Glaucoma and brain functional networks: a bidirectional Mendelian randomisation study

Abstract

Objective

Glaucoma is a complex neurodegenerative ocular disorder accompanied by brain functional abnormalities that extend beyond the visual system. However, the causal association between the two remains unclear at present. This study aimed to investigate the potential causal relationships between glaucoma and brain functional networks in order to provide novel insights into the neuropathic mechanism of glaucoma.

Methods and analysis

Based on the genome-wide association studies data of glaucoma and resting-state functional MRI (Rs-fMRI), a bidirectional Mendelian randomisation (MR) analysis was conducted between glaucoma and brain functional networks. Inverse variance weighting was applied as the primary method to estimate causality with false discovery rate correction. Additional sensitivity analyses were conducted to evaluate the robustness of the results.

Results

Forward MR analysis suggested that glaucoma was causally associated with two brain networks between the subcortical cerebellum and the attention or visual network (p=0.022), as well as the default mode and central executive network (p=0.008), but without significance after false discovery rate correction (q>0.1). Reverse MR analysis revealed 19 Rs-fMRI traits related to glaucoma risk, including the salience or central executive network in the frontal region (p=0.0005, q=0.08) and the motor network (p=0.0009, q=0.08) with significant causality.

Conclusions

This MR study revealed potentially causal relationships between glaucoma and brain functional networks. Especially, the functional connectivity of the motor network between the postcentral or precentral areas may potentially lead to increased risk of glaucoma.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Studies using resting-state functional MRI (Rs-fMRI) in patients with glaucoma have indicated altered brain functional measures in the visual pathway and other related brain regions.

• The causal relationships between glaucoma and brain functional networks remain unclear.

WHAT THIS STUDY ADDS

• Nineteen Rs-fMRI traits in resting state were potentially associated with the risk of glaucoma, and glaucoma could result in the alterations of two Rs-fMRI phenotypes.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• This study revealed the bidirectional causal relationships between glaucoma and brain functional networks, thus enhancing the current understanding of the neuropathic mechanism of glaucoma.

• These findings provide novel insights into potential diagnostic biomarkers and neurotherapeutic targets for glaucoma.

Introduction

As a neurodegenerative ocular disease, glaucoma is characterised by the progressive loss of retinal ganglion cells (RGCs) and structural changes in the optic disc, resulting in visual field defects.¹ The number of patients with glaucoma worldwide is projected to reach approximately 112 million by the year 2040,² leading to a heavy burden on both individual’s quality of life and society. The main risk factors for glaucoma mainly include elevated intraocular pressure, myopia and family history of glaucoma, but the pathogenesis mechanism remains unclear. Recent neuroimaging evidence has suggested that the degenerative damage of glaucoma involves the widespread brain regions beyond the entire visual pathway,^{3 4} and some researchers consider glaucoma as a neurodegenerative brain disorder with an unclear pathogenesis.^{3 5 6}

Functional MRI (fMRI) is a crucial neuroimaging technique used to investigate brain function by detecting alterations in blood oxygen levels within neurons.^{7 8} Furthermore, resting-state fMRI (Rs-fMRI) can estimate the functional connectivities between different brain regions in a resting state without the presence of visual stimuli.^{9 10} Studies using Rs-fMRI have indicated reduced neuronal activities in the visual cortex as well as altered brain functional measures in other non-visual related regions in patients with glaucoma.¹¹,¹⁵ However, although these cross-sectional studies have suggested significant associations between brain functional phenotypes and glaucoma, the causality could not be clarified.

Mendelian randomisation (MR) uses genetic variations, particularly single-nucleotide polymorphisms (SNPs), as natural instruments and is an exciting approach to estimate potential causal associations between exposure factors and outcomes.^{16 17} Recently, genome-wide association studies (GWASs) based on Rs-fMRI have been conducted¹⁸ and applied to analyse brain functional alterations in patients with various neurological and psychiatric disorders using MR.^{19 20} Besides, different diseases or phenotypes have also been identified as potential risk factors or outcomes of glaucoma in recent MR studies, such as mental disorder,²¹ cataract²² and type 2 diabetes,²³ revealing the potential mechanisms involved in the glaucomatous neurodegeneration.²⁴,²⁶ However, brain functional networks have not yet been included.

Indeed, our study was the first study to explore the causal associations between glaucoma and brain functional networks using MR analysis based on GWASs data of glaucoma and Rs-fMRI.¹⁸ We aimed to provide novel insights into the neurodegenerative pathogenesis of glaucoma, enhancing our understanding on both diagnosis and treatment of this irreversible ocular disease.

Materials and methods

Genome-wide association studies (GWASs) data and study design

We used publicly available GWASs data on glaucoma from 20 906 cases and 391 275 controls from the FinnGen database (R10) in our MR analysis. The GWASs data of Rs-fMRI was obtained from a previous study [(dataset) Zhao B, Li T, Smith SM, Xiong D, Wang X, Yang Y, Luo T, Zhu Z, Shan Y, Matoba N, Sun Q, Yang Y, Hauberg ME, Bendl J, Fullard JF, Roussos P, Lin W, Li Y, Stein JL, Zhu H. Data from: GWAS summary statistics for 191 resting-state functional MRI (rs-fMRI) traits. Zenodo Record, December 12, 2021. https://zenodo.org/records/5775047¹⁸]. Briefly, the Rs-fMRI GWASs data were collected from four independent studies, which contained 47 276 individuals. After spatial independent component analysis, 1777 neuroimaging phenotypes were obtained (consisting of 76 amplitude traits (nodes), 1695 functional connectivities (edges) and six global functional connectivity measures; online supplemental table 1). At a significance level of 2.8×10^-11 (Bonferroni-adjusted), 191 traits were selected from the 1777 traits for our MR analysis, including 75 node traits (reflecting regional spontaneous neuronal activity), 111 edge traits (quantifying interregional coactivity) and five global functional connectivities. Details of the 191 Rs-fMRI traits are provided in online supplemental table 2. These neuroimaging traits constitute seven fundamental network structures, including the default mode, salience, attention, limbic, central executive, visual and somatomotor networks. Details of the GWASs data applied in our study are shown in table 1.

Table 1. GWASs data sources and sample size of glaucoma and Rs-fMRI.

	GWASs data sources	Sample sizes	Ancestry
Glaucoma	Finngen_R10; https://storage.googleapis.com/finngen-public-data-r10/summary_stats/finngen_R10_H7_GLAUCOMA.gz	20 906 cases and 391 275 controls	EUR
191 Rs-fMRI traits	PubMed ID: 38724650; https://zenodo.org/records/5775047	47 276 individuals	EUR

EUR, European; GWASs, genome-wide association studies; Rs-fMRI, resting-state functional MRI.

MR analysis relies on three key principles: (1) genetic variations must have a significant association with the exposure factor; (2) genetic variations must impact the outcome only through the exposure factor; and (3) genetic variations must be independent of all confounding factors. The basic study design and the brief procedure of our MR analysis are shown in figure 1. We performed a bidirectional MR analysis between glaucoma and 191 Rs-fMRI traits to investigate the potential causal relationship between glaucoma and brain functional networks. Forward MR analysis was conducted with glaucoma as the exposure factor and 191 Rs-fMRI traits as the outcome. In comparison, the risk of glaucoma was regarded as the outcome, and 191 Rs-fMRI phenotypes were exposure factors in the reverse MR analysis.

Figure 1. Study design and procedure of MR analysis. (A) Study design of our bidirectional (forward and reverse) MR analysis between glaucoma and brain functional networks. (B) Procedure of MR analysis between glaucoma and 191 Rs-fMRI phenotypes briefly in three steps: IV selection, forward and reverse analysis, and sensitivity analyses. FDR, false discovery rate; IVs, instrumental variables; IVW, inverse variance weighted; LD, linkage disequilibrium; MR, Mendelian randomisation; Rs-fMRI, resting-state functional MRI.

Instrumental variable selection

According to the foundational principles of MR analysis, we performed the following SNPs filtering process to select instrumental variables (IVs): (1) significant SNPs were associated with exposure as p<1×10^-5; (2) independent SNPs were selected by performing clumping function with the following linkage disequilibrium parameters (r²= 0.001 and a window size of 1000 kb); (3) IVs related to the confounders that affect the outcomes (such as intraocular pressure, obesity, myopia)²⁷ according to the PhenoScanner v2 database were removed; and (4) strong SNPs were assessed as IVs with the F-statistic values≥10 to avoid bias.

Statistical analysis

Inverse variance weighting (IVW) based on multiplicative random effects was applied as the primary analytical method for estimating causal effects in our MR study. The false discovery rate (FDR) was used to correct the P value to a q value. Results were considered statistically significant when q <0.1, whereas p <0.05, but q >0.1 was regarded as suggestive of potential causality.²⁷ The OR indicated the degree of effect of the causal associations between exposure and outcome. The simple mode, MR-Egger, weighted mode and weighted median methods were also used for additional statistical analyses to provide robust causal estimates.

Further sensitivity analyses were performed in our forward and reverse MR analyses. Horizontal pleiotropy of IVs was evaluated via the MR-Egger intercept test (p<0.05), and heterogeneity of the IVs was detected by the Cochran’s Q test (p<0.05). The leave-one-out analysis was used to evaluate the reliability and stability of the MR outcomes, and the results were considered robust when the error bars were all on the same side of the zero bar. All analyses were performed with R software (V.4.3.2) and the ‘TwoSampleMR packages’.²⁸

Patient and public involvement

Patients and/or the public were not involved in the design, conduct, reporting or dissemination plans of this research.

Results

Overview of the study

A bidirectional MR study was performed to explore the causal relationship between glaucoma and the 191 resting-state fMRI phenotypes. Based on our IV selection process, a summary of the SNPs applied in the forward and reverse MR analyses is shown in onlinesupplemental tables 3 4. After performing forward and reverse MR analyses, we identified two and 19 Rs-fMRI traits potentially serving as outcomes and risk factors of glaucoma, respectively (IVW-P value <0.05).

Forward Mendelian randomisation (MR) of glaucoma on resting-state functional (Rs-fMRI) traits

The results of the forward MR study suggested the increased risk of glaucoma leads to brain functional network abnormalities, mainly in two Rs-fMRI traits, Edge-pheno1013 and Edge-pheno956, as shown in figure 2A. Our study showed that glaucoma exhibited a positive association with the brain functional activities located in the cerebellum and temporal or occipital regions, affecting the functional connectivities within subcortical cerebellum and attention or visual networks (IVW OR=1.035, 95% CI 1.005 to 1.065, p=0.022). In the default mode network (DMN) or central executive network (CEN), the risk of glaucoma was positively related to the functional connectivities within the precuneus, angular or cingulate area, and frontal areas (IVW OR=1.044, 95% CI 1.011 to 1.079, p=0.008). Additional analyses of the simple mode, MR-Egger, weighted mode and weighted median indicated consistent causal estimates in the same direction in figure 2B. However, no significant causality was observed after FDR adjustment (q >0.01) for either phenotype. All results of the forward MR analysis are provided in online supplemental table 5.

Figure 2. Causalities in the forward MR analysis between glaucoma and brain functional networks. (A) Two brain functional networks related to the risk of glaucoma in forward MR analysis. (B) Five analysis methods in MR test of Edge-pheno1013 and Edge-pheno956. p <0.05 was regarded as suggestive potential causality, and *q <0.1 (FDR corrected) represented significant causal relationship. FDR, false discovery rate; IVs, instrumental variables; IVW, inverse variance weighted; MR, Mendelian randomisation; OR, odds ratio.

Reverse Mendelian randomisation (MR) of resting-state functional (Rs-fMRI) traits on glaucoma

The results of the reverse MR study identified 19 neuroimaging phenotypes associated with the risk of glaucoma, including seven functional connectivities (edges) and 12 amplitudes (nodes), as shown in figure 3A. All results of the reverse MR analysis are provided in online supplemental table 6. Two node traits with an IVW-P value <0.05 (Node-pheno2 and Node-pheno28) were excluded because of inconsistent directions in our additional four statistical analysis methods. Additionally, Edge-Pheno55 and Node-Pheno16 in figure 3B both indicated a statistically causal relationship between the brain functional network and glaucoma after FDR correction, including the salience or CEN located in the frontal area (IVW OR=1.139, 95% CI 1.058 to 1.225, p=0.0005, q=0.08) and the motor network between the postcentral and precentral areas (IVW OR=1.190, 95% CI 1.074 to 1.319, p=0.0009, q=0.08).

Figure 3. Causalities in the reverse MR analysis between glaucoma and brain functional networks. (A) Nineteen brain functional networks associated with the risk of glaucoma in reverse MR analysis. (B) Five analysis methods in MR test of Edge-Pheno55 and Node-Pheno16. p <0.05 was regarded as suggestive potential causality, and *q <0.1 (FDR corrected) represented significant causal relationship. FDR, false discovery rate; IVs, instrumental variables; IVW, inverse variance weighted; MR, Mendelian randomisation; OR, odds ratio.

Sensitivity analysis

The results of the MR-Egger intercepts and Cochran’s Q test of the 21 identified traits from forward and reverse MR analysis are shown in table 2. No significant heterogeneity was revealed in the Cochran’s Q test (p>0.05), and the MR-Egger intercepts were 0.0013 (p=0.78) and 0.009 (p=0.08) in the forward MR analysis, representing no horizontal pleiotropy of the IVs. As for the reverse MR analysis, no horizontal pleiotropy was indicated in the MR-Egger intercept test (p>0.05). However, the Q statistics test revealed significant heterogeneity of IVs in 14 out of 19 traits (p<0.05), including the Node_Pheno16. Furthermore, after the leave-one-out analysis, none of the 21 traits showed significant IVs driving causality between glaucoma and brain functional networks (online supplemental PDF1).

Table 2. Egger intercepts and Cochran’s Q test in the forward and reverse MR analyses.

	IVs N	Egger intercept	Egger-P value	Cochran’s Q	Q-P value
Outcome (traits)
Edge_Pheno956	52	0.0090	0.081	64.20	0.1015
Edge_Pheno1013	52	−0.0013	0.787	51.94	0.4371
Exposure (traits)
Edge_Pheno1141	54	0.0050	0.379	37.07	0.3737
Edge_Pheno1142	23	0.0104	0.059	37.93	0.6079
Edge_Pheno1256	34	0.0029	0.771	52.79	0.0693
Edge_Pheno1257	40	−0.0061	0.430	48.23	0.0327*
Edge_Pheno286	57	−0.0012	0.899	37.10	0.0428*
Edge_Pheno55	40	−0.0083	0.460	12.06	0.9560
Edge_Pheno903	37	0.0013	0.905	23.18	0.1839
Node_Pheno1	38	0.0027	0.639	115.73	<0.0001***
Node_Pheno16	19	−0.0042	0.439	76.18	0.0202*
Node_Pheno20	60	−0.0040	0.529	99.33	0.0003***
Node_Pheno31	49	−0.0099	0.262	67.91	0.0010**
Node_Pheno32	42	−0.0129	0.084	86.46	<0.0001***
Node_Pheno36	53	−0.0049	0.481	73.53	0.0026**
Node_Pheno40	36	−0.0070	0.340	96.48	<0.0001***
Node_Pheno42	44	−0.0052	0.543	120.22	<0.0001***
Node_Pheno50	25	−0.0108	0.282	62.10	0.0060**
Node_Pheno56	50	−0.0070	0.372	101.92	<0.0001***
Node_Pheno65	43	−0.0068	0.361	115.83	<0.0001***
Node_Pheno68	47	−0.0002	0.979	66.93	0.0036*

*P<0.05; ** P**p<0.01; *** P***p<0.001.

IVs, instrumental variables; MR, Mendelian randomisation.

Discussion

It is widely recognised that glaucoma is a neurodegenerative disease involving alterations in the brain.^{1 3} As neuroimaging techniques advanced rapidly, alterations in functional brain networks have been widely reported in patients with glaucoma. However, these observational studies may have introduced confounding factors and failed to differentiate between outcomes and causes.²⁹ Through forward and reverse MR analyses, our study identified two different Rs-fMRI traits that revealed the impact of glaucoma on brain functional networks potentially, and we also identified 19 Rs-fMRI phenotypes that indicated the possible influence of brain functional networks on the risk of glaucoma.

Previous Rs-fMRI studies of glaucoma have demonstrated alterations in brain functional connectivity in the visual pathway, especially in the occipital area and related brain regions.³⁰,³³ In our forward MR analyses, we found that an elevated risk of glaucoma was possibly related to the increased functional connectivities in the visual network and occipital area among Edge-pheno1013, which may represent compensatory activation or reduced inhibitory influence.³⁰ Additionally, alterations of functional activity affected by glaucoma in non-visual pathways are located in the cerebellum and temporal regions according to Edge-pheno1013, and previous studies also demonstrated a decreased volume of grey matter in these areas among patients with glaucoma,^{34 35} which confirms that glaucomatous neurodegenerative damage may involve the entire brain beyond the visual pathway. We postulated that this might be attributed to anterograde trans-synaptic degeneration triggered by RGC death,³⁶ the key pathological mechanism of glaucomatous damage. It is worth noting that the functional and structural alterations in the visual pathways or related brain regions might be correlated with the severity of glaucomatous damage.^{11 32 37 38} Thus, the neuroimaging phenotypes we identified could be potential biomarkers for both diagnosis and prognosis of glaucoma.^{38 39} Furthermore, our findings suggested that treatment strategies focused on neuroplasticity, such as electrical stimulation of the visual cortex, may represent novel therapeutic approaches to promote neuroregeneration in visual pathways for glaucoma.^{36 40 41}

Additionally, our results indicated that the risk of glaucoma was potentially associated with dysfunction of two basic networks, the DMN and CEN, and different brain regions (precuneus, angular, cingulate and frontal area) according to the Node-pheno956. These brain regions and functional networks are primarily involved in activity inhibition and emotion regulation.⁴² Interestingly, the subcortical cerebellum and attention network in Edge-Pheno1013 also play similar roles.¹⁹ Previous research has shown a strong correlation between these brain areas or networks and an increased risk of psychological disorders, such as anxiety, post-traumatic stress disorder and autism spectrum disorder.⁴³ Abnormal neural activity in these relevant regions or networks has also been reported in patients with glaucoma by fMRI,³⁹⁴⁴,⁴⁶ which may affect their emotional and cognitive functions.^{13 32 47} Our results might explain why patients with glaucoma often experience mood disorders, such as anxiety and depression;^{48 49} nonetheless, more research is required to investigate the specific mechanisms. Unfortunately, our forward MR analysis did not show significance after FDR correction. This may be a result of the selection criteria of IVs and our sample size, but the subsequent sensitivity analysis confirmed the stability of results in forward MR. Therefore, we considered that our forward MR results still provided important evidence for the potential positive causal relationship between glaucoma and brain functional networks.

The results of reverse MR analyses showed that a broader range of neuroimaging abnormalities could affect the risk of glaucoma, including seven types of functional connectivity (edges) and 12 types of amplitude (nodes). In particular, two phenotypes (Edge-Pheno55 and Node-Pheno16) showed statistically significant association after FDR correction. In other words, an increase of 1 SD in the functional connectivity of the motor network between the postcentral or precentral areas was related to a 19.0% elevated risk of glaucoma, while an increase of 1 SD in the neuronal activity of the frontal area increased the risk of glaucoma by 13.9%. Several Rs-fMRI studies have also reported these altered regional neural activities or functional connectivities presented in patients with glaucoma.^{29 50 51} Interestingly, rather than directly involving the visual pathway, these altered brain functional phenotypes causal with glaucoma are also found in different neurodegenerative brain disorders, including Parkinson’s disease, amyotrophic lateral sclerosis, and Alzheimer’s disease.⁵²,⁵⁴ In fact, clinical studies have found that many patients with glaucoma are presented with various central neurodegenerative diseases.⁵⁵,⁵⁷ Moreover, central neurodegenerative diseases are also commonly accompanied by structural abnormalities in the retina or optic nerve.⁵⁸,⁶⁰ However, a recent study has revealed that the associated phenotypes of primary open angle glaucoma and neurodegenerative disorders do not exhibit a significant causal relationship; instead, they share genetic mechanisms and overlap in brain morphology.⁶¹ Thus, we supposed that the high correlation between glaucoma and other neurodegenerative brain diseases might be due to common pathological mechanisms,^{55 62 63} and glaucoma is a potential neurodegenerative disorder originating from the central nervous system in some aspects. This might help explain why glaucomatous damage occurs even within acceptable intraocular pressure values.⁶⁴ Further studies are still required to explore the specific neuropathological mechanisms underlying functional network-induced glaucoma, and our findings based on reverse MR analysis provided valuable directions and targets for future investigations.

However, Cochran’s Q test revealed mostly significant heterogeneity in the reverse MR analysis of the present study, showing potential weakness and invalidity of IVs,⁶⁵ included in Node-Pheno16. We speculate that it could be attributed to the large sample size and different population distributions, and we also further used weight median as the main statistical method to assess the reliability (online supplemental table 6, p<0.05). Considering that our MR analysis was based on multiplicative random effects and assessed by the leave-one-out sensitivity test, the heterogeneity has a relatively small effect overall on our MR results.

An advantage of our study was the non-invasive evaluation of the degree of functional activities and connectivities in specific brain regions with the conduction of fMRI. At the same time, the MR method overcame the financial constraints of employing fMRI and the inherent limitations of cross-sectional studies. However, there were also several limitations in our MR analyses. First, the population was limited to European individuals despite the regional distributions of different types of glaucoma being varied. Second, different types of glaucoma differ in pathogenesis, and further subgroup studies are required. Third, Rs-fMRI is based on anatomical localisation and functional partitions in the brain. However, there may still be an undefined neural network or component related to glaucoma. Fourth, we employed more liberal thresholds (eg, controlling for SNPs with p<1×10^-5 and the FDR of IVW<0.1) to perform our MR study, which may increase the false-positive rate, while enabling a more extensive and profound evaluation of the association between glaucoma and brain functional networks.⁶⁶ Last but not the least, robust causal inference needs triangulation of evidence than just MR analysis alone.⁶⁷ This is because as a high-level observational study, MR analysis still suffers from biases, such as population stratification.⁶⁸ Thus, additional natural experiments and prospective clinical studies on Rs-fMRI in glaucoma are still needed. Importantly, our findings can serve as a crucial component of the triangulation framework to establish conclusive causality between glaucoma and brain functional networks.

Overall, our MR analyses identified a total of 21 Rs-fMRI phenotypes that indicated a potential causal relationship between glaucoma and brain functional networks. Especially, the functional connectivity of the motor network between the postcentral or precentral areas may potentially lead to increased risk of glaucoma. These findings help to enhance our understanding of the neuropathic damage mechanism, improve diagnostic capabilities and advance neuroprotective strategies of glaucoma.

Data availability statement

Data are available in a public, open access repository. All data relevant to the study are included in the article or uploaded as supplementary information.