and infiltration [22], and it has been demonstrated that MCP-1 is highly expressed in AMD and enhances macrophage RAW 264.7 cells migration and cytokines TNF-α, IL-1β, and VEGF secretion to promote the progression of AMD [23]. This finding is consistent with our results. However, while several experimental studies have suggested that MCP-1 is positively correlated with glaucoma [24, 25], our study revealed that MCP-1 was negatively correlated with the glaucoma risk, probably because different types of glaucoma may be caused by a variety of pathogenetic mechanisms with distinct targets and pathways. The MR analysis results of disease typing verified our conjecture that MCP-1 was not associated with either POAG or PACG. We also found a negative correlation between cytokine IL-1ra and the risk of glaucoma, POAG, and senile cataract. Oribio-Quinto et al. [26] showed that IL-1ra was elevated in POAG and decreased in primary congenital glaucoma. However, it has also been shown that IL-1ra is not significantly associated with various ophthalmic diseases, including retinitis pigmentosa, AMD, glaucoma, and cataract disease [27]. The role of IL-1ra on ophthalmic disease was controversial and required more in-depth studies. It is worth mentioning that the MR results showed an association between IP-10 and an elevated risk of POAG, which is consistent with findings of Oribio-Quinto et al. [26]. In addition to some of these suggestive causal cytokines, we still need to focus on those cytokines, including VEGF, TNF-α, and TNF-β, which are essential for the development of the retina and optic nerve [28, 29].

In our study, we identified several disease-related genes, including ACKR2, RDM1, MACROD2, and the colocalized gene NT5DC3. Notably, these genes have not been intensively investigated in the pathogenesis of the three ophthalmic diseases, but our findings suggested the potential roles of these genes. ACKR2 is a chemokine binding protein that can regulate leukocyte infiltration and cause an inflammatory response [30]. The study revealed that ACKR2 plays a role in fine-tuning the inflammatory response and neovascularisation in herpes interstitial keratitis [31]. Genetic variants in MACROD2 have been shown to increase susceptibility to thyroid-related orbital disease [32]. RDM1 and NT5DC3 have not been studied in ophthalmic diseases. Studies have shown that RDM1 is associated with the MEK/ERK signaling pathway [33], which is associated with optic nerve and retinal vasculopathy [34, 35]. NT5DC3 is a target of lactoferrin action and is involved in the pathogenesis of cancer [36]. Extensive studies of lactoferrin in ophthalmic diseases [37, 38] have demonstrated the potential of NT5DC3 for research in ophthalmic diseases.

The strengths of the MR analysis include the large sample sizes of ophthalmic diseases and the involvement of the wide range of circulating cytokines. MR analysis of disease typing was performed to identify differences in circulating cytokines associated with disease heterogeneity. Moreover, we combine public databases to investigate cytokine-targeting drugs as well as key genes. However, it is essential to acknowledge the primary limitation of this MR analysis: the low number of IVs (nSNP<5) for certain circulating cytokines, e.g. TNF-β, TNF-α, MCP3, IL-8, and IL-1-β; this limitation could potentially impact our ability to refute the original hypothesis conclusively. Moreover, the effectiveness of MR sensitivity analysis (MR-Egger and MR-PRESSO) relies on a substantial number of independent SNPs. Although MR sensitivity analysis showed MR results to be stable, the number of SNPs in the included circulating cytokines ranged from 4 to 15. Additionally, we did not find an association between these circulating cytokines and ophthalmic diseases in the UKBB database, possibly due to fewer SNPs between ophthalmic diseases and circulating cytokines sourced from the UKBB database or differences in disease typing. Thus, caution should be exercised in interpreting these analytical results. Potential synergies between the circulating cytokines studied were not considered in this study, and some circulating cytokines without available genetic IVs (e.g. IL-3 [39], interferon-γ [40], and NLRP3 [21]) that might be associated with AMD, glaucoma, and senile cataract. To address these gaps, future endeavors involving larger cytokine concentration single-trait and multi-trait GWAS and MR studies with individual-level data hold promise in shedding light on these unexplored associations. In conclusion, this study revealed a suggestive causal relationship between several circulating cytokines and the risk of these ophthalmic diseases. Further studies are warranted to assess the feasibility of these circulating cytokines as potential candidates for disease-preventive therapeutics or lifestyle interventions.

Material and methods

Study design

The detailed study flowchart is shown in Supplementary Material, Fig. S8. First, we performed MR analysis of confounders and 41 circulating cytokines for the three ophthalmic diseases. Second, after removing confounders-related SNPs and outlier SNPs, we performed a new MR analysis of 41 circulating cytokines and the three ophthalmic diseases. Third, the direction of causal association was further determined by reverse MR. Finally, conducted the SNP-Gene mapping and joint analysis with the differential genes of three ophthalmic diseases from the Gene Expression Omnibus (GEO) database.

Genetic instrument variable selection

To construct a valid IV, three basic assumptions must be met: (1) a correlation between the IVs and exposure, (2) the IVs’ independence from any confounders associated with the relationship between exposure and outcome, and (3) the IVs’ effects on the outcome only through the exposure pathway. Due to the use of r² < 0.001 within a 10 000-kb window, P < 5 × 10⁻⁰⁸, and 1000G_EUR_Phase3 (GRCh37/hg19) as a reference panel for removing the cascading imbalance between SNPs, some circulating cytokines have no associated SNPs [41]. Consequently, we adjusted the significance threshold for these cytokines to P < 5 × 10⁻⁰⁶, r² < 0.001, kb = 10 000. Furthermore, we calculated the F-statistic for each IV to assess its strength and avoid weak instrumental bias [42]. An F-statistic value of > 10 indicates a strong correlation between the IV and the exposure factor [43]. In the current analysis, all IVs were robust instruments with F-statistics > 10.

Data sources

The summary-level GWAS datasets of 41 circulating cytokines were sourced from Ahola-Olli et al’s work, including three independent population cohorts (The Cardiovascular Risk in Young Finns Study, FINRISK1997, and FINRISK2002), involving up to 8293 Finnish individuals. All three cohorts were adjusted for age, sex, and the first 10 principal components (PCs) to examine genetic associations between SNPs and concentrations of 41 cytokines [17].

The summary-level GWAS datasets of primary outcomes (AMD, glaucoma, and senile cataract), and secondary outcome (dry AMD, wet AMD, primary open-angle glaucoma [POAG], and primary angle-closure glaucoma [PACG]) were obtained from the FinnGen database (Release 9), and the first 10 PCs, genetic factors, age, and sex were adjusted [44]. The FinnGen Study is a GWAS meta-analysis involving 9 biobanks with very limited overlap (< 4%) with the circulating cytokines GWAS study. Therefore, we considered the risk of bias due to sample overlap to be minimal [45]. The summary statistics of the three ophthalmic diseases used for validation and confounders included intraocular pressure (IOP), body mass index (BMI), diabetes, hypertension, alcohol consumption, and smoking can be found at the IEU OpenGWAS project (https://gwas.mrcieu.ac.uk/). Detailed information on these data is provided in Supplementary Material, Table S15. Inclusion criteria for AMD, glaucoma, and senile cataract were based on the International Statistical Classification of Diseases and Related Health Problems 10th Revision—the WHO Version for 2016 designation (Supplementary Material, Table S16). All GWAS studies included in the analysis had obtained relevant ethical review approvals. No additional ethical review approval was required for this study.

Colocalization analysis

The presence of shared causal variants between exposure and outcome was analysed using GWAS-GWAS colocalization. These analyses were performed using the packages “coloc”, and “locuscomparer” [46, 47], with a reference panel (GWAS gene window default of 50 kb, P < 5 × 10⁻⁰⁶). A posterior probability (PP) of H₄ (PPH₄) > 0.8 was identified as evidence in favor of colocalization.

Systematic review of publicly available databases

We used DrugBank (go.drugbank.com/, accessed on February 29, 2024) to review circulating cytokines associated with three ophthalmic diseases and identify cytokine-associated drug targets [48]. Subsequently, for the drugs that were identified, we conducted in-depth searches within clinical trial registry databases (clinicaltrials.gov, accessed on February 29, 2024) [49]. Additionally, the genes linked to important circulating cytokines were identified using GEO2R analysis of the GSE221042, GSE27276, and GSE213546 datasets within the GEO database (accessed on August 28, 2023) [50]. We performed SNP-gene mapping by Variant Effect Predictor (VEP), a powerful online tool from the ENSEMBL website (https://asia.ensembl.org/Tools/VEP, Ensembl release 111) that provides extensive genome annotations [51].

Statistical analysis

We used four MR analysis methods, inverse variance weighted (IVW), MR-Egger, weighted median, and weighted mode to assess the association of circulating cytokine levels with the risk of AMD, glaucoma, and senile cataract [52, 53]. The IVW was used as the primary analysis to combine causal estimates (the beta values and standard errors) of SNPs for circulating cytokines and ophthalmic diseases using a fixed-effects meta-analysis model. The F-statistic and the proportion of variance explained (r²) were used to determine the strength of each SNP associated with 41 circulating cytokines [54]. We conducted several sensitivity tests, including Cochran’s Q to test for heterogeneity among IVs of circulating cytokines [55], MR Egger regression [53] and MR-PRESSO [56] tests for horizontal pleiotropy assessment and outliers removal. Heterogeneity and pleiotropy were assumed not to exist if P > 0.05, with heterogeneity present, the multiplicative random effects model was chosen. SNPs suggesting opposite causal directions were removed by MR-Steiger filtering [57], and SNPs associated with confounders were eliminated by Phenoscanner [58].

We corrected for multiple testing using the Bonferroni method; associations with P-values below 0.0012 (0.05/41) were considered strong evidence of an association, and associations with P-values between 0.0012 and 0.05 were considered suggestive. All analyses were performed using the Two Sample MR package (version 0.5.7), MR package (version 0.8.0), and MR-PRESSO package (1.0) in R Software 4.3.1 (R-project.org) [56, 59].

Acknowledgements

We thank all the researchers and participants for publicly making the GWAS summary data available.

Author contributions

C.Y.H. and Z.X. contributed to the design of the study. Z.X., F.Q.Q., and C.Y.Y. obtained and analyzed the data. J.Y.P. and L.X.L. performed tabular and graphical presentations and interpreted the analyzed results. Z.X., F.Q.Q., and C.L. were the main contributors to the manuscript writing. C.Y.H., F.Q.Q., and J.Y.P. revised and reviewed the manuscript. All authors reviewed and approved the final version of the manuscript.

Conflict of interest statement: The author(s) report no conflicts of interest in this work.

Funding

This work was supported by the National Natural Science Foundation of China (82271050 & 82301175).

Data availability

All the data used in this work have been documented in supplementary materials accompanying the manuscript, and the raw data were obtained from public databases (IEU Open Gwas Project (gwas.mrcieu.ac.uk); FinnGen database (r9.finngen.fi); GEO dataset: GSE221042, GSE27276, and GSE213546; DrugBank (go.drugbank.com/)) as well as from published articles (PMID: 27989323).