Identification of systemic biomarkers and potential drug targets for age-related macular degeneration

Purpose Since age-related macular degeneration (AMD) is tightly associated with aging and cellular senescence, objective of this study was to investigate the association between plasma levels of senescence-related proteins (SRPs) and risk of AMD. Design The whole study was based on two-sample Mendelian randomization (MR) analysis. Methods For MR analysis, the primary approach for MR analysis was the inverse-variance weighted (IVW) method and the heterogeneity and pleiotropy of results were tested. The instrumental single-nucleotide polymorphisms (SNPs) associated with 110 SRPs were filtered and selected from a large genome-wide association study (GWAS) for plasma proteome involving 35,559 participants. The GWAS data of AMD was obtained from FinnGen consortium (6,157 AMD cases and 288,237 controls) and further validated by using data from UK Biobank consortium (3,553 AMD cases and 147,089 controls). Results The MR results at both discovery and validation stages supported the causality (IVW-P < 0.00045) between plasma levels of 4 SRPs (C3b, CTNNB1, CCL1, and CCL3L1) and the risk of AMD and supported potential causality (IVW-P < 0.05) between other 10 SRPs and risk of AMD. No heterogeneity or pleiotropy in these results was detected. Conclusion Our findings supported that high plasma levels of C3b, CTNNB1, CCL1, and CCL3L1 were associated with increased risk of AMD, thereby highlighting the role of systemic inflammation in AMD pathogenesis and providing the rationale for developing new preventative and therapeutic strategies.


Introduction
Age-related macular degeneration (AMD) is a neurodegenerative disease involving neuroretina and retinal pigment epithelium (RPE), thereby leading to visual impairment or even blindness (Wong et al., 2014).Studies have reported considerable health burden in patients with AMD, which mainly affects adults aged 40 years and older (Zhu et al., 2019).The number of patients with AMD is continuously increasing and estimated to be about 288 million worldwide by 2,040 (Congdon et al., 2004;Wong et al., 2014).
Intravitreal injection of anti-vascular endothelial growth factor (VEGF), such as ranibizumab (Blodi et al., 2023) and aflibercept (Wykoff et al., 2023), have been used to slow the progression of the neovascular or exudative subtype (nAMD or eAMD).However, for geographic or atrophic AMD, there is no effective treatment available.Therefore, it is necessary to identify risk factors for AMD to help prevent the incidence of AMD and alleviate the burden of this disease on public health.Some intraocular risk factors have been revealed, such as extracellular deposits (Chen et al., 2022).Notably, localized mechanistic studies in the eye have failed to fully elucidate pathogenesis of AMD and the systemic risk factors for AMD were still lacking, thereby limiting the interventions for reducing the risk of AMD.
As a hallmark of aging, cellular senescence is a significant contributor to aging and age-related diseases including Alzheimer's disease (AD) (Holloway et al., 2023).Previous studies have suggested that oxidative stress, inflammation and RPE senescence may all play a critical role (Kauppinen et al., 2016) in AMD initiation and development.Notably, Saul et al. (2022) has identified a gene set (senescence-related genes, SRGs) for predicting senescence-associated pathways across tissues, which has been used and cited by many researches about age-related diseases (Doolittle et al., 2023;Farr, 2023).
Observational study and randomized clinical trial (RCT) are useful for identifying risk factors of diseases.However, observational studies are vulnerable to reverse causation, residual confounding, and selective bias (Davey Smith and Hemani, 2014).A RCT allows reliable and robust causal inferences to be drawn, but it is costly, time-consuming, and sometimes impractical to conduct one.Mendelian randomization (MR) approaches have opened up opportunities to assess and determine clinically associated characters for multiple diseases (Davey Smith and Hemani, 2014), which examines causal relationships between exposures and outcomes using genetic variants significantly associated with an exposure as instrumental variables.By using MR approaches, the defects of observational study and RCT can be nicely overcome (Davey Smith and Hemani, 2014).In addition, this approach has been used to explore risk factors of multiple eye disorders (Patasova et al., 2021;Choquet et al., 2022).
In this study, we obtained the list of SRGs and explored the causality between the plasma levels of the proteins (senescencerelated proteins, SRPs) encoded by these SRGs and the risk of AMD by two-sample MR analysis.The design and findings of this study were summarized in Figure 1.As a results, plasma levels of 4 SRPs (C3b, CCL1, CCL3L1, and CTNNB1) were identified and confirmed to be associated with risk of AMD.It is hoped that these 4 SRPs will serve as new drug targets for intervention.

Theoretical foundation of MR analysis
The list of SRGs was obtained from Saul et al. (2022) (Supplementary Table 1).The theoretical basis and three basic assumptions of MR analysis (Emdin et al., 2017;Zheng et al., 2017) were shown in Figure 2. In this study, we implemented two-sample MR approaches to judge causation between plasma SRPs and AMD risk.

Data source for exposures and outcomes
The summary level genome-wide association studies (GWAS) data of SRPs were obtained from Ferkingstad et al. (2021), who have conducted a large-scale GWAS project on plasma proteome involving 35,559 participants and 4,907 plasma proteins.The GWAS data of AMD were obtained from FinnGen (6,157 cases and 288,237 controls) (Kurki et al., 2023) and UK Biobank consortium (3,553 cases and 147,089 controls) (Sudlow et al., 2015).The AMD cases were defined by H7 in International Classification of Disease-10 (ICD-10) and 3,625 in ICD-9.The FinnGen consortium data was used for discovery and data from UK Biobank consortium was for replication and validation.The flowchart of this study is presented in Figure 3.

Selection of genetic instruments
Instrumental variables (single-nucleotide polymorphisms, SNPs) were selected via the following criteria: (i) with genomewide significance (P < 5 × 10 −6 ) and (ii) pruned by linkage disequilibrium (r 2 < 0.01 and within 10 000 kb from the index variant).PhenoScanner (Staley et al., 2016) is an online platform with comprehensive information about genotype-phenotype association.We examined whether the obtained instrumental SNPs were associated with the outcomes and the potential confounders and subsequently remove the associated ones.

Causality estimated by MR analysis
Mendelian randomization's validity depends on the crucial assumption of no pleiotropy (Lawlor et al., 2008).Therefore, we used the random-effect inverse-variance weighted (IVW) method (Burgess et al., 2016) as the primary method and performed Cochran's Q test and MR-Egger intercept test to evaluate the heterogeneity and detect pleiotropy (Bowden et al., 2019).

Identification of druggable targets
A list of druggable genes were obtained from Freshour et al. (2021)  The summary of design and findings of this study.The theoretical basis and three basic assumptions of MR analysis.MR, Mendelian randomization.
(Supplementary Table 2).DGIdb provides information on druggene interactions and druggable genes from publications, databases and other web-based sources.We downloaded the "Categories Data" (released in February 2022), including all genes in the druggable categories in the DGIdb, from all sources.

Statistical analysis
We performed all the analyses in R (version 4.0.1)using the TwoSampleMR (Hemani et al., 2018) R packages.The code for MR analysis is accessible at https://mrcieu.github.io/TwoSampleMR/articles/index.html.All statistical tests are 2 sided.Results with IVW-P < 0.05 was considered nominally significant and IVW-P < 0.00045 was taken as statistically significant.

Discussion
Age-related macular degeneration is a neurodegenerative disease predominantly affecting the elders, which can cause vision loss and has a significant impact on the quality of life of affected individuals (Wood et al., 2011).Identifying systemic risk factors for AMD is important for preventing development of this disease.Therefore, this study aimed to explore the relationship between plasma SRPs and AMD risk by two-sample MR analysis, which is a useful tool for assessing and determining clinically associated characters for multiple diseases (Davey Smith and Hemani, 2014).As a results, high plasma levels of 4 SRPs (C3b, CCL1, CCL3L1, and CTNNB1) were identified to be associated with increased risk of AMD.
Blood supply of the retina consists of retinal microcirculatory system and the underlying choriocapillaris (Funk, 1997).Notably, systemic inflammation has been associated with the development of AMD (Sannan, 2023).In this study, the identified 4 SRPs associated with AMD risk are tightly associated with inflammation processes, which further highlighted the important role of systemic inflammation and the high plasma levels of inflammatory mediators in the pathogenesis of AMD.
Complement C3 is a gene that plays a crucial role in the activation of the complement cascade and major effector functions of complement are mediated through C3b (Haapasalo and Meri, 2019).In AMD, the injury of blood-retinal barrier allows leakage of serum proteins, including complement components, into the retina from the underlying choriocapillaris (Katschke et al., 2018).Local complement activation leads to the recruitment of microglia into the lesion, which then produce additional complement components, prune complement-coated synapses away from neurons, depriving neurons of trophic support, and cause neuroinflammation that adds to neuronal damage and loss (Stephan et al., 2012).The elevated level of plasma C3b may promote the development of AMD by facilitating the neuroinflammation in the retina, which explained the observed association between higher plasma level of C3b and the increased risk of AMD.Notably, C3b has been associated with AMD (Helgason et al., 2013) and taken as the therapeutic target for AMD (Yang et al., 2022;Jia et al., 2023)   Forest plot of the MR results in discovery stage.MR: Mendelian randomization; SRPs, senescence-related proteins; AMD, age-related macular degeneration; IVW, inverse-variance weighted.P < 0.05 was considered nominally significant and P < 0.00045 was considered statistically significant.
Frontiers in Aging Neuroscience 07 frontiersin.orgfusion protein (efdamrofusp alfa), which is capable of neutralizing both VEGF isoforms and C3b/C4b, in neovascular AMD (nAMD) treatment.However, recent studies focused on neutralizing local C3b instead of lowering circulating level of it.Considering the identified causal effect of C3b on AMD risk in this study, circulating C3b lowering strategy might be a potential way for reducing AMD risk, which calls for further studies.
The other three AMD-associated proteins identified in this study may also contribute to the development of AMD through the induction of neuroinflammation in the retina.CCL1 belongs to the C-C subfamily of chemokines, which are secreted proteins involved in immunoregulatory and inflammatory processes.It binds to the C-C motif receptor 8 (CCR8).In addition, it has been implicated in various inflammation-associated diseases, including asthma (Hurme et al., 2022), rheumatoid arthritis (White et al., 2013), and multiple sclerosis (Schropp et al., 2023).In asthma, CCL1 is thought to play a role in recruiting inflammatory cells to the airways, leading to airway inflammation and hyperresponsiveness.Targeted drug discovery efforts have focused on developing drugs that can block the interaction between CCL1 and its receptor, with the aim of reducing inflammation and disease progression (Connolly et al., 2012).CCL3L1 encodes a protein that binds to several chemokine receptors, including CCR5 (Urban et al., 2009).CCL3L1 has also been implicated in inflammatory diseases such as rheumatoid arthritis (Nordang et al., 2012).CTNNB1 encodes a protein that is part of a complex of proteins that make up adherens junctions (AJs) (van der Wal and van Amerongen, 2020), which are essential for creating and maintaining epithelial cell layers by regulating cell growth and adhesion between cells.In addition, CTNNB1 is related to cholesterol homeostasis (Chen et al., 2023), which might also be the mechanical basis of the association between plasma CTNNB1 level and AMD risk.
There were some limitations in this MR-designed investigation, despite its many advantages over conventional epidemiological studies.First, this study only included European-ancestry individuals, which suggests that our findings cannot be directly applied to other populations.Second, our findings only revealed the causality between plasma levels of several SRPs and AMD, not the underlying mechanisms, which call for further researches.
We obtained druggable genes from the DGIdb database (Freshour et al., 2021).Importantly, most of the SRGs are druggable genes, including the four genes identified as risk factors for AMD, i.e., C3b, CCL1, CCL3L1, and CTNNB1.Therefore, our study may provide some novel potential drug targets for AMD or the rationale of existing drugs.Still, the results need to be confirmed by further studies and it is hoped that these 4 SRPs will serve as new drug targets for preventing the incidence of AMD.Forest plot of the MR results in validation stage.MR: Mendelian randomization; SRPs, senescence-related proteins; AMD, age-related macular degeneration; IVW, inverse-variance weighted.P < 0.05 was considered nominally significant and P < 0.00045 was considered statistically significant.
TABLE 2 The 4 SRPs significantly associated with AMD risk.

Protein
Related GO annotations Related diseases CCL1 GO:0090026 positive regulation of monocyte chemotaxis; GO:0048245 eosinophil chemotaxis; GO:0032740 positive regulation of interleukin-17 production.
HIV infection/AIDS, rheumatoid arthritis C3b GO:0001970 positive regulation of activation of membrane attack complex; GO:0150064 vertebrate eye-specific patterning; GO:0001798 positive regulation of type IIa hypersensitivity.
Atypical hemolytic uremic syndrome (aHUS) and AMD CTNNB1 GO:0007403 glial cell fate determination; GO:0044336 canonical Wnt signaling pathway involved in negative regulation of apoptotic process; GO:0061324 canonical Wnt signaling pathway involved in positive regulation of cardiac outflow tract cell proliferation.
Several types of cancer, including uveal melanoma, colorectal cancer, and ovarian cancer SRPs, senescence-related proteins; AMD, age-related macular degeneration; GO, Gene Ontology.
In conclusion, we provided the genetic evidence that plasma levels of C3b, CTNNB1, CCL1, and CCL3L1 are causally associated with risk of AMD, which highlights the role of systemic inflammation in the pathophysiology of AMD.Given that the genes encoding these four proteins are all druggable targets, the findings may contribute to understanding the pathogenesis of AMD and the development of new therapeutic or preventive strategy for AMD.

FIGURE 4 P
FIGURE 4 P-values of IVW, heterogeneity test, and pleiotropy test of SRPs and AMD risk in discovery stage.MR: Mendelian randomization; SRPs, senescence-related proteins; AMD, age-related macular degeneration; IVW, inverse-variance weighted.

FIGURE 6 P
FIGURE 6 P-values of IVW, heterogeneity test and pleiotropy test of SRPs and AMD risk in validation stage.MR: Mendelian randomization; SRPs, senescence-related proteins; AMD, age-related macular degeneration; IVW, inverse-variance weighted.

TABLE 1 (
Continued) .Yang et al. (2022)reported the preclinical assessment and phase 1 clinical outcomes of a bispecific