The discovery of the link between variants in complement factors and age-related macular degeneration (AMD)1-8 has led to complement factor 3 (C3) and C5, among other complement pathway members, becoming high-priority pathways for investigation. C3 and C5 interact closely with complement factors H and B; early subretinal deposition of C3 and C5 has been found in humans and in mouse models of AMD.9 This suggests that these proteins specifically contribute to the disease, and additional reports have suggested a functional role of C3 in AMD.10
A tag single-nucleotide polymorphism (SNP) is a representative SNP in a region of the genome with high linkage disequilibrium. It is possible to identify genetic variation without genotyping every SNP in a chromosomal region. We selected tag SNPs across C3 and C5—including SNP rs2230199 in C3 in particular—as it has been reported to have P = 2.8 X 10-5 in single-marker tests using genotyping arrays from the US National Institutes of Health dbGAP database in a genome-wide association of 400 individuals with AMD and 200 controls from the National Eye Institute's AREDS (Age-Related Eye Disease Study).
STUDY POPULATION
The study population, made up of 2,172 unrelated individuals of primarily European descent, were aged ≥60 years and were diagnosed based on ocular examination and fundus photography. Advanced cases of AMD and controls were classified according to our Clinical Age-Related Maculopathy Grading System,11 which has been adapted by other investigators. We have previously described this cohort in detail.7 A single SNP in C3—rs2230199—showed significant association with AMD, with P < 10-12, and a minor allele frequency of 0.21 in control individuals and 0.31 in affected individuals (Table 1).1 An additive, rather than dominant or recessive, model of association was the best fit with the finding. This SNP creates a nonsynonymous coding change (R102G) in the third exon of C3. No individually statistically significant association was found with any other SNPs typed in C3.
MULTIMARKER HAPLOTYPE TESTS
Additionally, our group performed multimarker haplotype tests to evaluate association at untyped SNPs present on HapMap; we did not find any additional associations. The International HapMap Project, an international scientific effort to identify common genetic variations among people, represents a collaboration of scientists from public and private organizations in six countries. We tested for association at the untyped SNPs and haplotypes conditioning on the genotype at rs2230199, and again, did not observe any additional significant associations. We also tested for any difference in association between the neovascular and geographic atrophy forms of AMD. No statistically significant differences were observed. We found no SNPs in C5 that revealed significant association with AMD.
Next, we evaluated the role of epistasis between rs2230199 and the five variants in other genes previously showing association in this cohort, including a novel variant we discovered in the CFH gene.7 Two variants at the CFH locus, two variants at the CFB/C2 locus, and one at the LOC387715/HTRA1 locus were established as unequivocally associated with AMD risk in this cohort. Using logistic regression, we did not observe any statistically significant interaction terms between any pair of these SNPs, the two factor B rare protective SNPs as a category, or the three haplotypes formed by the two independently associated but tightly linked CFH SNPs. This result suggests that, despite targeting the same pathway, these variants largely confer risk in an independent, log-additive fashion.
ADDED TO PREVIOUS RISK MODEL
Because the new variant acts independently, we were able to add it to the multilocus risk model from our earlier work7 (Figure 1). The individual and combined effects of the AMD-associated variants are additive; therefore, we estimate that the overall proportion of population variance in risk (assuming a prevalence of late-stage AMD in this age group of 5%) explained by this locus is about 2%. For comparison, a comparable estimate of the effects of variation at CFH, LOC387715/HTRA1 and CFB are 16%, 10%, and 2.5%, respectively, indicating that the individual effects of these four identified genetic factors alone explain 30% of the population variation in risk for a late-onset complex disorder with known environmental covariates.12 Due to the frequencies and penetrances of these alleles, the combination of these independent effects creates predictive value for late-stage AMD in the population from which these cases and controls were drawn. In this age group, the prevalence of late-stage AMD is about 5%; variation at these four genes identifies 20% of the population that have <1% risk of disease, and at the opposite end it identifies 1% of the population with >50% risk. Among this latter category, we observed 154 affected individuals (out of 1,238) and only nine control individuals (out of 934).
LINKAGE DISEQUILIBRIUM
Although rs2230199 is the most associated observed SNP, it might simply be in linkage disequilibrium (LD) with an unassayed causal variant (rather than being causal itself). To make this determination, we investigated in depth the surrounding LD and variation. HapMap Phase II shows few proxies for rs2230199: two SNPs (rs2230203, a synonymous exonic polymorphism 7.6 kb downstream, and rs163494, a noncoding SNP 5.9 kb upstream) are correlated with r2 = 0.75, and no others have r2 > 0.4 with rs2230199. We found that the only substantial correlations (r2 > 0.35) with rs2230199 exist within a 14-kb interval centered on exon 3. This finding strongly suggests that the association is in this segment.
Although the minor allele frequency of rs2230199 in the HapMap CEPH sample, 0.175, matches closely the control frequency, the minor allele is not present in Chinese or Japanese population samples, and the frequency is <0.01 in the Yoruba from the Ibadan, Nigeria, sample. Therefore, C3 may not have a role in AMD risk among these populations.
OTHER VARIANTS IN REGION
We used C3 resequencing data from 23 CEPH samples publicly available from the SeattleSNPs project to evaluate further candidate polymorphisms in this region. We examined this data set for potentially relevant variants that do not appear in the HapMap. One additional nonsynonymous SNP not present in the HapMap (rs1047286) was correlated with rs2230199 at r2 > 0.4, with r2 = ~0.53. Added to the observed lack of a stronger association and the previously established function of this variant, we propose that rs2230199 is likely to be the associated variant at C3.
Previously, this associated R102G variant was established as the molecular basis of the two common allotypes of C3: C3F (fast) and C3S (slow).13 C3F—the risk variant for AMD—has been previously reported as being associated with other immune-mediated conditions such as immunoglobulin A nephropathy.14 The variant may also influence the long-term success of renal transplants,15 in which C3S homozygote recipients had much better graft survival and function when receiving a donor kidney with a C3F allotype than a matched homozygote C3S donor.
This discovery of a new association between variation in the complement system and AMD adds to the growing evidence that inflammatory and immune processes are involved. Genetic factors do play a large role, and our twin study estimated that up to 70% of AMD is explained by genetic factors, although environmental factors also contribute.16 We have previously shown that both smoking and higher body mass index increase risk of AMD among individuals with the same genetic susceptibility,12,17,18 and that these same genetic and nongenetic factors predict progression from the early and intermediate stages to the advanced dry and wet forms of AMD.19 As evidence grows, knowledge of both nature and nurture will be essential for managing and treating our patients in the future.
Johanna M. Seddon, MD, ScM, is Professor of Ophthalmology, Tufts University School of Medicine, and Director, Ophthalmic Epidemiology and Genetics Service, Tufts-New England Medical Center, New England Eye Center. She may be reached at johanna_seddon@yahoo.com.
This article was adapted from an article that appeared in Nature Genetics.1
_1773249222.png?auto=compress,format&w=75)
