Autophagy Genes for Wet Age-Related Macular Degeneration in a Finnish Case-Control Study

Age-related macular degeneration is an eye disease that is the main cause of legal blindness in the elderly in developed countries. Despite this, its pathogenesis is not completely known, and many genetic, epigenetic, environmental and lifestyle factors may be involved. Vision loss in age-related macular degeneration (AMD) is usually consequence of the occurrence of its wet (neovascular) form that is targeted in the clinic by anti-VEGF (vascular endothelial growth factor) treatment. The wet form of AMD is associated with the accumulation of cellular waste in the retinal pigment epithelium, which is removed by autophagy and the proteosomal degradation system. In the present work, we searched for the association between genotypes and alleles of single nucleotide polymorphisms (SNPs) of autophagy-related genes and wet AMD occurrence in a cohort of Finnish patients undergoing anti-VEGF therapy and controls. Additionally, the correlation between treatment efficacy and genotypes was investigated. Overall, 225 wet AMD patients and 161 controls were enrolled in this study. Ten SNPs (rs2295080, rs11121704, rs1057079, rs1064261, rs573775, rs11246867, rs3088051, rs10902469, rs73105013, rs10277) in the mTOR (Mechanistic Target of Rapamycin), ATG5 (Autophagy Related 5), ULK1 (Unc-51-Like Autophagy Activating Kinase 1), MAP1LC3A (Microtubule Associated Protein 1 Light Chain 3 α), SQSTM1 (Sequestosome 1) were analyzed with RT-PCR-based genotyping. The genotype/alleles rs2295080-G, rs11121704-C, rs1057079-C and rs73105013-T associated with an increased, whereas rs2295080-TT, rs2295080-T, rs11121704-TT, rs1057079-TT, rs1057079-T, rs573775-AA and rs73105013-C with a decreased occurrence of wet AMD. In addition, the rs2295080-GG, rs2295080-GT, rs1057079-TT, rs11246867-AG, rs3088051-CC and rs10277-CC genotypes were a positively correlated cumulative number of anti-VEGF injections in 2 years. Therefore, variability in autophagy genes may have an impact on the risk of wet AMD occurrence and the efficacy of anti-VEGF treatment.


Introduction
Age-related macular degeneration (AMD) is the most common cause of permanent sharp and color vision loss in the elderly which has a significant socioeconomic burden on patients and their Figure 1. Color fundus photographs, fluorecein angiography (FAG) and optical coherent tomography (OCT) images from healthy and wet AMD cases. Images retrieved from participants enrolled in this study, wet AMD patient had a TT allele of rs73105013. The clinical hallmarks of wet AMD are hemorrhages (blue arrow), fluorescein leaking in a late FAG (black arrow) and intra/subretinal fluids in OCT (white arrows). RPE means retinal pigment epithelium layer.
Clinical data included blood pressure, anti-cholesterol, anti-coagulant and anti-platelet drugs. Body mass index (BMI) and smoking history were also examined ( Table 1).
Ethics Committee of the Kuopio University Hospital (42/2014) has approved the study and the tenets of the Declaration of Helsinki were followed. All participants signed an informed consent form. Some 10 mL aliquots of whole blood were taken into a 3 mL tube containing EDTA, shaken gently 10 times, and kept at room temperature for 30-60 min prior to 3200× g centrifugation at 20 °C for 15 min. DNA was extracted from the peripheral blood leucocytes with QiAamp DNA Blood Midi kit (Qiagen, Valencia, CA, USA) according to the manufacturer's instructions. The quality and the amount of the extracted genomic DNA were measured with NanoDrop Microvolume spectrophotometer model ND-1000 (ThermoFisher Scientific, Wilmington, MA, USA).
The public domain of the single-nucleotide polymorphism database at the National Center for Biotechnology Information and the related literature to identify potentially functional polymorphisms in autophagy genes were used for SNP selections. The selected polymorphisms have known distributions in various European populations. SNP selection favored those with a minor allele frequency no less than 3%. Moreover, we selected polymorphisms at different positions in the Color fundus photographs, fluorecein angiography (FAG) and optical coherent tomography (OCT) images from healthy and wet AMD cases. Images retrieved from participants enrolled in this study, wet AMD patient had a TT allele of rs73105013. The clinical hallmarks of wet AMD are hemorrhages (blue arrow), fluorescein leaking in a late FAG (black arrow) and intra/subretinal fluids in OCT (white arrows). RPE means retinal pigment epithelium layer.
Clinical data included blood pressure, anti-cholesterol, anti-coagulant and anti-platelet drugs. Body mass index (BMI) and smoking history were also examined ( Table 1).
Ethics Committee of the Kuopio University Hospital (42/2014) has approved the study and the tenets of the Declaration of Helsinki were followed. All participants signed an informed consent form. Some 10 mL aliquots of whole blood were taken into a 3 mL tube containing EDTA, shaken gently 10 times, and kept at room temperature for 30-60 min prior to 3200× g centrifugation at 20 • C for 15 min. DNA was extracted from the peripheral blood leucocytes with QiAamp DNA Blood Midi kit (Qiagen, Valencia, CA, USA) according to the manufacturer's instructions. The quality and the amount of the extracted genomic DNA were measured with NanoDrop Microvolume spectrophotometer model ND-1000 (ThermoFisher Scientific, Wilmington, MA, USA).
The public domain of the single-nucleotide polymorphism database at the National Center for Biotechnology Information and the related literature to identify potentially functional polymorphisms in autophagy genes were used for SNP selections. The selected polymorphisms have known distributions in various European populations. SNP selection favored those with a minor allele frequency no less than 3%. Moreover, we selected polymorphisms at different positions in the relevant genes. There were polymorphisms located directly in the core promoter and thousands base pair upstream from the transition start site. We selected both exonic and intronic locations as well as the 3 untranslated region. This diversity in the locations of SNPs within genes ensures their potential various functionality. Genotyping was performed using TaqMan SNP Genotyping Assays (Thermo Fisher Scientific, Foster City, CA, USA) ( Table 2) according to the manufacturer's instructions. Briefly, genotyping was carried out in a total volume of 15 µL containing 1× TaqMan Genotyping Master Mix (Thermo Fisher Scientific), 1× SNP Genotyping Assay mix and 2 ng of genomic DNA in a 96-well plate format. PCR reactions were carried out with 2720 Thermal Cycler (Thermo Fisher Scientific) and the fluorescein labels were read using QuantStudio 5 (Thermo Fisher Scientific).

Statistical Analysis
The descriptive data are shown as the mean ± SD or as the number of patients in each category. The normality of the studied group was verified with the Shapiro-Wilk test, the homogeneity of variance was checked with Brown-Forsythe test. Accordingly, either the unpaired Student's t test or Mann-Whitney U test was used. In some bivariate and multivariate analyses, we used the approach of resampling with replacement (the bootstrap-boosted versions of the tests, 10,000 iterations) to make sure that the revealed differences were not detected by a pure chance.
An unconditional multiple logistic regression model was used to calculate the associations between the studied polymorphisms and the occurrence of a disease (with or without stratification for sex). The results are shown as odds ratios (ORs) with a 95% confidence interval (±95% CI). In addition, the significant outcomes were further validated with the use of two approaches: the bootstrap-boosted multiple logistic regression (resampling with replacement, 10,000 iterations) and the cross-validated logistic regression that corresponds to the d-jackknife technique with the patient group as the modelled class. This was intended to overcome any possible bias related to relatively low sample sizes. The goodness-of-fit of logistic regression models showing a significant discrimination between controls and patients was estimated with the Hosmer-Lemeshow test. The analysis of collected data was done in Statistica 12 (Statsoft, Tulsa, OK, USA), SigmaPlot 11.0 (Systat Software Inc., San Jose, CA, USA), Resampling Stats Add-in for Excel v.4 (Arlington, VA, USA) and StudSize3.02 (CreoStat HB, Västra Frölunda, Sweden; used for power analysis).

Results
Autophagy-related genes (mTOR, ATG5, ULK1, MAP1LC3A, and SQSTM1) SNPs cover the genes and 10 kbp flanking regions both up-and downstream ( Table 2). Based on their linkage disequilibrium (LD) structure, the seven SNPs cover 90% of the allelic variation within these regions, while three SNPs showed disequilibrium ( Table 3). The total genotyping rate in the individuals was >90%. Differences in call rates for each genotype between cases and controls were tested and no significant (p < 0.05) differences were detected in genes that showed agreement with Hardy-Weinberg equilibrium (HWE) ( Table 3). However, rs2295080, rs11121704 and rs573775 deviations were detected that significantly (p < 0.05) departed from the HWE disequilibrium. Table 3. Distributions of the genotypes of polymorphisms of autophagy-related genes in wet AMD patients and controls tested for agreement with Hardy-Weinberg equilibrium.

Polymorphism
Control AMD As shown in Table 4, the following genotypes/alleles were associated with an increased risk of wet AMD: rs2295080-G (adjusted), rs11121704-C (adjusted), rs1057079-C (crude and adjusted) and rs73105013-T (crude and adjusted). The protective associations were observed for rs2295080-TT (crude and adjusted), rs2295080-T, rs11121704-TT (adjusted), rs1057079-TT (crude and adjusted), rs1057079-T (crude and adjusted), rs573775-AA (crude) and rs73105013-C (crude and adjusted). In general, these associations showed ORs of about 1.5 in cases of positive associations and not less than 0.4 in their negative counterparts. A heatmap was generated for the visualization of different allele frequencies between healthy controls and wet AMD patients ( Figure 2).    Since the amount of anti-VEGF intravitreal injections varies between patients and with personalized demands, we were curious to analyze the connections of autophagy gene SNPs and the amount of cumulative injections during two years of treatment. Our results showed that the rs2295080-GG, rs2295080-GT, rs1057079-TT, rs11246867-AG, rs3088051-CC and rs10277-CC genotypes were associated with two years of a cumulative number of injections (0.13 < R < 0.14 Spearman's correlation coefficient) ( Table 5). Table 5. Multiple correlation between the cumulative number of injections (median of six injections) for two years in wet AMD patients and genotypes of polymorphisms in autophagy-related genes.

Discussion
AMD is a multifactorial, multi-compartmental complex eye disease. The importance of genetic factors and their variability were confirmed in many studies (reviewed in [6]). However, the penetrance of many gene variants is very low, and one of the challenges of complex disease is to capture the extent of processes that may be perturbed. Genome-wide association studies (GWAS) have shown that mutations in the complement system genes are strongly associated with the risk of AMD. However, those SNPs do not differentiate AMD phenotypes, progression, or treatment responses. Phenotypic, demographic, environmental, genetic, and molecular risk factors have been used to predict disease progression, but personalized risk may differ from that determined by commonly accepted risk factors [29]. Therefore, rare variants should be considered in the prediction of AMD progression and treatment responses. Moreover, it is assumed that the risk genes are expressed in the retinal tissue and are locally involved in AMD mechanisms [30].
Since AMD is strongly associated with disturbed proteostasis [16,23], we targeted our gene analysis to autophagy genes that are not strong hits in GWAS analyses [6]. To our knowledge, we showed for the first time that autophagy gene polymorphisms were moderately associated with wet AMD development and anti-VEGF treatment responses. The T alleles of the mTOR gene loci rs2295080, rs11121704 and rs1057079 indicated OR 0.6 protection against wet AMD. The ATG5 rs573775 A allele and the LC3 rs731105013 C allele SNPs showed a similar trend in the prevention of wet AMD. These SNP associations revealed that autophagy cascaded from the autophagy initiation to the flux termination and may have a regulatory role in the development of wet AMD. The rs73105013 T allele might be a useful genotyping marker in deeper wet AMD risk analysis. Interestingly, the mTOR rs2295080 G allele associated with both an increased wet AMD and weaker anti-VEGF treatment response once the cumulative amount of injections for two years was calculated. According to the Finnish wet AMD guidelines, a pro re nata protocol was used for anti-VEGF injections [28]. To show the complexity of the genetic association studies, the mTOR rs1057079 T allele had a protective sign to develop wet AMD, but was associated with a weaker anti-VEGF treatment response, as it was related to the number of given injections. Similar observations were made for the protective genotype of rs800292 variant of the complement H (CFH) gene with a poor anatomical response [31]. The p.69A > S polymorphism for the homozygous T risk allele required more anti-VEGF injections over the 48-month follow-up [14]. A cumulative effect of the high-risk alleles in CFH, ARMS2, and VEGFA was associated with a younger age of AMD onset in combination with a poor response rate to anti-VEGF treatment [32]. A limited contribution of common genetic variants to variability in wet AMD treatment response was observed, while rare protein-altering variants in the C10orf88 and UNC93B1 genes associated with a worse response to anti-VEGF therapy [15]. Rare variants, including SNPs in autophagy genes, may play an important role in personalized anti-VEGF treatment responses.
Numerous AMD-related risk factors have been documented, but the evidence and strength of association is variable depending on ethnicity and study design. Our Finnish cohort shows that smoking history and the use of anti-cholesterol drugs are associated with wet AMD. One systemic review and meta-analysis showed moderate and consistent risk associations with higher BMI, history of cardiovascular disease, and hypertension [33]. A recent French study indicated a high risk for the incidence of early AMD in individuals with high plasma high-density lipoprotein (HDL) cholesterol levels and confirmed the high risk for progression from early to advanced AMD in heavy smokers and carriers of the CFH Y402H at-risk genotypes [34]. In numerous studies, smoking is reported as a modifiable risk factor in AMD. It is associated with a 2-4-fold increased risk for any form of AMD [29]. Our results are in line with those which previously reported the role of smoking in AMD development. The use of anti-cholesterol drugs was also clearly associated with wet AMD in this study. The age-related eye disease study (AREDS) report reveals that statin use is not significantly associated with the progression to late AMD in the AREDS2 participants [35]. In our material, anti-cholesterol drugs were not classified and plasma HDL levels were not analyzed. Interestingly, autophagy has a key role in the clearance of cholesterol droplets and high cholesterol levels may decrease autophagy [36,37]. In addition, autophagy, glucose, lipid, and energy metabolism regulator adiponectin receptor 1 rs10753929 ADIPOR1 variant was associated with advanced AMD in the Finnish population [27].
An increased BMI is found to be associated with the increased chance of developing AMD in some studies, while some other studies show no association [29]. We did not observe association between increased BMI and wet AMD. Although the possible involvement of hypertension in AMD has been well documented in the literature, it is not always found to be a risk factor for AMD and anti-hypertensive medication is not proven to have a positive effect on AMD [29]. We analyzed only blood pressure and drug use without finding any association with wet AMD prevalence. A limitation of our study is that we did not analyze BMI development, hypertension disease history or its severity level and we did not classify anti-hypertensive drugs. In recent years, anti-coagulation and anti-platelet drugs have been considered to exacerbate wet AMD [38]. We found no association between the use of low-dose anti-coagulating or anti-platelet drug use and wet AMD, which is in line with the recent reports [39][40][41].
The number of injections, regimens and patient compliance changes between patients. We used Finnish national guidelines for modified PRN which allows a different anti-VEGF drugs and injection regimen. By identifying the prognostic biomarkers of AMD that could be used to select the individual drug treatment and follow-up protocol, the quality and safety of treatment and patient compliance could be improved. This would also more precisely allocate available resources, especially as the need for AMD patients care increases as the population ages. Variability in autophagy-related genes in the Finnish population may moderately contribute to wet AMD pathogenesis and therapy. The Finnish population has reduced diversity and homogeneity, which is beneficiary for gene mapping studies. Since we used RT-PCR-based SNP genotyping, further work should include the whole-genome sequencing of the samples with subsequent analysis and public deposition of the acquired datasets. We believed that autophagy is one important target to develop pharmacogenetics and personalized medicine for AMD.

Conclusions
Autophagy genes are associated with wet AMD and the efficacy of anti-VEGF treatment in a Finnish Case-Control Study.