Impact of the Common Genetic Associations of Age-Related Macular Degeneration upon Systemic Complement Component C3d Levels

Age-related macular degeneration (AMD) is a common condition that leads to severe vision loss and dysregulation of the complement system is thought to be associated with the disease. To investigate associations of polymorphisms in AMD susceptibility genes with systemic complement activation, 2655 individuals were genotyped for 32 single nucleotide polymorphisms (SNPs) in or near 23 AMD associated risk genes. Component 3 (C3) and its catabolic fragment C3d were measured in serum and AMD staging was performed using multimodal imaging. The C3d/C3 ratio was calculated and associations with environmental factors, SNPs and various haplotypes of complement factor H (CFH) genes and complement factor B (CFB) genes were analyzed. Linear models were built to measure the influence of genetic variants on the C3d/C3 ratio. The study cohort included 1387 patients with AMD and 1268 controls. Higher C3d/C3 ratios were found for current smoker (p = 0.002), higher age (p = 1.56×10−7), AMD phenotype (p = 1.15×10−11) and the two SNPs in the C3 gene rs6795735 (p = 0.04) and rs2230199 (p = 0.04). Lower C3d/C3 ratios were found for diabetes (p = 2.87×10−6), higher body mass index (p = 1.00×10−13), the SNPs rs1410996 (p = 0.0001), rs800292 (p = 0.003), rs12144939 (p = 4.60×10−6) in CFH, rs4151667 (p = 1.01×10−5) in CFB and individual haplotypes in CFH and CFB. The linear model revealed a corrected R-square of 0.063 including age, smoking status, gender, and genetic polymorphisms explaining 6.3% of the C3d/C3 ratio. After adding the AMD status the corrected R-square was 0.067. In conclusion, none of the evaluated genetic polymorphisms showed an association with increased systemic complement activation apart from two SNPs in the C3 gene. Major genetic and non-genetic factors for AMD were not associated with systemic complement activation.


Introduction
Age-related macular degeneration (AMD) is a neurodegenerative disease causing visual impairment and blindness in the elderly population. Accumulation of drusen between Bruchs membrane and the retinal pigment epithelium characterizes the early forms while the two advanced forms show geographic atrophy and choroidal neovascularization. Risk is multifactorial including environmental and genetic factors. Genetic variation accounts for up to 71% of the disease risk [1]. Many genetic polymorphisms were found in the alternative pathway of the complement system including complement factor H (CFH), complement component 3 (C3), Complement factor I (CFI), and complement factor B (C2/CFB locus) [2][3][4][5][6][7][8]. Complement proteins and their activation products have been identified in retinal deposits of AMD patients [3,[9][10][11].
The alternative complement pathway is constantly activated by the spontaneous hydrolysis of a thioester bond in C3 and a tight regulation including CFH is necessary to prevent excessive activation. It is hypothesized that a dysregulation of the complement system leads to tissue damage and finally AMD.
The dysregulation of the complement system or their activation fragments were also found systemically. In AMD patients, various components of the complement system were found at increased levels such as CFB, CFD, C3a, C5a, C3d, and Ba [12,13].
While the association of genetic polymorphisms with AMD is well established, only polymorphisms in the C3 gene and few haplotypes in the CFH and CFB/C2 gene were found to be associated with complement activation products including factor C3d in two small cohorts [12,13]. The impact of other AMD susceptibility genes on the regulation of systemic complement activation remains unclear. In our study, we analyzed the association of 32 single nucleotide polymorphisms (SNPs) in or near 23 AMD risk genes with the C3d/C3 ratio as a marker for chronic complement activation in a Caucasian cohort of 2655 participants.

Study population
2655 participants from the European Genetic Database (EUGENDA, www.eugenda.org) were included in the study. The study was performed in accordance with the tenets of the Declaration of Helsinki and the Medical Research Involving Human Subjects Act (WMO) and was approved by the local ethics committee of the University Hospitals in Cologne and Nijmegen. Written informed consent was obtained from all participants. AMD staging was performed by grading of retinal images including stereo fundus photographs (FPs), fluorescein angiograms (FAs) and spectral domain optical coherence tomograms (SDOCTs) according to the standard protocol of the Cologne Image Reading Center (CIRCL) by certified graders (TR, LE). AMD was classified by the presence of pigmentary changes together with at least 10 small drusen (,63 mm) or the presence of intermediate (63-124 mm) or large drusen ($125 mm diameter) in the Early Treatment Diabetic Maculopathy Study (ETDRS) grid or geographic atrophy and/or choroidal neovascularisation (CNV) secondary to AMD in at least one eye.
Demographic data and non-genetic parameters including history of smoking (current/past/never), regular alcohol intake (yes/no), body mass index (BMI), arterial hypertension (yes/no), diabetes (yes/no), rheumatoid arthritis (yes/no), thyroid disease (yes/no), kidney disease (yes/no) and history of allergy (yes/no) were obtained by standardized interviewer-assisted questionnaires.

Complement component measurements and genetic analysis
Serum samples were used for C3d and C3 measurements. Serum was prepared by coagulation at room temperature. After centrifugation, the samples were stored at 280uC within 1 hour after collection. Complement component C3 and the activation fragment C3d were measured in serum samples as described previously [14]. The C3d/C3 ratio was calculated as a measure of C3 activation.

Haplotype analysis
In order to analyze the influence of haplotypes on C3d/C3 ratios, the posterior probability of each haplotype in the CFH gene including rs1061170, rs800292 and rs12144939 and in the CFB gene including rs4151667 and rs641153 was calculated using PHASE software, version 2.1 [16,17].

Statistical analysis
All calculations were performed using SPSS software version 21.0 (IBM Software and Systems, Armonk, NY, USA). C3d/C3 ratios are given as median and interquartile range (1 st quartile -3 rd quartile). Due to the skewed nature of the data, the logarithm (log 10 ) of the C3d/C3 ratios was used for analysis. Associations between logarithmic C3d/C3 ratios and genetic polymorphisms, haplotypes, phenotype and environmental factors were analyzed using t-tests or univariate analysis of variance (ANOVA) depending on number of variables. Associations between AMD phenotype and genetic polymorphisms were evaluated using logistic regression analysis. Linear models were performed to illustrate the influence of the genetic factors on complement activation. P-Values ,0.05 were considered statistically significant.

Demographics and non-genetic factors
Mean age of the study population was 73.268.0 years (75.868.1 years for AMD patients and 70.466.8 years for controls, p,0.001). Demographic data, phenotype and environmental factors are summarized in Table 1. C3d/C3 ratios showed significant differences for age with increasing levels (except the youngest group from 50-59 years) and phenotype with higher values for AMD patients. A significant association was also found for diabetes, smoking, and BMI.

Associations of C3d/C3 levels with genetic polymorphisms
Significant associations of C3d/C3 levels were found in the CFH gene for the SNPs rs1410996, rs800292 and rs12144939, in the CFB gene for rs4151667 and in the C3 gene for rs6795735 and rs2230199. In all SNPs of the CFH and CFB gene, these variants showed lower C3d/C3 ratios than the reference alleles, only variants in C3 revealed higher values. After stratification in AMD cases and controls, associations for the major risk variants in rs1061170 (CFHY402H, p = 0.35 for no AMD; p = 0.55 for AMD) and rs10490924 (ARMS2, p = 0.75 for No AMD, p = 0.25 for AMD) with the C3d/C3 ratio could not be observed. A detailed analysis is outlined in Table 2.
For the SNPs rs1061170, rs800292 and rs12144939 in the CFH gene and the SNPs rs4151667 and rs641153 in the CFB gene, haplotypes were associated with C3d/C3 levels ( Table 3). All haplotypes were associated with lower C3d/C3 levels than the reference haplotype (Figure 1 and 2).

Associations of genetic polymorphisms with AMD
Performing logistic regression analysis, protective effects were found for variants in CFH rs1410996, CFH rs800292, CFH rs12144939, CFB rs641153 and FADS1 rs174547. Variants in CFB rs4151667, TIMP3 rs9621532 and APOE4 rs429358 showed a trend for a protective effect on AMD without reaching statistical significance, which may be due to low minor allele frequencies or smaller effects of those SNPs.
Variants in ARMS2 rs10490924, CFH rs1061170, C3 rs2230199, C3 rs6795735 and CETP rs2230199 were found to be associated with significantly higher risk for AMD (Table 4). Variants in VEGFA rs699946, SLC16A8 rs8135665 and ADAMTS9-AS2 rs6795735 also showed a trend for a higher AMD risk without statistical significance, which also may be due to a smaller effect on AMD development for each of those SNPs compared to ARMS2 rs10490924 or CFH rs1061170.

Linear Models
Linear models were composed based on univariate analysis of covariance (ANCOVA) with the logarithmic C3d/C3 ratio as dependent variable to evaluate the influence of various factors on the C3d/C3 ratio. In the first model, we included all SNPs (rs1410996, rs800292, rs12144939 in CFH, rs4151667 in CFB, rs6795735 and rs2230199 in C3) that had reached statistical signifcance in the individual analysis and the two major AMD risk SNPs risk ARMS2 rs10490924 and CFH rs1061170. Additionally, age, gender, and smoking status was included. The corrected Rsquare was 0.063. Adding the AMD status to the model, the corrected R-square was 0.067. In the second model, CFH haplotypes, age, gender, and smoking status were included. The corrected R-square was 0.038.

Discussion
Dysregulation of the alternative complement pathway is thought to play a key role in AMD pathogenesis, which is also reflected by increased systemic complement levels.
In this study we analyzed the association of genetic AMD risk polymorphisms with systemic complement activation. We identified only a few variants in the CFH, CFB, and C3 gene that showed an association with systemic complement activation, while for all other genetic polymorphisms associations were not observed.
While the association with genetic polymorphisms was weak, we found a significant association of the phenotype AMD with an increased C3d/C3 ratio which is in line with other smaller studies. [12,13] Our linear model including the AMD phenotype, the two major non-genetic risk factors age and smoking, and eight relevant SNPs could only explain 6.7% of the variation in the C3d/C3 ratio, indicating that these AMD risk polymorphisms do not explain sufficiently increased systemic complement activation found in AMD patients. The inclusion of CFH haplotypes in the model revealed an even lower explanation of the C3d/C3 ratio of only 3.8%. Our analysis concentrated on C3d as a marker of chronic complement activation because it is a relatively stable protein with a long half-life. To correct for differences in the concentration of the precursor protein, we also measured C3 in the plasma.
CFH, CFB, and C3 influence the regulation of the alternative pathway of the complement system. CFH acts as the major regulator of complement activation controlling the alternative pathway in blood and on cell surfaces [18], and accelerates the decay of the alternative C3 convertase (C3bBb) [19].
CFH is also a cofactor of CFI-mediated cleavage and inactivation of C3b [20]. The formation of C3d, a polypeptide fragment generated during alternative C3 convertase cleaves C3 to C3b, is also CFH dependent [21]. Alterations in the CFH gene may change the regulating characteristics of CFH resulting in an up or down regulation of the CFH dependent elements of the alternative complement pathway. In our cohort, SNPs rs1410996, rs800292, and rs12144939 in the CFH gene were associated with lower C3d/C3 ratios and a lower risk for AMD, whereas the most common AMD risk variant rs1061170 was not associated with the C3d/C3 ratio even after stratification in AMD patients and controls. Additionally, CFH haplotypes showed lower C3d/C3 ratios in all cases compared to the reference haplotype. Therefore, CFH SNPs were not associated with increased systemic complement activation.
CFB is an acute phase protein involved in the alternative complement pathway as a precursor of C3 convertase. CFB is cleaved to Bb which combines with C3b to form the alternative pathway C3 convertase C3bBb. An acute phase responsemediated up-regulation may result in elevated systemic plasma levels of CFB in AMD patients and may contribute to an enhanced systemic complement activity [12,13,22]. In our study we observed lower C3d/C3 ratios for the operatively protective CFB variants for AMD indicating that individuals with these polymorphisms show less complement activation.
Among all analyzed SNPs, only variants in the C3 gene were associated with higher systemic C3d/C3 ratios which aligns with the results by Hecker et al. 11 The alternative pathway of the complement system starts with spontaneous hydrolysis of C3 and variants with a higher risk for AMD seem to influence this part of the complement cascade resulting in elevated systemic complement activation. Scholl et al did not observe a correlation between genetic variants in C3 and systemic C3d levels 12 underlining that the systemic effects of AMD susceptibility genes on complement activation are only weak.
In our study, slightly higher C3d/C3 levels were found in AMD patients. In order to not miss a combined effect of multiple SNPs, we performed linear models to illustrate the effect of the combination of SNPs on complement activation. These models could not explain the C3d/C3 ratio, showing that there have to be other systemic effects than AMD phenotype or genetic variants influencing systemic complement activation. Hecker et al also showed in a small cohort that risk haplotypes in CFH did not alter complement levels, whereas protective haplotypes reduced complement levels including C3d [12].
A limitation of our study is the analysis of only two components of the complement system, which is accompanied by several strengths including a large cohort of well-balanced AMD patients and controls, a high number of investigated SNPs and the use of multimodal imaging that avoids misclassification of phenotypes.
In summary, we showed that the major AMD risk polymorphisms in CFH and ARMS2 are not associated with increased systemic complement activation as measured by the C3d/C3 ratio. Few SNPs were associated with lower levels of systemic complement activation, particularly the CFH and CFB polymorphisms that are protective against AMD. Only variants in C3 were associated with elevated complement levels. Furthermore, a model including major genetic and non-genetic factors for AMD was not able to explain complement activation.