2011; 17:983-988 <http://www.molvis.org/molvis/v17/a109>
Received 2 December 2010 | Accepted 16 April 2011 | Published 20 April 2011
Anne E. Hughes, Gemma M. Mullan, Declan T. Bradley
Centre for Public Health, Queen’s University Belfast, UK
Correspondence to: Anne E. Hughes, Centre for Public Health, Institute of Clinical Sciences, Block B, Queen’s University Belfast, Royal Victoria Hospital, Grosvenor Road, Belfast, BT12 6BA, UK; Phone: +442890632718; FAX: +442890235900; email: firstname.lastname@example.org
Purpose: The 32Q (rs641153; A) and 32W (rs12614; T) variants of complement factor B (CFB) cause less efficient complement activation in vitro than the common 32R variant. This is thought to be the reason that the 32Q variant is associated with decreased risk of age-related macular degeneration (AMD). We investigated whether the 32W variant was also associated with decreased risk of AMD.
Methods: We genotyped 367 cases with neovascular AMD and 251 disease-free controls. Association with the disease phenotype was assessed by logistic regression for polymorphisms of CFB alone and in combination with smoking status and genetic risk markers of complement factor H (CFH) and HtrA serine peptidase 1 (HTRA1). We performed meta-analysis of all previously published reports of 32W allele frequency in AMD cases and controls.
Results: The CFB variant 32W was associated with protection against neovascular AMD, compared to the common 32R variant (odds ratio 0.64, p<0.05, in logistic regression with CFB variants; odds ratio 0.53, p<0.05, in logistic regression with CFB variants, CFH haplotypes, HTRA1 rs10490924 genotype, and smoking status). Meta-analysis (n=1,795) including this study and two others of neovascular AMD showed a combined odds ratio of 0.75 (p<0.05) for 32W, compared to 32R. Meta-analysis (n=2,600) of all reported studies of all types of AMD showed a combined odds ratio of 0.79 (p<0.01).
Conclusions: Our study shows that the 32W variant of CFB is associated with protection against AMD, in keeping with evidence of its functional effect on the complement system. The protective effect is less strong than that associated with 32Q.
Age-related macular degeneration (AMD) is a major cause of visual impairment and blindness among older people . The alternative pathway of the complement system, an ancient defense against infectious microbes [2,3], is implicated in the etiology and pathogenesis of the disease. Local and systemic complement activation occurs in individuals with AMD [4,5]. The proteins of the complement system and products of their activation are found to be raised in the circulation and in local deposits (drusen) in the retina . Genome-wide association studies first revealed the association between polymorphisms in the complement factor H (CFH) gene and susceptibility to the disease [6-8]. Since then, risk-associated polymorphisms in complement factor I (CFI) , complement factor B (CFB) , complement component 3 (C3) , complement factor H-related 3 (CFHR3), and complement factor H-related 1 (CFHR1)  genes have been discovered and replicated. Progress has been made toward understanding the functional effects of the I62V and Y402H [13-15] polymorphisms in CFH, the deletion of CFHR3 and CFHR1 , and the R32Q polymorphism in CFB .
The central step of all three complement pathways is activation of the C3 molecule to C3b by a C3 convertase enzyme . In the alternative pathway, the C3 convertase is made by the binding of C3b to CFB, forming the proenyzme C3bB. An enzyme, complement factor D (CFD), cleaves the CFB part of the proenzyme, separating the Ba fragment of the molecule from the active C3bBb complex. Montes et al.  studied the effect of two adjacent genetic polymorphisms on the function of human and recombinant CFB in vitro. These single nucleotide polymorphisms, rs641153 (R32Q) and rs12614 (R32W), affect the same codon and amino acid residue. Of the four possible combinations of bases, three are found to exist, with the combination of both major alleles resulting in an arginine (R), the minor allele of rs641153 with the major allele of rs12614 resulting in a glutamine (Q), and the minor allele of rs12614 with the major allele of rs641153 resulting in a tryptophan (W) at residue 32 (Table 1). This residue is in the Ba fragment of CFB, and is part of the binding site for C3b . Compared to 32R, the 32Q variant of CFB binds with fourfold reduced affinity to C3b, reducing formation of the C3bB proenzyme and reducing lytic activity . The 32W variant of CFB also exhibits reduced binding, proenzyme formation and lytic activity relative to 32R, but the reduction is less marked than that found with 32Q . The 32Q variant is associated with a reduced risk of developing AMD . The 32W variant has been studied by several groups [5,10,18,19], but has not been found to be associated with AMD risk. In this study, we investigate the risk of AMD associated with the 32R, 32W, and 32Q variants of CFB. Statistical analysis is presented for these variants alone, and in a logistic regression incorporating the other major genetic risk factors, CFH and HTRA1, and smoking status. We also perform a meta-analysis of data from all identified studies that have previously investigated rs12614 (R32W) in AMD.
The characteristics of the study group are shown in Table 2. Individuals with neovascular AMD in at least one eye were recruited between June 2002 and September 2006 from ophthalmology clinics in the Royal Victoria Hospital, Belfast, UK, the regional referral center for Northern Ireland. Diagnosis was made by clinical examination and fluorescein angiography. Stereoscopic digital fundus photographs were graded according to the Wisconsin age-related maculopathy grading system . Cases had grade 4 neovascular AMD with or without geographic atrophy. Control participants were recruited from the local population, and they underwent retinal photography. Those included had grade 1a retinas (free of drusen or fewer than five hard drusen with a diameter of <63 µm). All participants self-reported European ancestry.
The study conformed to the tenets of the declaration of Helsinki, and it was approved by the Office for Research Ethics Committees, Northern Ireland. All participants gave informed consent.
DNA was extracted from peripheral blood leucocytes or frozen buffy coat samples. We performed SNP genotyping of rs12614 and rs641153 by PCR (forward primer: 5′-ACA CAC CAT CCT GCC CCA G-3′; reverse primer: 5′-TAC CCC CTC CAG AGA GCA GG-3′) followed by DNA sequencing using the forward primer. All samples showing the minor allele of rs12614 were confirmed using SNaPshot (Applied Biosystems, Warrington, UK). Briefly, primer (5′-CAG GTG TGA CCA CCA CTC CAT GGT CTT TGG CC-3′) extension was performed on the PCR product using a fluorescent detection system. Data for the smoking status and CFH and HTRA1 haplotypes of these participants were available from our previous studies [12,21].
Deviation from the Hardy–Weinberg equilibrium was investigated by use of the exact method of Wigginton et al.  as used in PLINK v1.07 . Logistic regression was performed using PASW v18. The numbers of copies (0, 1, or 2) of each CFH haplotype and CFB codon 32 variant carried by an individual were included as variables, omitting a reference haplotype from each gene. CFH haplotypes were tagged by rs6677604, rs3753396, rs419137, and rs2284664 (1: GACG; 2: GAAG; 3: GGAG; 4: GAAA; 5: AAAG), with haplotype 3 omitted as the reference. The heterozygous or homozygous state for the HTRA1 risk allele (rs10490924; T) was included, with homozygosity for the major allele (rs10490924; G) omitted as the reference. Current smoking and past smoking were included in the model as binary variables, with having never smoked omitted as the reference. Associations between disease status and allele frequency in the studies performed by ourselves and other groups were investigated using Pearson’s χ2 test in Epi-Info v6. To enable comparison of data sets from other groups, we used their published data, multiplying allele frequencies by the number of cases or controls to give the allele count if these were not published. For the meta-analysis, we compared the counts of 32W and 32Q alleles to the reference variant, 32R. Statistical significance was accepted at p<0.05 for all tests.
The 32Q and 32W variants of CFB were found to be protective against the disease when compared to the 32R variant (Table 3). The 32Q variant was associated with an odds ratio of 0.43 (p=8.1×10−5), and the 32W variant was associated with an odds ratio of 0.64 (p=0.04).
The 32Q and 32W variants of CFB were found to be protective against AMD when smoking status and other known genetic risk loci were integrated into a logistic regression (Table 4). The gradation of effect was the same as in the individual regression, with 32Q being more protective than 32W (odds ratios 0.33 and 0.53, respectively). CFH haplotypes 1 and 2, which carry the Y402H risk allele , were associated with increased risk relative to haplotype 3 (odds ratios 1.91 and 2.41, respectively). CFH haplotype 5, which carries the deletion of CFHR3 and CFHR1 , was protective against AMD (odds ratio 0.47). The HTRA1 risk allele was associated with increased risk, particularly in homozygosity (odds ratio 25.04). Past and current smoking were associated with increased risk (odds ratios 1.73 and 3.61, respectively) compared to having never smoked.
The 32Q variant of CFB was significantly associated with protection from AMD in all of the studies that also reported data for 32W (Table 5). The 32W variant was associated with moderate protection from neovascular AMD in the current study and the Gold  and Spencer  studies, though the finding was not statistically significant in the latter two studies. Meta-analysis of these three neovascular AMD studies showed an odds ratio of 0.75 (p=0.02) for the 32W variant, compared to the 32R variant. Gold et al.  reported data for other AMD phenotypes, while Kaur  and Scholl  reported studies of mixed phenotypes. The trend in these studies was less consistent, with Gold reporting no association for geographic atrophy (though with wide confidence intervals)  and Scholl reporting a nonsignificant increase in risk (in a small study of mixed phenotype, with wide confidence intervals) . Meta-analysis of all the studies that included all phenotypes showed that the 32W variant of CFB was associated with a significant reduction in risk of AMD (odds ratio 0.79, p=0.01).
Significant association between the minor allele at rs12614 in CFB (32W) and decreased risk of neovascular AMD has not been reported previously. The study by Montes et al.  provided evidence that this polymorphism was associated with less efficient alternative complement pathway activation due to weakened binding between CFB and C3b. Their study also showed that the established protective 32Q variant (associated with the minor allele at rs641153) was associated with a more extreme reduction in complement activation relative to 32R. We investigated the hypothesis that the genetic variant causing 32W was associated with protection from AMD, and found this to be the case when compared to 32R by Pearson’s χ2, by logistic regression with rs641153 and by logistic regression that included other known risk factors for AMD. The two other studies of neovascular AMD (Gold  and Spencer ) showed nonsignificant protective effects (odds ratios 0.87 and 0.77, respectively). When all neovascular cases were examined in a meta-analysis, a statistically significant, moderately protective odds ratio (0.75; p=0.02) was revealed. A more significant, less strong odds ratio (0.79; p=0.01) was revealed when meta-analysis of all AMD cases was undertaken.
This association has not been detected previously. This discord may reflect chance findings of either our study or the studies of others. However, combined analysis of all available data supports the conclusion that CFB 32W is associated with moderate protection against neovascular AMD, a finding that is in keeping with the study by Montes et al., which showed decreased complement activation relative to CFB 32R .
Our study provides further evidence that an individual’s risk of AMD is affected by gene polymorphisms that affect the function and interaction of the alternative complement pathway proteins.
We are grateful to the participants of the study for their involvement. We thank Usha Chakravarthy, Vivienne McConnell and Giuliana Silvestri for recruitment and provision of clinical data and blood samples, Ruth Hogg for validation of the AMD database, David McGibbon for extraction of DNA, Chris Patterson for statistical assistance and Gareth McKay for helpful discussions. D.T.B. is supported by a research and development doctoral fellowship from the Public Health Agency, Northern Ireland. Part of this work was presented at the annual meeting of the American Society of Human Genetics, November 2010.