Our study assessed the influence of genes in the CFH and LOC387715/HTRA1 regions and smoking status on the etiology of neovascular AMD. We developed a risk model for disease based on comprehensive haplotyping of the genes followed by analysis using logistic regression that also included smoking status. Our model is the first that we know of to take account of haplotypes within the CFH and LOC387715/HTRA1 regions, and allows early identification of those at high risk of developing neovascular AMD in old age. AMD was not inevitable for those in the highest-risk categories, hence exposure to a trigger may be necessary for development of disease.
AMD is one of the most common multifactorial diseases of the elderly. For several years, genes have been thought to play a role in susceptibility. The model generally applied to such complex disorders has been of multiple genes each contributing a small effect to overall risk, in conjunction with environmental factors. This model has been further complicated by the possibility of genes acting with either dominant or negative effects, of gene–gene or gene–environment interactions, and of genetic heterogeneity in which mutations in different genes may have a major influence in different individuals. As is often the case in clinical genetics, a much more straightforward pattern emerges when the possibilities are investigated and some ruled out. Credit must be given to the firm scientific base provided by well-designed linkage studies of families with more than one affected patient from which AMD susceptibility loci have been identified. Although individual studies often identified linkages that have not been confirmed, meta-analyses [13
] clearly revealed the regions of Chromosomes 1 and 10 (from which the CFH
gene cluster and LOC387715/PLEKHA1/HTRA1
have recently emerged) as major susceptibility loci for all forms of AMD. It is increasingly likely that the genetic component of AMD should be viewed as oligogenic rather than polygenic.
gene cluster comprises five related genes that are all expressed. These genes share a very high level of homology arising from gene duplication, and care must be taken in typing and identifying polymorphic variants at the C-terminal region of CFH and in the CFH
-related genes. We chose to score risk on the basis of five haplotypes of the most central block, which was in strong LD with the adjacent haplotype blocks in CFH
, and CFHR1
. Our earlier studies showed strong association between AMD and markers extending across this region [19
]. Any coding variants or polymorphisms regulating splicing or expression of these genes may affect risk, and may be based within different haplotype blocks. We also typed Y402H and I62V, but analysis of these common coding polymorphisms did not confer additional information affecting AMD risk. Other studies have limited their analysis to the effect of Y402H, which has been regarded as the main determinant of disease, but this analysis gives an incomplete and inaccurate representation of the region. Only haplotype analysis can assess the effect of combinations of all the coding variants in CFH
that may contribute to AMD susceptibility. The C allele of rs1061170, which types histidine at codon 402 within the first haplotype block, is associated with a haplotype conferring high risk of AMD, and is coupled with our haplotypes 1 and 2 in the second haplotype block. It is also in complete LD with several other SNPs, one of which has the potential to create a novel splice site and to introduce a new exon containing a premature stop codon. This variant would be expected to result in nonsense-mediated decay of the transcript and reduced CFH levels. Despite being associated with the T allele of rs1061170, haplotype 3 is only marginally less severe in risk than haplotype 1. Haplotypes 4 and 5 are relatively protective. We previously identified a deletion of 85 kb encompassing CFHR1
, which plausibly confers the protective effect of haplotype 5 [19
]. The coding variant most strongly associated with haplotype 4 may be the A allele of rs800292, which encodes I62, though this represents a conservative amino acid change. Further work is needed to clarify the molecular bases of the haplotypic effects.
Much interest surrounds the identity of the AMD susceptibility locus on Chromosome 10. Very strong data focus on the protein encoded by the hypothetical gene LOC387715
, with the alanine-to-serine polymorphism detected by rs10490924 as the most probable causative variant [23
]. Alternatives are either PLEKHA1
, which flank the locus; HTRA1
, which encodes a serine protease, is the more attractive on functional grounds. Another possibility is that the putative exons of LOC387715
or additional open reading frames in the HTRA1
promoter region may be included in alternative transcripts of HTRA1
. The LOC387715
transcript is exceedingly rare, but has been reported to be present at very low levels in the retina [23
]. The markers we typed for this locus identified all common haplotypes of a block that encompassed LOC387715
to the HTRA1
promoter. In contrast to the CFH
gene cluster, the effect of the LOC387715/HTRA1
region on AMD susceptibility appeared to be mediated through a single detrimental haplotype, with all others being generally protective. In addition to carrying the rare coding variant of rs10490924, the detrimental haplotype was also very strongly associated with several variants in the untranslated region of LOC387715
and in the promoter of HTRA1
, and we cannot at present discriminate between the relative merits of variation in LOC387715
or the HTRA1
promoter as effectors of AMD susceptibility at this locus.
Further genetic or environmental factors, or their interactions, may be found to affect AMD risk. These factors should be assessed for effect using a regression model that includes all proven genetic and environmental factors, rather than in isolation. Initial reports are prone to inflation to some extent by type 1 error. Conflicting data have been reported about a possible interaction between LOC387715
and smoking [35
]. Our evidence does not support an interaction, and we find limited support for an interaction recently reported between CFH
]. Interaction between CFH
and the genes encoding complement C2 or complement factor B [38
], and CFH
and an excision repair gene named ERCC6
] await confirmation.
Options for reducing the future burden of AMD appear as our knowledge increases. Although smoking status is of less importance than the two main genetic factors, it is a relatively straightforward risk to target in individuals with high genetic load. Our data show that if none of our population had smoked, there would have been an overall reduction of 33% in AMD today. Our current and ex-smokers lost vision at an earlier age, so it is probable that our calculations underestimate the benefit to the population from nonsmoking. We did not assess the importance of smoking on earlier stages of AMD; however, delaying either onset of disease or advancement to the blinding end-stage by 5–10 y would have a major impact on prevalence of blindness. It is now possible to identify those at high genetic risk for whom smoking is particularly unwise. In the past, smokers have been prepared to ignore greatly increased risks of lung cancer and other smoking-related diseases, but attitudes to smoking are changing. Relatives of AMD patients are more likely to carry an excess of LOC387715/HTRA1 and CFH-related genetic risk factors. Those who are shown to be at high risk, with relatives who have lost their vision, may be more prepared to heed advice to refrain from smoking than the general population.
Gene therapy to repair genetic defects is still in the experimental phase and is largely unproven. Of great promise are methods aimed at gene silencing by degradation of RNA using short interfering double-stranded RNA molecules. Deletion of CFHR1 and CFHR3 from the CFH-related gene cluster is strongly protective against AMD, hence these genes make excellent targets for silencing in future studies aimed at reducing the burden of disease.
We assessed the genetic factors and smoking history, which are generally accepted as conferring high population-attributable risk for AMD. The model we developed to estimate risk of neovascular AMD best fits our data, and it is important that it should be validated in independent datasets of similar phenotypes. It is possible that rare variants in other genes may be found to be highly influential in a small proportion of the population, and may cause AMD in a few individuals predicted to be at low risk using our present model. Genes with a modifying effect on AMD risk may also be identified. In either case, our model for prediction of AMD risk may be modified easily. CFH
may also affect risk of myocardial infarction [40
], and we made no attempt to adjust for this condition or for smoking-related diseases that affect longevity.