Latest Development on Genetics of Common Retinal Diseases

Abstract

Many complex forms of retinal diseases are common and pan-ethnic in occurrence. Among them, neovascular age-related macular degeneration, polypoidal choroidal vasculopathy, and central serous choroid retinopathy involve both choroidopathy and neovascularization with multifactorial etiology. They are sight-threatening and potentially blinding. Early treatment is crucial to prevent disease progression. To understand their genetic basis, candidate gene mutational and association analyses, linkage analysis, genome-wide association studies, transcriptome analysis, next-generation sequencing, which includes targeted deep sequencing, whole-exome sequencing, and whole genome sequencing have been conducted. Advanced genomic technologies have led to the identification of many associated genes. But their etiologies are attributed to complicated interactions of multiple genetic and environmental risk factors. Onset and progression of neovascular age-related macular degeneration and polypoidal choroidal vasculopathy are affected by aging, smoking, lifestyle, and variants in over 30 genes. Although some genetic associations have been confirmed and validated, individual genes or polygenic risk markers of clinical value have not been established. The genetic architectures of all these complex retinal diseases that involve sequence variant quantitative trait loci have not been fully delineated. Recently artificial intelligence is making an impact in the collection and advanced analysis of genetic, investigative, and lifestyle data for the establishment of predictive factors for the risk of disease onset, progression, and prognosis. This will contribute to individualized precision medicine for the management of complex retinal diseases.

INTRODUCTION

Neovascular age-related macular degeneration (nAMD), polypoidal choroidal vasculopathy (PCV), and central serous choroid retinopathy (CSCR) are common and complex retinal diseases that involve choroidopathy and neovascularization. They mainly affect the macular region, and their occurrences are pan-ethnic. Prevalence of nAMD is similar in white and Asian populations, but PCV is much more prevalent in Asians than other populations including whites.^1,2 There are overlaps in clinical features of nAMD and PCV. Questions remain whether they are independent disease entities or subtypes of the same disease. Both PCV and nAMD can have subretinal exudation, hemorrhage, and scarring (nAMD, Figs. 1A, B; PCV, Figs. 1C, D), with consequential loss of central vision. PCV has been estimated to contribute to 22%–54% of age-related macular degeneration (AMD) prevalence in Asians, compared with that of <10% in whites.³ Another major macular disease, central serous chorioretinopathy (CSCR), also overlaps with PCV and nAMD in clinical presentations, especially in microvasculature of the retina (Figs. 1E, F).⁴ Overall CSCR is a less severe and less progressive disease. It tends to happen in young adulthood and mid-age, whereas nAMD and PCV affect more in elderly people older than 60 years. Their responses to treatment are different, with PCV more responsive to combined anti-vascular endothelial growth factor (VEGF) and photodynamic therapy (PDT) and nAMD to anti-VEGF monotherapy. In contrast, acute CSCR usually resolves spontaneously over a few months, whereas chronic CSCR, with a disease duration >3 months, usually requires treatment, such as half-dose PDT. However, once secondary choroidal neovascularization occurs, chronic CSCR patients should be treated with anti-VEGF. AMD and related diseases are major causes of irreversible blindness, especially in Asia. Its prevalence is increasing.^1,5 There is a need to differentiate nAMD, PCV, and CSCR for the most appropriate treatment to prevent irreversible visual deterioration. Delineation of their genetic architectures should throw light on the differential molecular mechanisms of these diseases. In recent years there have been active investigations in their molecular genetics.

FIGURE 1: Clinical images of neovascular age-related macular degeneration [(A) fundus photo; (B) optical coherence tomography image], polypoidal choroidal vasculopathy [(C) fundus photo; (D) optical coherence tomography image] and central serous choroid retinopathy [(E) < fundus photo; (F) optical coherence tomography image]. White arrow indicates subretinal hemorrhage in neovascular age-related macular degeneration; blue arrow indicates subretinal exudate, and green arrow indicates a polyp in polypoidal choroidal vasculopathy; yellow arrow indicates the subretinal fluid in central serous choroid retinopathy; red arrow indicates the subretinal fluid in neovascular age-related macular degeneration, polypoidal choroidal vasculopathy, and central serous choroid retinopathy. Red “*” indicates the pigment epithelial detachment in neovascular age-related macular degeneration and polypoidal choroidal vasculopathy; black “*” indicates the subretinal exudate inneovascular age-related macular degeneration and polypoidal choroidal vasculopathy; and white “*” indicates the intraretinal fluid in neovascular age-related macular degeneration.

GENETIC EPIDEMIOLOGY

Early evidence for genetic influences on the development of AMD was provided by familial aggregation and twins studies. First-degree relatives of patients with late AMD have been found to develop AMD at an increased rate at a relatively young age,^6,7 and significantly higher concordance of AMD was found in monozygotic than in dizygotic twins,⁸ suggesting a role of genetic susceptibility in the disease risk. Later, familial linkage studies confirmed the genetic involvement in AMD. The first AMD locus was mapped to chromosome 1q25–q31 by genome-wide linkage analysis in a pedigree of AMD.⁹ It was later confirmed as the locus that harbors an important AMD-associated gene, complement factor H (CFH), which mainly involved in the complement pathway and alternative pathway.^10,11 This finding laid the foundation for future genetic analysis of AMD.

In contrast to AMD, family aggregation and twins studies on PCV have been less reported. In 2 brothers of West Indian origin, bilateral PCV was confirmed based on clinical features and indocyanine green angiography.¹² In a cohort of monozygotic twins with typical PCV, concordances in disease progression and response to treatment between the twins were identified.¹³ Thus the role of genetic factors in affecting the occurrence of PCV and its clinical manifestations was evident. However, no linkage locus has been reported for PCV from familial linkage analysis. This is likely due to the late disease onset and rarity of the large pedigree of PCV that are eligible for linkage analysis. The complex and overlapping clinical features of PCV and nAMD also result in phenotypic diversities among patients and thus difficulties in linkage analysis.

In CSCR, aggregation of the disease in core families¹⁴ and high frequency (around 50%) of affected subjects among the relatives of CSCR patients^15,16 have suggested the involvement of genetic factors in its etiology. So far, no linkage locus has been reported for CSCR by familial linage analysis, likely due to the rarity of large disease pedigrees. Instead, an association study, which involves the comparison of certain genetic variants between unrelated patients and control subjects, is a useful strategy for identifying genetic loci for CSCR.

GENE MAPPING STRATEGIES AND TECHNOLOGIES

With the advent and advancement of next-generation sequencing platforms, the whole genome can be sequenced efficiently. The large data sets of genomic variants can be analyzed with different bioinformatics and biostatistics programs. Hence, a whole tier of genetic technologies and approaches has been formed, mainly including the genotyping of single, multiple, or genome-wide single-nucleotide polymorphisms (SNPs) or microsatellite markers; sequencing of single or multiple amplicons of a gene, a targeted chromosomal region (targeted sequencing), the whole exome [whole-exome sequencing (WES)], or the whole genome [whole genome sequencing (WGS)]; and determination of copy number variants and large structural variants in specific chromosomal regions or the whole genome.

The selection of strategies, that is, the combination of technologies and approaches, for genetic analysis of a targeted disease, is affected by multiple factors. On the clinical side, factors to be considered include the clinical homogeneity of the disease (eg, heterogeneities in clinical manifestations and existence of clinical subtypes), age of disease onset (eg, early or late onset), sex predilection (eg, female or male prominent), prevalence (eg, common or rare), apparent inheritance pattern (Mendelian or sporadic), complexity of endophenotypes (eg, ocular parameters that are relevant to the disease manifestations and risks), systemic associations (eg, neurological stress and cortical steroid intake in CSCR), and involvement of risks from life styles and/or environmental exposure (eg, smoking for AMD and PCV). While at the technology side, what are considered include the aims and scale of the intended study (eg, hypothesis-based or hypothesis-free), availability of equipment, manpower and expertise, and the budget (Fig. 2).

FIGURE 2: Strategies and technologies used in the gene mapping for retinal diseases of Mendelian inheritance (left panel) and multifactorial etiology (right panel). Apart from the genetic factors, inherent and environmental risk factors also play a role in the both forms of retinal diseases, with more in the multifactorial form.

In genetic studies, strategies combining different approaches may be adopted, as reflected in variable study designs and consequently different outputs. For example, for multigeneration large pedigrees of disease following Mendelian inheritance, linkage analysis using genome-wide microsatellite markers or SNPs can be used to map the linkage locus. This is usually followed by a series of candidate gene sequencing studies to identify the responsible gene(s) and variant(s) in this locus. However, the loci identified by traditional linkage analysis usually contain a large number of genes, mostly hundreds. Pinpointing the causal genes is difficult. In recent years, WES and WGS with combined linkage analysis and regional deep sequencing in large pedigrees followed by pipeline filtering strategies have been conducted to identify the disease-causing or associated variants with unprecedented efficiency.

In contrast, for sporadic diseases, which are mostly multifactorial in etiology, case-control association analysis is an appropriate strategy, where allelic/genotypic distributions of the gene variants of interest are compared between cases and controls. Significant differences in the distributions suggest a positive association. With information from the HapMap project¹⁷ and the availability of array-based high-throughput genotyping platforms, genome-wide association studies [genome-wide association studies (GWAS)] had become possible. The first GWAS on eye disease, published in 2005, identified a strong association of the Y402H (rs1061170) polymorphism in the CFH gene with advanced AMD in a white cohort.¹⁸ Since then, GWAS—a hypothesis-free approach—had become very prominent and efficient in identifying novel disease-associated genes/loci and variants. This has resulted in a new understanding of the disease pathways and mechanisms. These variants are usually common SNPs with a minor allele frequency (MAF) >1% and confer an odds ratio (OR) typically within the range of 1.1–2.0.¹⁹ Therefore, the contributions of individual variants in the disease heritability and pathogenesis are small. Moreover, the initial hits of association signals from GWAS are usually located in intronic or intergenic regions, of which the functions are not well-known. Also, whether the gene upstream or downstream to the intergenic SNP is responsible for the disease needs to be confirmed by candidate gene analyses. With the development of WES, exome-wide association studies (EWAS) have become an important hypothesis-free approach to identify disease-associated coding exonic variants, which are more likely to affect gene function. These variants can be common or rare. The first WGS study in AMD was published in 2013, and a disease-associated, rare nonsynonymous SNP in Complement C3 (C3) was identified.²⁰

In genomic studies, newly identified candidate genes are usually directly sequenced in a larger cohort of patients and controls to further estimate their contributions to the population. In candidate gene studies, one or more genes can be sequenced to identify disease-causing or associated variants, and one or more SNPs are compared in case and control cohorts to determine the association profiles. There are strategies to select the candidate genes, such as (1) genes that are located in the linkage loci of the targeted disease, (2) genes that have been linked to another disease with similar clinical features, (3) genes with functional roles in a potential disease pathway, (4) genes identified in previous GWAS, EWAS, WES, and/or WGS as putatively disease-associated or disease-causing, (5) genes, which are associated with the disease in the initial cohort, and (6) genes, which are associated with the disease based on a limited number of SNPs, to be fine-mapped using more markers with full coverage of the genes.

Candidate gene analysis as a major hypothesis-based approach has more specific targets and advantages, including (1) addressing specific questions about one or more genes on the disease of interest in specific study cohorts, (2) requiring a less stringent significance threshold (eg, P value) to achieve a significant association, thereby requiring smaller sample sizes, (3) useful and required for validating and fine-mapping the initial association signals from GWAS and EWAS, (4) needed for validating the initially-identified disease-causing genes and depict the mutation spectra and frequencies, and (5) requiring shorter study period, less labor and laboratory consumptions, and lower budget. Therefore, the candidate gene approach is an important complementary strategy to hypothesis-free approaches in depicting the genetic landscape of eye diseases.

MOLECULAR GENETICS OF SPECIFIC RETINAL DISEASES

Molecular Genetics of Neovascular Age-Related Macular Degeneration

Combing different genetic strategies have led to the identification of a large number of genes and variants for common retinal diseases, especially nAMD, PCV, and CSCR (Table 1). So far, the mapping for AMD-associated genes has been the most fruitful. Some genes identified for inherited macular dystrophies were proposed to be candidate genes for AMD. One example is the Adenosine triphosphate binding cassette subfamily A member 4 (ABCA4, also known as ABCR) gene for autosomal recessive Stargardt type 1 macular dystrophy. By screening the ABCR gene in 167 unrelated AMD patients and 220 controls, Allikmets et al²¹ identified 13 AMD-associated alterations in 26 patients (16%) and 1 control (0.45%). In a cohort of 1176 AMD patients and 1444 controls, 2 variants, G1961E and D2177N, were identified in 40 patients (around 3.4%) and 13 control subjects (around 0.95%, P < 0.0001), further supporting ABCR as a susceptibility gene for AMD.²² However, a lack of association between ABCR and AMD was also reported.^23,24 Similarly, both positive²⁵ and negative²⁶ associations with AMD had been reported for the ELOVL Fatty Acid Elongase 4 (ELOVL4) gene of autosomal dominant Stargardt type 3 macular dystrophy.

TABLE 1 - Genes Associated With AMD, PCV and/or CSCR

				AMD			PCV			CSCR
No.	Gene/loci	Representative variants	Effect allele	Effect	Population	Method	Effect	Population	Method	Effect	Population	Method
Common variants shared by AMD, PCV and CSCR
1	*CFH	rs1061170	C	+	Multipopulations	GWAS1	+	EuropeanAmerican	Candidate gene study2	−	Multipopulations	Candidate gene study3
		rs1329428	C	+	Multipopulations	GWAS1	+	Chinese	Candidate gene study4	−	Japanese	Candidate gene study5
		rs800292	G	+	Chinese	GWAS6	+	Chinese	Candidate gene study4	−	Japanese	Candidate gene study5
		rs3753394	T	+	Chinese	Candidate gene study7	+	Chinese	Candidate gene study4	−	Japanese	Candidate gene study5
2	*ARMS2-HTRA1	rs10490924	T	+	German	Candidate gene study8	+	Japanese	Candidate gene study9	−	Dutch	Candidate gene study10
		rs11200638	A	+	Chinese	GWAS6	+	Japanese	Candidate gene study9	/
		rs3750848	G	+	Multipopulations	Candidate gene study11	+	Japanese	Candidate gene study12	/
3	*TNFRSF10A	rs13278062	T	+	Japanese	GWAS13	+	Japanese	GWAS13	+	Multipopulations	GWAS14
4	*COL4A3	rs11884770	T	−	Multipopulations	GWAS15	/			/
		Multiple rare variants in gene-based analysis		/			NA	Multipopulations	Candidate gene study16	/
		rs4276018	G	/			/			−	Japanese	Candidate gene study17
5	*B3GALTL	rs9542236	C	+	Multipopulations	GWAS18	/			/
		Multiple rare variants in gene-based analysis		/			NA	Multipopulations	Candidate gene study16	/
		rs9564692	T	/			/			−	Japanese	Candidate gene study17
6	*LIPC	rs10468017	T	−	Multipopulations	GWAS19	/			/
		rs1532085	A	/			+	Chinese	Candidate gene study20	/e
		rs2043085	A	/			/			+	Chinese	Candidate gene study21
Common variants shared by AMD and PCV
7	*ADAMTS9	rs6795735	T	+	Multipopulations	GWAS18	/			/
		rs6775974	C	/			+	Multipopulations	Candidate gene study16	/
8	*TGFBR1	rs334353	T	+	Multipopulations	GWAS18	/			/
		rs44466	C	/			+	Multipopulations	Candidate gene study18	/
9	*VEGFA	rs4711751	T	+	Multipopulations	GWAS22	/			/
		rs833069	G	/			+	Korean	Candidate gene study23	/
10	*CETP	rs3764261	A	+	Multipopulations	GWAS22	+	Japanese	Candidate gene study24	/
		rs5882	G	/			+	Chinese	Candidate gene study25	/
11	*RDBP-CFB	rs522162	G	−	American	GWAS26	/			/
		rs3880457	C	/			−	Japanese	Candidate gene study27	/
12	*C2-CFB-SKIV2L	rs401775	T	+	Multipopulations	Candidate gene study16	+	Multipopulations	Candidate gene study16	/
13	*ANGPT2	rs4455855	A	+	Multipopulations	Candidate gene study28	+	Multipopulations	Candidate gene study28	/
		rs13269021	G	+	Multipopulations	Candidate gene study28	+	Multipopulations	Candidate gene study28	/
14	*TIE2	rs625767	C	−	Multipopulations	Candidate gene study29	−	Multipopulations	Candidate gene study29	/
15	*ABCG1	rs225396	T	+	Multipopulations	Candidate gene study30	+	Multipopulations	Candidate gene study30	/
16	FPR1	rs867229	T	−	Chinese	Candidate gene study31	/			/
		rs78488639	A	+	Chinese	Candidate gene study31	+	Chinese	Candidate gene study31	/
17	GDF6	rs6982567	A	+	EuropeanAmerican	Candidate gene study32	+	Chinese	Candidate gene study33	/
18	ELN	rs2301995	C	+	Japanese	Candidate gene study34	−	Japanese	Candidate gene study35	/
19	*SKIV2L	rs438999	C	−	Irish	Candidate gene study36	/			/
		rs429608	A	−	Chinese	Candidate gene study37	−	Japanese	Candidate gene study38	/
		rs2075702	C	/			−	Japanese	Candidate gene study27	/
20	*C3	rs2230199	G	+	English	Candidate gene study39	/			/
		rs2241394	C	−	Japanese	Candidate gene study40	−	Japanese	Candidate gene study12	/
21	*C2	rs9332739	C	−	European American	Candidate gene study41	/			/
		rs547154	T	−	European American	Candidate gene study41	−	Japanese	Candidate gene study42	/
22	*CFB	L9H	A	+	European American	Candidate gene study41	/			/
		R32Q	A	+	European American	Candidate gene study41	/			/
		rs541862	C	−	Japanese	Candidate gene study42	−	Japanese	Candidate gene study42	/
Common variants associated with CSCR
23	*GATA5	rs6061548	T	/			/			+	Japanese	GWAS14
24	*VIPR2	rs3793217	G	/			/			+	Japanese	GWAS43
25	SLC7A5	rs11865049	A	/			/			+	Japanese	GWAS44
26	MCUB-PLA2G12A-CFI	rs4698775	G	/			/			+	Chinese	Candidate gene study21
27	NR3C2	rs2070951	G	/			/			+	Multipopulations	Candidate gene study45
28	CDH5	rs7499886	A	/			/			+	Multipopulations	Candidate gene study46
Common variants associated with PCV
29	HERPUD1	rs2217332	A	/			−	Chinese	Candidate gene study47	/
30	9p21	rs10757278	A	/			+	Chinese	Candidate gene study48	/
Common variants associated with AMD
31	*RLBP1	rs3825991	A	+	Multipopulations	GWAS49	/			/
32	CFHR4	rs6685931	C	+	Multipopulations	GWAS50	/			/
33	*ATF7IP2	rs28368872	A	+	Multipopulations	GWAS51	/			/
34	*TNR	rs58978565	CAGAGT	+	Multipopulations	GWAS51	/			/
35	STON-GTF2A1L/LHCGR/FSHR	rs4482537	C	+	Multipopulations	GWAS52	/			/
36	CD46	1:207980901	G	+	Multipopulations	GWAS53	/			/
37	C20orf85	rs201459901	TA	−	Multipopulations	GWAS15	/			/
38	CNN2	rs67538026	T	−	Multipopulations	GWAS15	/			/
39	NPLOC4-TSPAN10	rs6565597	T	+	Multipopulations	GWAS15	/			/
40	TMEM97/VTN	rs11080055	A	−	Multipopulations	GWAS15	/			/
41	*CTRB2-CTRB1	rs72802342	A	−	Multipopulations	GWAS15	/			/
42	ACAD10	rs61941274	A	+	Multipopulations	GWAS15	/			/
43	ARHGAP21	rs12357257	A	+	Multipopulations	GWAS15	/			/
44	MIR6130-RORB	rs10781182	T	+	Multipopulations	GWAS15	/			/
45	TRPM3	rs71507014	G	+	Multipopulations	GWAS15	/			/
46	*KMT2E-SRPK2	rs1142	T	+	Multipopulations	GWAS15	/			/
47	PILRB-PILRA	rs7803454	T	+	Multipopulations	GWAS15	/			/
48	PRLR-SPEF2	rs114092250	A	−	Multipopulations	GWAS15	/			/
49	MMP20	rs10895322	G	+	Japanese	GWAS54	/			/
50	*FGD6	rs10507047	G	−	Multipopulations	EWAS55	/			/
51	*SLC44A4	rs12661281	T	+	Multipopulations	EWAS55	/			/
52	*C6orf223	rs2295334	A	−	Multipopulations	EWAS55	/			/
53	RAD51B	rs8017304	A	+	Multipopulations	GWAS18	/			/
54	*SLC16A8	rs8135665	T	+	Multipopulations	GWAS18	/			/
55	IER3-DDR1	rs3130783	A	+	Multipopulations	GWAS18	/			/
56	*COL8A1-FILIP1L	rs13081855	T	+	Multipopulations	GWAS18	/			/
57	SOD2	rs2842992	A	−	Multipopulations	GWAS56	/			/
58	MBP	rs1789110	A	−	Multipopulations	GWAS56	/			/
59	PECI	rs582301	A	+	Multipopulations	GWAS56	/			/
60	C8orf42	rs722782	A	−	Multipopulations	GWAS56	/			/
61	TRIB1	rs35691538	A	+	Multipopulations	GWAS56	/			/
62	BRUNOL4	rs8091635	T	+	Multipopulations	GWAS56	/			/
63	TRA2A	rs10262213	T	−	Multipopulations	GWAS56	/			/
64	ADAM19	rs59795197	T	−	Multipopulations	GWAS56	/			/
65	*SERPINB8-CDH7	rs17073641	C	+	European American	GWAS26	/			/
66	NOTCH4	rs2071277	C	+	English	GWAS57	/			/
67	TNXB-FKBPL	rs12153855/rs9391734	C	+	English	GWAS57	/			/
68	FRK/COL10A1	rs1999930	T	−	Multipopulations	GWAS22	/			/
69	*TIMP3	rs9621532	C	−	Multipopulations	GWAS22	/			/
70	*REST-C4orf14-POLR2B-IGFBP7	rs1713985	G	+	Japanese	GWAS13	/			/
71	RREB1	rs11755724	A	−	Multipopulations	GWAS19	/			/
72	*CFI	rs2285714	T	+	Multipopulations	GWAS58	/			/
73	MYRIP	rs2679798	G	−	Multipopulations	GWAS59	/			/
74	CTRB1	rs8056814	A	−	Multipopulations	WES60	/			/
75	*UBE3D	rs7739323	T	−	Multipopulations	WES61	/			/
		rs141853578	T	+	Chinese	Candidate gene study62	/			/
76	SIRT1	rs7895833	G	+	Lithuanian	Candidate gene study63	/			/
77	TR	rs8177178	T	+	Chinese	Candidate gene study62	/			/
78	RP1L1	rs3924612	G	+	Lithuanian	Candidate gene study64	/			/
79	PRPH2	rs3818086	A	+	Polish	Candidate gene study65	/			/
80	TRF1	rs10107605	C	+	Lithuanian	Candidate gene study66	/			/
81	mTOR	rs2295080	G	+	Finnish	Candidate gene study67	/			/
82	IL1RL1	rs1041973	A	+	Lithuanian	Candidate gene study68	/			/
83	IL1RAP	rs4624606	A	+	Lithuanian	Candidate gene study68	/			/
84	CYP2J2	rs890293	T	−	Lithuanian	Candidate gene study69	/			/
85	TYR	rs621313	A	−	Multipopulations	Candidate gene study53	/			/
86	PVRL2	rs6857	T	−	Multipopulations	Candidate gene study53	/			/
87	LOXL1	rs1048661	T	+	Chinese	Candidate gene study70	/			/
88	KCTD10	rs56209061	A	−	Lithuanian	Candidate gene study71	/			/
89	MIR146A	rs2910164	G	+	Italian	Candidate gene study72	/			/
90	MIR27A	rs11671784	A	+	Italian	Candidate gene study72	/			/
91	SLCO1B1	rs4149056	C	+	Lithuanian	Candidate gene study73	/			/
92	FILIP1L	rs144351944	G	+	Chinese	Candidate gene study74	/			/
93	BBX	rs189132250	G	+	Chinese	Candidate gene study74	/			/
94	SGCD	rs931798	A	+	Mexican	Candidate gene study75	/			/
95	RAGE	rs1800624	T	−	Lithuanian	Candidate gene study76	/			/
		rs1800625	G	+	Lithuanian	Candidate gene study76	/			/
96	NRTN-FUT6-C3	rs12019136	G	+	Multipopulations	Candidate gene study16	/			/
97	MT2A	rs28366003	G	+	Spanish	Candidate gene study77	/			/
98	TOMM40	rs2075650	G	+	Multipopulations	Candidate gene study78	/			/
99	*PGF	rs2268615	G	+	Chinese	Candidate gene study79	/			/
100	*VEGFR1	rs9554322	C	+	Chinese	Candidate gene study80	/			/
101	TLR2	R753Q	A	+	Istanbul	Candidate gene study81	/			/
102	MMP-2	-1306 C>T	C	+	Lithuanian	Candidate gene study82	/			/
103	*DAPL1	rs17810398	T	+	Multipopulations	Candidate gene study83	/			/
104	IL17	rs2275913	A	+	Chinese	Candidate gene study84	/			/
105	ADIPOQ	rs822396	G	+	Chinese	Candidate gene study85	/			/
106	FBN2	rs154001	C	+	American	Candidate gene study86	/			/
107	SMUG1	c.-31A>G	A	+	Polish	Candidate gene study87	/			/
		g.4235T>C	T	+	Polish	Candidate gene study87	/			/
108	COL1A2	rs42524	G	−	Chinese	Candidate gene study32	/			/
109	GSTP1	rs1695	G	+	Chinese	Candidate gene study88	/			/
110	IRP1	rs17483548	G	+	Polish	Candidate gene study89	/			/
		rs867469	G	+	Polish	Candidate gene study89	/			/
111	ADIPOR1	rs10753929	T	+	Finnish	Candidate gene study90	/			/
112	HMOX1	19G>C	G	−	Polish	Candidate gene study91	/			/
113	HMOX2	−42+1444A>G	A	−	Polish	Candidate gene study91	/			/
114	*C9	R95X	NA	−	Japanese	Candidate gene study92	/			/
115	Gas6	834+7G>A	A	−	Hungarian	Candidate gene study93	/			/
116	CCR2	rs4586	T	+	Indian	Candidate gene study94	/			/
117	CCL2	rs1799865	T	+	Indian	Candidate gene study94	/			/
118	KDR	rs2071559	T	+	Italian	Candidate gene study95	/			/
119	IGF1R	rs2872060	G	+	Multipopulations	Candidate gene study96	/			/
120	CFD	rs3826945	C	+	Multipopulations	Candidate gene study97	/			/
121	FADS1-3	rs174547	T	+	Multipopulations	Candidate gene study98	/			/
122	ABCA1	rs1883025	C	+	Multipopulations	Candidate gene study98	/			/
123	TLR3	rs3775291	T	−	European American	Candidate gene study99	/			/
124	TFR2	−576	A	+	Polish	Candidate gene study100	/			/
125	ABCA4	rs3112831	A	+	Spanish	Candidate gene study101	/			/
126	FGF2	rs6820411	A	+	Spanish	Candidate gene study101	/			/
127	*ELOVL4	M299V	NA	−	American	Candidate gene study102	/			/
128	MTHFR	rs1801133	C	+	Japanese	Candidate gene study34	/			/
129	GSTM1	Null genotype	NA	+	Turkish	Candidate gene study103	/			/
130	TNF-a	−1031	C	+	Chinese	Candidate gene study104	/			/
131	RORA	rs12900948	G	+	European American	Candidate gene study105	/			/
132	XPD751	K751Q	A	+	Turkish	Candidate gene study106	/			/
133	SERPING1	rs1005510	A	+	American	Candidate gene study107	/			/
		rs2511989	G	+	American	Candidate gene study107	/			/
134	*NOS2A	rs8072199	T	+	American	Candidate gene study108	/			/
135	SCARB1	rs5888	T	+	Multipopulations	Candidate gene study109	/			/
136	CD36	rs3173798	C	−	Japanese	Candidate gene study110	/			/
		rs3211883	A	−	Japanese	Candidate gene study110	/			/
137	TNMD	rs1155974	T	+	Finnish	Candidate gene study111	/			/
138	*mtDNA	A4917G	G	+	American	Candidate gene study112	/			/
139	*IL8	−251	A	+	English	Candidate gene study113	/			/
		rs2227306	T	+	Italian	Candidate gene study114	/			/
		rs2227543	T	+	Chinese	Candidate gene study62	/			/
140	PEDF	rs1136287	T	+	Chinese	Candidate gene study115	/			/
141	VEGF	674	C	+	English	Candidate gene study116	/			/
		−460	C	+	Polish	Candidate gene study117	/			/
		−634	C	+	Polish	Candidate gene study117	/			/
142	ERCC6	−6530C>G	G	+	American	Candidate gene study118	/			/
143	CFHR3-CFHR1	Deletion	NA	−	Multipopulations	Candidate gene study119	/			/
144	CFHR1	CFHR1*A	Allotype	+	Spanish	Candidate gene study120	/			/
145	LRP6	hCV345771	T	+	American	Candidate gene study121	/			/
146	TLR4	rs4986790	G	+	American	Candidate gene study122	/			/
147	PLEKHA1	rs4146894	T	+	German	Candidate gene study8	/			/
148	HLA	B*4001	NA	−	English	Candidate gene study123	/			/
		DRB1*1301	NA	−	English	Candidate gene study123	/			/
		Cw*0701	NA	+	English	Candidate gene study123	/			/
149	*MMP9	CA allele	>=22 repeats	+	Italian	Candidate gene study124	/			/
150	CX3CR1	V249I	NA	+	American	Candidate gene study125	/			/
151	CST3	rs1064039	A	+	Multipopulations	Candidate gene study126	/			/
152	ACE	Alu insertion	Alu+	+	American	Candidate gene study127	/			/
153	*APOE	NA	ε2	+	Dutch	Candidate gene study128	/			/
Rare variants associated with CSCR
154	CRH	rs562792458	c.154_155insCGC	/			/			+	Chinese	Candidate gene study129
155	PTPRB	4145C>T	T	/			/			+	Dutch	WES130
156	SGK1	M32V	G	/			/			+	Turkish	Candidate gene study131
157	*PIGZ	Multiple rare variants in gene-based analysis		/			/			NA	Dutch	Exome sequencing study132
158	*RSAD1	Multiple rare variants in gene-based analysis		/			/			NA	Dutch	Exome sequencing study132
159	*DUOX1	Multiple rare variants in gene-based analysis		/			/			NA	Dutch	Exome sequencing study132
160	*LAMB3	Multiple rare variants in gene-based analysis		/			/			NA	Dutch	Exome sequencing study132
161	C4B	3 copies	NA	/			/			−	Multi-populations	Candidate gene study3
Rare variants associated with PCV
162	*FGD6	K329R	G	/			+	Chinese	WES133	/
163	*IGFN1	6196A>G	G	/			+	Chinese	Candidate gene study134	/
Rare variants associated with AMD
164	*SLC16A8	214+1G>C	C	+	Multipopulations	GWAS15	/			/
165	*TIMP3	S38C	NA	+	Multipopulations	GWAS15	/			/
166	*CETP	D442G	NA	+	Multipopulations	EWAS55	/			/
167	*UBE3D	rs7739323	T	−	Multipopulations	WES61	/			/
168	TACC2	rs112188313	A	+	Multipopulations	WES135	/			/
169	PHF12	rs148347485	G	+	Multipopulations	WES135	/			/
170	RXFP2	rs121918303	C	+	Multi-populations	WES135	/			/
171	PUS7	rs139058270	C	+	Multipopulations	WES135	/			/
172	SPEF2	rs80010329	G	+	Multipopulations	WES135	/			/
173	BCAR1	rs74024754	C	+	Multipopulations	WES135	/			/
174	PLEKHA1	S177N	NA	+	Jewish	WES136	/			/
175	*COL8A1	V58A	NA	+	Multipopulations	WES137	/			/
176	PELI3	A307V	NA	−	Multipopulations	WES60	/			/
177	*CFI	V412M	NA	+	Israelis	WES138	/			/
178	HMCN1	4162delC	NA	+	Israelis	WES138	/			/
179	*C9	P167S	NA	+	Multipopulations	WES139	/			/
180	C3	K155Q	NA	+	Multipopulations	WES139	/			/
181	FBN2	E1144K	NA	+	American	WES86	/			/
182	C4A	Increased copy number	NA	−	Multipopulations	Candidate gene study119	/			/
183	P2X7	G150R	NA	+	Tunisian Jewish	Candidate gene study141	/			/
184	P2X4	Y315C	NA	+	Tunisian Jewish	Candidate gene study141	/			/
185	CNP147	<2 copies	NA	−	Multipopulations	Candidate gene study142	/			/
186	CFH	R1210C	NA	+	Multipopulations	Candidate gene study143	/			/
187	FBLN5	V60L	NA	+	American	Candidate gene study144	/			/
188	HEMICENTIN-1	Q5345R	NA	+	American	Candidate gene study145	/			/
189	VMD2	K149stop*	NA	+	Multipopulations	Candidate gene study146	/			/
190	*ABCR	E471K	NA	+	American	Candidate gene study147	/			/

AMD indicates age-related macular degeneration; CSCR, central serous choroid retinopathy; EWAS, exome-wide association study; GWAS, genome-wide association study; NA, data not available from the original report; PCV, polypoidal choroidal vasculopathy; WES, whole-exome sequencing.

“+” indicates an increased risk; “−” , an decreased risk; “/” , no previous report; “*” , the genes mentioned in the text.

Please refer to (Supplementary Digital Content 1https://links.lww.com/APJO/A233) for the reference list of studies included in this table.

With the completion of the International HapMap Project and the advancement of genotyping platforms, regional fine-mapping using haplotype-tagging SNP and GWAS became a feasible and effective approach for mapping new AMD genes. In 2005, a GWAS reported a strong association of advanced AMD with the CFH Y402H polymorphism in whites.¹⁸ The association was confirmed by another 2 studies using regional mapping^27,28 and subsequently replicated in multiple white populations. The Y402H polymorphism was also associated with the progression of AMD.²⁹ However, the MAF of Y402H is low in East Asians with no association with AMD.^30,31 Instead, another coding variant in CFH, I62V (rs800292), is a major associated SNP with nAMD in Asians,^31–33 showing allelic diversities in the association profiles among different ethnicities. Subsequently, other genes in the complement pathway have been associated with AMD, including Complement C2 (C2),³⁴Complement Factor B (CFB),³⁴C3,³⁵ and Complement Factor I (CFI),³⁶ initially in whites. C2/CFB and C3 were also independently related to progression from early/intermediate to advanced AMD.³⁷ However, in a Chinese cohort, the SKI2 subunit of superkiller complex (SKIC2, also known as SKIV2L) gene, rather than C2 or CFB, was the gene variant associated with nAMD in the C2-CFB-negative elongation factor complex member E (NELFE, also known as RDBP)-SKIV2L locus.³⁸ Also, C3³⁹ and CFI⁴⁰ were not associated with nAMD in other Chinese cohorts, suggesting ethnic diversities.

Subsequent GWAS on AMD have been reported in different populations. In Chinese, a HtrA serine peptidase 1 (HTRA1) promoter polymorphism, rs11200638, was found to be strongly associated with nAMD in a GWAS.⁴¹ This association was confirmed in whites and other populations.^42,43 In a GWAS of nAMD in Japanese, 2 susceptibility loci were identified: Tumor necrosis factor receptor superfamily member 10a (TNFRSF10A)-LOC389641 and RE1 silencing transcription factor (REST)-nitric oxide associated 1 (also known as C4orf14)-RNA polymerase II subunit B (POLR2B)-insulin-like growth factor binding protein 7 (IGFBP7).⁴⁴ In other GWAS, which were mainly conducted in whites, new loci were identified for AMD (Table 1), including TIMP metallopeptidase inhibitor 3 (TIMP3),⁴⁵Lipase C, Hepatic type (LIPC),⁴⁶Collagen Type VIII Alpha 1 Chain (COL8A1)- Filamin A Interacting Protein 1 Like (FILIP1L), ADAM Metallopeptidase With Thrombospondin Type 1 Motif 9 (ADAMTS9), and beta 3-glucosyltransferase (B3GLCT, also known as B3GALTL).⁴⁷ In a GWAS of AMD progression, age-related maculopathy susceptibility 2 (ARMS2)-HTRA1, CFH, C2-CFB-SKIV2L, C3, LIPC, and Chymotrypsinogen B2 (CTRB2)-Chymotrypsinogen B1 (CTRB1) were associated with AMD progression, and SNPs rs58978565 near tenascin R (TNR), rs28368872 near activating transcription factor 7 interacting protein 2 (ATF7IP2), and rs142450006 near matrix metallopeptidase 9 (MMP9) with progression to nAMD but not geographic atrophy.⁴⁸

In a large-scale GWAS that involved over 16,000 AMD patients and 17,000 controls, 52 independently associated common and rare variants distributed across 34 loci were identified.¹⁹ A locus near MMP9 was found specifically for nAMD, not dry AMD. Very rare coding variants (frequency <0.1%) in CFH, CFI, TIMP3, and solute carrier family 16 member 8 (SLC16A8) were also found with a significant association.¹⁹ These findings suggested that both common and rare variants contribute to the genetic architecture of AMD.

Rare disease-associated variants are mostly located in exonic regions and are more likely to affect protein structure and function. They are usually of larger effect sizes than common SNPs. High-throughput sequencing of the CFH gene identified a strong association of a rare variant, R1210C, with AMD and a 6-year-earlier onset of disease.⁴⁹ Subsequently, a WGS study reported the association of a rare nonsynonymous SNP (K155Q, MAF=0.55%, OR=3.45) in the C3 gene with AMD.²⁰ Later, WES in 9 families with early-onset AMD identified 2 rare and functional CFH variants (R53C and D90G).⁵⁰ The first EWAS on AMD identified an East Asian–specific mutation, cholesteryl ester transfer protein (CETP) D442G, which increased the risk of AMD, along with another 3 novels AMD loci: long intergenic nonprotein coding RNA 3040 (LINC03040, also known as C6orf223), solute carrier family 44 member 4 (SLC44A4) and FYVE, RhoGEF and PH domain containing 6 (FGD6).⁵¹ Rare variants in other genes have also been identified for AMD by using regional targeted deep sequencing and WES (Table 1), such as complement C9 (C9),⁵²CFI,⁵²ubiquitin protein ligase E3D (UBE3D),⁵³ and collagen type VIII alpha 1 chain (COL8A1).⁵⁴

Apart from GWAS, WGS, and WES, the candidate gene approach has also identified new candidate genes for AMD. Fine-mapping a candidate gene can pinpoint the causal variant and SNP analysis and help identify rare disease-associated coding variants. For example, the placental growth factor (PGF), angiopoietin 2 (ANGPT2), and TEK receptor tyrosine kinase (TEK, also known as TIE2) genes in the angiogenesis pathway have been identified as candidate genes for nAMD in Chinese,^55–57 so were the CETP and Adenosine triphosphate binding cassette subfamily G member 1 (ABCG1) genes in the high-density lipoprotein metabolic pathway.⁵⁸ By sequencing analysis of PGF, a novel 18-base-pair deletion mutation in the promoter was found to confer over 5-fold of risk to nAMD.⁵⁹ It is noteworthy that many candidate genes have been reported just once in specific study cohorts (Table 1). Further replications in other study populations are warranted.

Molecular Genetics of Polypoidal Choroidal Vasculopathy

As for AMD, candidate gene analysis is a major approach used to identify susceptibility genes for PCV. In view of the clinical similarities between nAMD and PCV, some major genes for nAMD have been tested for their associations with PCV. In a white cohort, SNPs in 3 major AMD loci, CFH, CFB/C2, and ARMS2-HTRA1, were associated with PCV.⁶⁰ In 2007, the ARMS2-HTRA1 variants were associated with both PCV and wet AMD in a Japanese population.⁶¹ In a Singaporean Chinese cohort, SNPs rs3753394 and rs800292 of CFH and rs11200638 of HTRA1 were associated with PCV.⁶² In Chinese cohorts from different regions of China, SNPs in CFH and ARMS2-HTRA1 have also been associated with PCV. In Hong Kong Chinese, ARMS2 rs10490924, HTRA1 rs11200638, CFH rs800292, and CETP rs3764261 were associated with PCV,^58,63 and ABCG1 rs225396, ANGPT2 rs13269021, and TIE2 rs625767 were first identified as genetic factors for PCV.^56,57,64 In a systematic review and meta-analysis, 31 polymorphisms in 10 AMD-associated genes/loci: ARMS2, HTRA1, CFH, C2, CFB, RDBP, SKIV2L, CETP, 8p21, and 4q12, were confirmed to be associated with PCV. However, 12 polymorphisms at the ARMS2-HTRA1 locus showed significant differences between PCV and nAMD.⁶⁵ Later in a study in multiple East Asian cohorts, SNPs at 8 out of 34 known AMD loci were associated with PCV, including ARMS2- HTRA1, CFH, C2-CFB-SKIV2L, CETP, vascular endothelial growth factor A (VEGFA), ADAMTS9-AS2, transforming growth factor beta receptor 1 (TGFBR1), and collagen type IV alpha 3 chain (COL4A3). Weaker association for PCV was observed at ARMS2-HTRA1 and lysine methyltransferase 2E (KMT2E)- SRSF protein kinase 2 (SRPK2), compared with nAMD.⁶⁶ In contrast, WES analysis showed an association of rare coding variants c.986A>G in FGD6,⁶⁷ and c.6196A>G in immunoglobulin-like and fibronectin type III domain containing 1 (IGFN1),⁶⁸ with PCV but not with nAMD. These findings together suggested that PCV and nAMD have shared and distinct genetic components, and certain gene variants that could affect the phenotypic expressions of PCV and nAMD.

Molecular Genetics of Central Serous Choroid Retinopathy

In CSCR, CFH was the first candidate gene assessed. A significant association has been reported between Japanese⁶⁹ and Europeans.^70,71 In a GWAS, two SNPs, CFH rs800292 and vasoactive intestinal peptide receptor 2 (VIPR2) rs3793217, were associated with CSCR in Japanese.⁷² The involvement of the complement system in cCSCR was also confirmed by a GWAS in Europeans.⁷³ In addition, a GWAS in Japanese and European cohorts identified 2 loci for CSCR, rs13278062 in TNFRSF10A-LOC389641 and rs6061548 near GATA binding protein 5 (GATA5).⁷⁴ In a meta-analysis study, significant associations of ARMS2, CFH, and TNFRSF10A with CSCR were confirmed.⁷⁵ Notably, SNP rs13278062 in TNFRSF10A had the same trend of effect on CSCR, nAMD, and PCV, whereas SNPs in ARMS2 and CFH showed an opposite effect on CSCR as compared with nAMD and PCV.⁷⁵ Of note, the same trend of effect means that a certain allele of an SNP has a risk or protective effect on one disease and a consistent effect on another disease. On the contrary, an opposite effect means a specific allele shows a risk effect on one disease but a protective effect on another disease. In a genome-wide survival analysis of macular neovascularization (MNV) development in CSCR, 4 known AMD/PCV genes, ARMS2, CFH, COL4A3, and B3GALTL, were found of association with MNV development.⁷⁶ Of note, COL4A3 and B3GALTL were associated stronger with PCV, suggesting certain shared genetic susceptibility between PCV and MNV in the central serous choroid.

AMD, PCV, and CSCR showed both similarities and differences in genetic associations with ethnic diversities. Further larger-scale genomic studies are warranted to identify more susceptibility genes for PCV and CSCR, and differentiate the genetic architectures among nAMD, PCV, and CSCR. Moreover, these diseases are multifactorial. Single gene only explains a small part of the disease heritability. Thus, studying the interactions among the genes and other factors, such as inherent systemic conditions, lifestyle, environmental exposures, and interventional modalities, will provide further insights into the roles of the genes in the pathogeneses and management of the diseases.

JOINT EFFECTS OF GENES, ENVIRONMENTAL AND INHERENT FACTORS

AMD, PCV, and CSCR are complex diseases that resulted from the additive and interactive effects of multiple environmental, inherent, and genetic risk factors. While aging, smoking, higher body mass index, and higher high-density lipoprotein-cholesterol pose risks for AMD and PCV,^3,77 risk factors for CSCR include exogenous glucocorticoid usage, endogenous hypercortisolism, psychological stress, and type A personality.⁷⁸ How they act on the pathogeneses of these maculopathies is complicated and not exactly known.

The joint effects of 2 or more genes, most frequently between CFH and ARMS2-HTRA1, have been reported in nAMD and PCV, but not in CSCR. In study subjects from the United States, significant gene-gene interaction was detected between CFH rs10801575 and HTRA1 rs2014307.⁷⁹ Subjects homozygous for both risk alleles of CFH Y402H and ARMS2 A69S had a 50-fold increased risk of AMD.⁸⁰ Individuals homozygous for both the G allele of RORA rs12900948 and ARMS2 A69S had a 40.8-fold increased risk of nAMD.⁸¹ In the International AMD Genomics Consortium data set, intronic and intergenic nuclear SNPs in abhydrolase domain containing 2, acylglycerol lipase (ABHD2)/retinaldehyde binding protein 1 (RLBP1) demonstrated significant joint effects and nominally significant interaction effects with the synonymous mitochondrial SNP MT-ND5 G12771A.⁸² In a Finnish cohort, carriers of at least 1 risk allele in each of the 3 susceptibility loci (ARMS2, CFH, and C3) had an 18-fold risk of AMD when compared with homozygote noncarriers in all 3 loci.⁸³ In a Korean cohort, joint effects for CFH and ARMS2 variants were shown on nAMD, with the ORs for individuals carrying 1, 2, and 3-copy risk alleles being 1.08, 3.49, and 3.64, respectively.⁸⁴ In a Hong Kong Chinese cohort, combined CFH rs800292 and HTRA1 rs11200638 caused a 23.3-fold increased risk of nAMD.⁸⁵ Moreover, CFH rs800292 had a significant interaction with ANGPT2 rs13269021 on nAMD and PCV.⁵⁶ Polygenic risk score, which summarizes the estimated effect of multiple variants on a phenotype, would be an excellent tool to further study the multi-SNP effects of more than 2 SNPs on nAMD and PCV, including particularly the significant SNPs identified in GWAS.

Although the genetic architecture of nAMD and PCV involve gene-gene interactions, there are also gene-environment interactions. A dominant environmental risk factor is smoking. A study in the United States has detected significant interaction between ARMS2 A69S and cigarette smoking history. ARMS2, CFH, and smoking together account for 61% of the population-attributable risk of AMD, with a population-attributable risk percentage of 20% for smoking, 36% for ARMS2, and 43% for CFH.⁸⁶ In a genome-wide gene-environment interaction analysis, an intergenic SNP rs17073641 between serpin family B member 8 (SERPINB8) and cadherin 7 (CDH7) associated strongly with AMD in nonsmokers (OR=0.57) but inversely among smokers (OR=1.42), indicating genetic modifying effects of smoking.⁸⁷ In another cohort from the United States, a significant interaction between nitric oxide synthase 2 (NOS2, also known as NOS2A) rs2248814 and smoking was detected on AMD risk, with the stronger association between AMD and smoking in carriers of AA genotypes (OR=35.98) than in carriers of the AG genotype (OR=3.05) or GG genotype (OR=2.1), suggesting that NOS2A might modulate the effect of smoking on AMD.⁸⁸ In an Australian cohort, the association of the apolipoprotein E (APOE) with early AMD was varied by smoking status, with the ε2-containing genotypes positively associated with early AMD for never and previous smokers but not for current smokers, whereas the ε4-containing genotype group (ε3ε4/ε4ε4) had an inverse association with early AMD among current smokers only.⁸⁹ In a Hong Kong Chinese cohort, combined HTRA1 rs11200638 and smoking caused a 15.7-fold increased risk to exudate AMD.⁸⁵ In a Japanese cohort, there was a significant interaction between CFH Y402H and smoking on PCV, with a synergy index of 2.41.⁹⁰ These findings together provided strong evidence for the involvement of gene-smoking interaction in the risk of AMD and PCV.

Certain inherent factors, such as sex, also interact with genetic factors and expresses a specific association with AMD, PCV, and/or CSCR. A joint analysis of 3229 cases and 2835 controls from Germany, UK, and United States, revealed the association of a synonymous SNP rs17810398 in death-associated protein-like 1 (DAPL1) with AMD in females (P = 2.62×10^-8, OR = 1.54) but not in males (P = 0.38, OR = 1.08).⁹¹ In an Australian cohort, a significant association of the ε2 genotype of APOE was found only in females who progressed with AMD.⁹² Apart from gene variants, the relative telomere length was also associated with AMD, with significantly shorter relative telomere length in AMD patients compared with controls in a European cohort, and this association occurred only in women (OR = 1.14; P < 0.001), not in men (OR = 1.01; P = 0.76).⁹³ These findings indicate certain sex-differential pathways for AMD, especially in females.

An interaction between C3 rs17030 and sex were identified in PCV in a Hong Kong Chinese cohort, with the rs17030-G allele conferring an increased risk for PCV in males (P = 0.010, OR = 1.56) but not in females. C3 may cast an epistatic effect with sex in the pathogenesis of PCV.³⁹ As for CSCR, the number of carriers in a Netherlands cohort of rare variants in the phosphatidylinositol glycan anchor biosynthesis class Z (PIGZ), dual oxidase 1 (DUOX1), radical S-adenosyl methionine domain containing 1 (RSAD1), and laminin subunit beta 3 (LAMB3) genes was significantly higher in the female chronic CSCR patients compared with female controls, whereas no associations were identified in the male patients.⁹⁴ Of note, such sex-differential associations of gene variants with AMD, PCV, and cCSCR were reported in single studies and required replications.

PHARMACOGENETICS IN AMD, PCV, AND CSCR

Different responses to treatments, mainly the anti-VEGF therapy and PDT, have been observed in nAMD, PCV, and CSCR patients. There is a putative genetic background underlying individuals’ response to the treatments. In fact, some gene variants have been reported to cast pharmacogenetic effects on AMD, PCV, and CSCR.

PDT has been a major treatment for nAMD and PCV, especially before the advent of anti-VEGF therapy. Half-dose PDT is the major treatment for chronic CSCR. The effects of PDT treatments for nAMD, PCV, and CSCR have been investigated in patients carrying different genotypes. Pharmacogenetic studies of PDT for CSCR are limited. In a Chinese cohort, none of the 7 SNPs examined in CFH were associated with 1-month outcomes after PDT in patients with CSCR.⁹⁵ In contrast, in a Japanese cohort with a 12-month follow-up, the minor T allele of ARMS2 A69S was associated with poor response to PDT in patients with chronic CSCR.⁹⁶

In nAMD, pharmacogenetic studies on PDT have been focused on the ARMS2-HTRA1 and CFH genes, with variable results. In a Japanese cohort of wet AMD, the best-corrected visual acuity (BCVA) 1 year after PDT was significantly increased in patients with HTRA1 rs11200638 GG as compared with patients with GA or AA.⁹⁷ In another Japanese cohort, ARMS2 rs10490924 was associated with the 3-year outcomes of PDT in patients with wet AMD, with the G allele associated with greater improvement in BCVA at 36 months after the first PDT.⁹⁸ In contrast, in a wet AMD cohort from the United States, ARMS2 A69S (rs10490924) was not associated with the response to PDT treatment.⁹⁹ In a cohort from Israel, ARMS2 rs10490924 and HTRA1 rs11200638 were also not associated with the response to PDT in nAMD.¹⁰⁰ Therefore, the pharmacogenetic association of ARMS2-HTRA1 with PDT in nAMD seemed to be population specific. Further studies in more populations are warranted.

In a wet AMD cohort from the United States, CFH Y402H was associated with the response to PDT, with patients with the TT genotype having worse outcomes than those with TC and CC.⁹⁹ In a UK cohort, wet AMD patients homozygous for CFH Y402H also had worse visual acuity after PDT.¹⁰¹ However, in a Finland cohort of nAMD, CFH Y402H did not affect the outcome of PDT.¹⁰² Also, in an Australian cohort of nAMD, CFH Y402H had no significant association with the response to PDT;¹⁰³ instead, 2 SNPs in C-reactive protein, rs2808635, and rs876538, were associated with a positive response to PDT.¹⁰³ Further studies in more cohorts are needed to elucidate the effects of these candidate genes in the pharmacogenetics of nAMD.

Pharmacogenetic studies on PDT for PCV were mainly reported in Asian populations. In a Korean cohort, PCV patients with the T allele of ARMS2 A69S had poorer BCVA than those with the G allele after combined PDT and intravitreal bevacizumab treatment at 12 months.¹⁰⁴ In Japanese patients, a pharmacogenetic association was found between ARMS2 A69S and the long-term results after PDT in eyes with PCV, with patients carrying the T allele having poorer visual acuity at the 12-month visit.¹⁰⁵ Also, the G allele of ARMS2 A69S was associated with a lower chance of retreatment after intravitreal aflibercept combined with PDT in PCV patients.¹⁰⁶ Besides, the serpin family F member 1 (SERPINF1) rs12603825 was shown to influence the initial response and visual prognosis after PDT for PCV. Patients with the AA genotype received additional treatment after PDT within a significantly shorter time required more retreatment within 3 months, and showed worse visual prognosis after PDT.¹⁰⁷

Anti-VEGF therapy is the mainstay treatment for nAMD and PCV, and chronic CSCR with secondary choroidal neovascularization. The response of nAMD patients to anti-VEGF treatments has been associated with polymorphisms in different genes, mainly CFH, ARMS2-HTRA, and VEGFA. In study cohorts from the United States, wet AMD patients with the CC genotype of CFH Y402H had worse visual outcomes with intravitreal bevacizumab than those with TC and TT.¹⁰⁸ Wet AMD patients homozygous for the CFH Y402H risk allele had a 37% higher risk of requiring additional ranibizumab injections over 9 months.¹⁰⁹ In a cohort from Austria, the CFH 402HH genotype was correlated with lower visual acuity outcomes after treatment with bevacizumab.¹¹⁰ For other CFH polymorphisms, the CFH rs800292 (I62V) was associated with a poor response to anti-VEGF treatment for nAMD.¹¹¹ These findings showed that the response to anti-VEGF treatment differed according to CFH genotypes in nAMD patients.

There are ethnic differences in pharmacogenetic associations for the AMRS2-HTRA1 locus. In an Australian cohort of nAMD, the homozygote risk genotypes AA at HTRA1 rs11200638 and GG at ARMS2 A69S were associated with poorer VA outcomes at 12 months after ranibizumab or bevacizumab treatment.¹¹² In a Spanish cohort of nAMD, patients homozygous for the risk T allele of ARMS2 A69S required more retreatments of anti-VEGF drugs over the 48-month follow-up.¹¹³ Notably, however, in a Korean cohort of nAMD, patients with the risk alleles of ARMS2-HTRA1 had improved treatment response and less need for additional injections.¹¹⁴

There is also evidence of pharmacogenetic responses to the VEGFA gene in anti-VEGF treatment for nAMD. In a United States cohort of nAMD, the poor-responders to anti-VEGF therapy had a higher frequency of the risk T allele of VEGFA rs943080.¹¹⁵ In an Australian cohort of nAMD, patients with the T allele of VEGFA rs3025000 had better visual outcomes at 6 months and a higher likelihood to respond to anti-VEGF treatment at 3, 6, and 12 months.¹¹⁶ In a Spanish cohort, the VEGFA rs699947 polymorphism predisposed nAMD patients to a good response to ranibizumab treatment.¹¹¹

Other genes have also been shown with pharmacogenetic associations with anti-VEGF treatment effects in nAMD. In a Spanish cohort, genetic variants in CFB and fms-related receptor tyrosine kinase 1 (FLT1, also known as VEGFR1) predisposed patients to good response, whereas variants in SERPINF1 were associated with poor response.¹¹¹ In another Spanish cohort, the BCVA response after 52-week aflibercept treatment in nAMD patients was associated with vascular endothelial growth factor B (VEGFB) rs12366035 and complement C5 (C5) rs12366035.¹¹⁷ In a cohort of nAMD from Finland, the interleukin 8 (IL8) promoter polymorphism 251A/T was associated with anatomic nonresponse to bevacizumab treatment.¹¹⁸ Moreover, in a cohort of European descendants, rare variants in the chromosome 10 open reading frame 88 (C10orf88) and unc-93 homolog B1 (UNC93B1) genes had a worse response to anti-VEGF therapy in nAMD patients.¹¹⁹ These findings suggested that both rare and common gene variants affect the pharmacogenetic responses of nAMD patients to anti-VEGF treatment.

Of note, however, negative findings have been reported. In a United States cohort of nAMD, SNPs CFH Y402H, ARMS2 A69S, and HTRA1 rs11200638 were not associated with the response to anti-VEGF therapy.¹²⁰ In another United States cohort, the CFH, ARMS2-HTRA1, C3, VEGFA, and kinase insert domain receptor (KDR, also known as VEGFR2) gene variants were not associated with the response to anti-VEGF therapy in nAMD patients.^121–123 In a randomized controlled trial from the UK, no significant pharmacogenetic association was detected for nAMD with ARMS2 A69S.¹²⁴

Pharmacogenetic reports of anti-VEGF treatment for PCV are mainly on Asian populations. In a Korean cohort, PCV patients with the ARMS2 A69S-TT and HTRA1 rs11200638-AA genotypes had poorer anatomic and visual outcomes at 12 months after combined PDT with intravitreal bevacizumab treatment.¹⁰⁴ In another Korean cohort, a significant pharmacogenetic association was found between ARMS2 A69S and the regression of retinal pigment epithelial detachment in PCV patients who completed the 12-month anti-VEGF monotherapy.¹²⁵ In a Japanese cohort of PCV patients receiving PDT combined with intravitreal ranibizumab or aflibercept, the absence of retreatment over 24 months was associated with the nonrisk genotype of ARMS2 A69S.¹²⁶

As of to-date, genetic polymorphisms have been related to variable responses to anti-VEGF therapy in nAMD and PCV patients. Further studies in more study cohorts are needed, taking into account of the differences in disease subtypes, treatments, follow-up time, selection of candidate genes and variants, genetic models, and the combined effects of different genetic variants.

The gene-gene, gene-environment, and pharmacogenetic interaction analyses, especially at a genome-wide scale, will involve and generate large data sets. Advanced and automatic analytical technologies are needed. Machine learning has been used to study epistasis, especially to identify gene-gene interactions from GWAS data.¹²⁷ Artificial intelligence (AI) with deep learning will play an important role in analyzing big data.

CONTRIBUTIONS OF ARTIFICIAL INTELLIGENCE

In recent years, AI has been actively applied to investigations of retinal diseases, advancing phenotyping, and refining structural features. In nAMD, deep learning has provided reliable algorithms in objective quantification of the microvasculature in the retina.¹²⁸ Optical coherence tomography (OCT) images have been analyzed by AI technology for automated segmentation, extraction, and quantification. In nAMD, such analysis can give predictions of disease progression. Applications have been attempted for AI systems in clinical practice utilizing OCT by automated whole-volume segmentation for differential diagnosis of AMD and diabetic macular edema.¹²⁹ Fundus photos can be analyzed by deep learning algorithms with deep convolution and network architectures to predict disease severity¹³⁰ and differentiate a multitude of retinal complications.¹³¹ However, responses to treatment have not been reported.¹³²

AI investigations on OCT images and fundus photos of retinal diseases have advanced phenotyping and refined structural features including minute temporal and microvasculature changes. Attempts have also been to correlate to genotypes. GWAS data sets involving 32,215 whites aged over 50 years from the International AMD Genomics Consortium have been tested by the neural network, lasso regression, support vector machine, and random forest to develop risk and prediction models for AMD.¹³³ These machine learning-based methods are capable to predict AMD at all stages with acceptable areas under the curve and Brier scores. A study has involved a huge amount of data for developing an automatic and comprehensive prediction model for AMD by a machine learning algorithm.¹³⁴ The data analyzed contained detailed demographic and ophthalmic information including age, drusen, hyper-pigmentation, and fundus photos. Lifestyle factors included smoking, diet, and education. The genetic information was 49 AMD-associated SNPs. Notably, an acceptable estimation of the area under the curve at around 0.92 was obtained.¹³⁴ Meanwhile, a deep learning approach of a combination of convolution neural networks has been used for the automatic classification of AMD in different stages, which were analyzed with for over more than 170,000 photographs with 10 genetic loci obtained by GWAS to reveal the detailed links of genetic variants with advancements of AMD.¹³⁵ The results from such big studies prove that the utilization of genetic information to establish genetic risk scores by advanced AI methodologies should provide a robust platform for reliable presymptomatic diagnosis and for prognostic prediction.¹³⁶ A recent bibliometric analysis of genetic retinal diseases in India has revealed available complex and non-Mendelian genetic data of AMD.¹³⁷ Combination of such large amount of genetic data with refined phenotyping can lead to the establishment of disease prediction models by advanced AI technologies, which should pave the way to precision medicine for AMD.

CLINICAL PERSPECTIVES

Complex retinal diseases are difficult to cure. Their clinical presentations and course are complicated especially with the involvements of choroidopathy and microvasculature. Early diagnosis is crucial for management to prevent or inhibit disease progression. Prediction of prognosis is vital too. There is also a need to differentiate nAMD, PCV, and CSCR for effective management. They overlap in some features but are different in clinical course, prognosis, and responses to treatment.³ Clinical assessments, ophthalmoscopic examinations, OCT imaging, and fundus photography have been exhaustively used and advanced for diagnostic confirmation and prognostic prediction. There can be prediction scores, but individualized assessments are required. Such individualized precision medicine is made possible with the advent and application of AI technologies to deal with the extremely larger volume of data and complicated data analysis. AI in recent years has been developed for the reliable and meticulous analysis of genetic data with clinical implications.¹³⁸ Genetic data, with big advancements in recent years particularly by GWAS, has been proven useful for clinical prediction, as well exemplified in a recent report of the association of 30 genetic variants with responses to anti-VEGF treatment in a big study cohort of over 15,000 study subjects.¹³⁹ In a study of GWAS variants in totally 2058 study subjects in discovery and validation cohorts, SNP rs12138564 in the chaperonin containing TCP1 subunit 3 (CCT3) gene has a mild association with anti-VEGF treatment outcome.¹¹⁹

Accurate and informative prediction of risk, progression, and prognosis of AMD, PCV, and CSCR required individual information for the clinical management to be tailored made. Such individualized precision medicine should be possible for general applications in most populations in the near future with the advancements of AI technologies and genetic information.

CONCLUSIONS

Neovascular AMD, PCV, and CSCR represent a major group of complex maculopathies and a leading cause of central visual impairment. They are multifactorial in etiology, resulted from joint and interactive effects of multiple risk factors. In recent years, the application of different strategies of genetic approaches has led to the identification of a large number of candidate genes for these macular diseases, especially for nAMD. Also, the shared and distinct genetic components and profiles among the 3 maculopathies have been recognized, indicating the importance of comparative genetics in further understanding of their pathogeneses. Apart from the main effect of single genes, gene-gene, and gene-environment interactions also play a role in the genetic architectures of maculopathies. In addition, pharmacogenetic interactions between gene variants and treatment responses will make personalized treatment possible. Analysis of these interactions, in the context of genome-wide data and a complete set of refined phenotyping ocular parameters, requires advanced statistical programs and computer power. AI with deep learning will thus play an essential role in automating these analyses in a more efficient manner. These, in the long run, will facilitate the identification of more genetic biomarkers with translational values, leading to advanced and more precise diagnostics, subgrouping, treatment options, and prognostic assessment and prediction, eventually toward precision medicine for complex retinal diseases.