November 2023
Volume 64, Issue 14
Open Access
Genetics  |   November 2023
Whole Genome Sequencing Identifies Novel Common and Low-Frequency Variants Associated With Age-Related Macular Degeneration
Author Affiliations & Notes
  • Ilhan E. Acar
    Roche Pharma Research and Early Development, Roche Innovation Center Basel, Basel, Switzerland
    Department of Ophthalmology, Donders Institute for Brain, Cognition and Behaviour, Radboud University Medical Center, Nijmegen, The Netherlands
  • Tessel E. Galesloot
    Radboud University Medical Center, Radboud Institute for Health Sciences, Department for Health Evidence, Nijmegen, The Netherlands
  • Ulrich F. O. Luhmann
    Roche Pharma Research and Early Development, Roche Innovation Center Basel, Basel, Switzerland
  • Sascha Fauser
    Roche Pharma Research and Early Development, Roche Innovation Center Basel, Basel, Switzerland
  • Javier Gayán
    Roche Pharma Research and Early Development, Roche Innovation Center Basel, Basel, Switzerland
  • Anneke I. den Hollander
    Department of Ophthalmology, Donders Institute for Brain, Cognition and Behaviour, Radboud University Medical Center, Nijmegen, The Netherlands
  • Everson Nogoceke
    Roche Pharma Research and Early Development, Roche Innovation Center Basel, Basel, Switzerland
  • Correspondence: Everson Nogoceke, Roche Pharma Research and Early Development, Roche Innovation Center Basel, F. Hoffmann-La Roche Ltd., Grenzacherstrasse 124, 4070 Basel, Switzerland; [email protected]
Investigative Ophthalmology & Visual Science November 2023, Vol.64, 24. doi:https://doi.org/10.1167/iovs.64.14.24
  • Views
  • PDF
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Ilhan E. Acar, Tessel E. Galesloot, Ulrich F. O. Luhmann, Sascha Fauser, Javier Gayán, Anneke I. den Hollander, Everson Nogoceke; Whole Genome Sequencing Identifies Novel Common and Low-Frequency Variants Associated With Age-Related Macular Degeneration. Invest. Ophthalmol. Vis. Sci. 2023;64(14):24. https://doi.org/10.1167/iovs.64.14.24.

      Download citation file:


      © ARVO (1962-2015); The Authors (2016-present)

      ×
  • Supplements
Abstract

Purpose: To identify associations of common, low-frequency, and rare variants with advanced age-related macular degeneration (AMD) using whole genome sequencing (WGS).

Methods: WGS data were obtained for 2123 advanced AMD patients (participants of clinical trials for advanced AMD) and 2704 controls (participants of clinical trials for asthma [N = 2518] and Alzheimer's disease [N = 186]), and joint genotype calling was performed, followed by quality control of the dataset. Single variant association analyses were performed for all identified common, low-frequency, and rare variants. Gene-based tests were executed for rare and low-frequency variants using SKAT-O and three groups of variants based on putative impact information: (1) all variants, (2) modifier impact variants, and (3) high- and moderate-impact variants. To ascertain independence of the identified associations from previously reported AMD and asthma loci, conditional analyses were performed.

Results: Previously identified AMD variants at the CFH, ARMS2/HTRA1, APOE, and C3 loci were associated with AMD at a genome-wide significance level. We identified new single variant associations for common variants near the PARK7 gene and in the long non-coding RNA AC103876.1, and for a rare variant near the TENM3 gene. In addition, gene-based association analyses identified a burden of modifier variants in eight intergenic and gene-spanning regions and of high- and moderate-impact variants in the C3, CFHR5, SLC16A8, and CFI genes.

Conclusions: We describe the largest WGS study in AMD to date. We confirmed previously identified associations and identified several novel associations that are worth exploring in further follow-up studies.

Age-related macular degeneration (AMD) is a progressive disease that causes central vision loss when it progresses toward advanced stages.1 AMD is the most common cause of vision loss in the elderly population in Western countries with limited therapy options.2 Therefore numerous studies have investigated the disease pathogenesis to discover potential drug targets. AMD is a multifactorial disease to which both genetic and environmental factors contribute. Family and twin studies have demonstrated a strong genetic component in AMD, with a heritability estimate of up to 71%.3 
Since 2005 several genome-wide association studies (GWASs) have been performed in AMD to identify genetic risk factors of the disease.4 Initial GWASs involved relatively small sample sizes and used genotyping microarrays that measure common variants only.5,6 Over the years the sample sizes of GWASs increased and were extended to microarray platforms that query both common and rare variants. The largest GWAS that was performed for AMD to date reported 52 genetic variants at 34 loci that are independently associated with AMD.7 The strongest associations were reported for variants at the CFH, ARMS2/HTRA1, C2/CFB, C3, and APOE loci. The genes in the identified loci can be grouped into three main pathways: the complement pathway, lipid metabolism, and extracellular matrix remodeling. The 52 variants were estimated to account for 27.2% of the disease variability, which is more than half of the total genomic heritability, which was estimated to be 46.7% based on genotyped markers of the study. In addition to these 52 variants, the same study detected a burden of rare coding variants in four genes: CFH, CFI, TIMP3 and SLC16A8
More recently, exome-wide association was investigated by whole-exome sequencing (WES), which involves sequence analysis of all exons in the genome, and can thus detect associations with common, low-frequency, and rare variants in the coding regions.810 These GWASs using WES identified additional associations of coding variants in UBE3B, FGD6, and COL8A1 with AMD. In addition, WES studies have been performed in AMD families and in cases with extreme phenotypes, which mainly identified rare coding variants in the CFH gene and pinpointed several other potential candidate genes for AMD (e.g., SCN10A and KIR2DL4).1120 
To date, one study performed whole genome sequencing (WGS)—involving sequencing of the entire genome—in AMD.21 This study included a total of 1689 AMD cases and 1518 controls, and specifically focused on rare loss-of-function variants in the previously identified AMD loci. An enrichment of loss-of-function variants in the complement system was observed in cases compared to controls. However, a main limitation of the study was that the WGS data was limited to 6 × coverage, which impacts the reliability of the identified variants. Additional studies with a higher sequence coverage and even larger populations are necessary to confirm the identified associations and to detect novel associations. 
Here, we performed a GWAS for AMD using WGS data with an average of 30 × coverage to accurately genotype low-frequency and rare variants across the entire genome. With this approach, we aimed to better cover regions that were not adequately investigated by previous studies. We included 2123 advanced AMD patients and 2704 controls from clinical trials. Our study comprised the following aims: (1) to confirm previously identified AMD variants, (2) to identify new AMD-associated variants in the whole genome, including low-frequency and rare variants in coding, intronic and intergenic regions, and (3) to perform gene-based association analyses to increase power by aggregating rare and low-frequency variants for each gene region. 
Methods
Study Population
The data were obtained through clinical trials performed by F. Hoffmann-La Roche Ltd., Basel, Switzerland. These trials included advanced-stage AMD patients with either geographic atrophy secondary to AMD (i.e., dry AMD; CHROMA phase III clinical trial for lampalizumab, SPECTRI phase III clinical trial for lampalizumab, Proxima A and B observational studies for geographic atrophy progression; N = 1272), and with neovascular AMD (i.e., wet AMD, HARBOR phase III clinical trial for ranibizumab; n = 937). The control samples were chosen from a limited pool of samples available in the ROCHE WGS database from other clinical trials. We selected individuals from clinical trials for asthma and Alzheimer's disease as controls (N = 2865), because these diseases were reported to have no or limited shared genetic etiology with AMD (Table 1).22 All included cases, and controls were self-reported to be of European ancestry, which was confirmed by principal component analysis. 
Table 1.
 
Demographics of the Data Set After Quality Control
Table 1.
 
Demographics of the Data Set After Quality Control
All patients in the clinical trials provided written informed consent to Roche, including consent to perform WGS. All AMD patients were evaluated with thorough eye examinations; however, no eye examination data were available for the asthma and Alzheimer's patients. 
Sample Sequencing
WGS was performed in a similar manner as described by Tom et al.23 Illumina 30 × WGS was performed on a total of 5074 samples using paired-end 150 base pair sequencing. Library preparation was performed using a combination of the Illumina TruSeq Nano and a custom Kapa preparation and then sequenced on an Illumina X10 machine. Read alignment and genotyping followed the GATK best practices pipeline for germline whole genome calling with haplotype caller.24 Sequences were aligned to the human reference genome build GRCh38. The gVCFs were jointly genotyped together into a single VCF file. The pass filter for GATK was set to the 99.0th truth tranche to include as many true positives as possible. After the GATK quality control step, sites that passed GATK VQSR were retained. The HLA region was excluded. 
Quality Control of WGS Data
Variants were further quality controlled according to their read depth and mapping quality. The threshold for read depth was set at ≥10, and mapping quality was set at ≥15. Variants were removed from the dataset if the genotyping call rate was <95% or if a deviation from Hardy-Weinberg equilibrium was detected (P value <1 × 10−6). At the sample level, we removed (1) samples with a call rate <95%, (2) non-European individuals based on principal component analysis of the WGS data from our current study together with the WGS data from the 1000 Genomes Project and subsequent visualization of PC1 versus PC2, (3) related individuals based on a PI-HAT 0.25 threshold for cryptic relatedness, and (4) gender mismatches based on discrepancies between chromosome X sequencing data and recorded sex. After the quality control steps, variants were divided into three groups depending on their minor allele frequencies (MAF): (1) common variants (MAF > 5%), (2) low frequency variants (1% ≤ MAF ≤ 5%), and (3) rare variants (MAF <1%). After quality control, 86 cases and 161 controls were filtered out. Approximately 16 million variants with an average coverage of ×30 reads remained in the data, of which 8.06 million were common and low- frequency variants. 
Statistical Analyses
Single-variant analysis was performed using logistic regression models in REGENIE software including a Firth bias correction to account for sparse data for the low-frequency and rare variants.25 For the chromosome X, heterozygous males were coded as heterozygous females (i.e., males were 0 or 1, females were 0, 1 or 2). Indels larger than 2 bp were excluded from the dataset because of quality measure concerns, such as large frequency differences or not being found in databases and because of their presence in fewer than five cases or controls. Gender and the first two principal components were added as covariates to the association models to account for the effect of gender on AMD risk and to adjust for potential residual population stratification bias, respectively. Age was not added as a covariate due to the difference in age distribution between cases and controls (Table 1). Effects of the variants were reported for the alternate alleles. All single variants were annotated to ascertain their functional effect and nearest genes, using the snpEff tool.26 Significance thresholds were adapted from Fadista et al.27 Common variants were considered genome-wide significantly associated if the p-value was below 5 × 10−8. For low frequency variants 3 × 10−8 and for rare variants 1 × 10−8 was considered to be genome-wide significant; however, variants that pass the 5 × 10−8 threshold were kept as suggestive significant associations. 
Gene-based analyses were performed for low-frequency and rare variants using SKAT-O as implemented in EPACTS.28,29 Genes were annotated using snpEff, and the gene and intergenic regions were predetermined from the GENCODE reference. For each gene, we performed three analyses using a selection of single variants based on the putative impact information from snpEff annotation: (1) all variants, (2) modifier impact variants, and (3) high and moderate impact variants (Supplementary Table S1). These three groups were made for rare and low-frequency variants combined (MAF ≤ 5%). Similarly as for the single variant analyses, gender and the first two principal components were added as covariates. The significance level for the gene-based analyses was set at 7 × 10−7 to correct for the evaluation of 69,934 features, including both gene and intergenic regions. A significance threshold of 1 × 10−5 was chosen to determine suggestive associations. 
To ascertain independence of our identified associations from previously reported AMD loci (N = 527) and from known asthma loci (N = 167),30 we performed conditional analyses to adjust for these variants. For the variants that were not present in our dataset (see Supplementary Table S2), we included linkage disequilibrium (LD) proxies (r2 > 0.7) if available. To limit the number of covariates in the association models, we only conditioned for the variants that were present on the same chromosome as the variant that was tested for association. The same approach was adopted for the gene-based analyses. All regional association results, with and without conditioning on AMD- and asthma-related variants, were uploaded on the LocusZoom website to allow further exploration (see Supplementary File S1 for the links).31 The loci definitions were adopted from Fritsche et al.,7 where each locus comprises an index variant ± 500kb region and is named after the nearest gene(s) to the variant.7 
Results
Characteristics of the Study Population
The final dataset contained 4827 samples, of which 2123 were advanced AMD cases (1251 dry AMD and 872 wet AMD) and 2,704 were controls (Table 1). The mean age of the cases was 78.8 years (standard deviation [SD] 8.1), and 60% of the cases were female participants. The mean age of the controls was 55.3 years (SD 10.4), and the percentage of females was 67%. 
Confirmation of Previously Identified AMD Variants
We identified 22 variants that were independently associated with AMD at the genome-wide significance level (P value < 5 × 10−8 and < 3 × 10−8, for common and low-frequency variants respectively) (Fig. 1). LocusZoom plots for each of the identified variants are presented in Supplementary File S2
Figure 1.
 
Manhattan plots of GWAS results (A) from the initial GWAS with no conditioning, followed by (B) Asthma and (C) AMD conditioning in the next two panels are shown (genes associated after AMD conditioning are listed in Supplementary Table S8). The peaks that reached genome-wide significance are annotated with the gene they belong to.
Figure 1.
 
Manhattan plots of GWAS results (A) from the initial GWAS with no conditioning, followed by (B) Asthma and (C) AMD conditioning in the next two panels are shown (genes associated after AMD conditioning are listed in Supplementary Table S8). The peaks that reached genome-wide significance are annotated with the gene they belong to.
First, we compared our results to a previously published GWAS for AMD.7 Of the 52 AMD variants that were previously associated with AMD,7 four variants were not present in our WGS dataset because of exclusion of the HLA region, four variants did not pass the GATK filter, and four variants were removed in the quality control steps (Supplementary Table S2). For two variants that were excluded from the final dataset (i.e., rs3750846 and rs61985136), variants in high LD (rs2284665, r2 = 0.96; and rs138361575, r2 = 0.97, respectively) were present in the WGS dataset. In our GWAS, genetic variants at the CFH, ARMS2/HTRA1, APOE, and C3 loci were genome-wide significantly associated with AMD (Table 2), and these loci also showed the strongest associations in the GWAS by Fritsche et al.7 The strongest association was detected for rs6677089 at the CFH locus with an odds ratio (OR) of 0.37 for allele A vs C (95% CI = 0.36–0.38, P value 1.08 × 10−102), which is in perfect LD (r2 = 1.0) with the top-associated variant at the CFH locus (rs10922109) reported by Fritsche et al.7 Similarly, the top-associated variant at the ARMS2/HTRA1 locus in our GWAS (rs2284665) was in perfect LD (r2 = 1.0) with the top-associated variant at this locus in the GWAS by Fritsche et al.7 (rs3750846). At the APOE and C3 loci, the same top-associated variants were detected in both GWASs (rs429358 and rs2230199, respectively). The remaining 38 variants previously reported by Fritsche et al.7 did not reach genome-wide significance in our GWAS. However, 33 of the variants showed similar odds ratios as in the study of Fritsche et al.,7 whereas five variants (rs62247658, rs114092250, rs67538026, rs73036519, rs191281603) showed a different direction of effect, although confidence intervals of these variants spanned both effect directions in our study (i.e., the odds ratios ranged between <1 to >1) (Supplementary Table S3). A summary of the AMD-associated variants, in comparison to the results of Fritsche et al.,7 is presented in Supplementary Table S3. AMD patients that were included in our study were not part of any previously published AMD-related GWAS studies. 
Table 2.
 
Known AMD Loci That Reached Genome-Wide Statistical Significance in the Current Study
Table 2.
 
Known AMD Loci That Reached Genome-Wide Statistical Significance in the Current Study
Because the control dataset mainly consisted of individuals who participated in clinical trials for asthma, we also compared our results to a previously reported GWAS for asthma.30 Of the 167 asthma-associated variants,30 seven were not sequenced, four did not pass the GATK filter, and seven were removed in the quality control steps. No variants were present in our GWAS dataset that were in high LD (r2 > 0.7) with the missing variants. Associations for the remaining 149 asthma-associated variants are presented in Supplementary Table S4. In our GWAS, none of the 149 variants reached the genome-wide significance level. The most significant associations were detected for variants at the TSLP (rs1837253, P value 4.66 × 10−7), IL13 (rs848, P value 9.51 × 10−7), and RAD51B (rs911263, P value 6.80 × 10−7) loci, which showed similar odds ratios in both GWASs. Of note, the rs911263 variant at the RAD51B locus is in substantial LD (r2 = 0.71) with the AMD-associated variant rs61985136 at the same locus. 
Identification of Novel AMD-Associated Variants
Common and Low-Frequency Variants
We identified 18 novel variants that reached genome-wide significance, 12 of which were common variants and six were low-frequency variants (Table 3, Supplementary File S2). All newly identified variants were located outside of the previously described AMD loci,7 and were also not located in or near the previously described asthma loci.30 The novel variants comprised 8 intronic variants, 7 intergenic variants, and 3 variants in long non-coding RNAs (lncRNA). The association signals of variants rs226251 in PARK7 and rs143255652 in AC103876.1 were supported by associations of variants that are in high LD with these top variants (Fig. 2). However, the remaining 16 variants were singletons, not supported by other variants, and might represent artifacts (Supplementary File S2). Of these 16 variants, six variants (rs669953, rs10740759, rs1733807, rs12145913, rs375015898, rs201405032) were low-frequency variants, and therefore they might show a signal independent of other variants because of their low frequency. The full list of variants that passed the genome-wide and suggestive significance threshold is presented in Supplementary Table S5
Table 3.
 
Novel, Most Significantly Associated Common and Low-Frequency Variants
Table 3.
 
Novel, Most Significantly Associated Common and Low-Frequency Variants
Figure 2.
 
LocusZoom plots of (A) intronic rs226251 variant on PARK7, and (B) intergenic rs143255652 near AC103876.1 are shown. Diamond-shaped variant shows the variants of interest (rs226251 and rs143255652, respectively), and dots show other variants in the region; red shows high LD, and blue shows no LD.
Figure 2.
 
LocusZoom plots of (A) intronic rs226251 variant on PARK7, and (B) intergenic rs143255652 near AC103876.1 are shown. Diamond-shaped variant shows the variants of interest (rs226251 and rs143255652, respectively), and dots show other variants in the region; red shows high LD, and blue shows no LD.
To determine whether all identified novel single nucleotide variants were independent of previously identified asthma- and AMD-associated variants, we repeated the GWAS conditioning on these variants (Figs. 1B, 1C). In our conditional analysis for asthma-associated variants, all 22 identified independent variants remained significantly associated with AMD, indicating that these associations are independent of the asthma loci. In addition, we discovered two other independent variants that were significantly associated with AMD after conditioning for asthma loci: rs143606602 near RPL21P91 and rs4988183 in MCM6 (P value 4.36 × 10−8 and 4.60 × 10−8, respectively). This brought the total number of known and novel genome-wide significant, independent associations to 24 (Supplementary Table S6). A full list of genome-wide and suggestive significant variants after conditioning for asthma-associated variants is provided in Supplementary Table S7
In a conditional analysis for AMD-associated variants, 15 of the newly identified variants remained significantly associated with AMD, including the associations identified at the PARK7 and AC103876.1 loci (top hit in conditional analysis shifted to rs374945368 instead of rs6677089), indicating that these represent novel associations independent of the previously identified AMD loci (Supplementary Tables S8, S9). Six of these 15 variants were also analyzed by Fritsche et al.,7 but although none of the six variants were significantly associated with AMD, five variants (including the rs143255652 variant in lncRNA AC103876.1) showed the same direction of effect as in our GWAS. Only rs226251 near PARK7 did not show the same direction of effect (Supplementary Table S10). 
Rare Variants
Three rare variants reached genome-wide significance (p-value < 1 × 10−8) in the single-variant association analysis, all of which were in non-coding regions (Supplementary Table S11). Two of these variants lost their significance after conditioning on asthma variants, while the intergenic variant rs1358927744 near the TENM3 gene remained suggestive significant (OR = 0.19, 95% CI = 0.09–0.39, P value 4.74 × 10−8). When conditioned for AMD variants, rs1358927744 remained suggestive significant, and additional associations were identified for an intronic variant in the KCNT2 gene (rs561508452, OR = 74.3, 95% CI = 3.87–1426.2, P value 3.34 × 10−8), a coding variant in the C3 gene (rs147859257; p.Lys155Gln, OR = 3.84, 95% CI 2.33–6.32, P value 1.79 × 10−8), and an intronic variant in the long non-coding RNA AL078604.4 (rs116216433, OR = 8.27 × 10−14, 95% CI 2.90 × 10−17–2.36 × 10−10, P value 2.22 × 10−12). For the variants rs561508452 and rs116216433, carriers were either only cases or controls; therefore, even though Firth's bias corrected regression was used, the effect estimates were inflated. 
Gene Level Association Analysis for Rare And Low-Frequency Variants
We identified 34 gene regions that contained a burden of rare and low-frequency variants in AMD cases compared to controls (Supplementary Table S12). Among these 34 gene regions, 20 localized in or near the CFH locus, four localized in or near the APOE locus, and one localized near the ARMS2/HTRA1 locus. After conditioning for known AMD variants to identify novel AMD associated variants, only five associated gene regions remained, none of which localized in or near established AMD loci. When conditioning on asthma variants, all 34 gene regions remained associated, including the five gene regions that were independent of the established AMD loci. 
We then performed a separate analysis for rare and low-frequency variants categorized as modifiers and for variants categorized as high- and moderate-impact variants. We identified 34 gene regions that contained a burden of modifier variants in cases compared to controls (Supplementary Table S13), which almost entirely overlapped with the 34 gene regions identified when including all variants in the analysis (Supplementary Table S12). After conditioning for AMD variants, eight gene regions remained associated, none of which localized in or near previously established AMD loci. These regions included four intergenic regions (RP13-146A14.1 on chromosome X, RNU6-354P on chromosome 15, AC073464.4 on chromosome 2 and RP11-549K20.1 on chromosome 5) and four genomic regions encompassing the RN7SL124P (or TNFAIP6 within same region), FAM69C, ZRANB3, and TMEM133 genes. 
In the analysis of high and moderate impact variants, a burden of rare and low-frequency variants was identified in the C3 and SLC16A8 genes (Supplementary Table S14). After conditioning for AMD variants, the burden of rare and low-frequency variants in the C3 and SLC16A8 genes remained associated (P = 5.70 × 10−9 and P = 3.99 × 10−6, respectively). In addition, a burden of rare and low-frequency variants was identified in the CFHR5 and CFI genes (P = 1.67 × 10−8 and P = 5.16 × 10−6, respectively). 
We zoomed in to the single variants present in the genes that showed associations: 58 variants were identified in the C3 gene, 37 in the CFHR5 gene, 39 in the SLC16A8 gene, and 61 in the CFI gene (Supplementary Table S15). Among these variants, the majority were singletons, (i.e., variants that were carried by only one individual). Within the single variant analysis, the rs147859257 (p.Lys155Gln) variant in the C3 gene had a p-value of 2.93 × 10−7, and the splice variant rs77968014 in the SLC16A8 gene had a p-value of 9.72 × 10−6. However, for CFI and CFHR5 no single variants showed significant associations. The majority (N = 183) of the variants were exonic missense, nonsense, and frameshift variants; six variants were intronic, three were splice site variants, and three variants were located in the untranslated region. 
Discussion
In this study, we performed a GWAS for AMD using WGS data of 2123 advanced AMD patients and 2704 controls. To our knowledge, this is the largest WGS study performed in AMD to date. We confirmed previously reported AMD variants, reassuring us of the analytical strategy, and identified new single variant associations near PARK7, AC103876.1 and TENM3 that were independent of the known AMD loci. In addition, gene-based association analyses identified a burden of rare and low-frequency modifier variants in several intergenic and gene-spanning regions, and of high- and moderate-impact coding variants in the C3, CFHR5, SLC16A8 and CFI genes. 
We detected significant associations of previously identified AMD variants at the CFH, ARMS2/HTRA1, APOE, and C3 loci. The C2/CFB/SKIV2L locus, which was previously also highly associated with AMD, was not included in our WGS dataset because of sequencing difficulties in this region. Our study did not have sufficient power to detect the remaining known AMD variants, but for the majority of the variants we found the same direction of effect as previously reported. In addition to the known AMD variants, we report 19 novel variants that reached genome-wide significance, all residing in intronic or intergenic regions. Previous studies in AMD focusing on low-frequency and rare variants were mainly restricted by the sequenced region (exome chip or sequencing of the exons only), and/or coverage and statistical power (<30 × coverage, and <5000 samples).10,32 Therefore these novel intergenic and intronic regions may represent true associations that may not have been captured in previous studies. Nevertheless, we would like to emphasize that these should be confirmed in follow-up studies to confirm that these are not false-positive findings. 
Two of the novel variants, the intronic rs226251 variant in PARK7 and the intergenic rs143255652 variant in AC103876.1, showed the association signals. The PARK7 gene has previously been associated with early-onset Parkinson's disease.33 It encodes the 189 amino acids-long protein deglycase DJ-1 protein that is expressed in multiple tissues, including the eye.34 DJ-1 has been described to have multiple functions but it is better characterized as an antioxidant and redox-sensitive molecular chaperone that protects cells from oxidative stress. In AMD, the retinal pigment epithelium (RPE) degenerates, which is in part due to oxidative stress.35 In vitro studies showed that DJ-1 levels were higher in oxidative stressed RPE cultures and in RPE from AMD donors. Furthermore, DJ-1 knockout mice showed higher levels of oxidative stress in the retina and some signs of retinal degeneration that appear to be more pronounced under oxidative stress.36,37 In our association analysis, the rs226251 variant in the PARK7 gene conferred a protective effect on AMD for the T allele (OR = 0.79). After successful replication of this signal, we suggest that further studies should be performed to determine the role of this variant in AMD. 
Results of the single-variant analysis for rare variants pointed towards the genes TENM3, KCNT2, C3, and long noncoding RNA AL078604.4. The KCNT2 gene is located adjacent to the CFH gene, which was previously shown to harbor rare variants in AMD,7 and the rs147859257 (p.Lys155Gln) variant in the C3 gene was previously associated with AMD,3840 corroborating the previous findings. TENM3 encodes the cell-adhesion molecular teneurin-3, which specifies morphological and functional connectivity of retina ganglion cells.41 Further studies are needed to confirm the role of TENM3 in AMD. 
In gene-based analysis of rare and low-frequency variants, we identified eight genomic intervals with a potential burden of variants with a potential modifying effect. None of these eight regions were previously associated with AMD. Four regions were intergenic, and four encompassed coding genes: TNFAIP6, FAM69C, ZRANB3, and TMEM133 genes. TNFAIP6 (TSG-6) is an anti-inflammatory protein that was shown to stabilize progression of retinal degeneration in Ccl2−/− / Cx3cr1−/−/Crb1rd8/rd8 mice that is dependent on the homozygous rd8 mutation42 and suppressed the development of choroidal neovascularization in a laser-induced rat model for AMD.43 In a gene-based analysis of high- and moderate-impact variants, a burden of rare and low-frequency variants was identified in the C3, SLC16A8, CFHR5, and CFI genes. The C3, SLC16A8, CFHR5 and CFI genes all lie in known AMD loci, and a burden of rare variants has previously been reported for these genes.7,44,45 Of the variants that were identified in these genes, many were not reported in previous AMD studies,46 and included intronic variants that map in cis-regulatory elements of the C3 gene, and variants located in the 3ʹ-UTR of the CFI gene. This underscores the strength of using WGS in association analysis for AMD, as new variants and variants located outside of the coding regions are also assessed. A recent study showed that more than half of rare coding variants in the CFI gene lead to reduced Factor I expression, supporting that the majority of CFI variants have a detrimental effect on the protein and lead to an increased risk for AMD.47 A higher frequency of CFHR5 variant alleles is observed in controls compared to AMD cases, which confirms a recent study that reported a protective effect of coding CFHR5 variants in AMD.45 
Two strengths of our study are the use of WGS data with 30 × average coverage and that it included only advanced AMD patients. These factors increase the precision and the power of our study, since genetic effects are stronger in advanced AMD patients compared to earlier stages of the disease.48 In addition, WGS has the advantage that variants are directly genotyped and WGS captures variants that are not covered by genotyping arrays and therefore can cover non-coding regions which may have an impact on AMD that may not have been found in prior studies.49 Even with imputation based on reference panels, capturing rare variants with genotyping arrays is challenging.50 Moreover, for the novel variants presented in Table 3, we compared the frequencies observed in our study with those in public databases (i.e., dbSNO and gnomAND [Supplementary Table S16]) and found them highly concordant except for a few variants. In any case, in our study, all samples were genotyped with the same technology, all genotypes were called with the same pipeline using all samples simultaneously, so we can be confident that allele frequency differences between cases and controls in our study are not due to technical or analytical batch effects. 
On the other hand, there is a number of limitations in the present study. An important limitation was the lack of healthy control subjects, and consequently potential confounding effect because of age or shared genetic background. Because of the nature of the WGS database available to us (patients from clinical trials), we had to select controls from patients with other diseases. To minimize potential confounding factors, we selected controls from diseases that were as similar as possible in demographics (i.e., age) and with as little shared genetic background, which resulted in the combination of asthma (N = 2518) and Alzheimer's disease (N = 186) patients used as our control cohort. To further control for the potential shared genetic effects, we adjusted our statistical analyses for known asthma loci, by performing conditional genetic analyses with asthma variants. Because of the low number of Alzheimer's cases, we did not perform conditional analyses on AD variants, so we cannot exclude the possibility that our novel findings are Alzheimer's loci rather than AMD loci. However, this is unlikely, because our novel findings are not listed in the results of one of the largest AD genetic studies to date.51 It might also be possible that the Alzheimer's controls have led to false-negative findings, because AMD progression was recently shown to be associated with cognitive impairment.52 We would like to emphasize that the shared genetic background may only cause false-negative findings (i.e., APOE not detected as a risk gene), but not false positives. 
Another limitation is that the control cohort was significantly younger than the AMD cases, even though Alzheimer's patients were included in the control cohort to partly account for this difference (Table 1). Because of the age difference between cases and controls, we checked our findings with regard to longevity. In a recent meta-analysis with over 30,000 samples, two SNPs were shown to be related to longevity: rs7412 and rs429358, both in the APOE gene.53 Interestingly, we identified rs429358 as one of the most significantly associated variants in our study, which is a replication of the results of Fritsche et al.7 The variant rs7412 reaches genome-wide significance (Supplementary Table S5), but this finding seems to be affected and loses its significance when conditioned for rs429358 (Supplementary Table S6). It is difficult for us to come to any conclusion about rs429358 regarding whether it is longevity related or not, due to lack of available data in our study for this purpose. 
An additional important limitation of our study is that our control cohort did not have eye examination and therefore was not assessed for AMD status. Moreover, it is possible that our control cohort, being relatively younger, may still develop AMD in the future. This potential misclassification bias may have decreased the power of our study if a part of the asthma patients used as controls would have AMD, but the reduction of power because of this bias is estimated to be low.54 Notably, this would have led again to false-negative findings and not false positives. Another limitation is the inconsistencies in allele frequencies for indels larger than 2 bp between our study and previously published GWASs, which made us exclude these indels to prevent spurious results. 
A final limitation is the relatively small sample size for WGS, even though it is the largest WGS in AMD to date. Nonetheless, based on power calculations assuming a prevalence of 9.8%,52 a MAF of 1% and a statistical significance level of 5 × 10−8, we had 13% power to detect a genotype relative risk of 2 (Supplementary Fig. S1) and we would need >8000 individuals to reach 80% power. With the current sample size, we were only able to detect variants with a genotype relative risk of 2.6 with 80% power. Thus larger WGS studies are needed to detect rare variants with smaller effect sizes, as well as to replicate the findings reported here. This will also allow stratified analyses for dry and wet AMD, which will increase the insights in the differences in genetic etiology between these two advanced AMD stages, because some previous studies indicated that most genetic effects are shared between the two AMD types.7 In conclusion, we confirmed previously identified associations for AMD, and identified several novel associations that are worth exploring in additional, independent cohorts. 
Acknowledgments
The authors thank the Genentech Human Genetics Initiative, including Tushar R. Bhangale, Matthew J. Brauer, Julie Hunkapiller, Jens Reeder, Kiran Mukhyala, Karen Cuenco, Jennifer Tom, Amy Cowgill, Jan Vogel, William F. Forrest, Natalie Bowers, and Suresh Selvaraj; Laurent Essioux, for relevant discussions of analysis and results, and the Roche-wide Enhanced Data and Insights Sharing (EDIS) network, particularly Neil Jones, Joshua Bernal, and Doug Kelkoff. 
Supported by the Roche Internships for Scientific Exchange (RiSE) program (to IEA, AIdH, and EN) and by the Dutch Research Council (016.Vici.170.024 to AIdH). The funding organizations had no role in the design or conduct of this research. 
Disclosure: I.E. Acar, None; T.E. Galesloot, None; U.F.O. Luhmann, F. Hoffmann-La Roche Ltd. (E, I); S. Fauser, F. Hoffmann-La Roche Ltd. (E, I); J. Gayán, F. Hoffmann-La Roche Ltd. (E, I); A.I. den Hollander, Dutch Research Council (F), Ionis Pharmaceuticals (C), Gyroscope Therapeutics (C), Gemini Therapeutics (C), F. Hoffmann - La Roche (C); E. Nogoceke, F. Hoffmann-La Roche Ltd. (E, I) 
References
Mitchell P, Liew G, Gopinath B, Wong TY. Age-related macular degeneration. Lancet. 2018; 392(10153): 1147–1159. [CrossRef] [PubMed]
Wong WL, Su X, Li X, et al. Global prevalence of age-related macular degeneration and disease burden projection for 2020 and 2040: a systematic review and meta-analysis. Lancet Glob Health. 2014; 2(2): e106–e116. [CrossRef] [PubMed]
Seddon JM, Cote J, Page WF, Aggen SH, Neale MC. The US twin study of age-related macular degeneration: relative roles of genetic and environmental influences. Arch Ophthalmol. 2005; 123: 321–327. [CrossRef] [PubMed]
Lauwen S, de Jong EK, Lefeber DJ, den Hollander A. Omics biomarkers in ophthalmology. Invest Ophthalmol Vis Sci. 2017; 58(6): BIO88–BIO98. [CrossRef] [PubMed]
Klein RJ, Zeiss C, Chew EY, et al. Complement factor H polymorphism in age-related macular degeneration. Science. 2005; 308(5720): 385–389. [CrossRef] [PubMed]
Dewan A, Liu M, Hartman S, et al. HTRA1 promoter polymorphism in wet age-related macular degeneration. Science. 2006; 314(5801): 989–992. [CrossRef] [PubMed]
Fritsche LG, Igl W, Bailey JN, et al. A large genome-wide association study of age-related macular degeneration highlights contributions of rare and common variants. Nat Genet. 2016; 48: 134–143. [CrossRef] [PubMed]
Huang LZ, Li YJ, Xie XF, et al. Whole-exome sequencing implicates UBE3D in age-related macular degeneration in East Asian populations. Nat Commun. 2015; 6: 6687. [CrossRef] [PubMed]
Huang L, Zhang H, Cheng CY, et al. A missense variant in FGD6 confers increased risk of polypoidal choroidal vasculopathy. Nat Genet. J2016; 48: 640–647. [CrossRef]
Corominas J, Colijn JM, Geerlings MJ, et al. Whole-exome sequencing in age-related macular degeneration identifies rare variants in COL8A1, a component of Bruch's membrane. Ophthalmology. 2018; 125: 1433–1443. [CrossRef] [PubMed]
Hoffman JD, Cooke Bailey JN, D'Aoust L, et al. Rare complement factor H variant associated with age-related macular degeneration in the Amish. Invest Ophthalmol Vis Sci. 2014; 55: 4455–4460. [CrossRef] [PubMed]
Yu Y, Triebwasser MP, Wong EK, et al. Whole-exome sequencing identifies rare, functional CFH variants in families with macular degeneration. Hum Mol Genet. 2014; 23: 5283–5293. [CrossRef] [PubMed]
Ratnapriya R, Zhan X, Fariss RN, et al. Rare and common variants in extracellular matrix gene Fibrillin 2 (FBN2) are associated with macular degeneration. Hum Mol Genet. 2014; 23: 5827–5837. [CrossRef] [PubMed]
Pras E, Kristal D, Shoshany N, et al. Rare genetic variants in Tunisian Jewish patients suffering from age-related macular degeneration. J Med Genet. 2015; 52: 484–492. [CrossRef] [PubMed]
Duvvari MR, van de Ven JP, Geerlings MJ, et al. Whole exome sequencing in patients with the cuticular drusen subtype of age-related macular degeneration. PLoS One. 2016; 11(3): e0152047. [CrossRef] [PubMed]
Sardell RJ, Bailey JN, Courtenay MD, et al. Whole exome sequencing of extreme age-related macular degeneration phenotypes. Mol Vis. 2016; 22: 1062–1076. [PubMed]
Wagner EK, Raychaudhuri S, Villalonga MB, et al. Mapping rare, deleterious mutations in Factor H: association with early onset, drusen burden, and lower antigenic levels in familial AMD. Sci Rep. 2016; 6: 31531. [CrossRef] [PubMed]
Geerlings MJ, Kremlitzka M, Bakker B, et al. The functional effect of rare variants in complement genes on C3b degradation in patients with age-related macular degeneration. JAMA Ophthalmol. 2017; 135: 39–46. [CrossRef] [PubMed]
Taylor RL, Poulter JA, Downes SM, et al. Loss-of-function mutations in the CFH gene affecting alternatively encoded Factor H-like 1 protein cause dominant early-onset macular drusen. Ophthalmology. 2019; 126: 1410–1421. [CrossRef] [PubMed]
Ratnapriya R, Acar IE, Geerlings MJ, et al. Family-based exome sequencing identifies rare coding variants in age-related macular degeneration. Hum Mol Genet. 2020; 29: 2022–2034. [CrossRef] [PubMed]
Pietraszkiewicz A, van Asten F, Kwong A, et al. Association of rare predicted loss-of-function variants in cellular pathways with sub-phenotypes in age-related macular degeneration. Ophthalmology. 2018; 125: 398–406. [CrossRef] [PubMed]
Grassmann F, Kiel C, Zimmermann ME, et al. Genetic pleiotropy between age-related macular degeneration and 16 complex diseases and traits. Genome Med. 2017; 9: 29. [CrossRef] [PubMed]
Tom J, Chang D, Wuster A, et al. Enabling genome-wide association testing with multiple diseases and no healthy controls. Gene. 2019; 684: 118–123. [CrossRef] [PubMed]
Poplin R, Ruano-Rubio V, DePristo MA, et al. Scaling accurate genetic variant discovery to tens of thousands of samples. BioRxiv. 2017:201178.
Mbatchou J, Barnard L, Backman J, et al. Computationally efficient whole genome regression for quantitative and binary traits. Nat Genet 2021; 53: 1097–1103.
Cingolani P, Platts A, Wang le L, et al. A program for annotating and predicting the effects of single nucleotide polymorphisms, SnpEff: SNPs in the genome of Drosophila melanogaster strain w1118; iso-2; iso-3. Fly (Austin). 2012; 6: 80–92. [CrossRef] [PubMed]
Fadista J, Manning AK, Florez JC, Groop L. The (in)famous GWAS P-value threshold revisited and updated for low-frequency variants. Eur J Hum Genet. 2016; 24: 1202–1205. [CrossRef] [PubMed]
Lee S, Wu MC, Lin X. Optimal tests for rare variant effects in sequencing association studies. Biostatistics. 2012; 13: 762–775. [CrossRef] [PubMed]
Kang HM . Efficient and Parallelizable Association Container Toolbox (EPACTS). Available at: https://genome.sph.umich.edu/wiki/EPACTS. Accessed September 25, 2020.
Han Y, Jia Q, Jahani PS, et al. Genome-wide analysis highlights contribution of immune system pathways to the genetic architecture of asthma. Nat Commun. 2020; 11: 1776. [CrossRef] [PubMed]
Pruim RJ, Welch RP, Sanna S, et al. LocusZoom: regional visualization of genome-wide association scan results. Bioinformatics. 2010; 26: 2336–2337. [CrossRef] [PubMed]
Cheng CY, Yamashiro K, Chen LJ, et al. New loci and coding variants confer risk for age-related macular degeneration in East Asians. Nat Commun. 2015; 6: 6063. [CrossRef] [PubMed]
Bonifati V, Rizzu P, van Baren MJ, et al. Mutations in the DJ-1 gene associated with autosomal recessive early-onset parkinsonism. Science. 2003; 299(5604): 256–259. [CrossRef] [PubMed]
Liu C, Liu X, Qi J, Pant OP, Lu CW, Hao J. DJ-1 in ocular diseases: a review. Int J Med Sci. 2018; 15: 430–435. [CrossRef] [PubMed]
Datta S, Cano M, Ebrahimi K, Wang L, Handa JT. The impact of oxidative stress and inflammation on RPE degeneration in non-neovascular AMD. Prog Retin Eye Res. 2017; 60: 201–218. [CrossRef] [PubMed]
Bonilha VL, Bell BA, Rayborn ME, et al. Loss of DJ-1 elicits retinal abnormalities, visual dysfunction, and increased oxidative stress in mice. Exp Eye Res. 2015; 139: 22–36. [CrossRef] [PubMed]
Shadrach KG, Rayborn ME, Hollyfield JG, Bonilha VL. DJ-1-dependent regulation of oxidative stress in the retinal pigment epithelium (RPE). PLoS One. 2013; 8(7): e67983. [CrossRef] [PubMed]
Seddon JM, Yu Y, Miller EC, et al. Rare variants in CFI, C3 and C9 are associated with high risk of advanced age-related macular degeneration. Nat Genet. 2013; 45: 1366–1370. [CrossRef] [PubMed]
Helgason H, Sulem P, Duvvari MR, et al. A rare nonsynonymous sequence variant in C3 is associated with high risk of age-related macular degeneration. Nat Genet. 2013; 45: 1371–1374. [CrossRef] [PubMed]
Zhan X, Larson DE, Wang C, et al. Identification of a rare coding variant in complement 3 associated with age-related macular degeneration. Nat Genet. 2013; 45: 1375–1379. [CrossRef] [PubMed]
Antinucci P, Nikolaou N, Meyer MP, Hindges R. Teneurin-3 specifies morphological and functional connectivity of retinal ganglion cells in the vertebrate visual system. Cell Rep. 2013; 5: 582–592. [CrossRef] [PubMed]
Tuo J, Cao X, Shen D, et al. Anti-inflammatory recombinant TSG-6 stabilizes the progression of focal retinal degeneration in a murine model. J Neuroinflammation. 2012; 9: 59. [CrossRef] [PubMed]
Kim SJ, Lee HJ, Yun JH, Ko JH, Choi DY, Oh JY. Intravitreal TSG-6 suppresses laser-induced choroidal neovascularization by inhibiting CCR2+ monocyte recruitment. Sci Rep. 2015; 5: 11872. [PubMed]
de Breuk A, Acar IE, Kersten E, et al. Development of a genotype assay for age-related macular degeneration: the EYE-RISK Consortium. Ophthalmology. 2021; 128: 1604–1617. [CrossRef] [PubMed]
Lores-Motta L, van Beek AE, Willems E, et al. Common haplotypes at the CFH locus and low-frequency variants in CFHR2 and CFHR5 associate with systemic FHR concentrations and age-related macular degeneration. Am J Hum Genet. 2021; 108: 1367–1384. [CrossRef] [PubMed]
Geerlings MJ, de Jong EK, den Hollander AI. The complement system in age-related macular degeneration: A review of rare genetic variants and implications for personalized treatment. Mol Immunol. 2017; 84: 65–76. [CrossRef] [PubMed]
de Jong S, Volokhina EB, de Breuk A, et al. Effect of rare coding variants in the CFI gene on Factor I expression levels. Hum Mol Genet. 2020; 29: 2313–2324. [CrossRef] [PubMed]
Colijn JM, Meester-Smoor M, Verzijden T, et al. Genetic Risk, Lifestyle, and Age-Related Macular Degeneration in Europe: The EYE-RISK Consortium. Ophthalmology. 2021; 128: 1039–1049. [CrossRef] [PubMed]
Zhang C, Owen LA, Lillvis JH, Zhang SX, Kim IK, DeAngelis MM. AMD genomics: non-coding RNAs as biomarkers and therapeutic targets. J Clin Med. 2022; 11: 1484. [CrossRef] [PubMed]
Si Y, Vanderwerff B, Zollner S. Why are rare variants hard to impute? Coalescent models reveal theoretical limits in existing algorithms. Genetics. 2021; 217(4): iyab011. [CrossRef] [PubMed]
Jansen IE, Savage JE, Watanabe K, et al. Genome-wide meta-analysis identifies new loci and functional pathways influencing Alzheimer's disease risk. Nat Genet. 2019; 51: 404–413. [CrossRef] [PubMed]
Le JT, Agron E, Keenan TDL, et al. Assessing bidirectional associations between cognitive impairment and late age-related macular degeneration in the Age-Related Eye Disease Study 2. Alzheimers Dement. 2022; 18: 1296–305. [CrossRef] [PubMed]
Deelen J, Evans DS, Arking DE, et al. A meta-analysis of genome-wide association studies identifies multiple longevity genes. Nat Commun. 2019; 10: 3669. [CrossRef] [PubMed]
Wellcome Trust Case Control C. Genome-wide association study of 14,000 cases of seven common diseases and 3,000 shared controls. Nature. 2007; 447(7145): 661–678. [CrossRef] [PubMed]
Figure 1.
 
Manhattan plots of GWAS results (A) from the initial GWAS with no conditioning, followed by (B) Asthma and (C) AMD conditioning in the next two panels are shown (genes associated after AMD conditioning are listed in Supplementary Table S8). The peaks that reached genome-wide significance are annotated with the gene they belong to.
Figure 1.
 
Manhattan plots of GWAS results (A) from the initial GWAS with no conditioning, followed by (B) Asthma and (C) AMD conditioning in the next two panels are shown (genes associated after AMD conditioning are listed in Supplementary Table S8). The peaks that reached genome-wide significance are annotated with the gene they belong to.
Figure 2.
 
LocusZoom plots of (A) intronic rs226251 variant on PARK7, and (B) intergenic rs143255652 near AC103876.1 are shown. Diamond-shaped variant shows the variants of interest (rs226251 and rs143255652, respectively), and dots show other variants in the region; red shows high LD, and blue shows no LD.
Figure 2.
 
LocusZoom plots of (A) intronic rs226251 variant on PARK7, and (B) intergenic rs143255652 near AC103876.1 are shown. Diamond-shaped variant shows the variants of interest (rs226251 and rs143255652, respectively), and dots show other variants in the region; red shows high LD, and blue shows no LD.
Table 1.
 
Demographics of the Data Set After Quality Control
Table 1.
 
Demographics of the Data Set After Quality Control
Table 2.
 
Known AMD Loci That Reached Genome-Wide Statistical Significance in the Current Study
Table 2.
 
Known AMD Loci That Reached Genome-Wide Statistical Significance in the Current Study
Table 3.
 
Novel, Most Significantly Associated Common and Low-Frequency Variants
Table 3.
 
Novel, Most Significantly Associated Common and Low-Frequency Variants
×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×