In a matched sample set from the AREDS cohort, we observed associations between haplotypes in genes involved in the complement cascade, and also in two genes on chromosome 10, with AMD. Expectations about the risk or protective effects of each were consistently met in our analysis, which reliably validates and supports numerous other research studies of AMD. Hypotheses about the nature of the associations of these SNPs and AMD are discussed below.
The physiological role of CFH is to downregulate and control initiation and magnitude of the complement cascade, and its function is dependent on how well it binds to cell surfaces. CFH consists of 20 short consensus repeat (SCR) modules, and SCR7 and SCR20 specifically bind to polyanions on the surface of cells.22
It is the SCR7 that harbors the polymorphic residue Y402H, and the His402 allele alters the precise specificity and affinity of its interaction with the cell surface, leading to altered levels of CFH retention on the retinal/macular surface, and thereby affecting complement activity in the area.22
This effect is supported by our study of polymorphism rs1061170 (Y402H) in which we see an OR of 2.0 in patients with one copy and 6.2 in patients with two copies of the His402 variant. Other studies have suggested that this polymorphism accounts for up to 50% of the population attributable risk for AMD.7, 8, 9, 10, 12
The other missense mutation in CFH
, rs800292, changes protein residue 62 from Ile to Val, but the resulting structural changes and effects are understudied. Both rs800292 and the non-coding rs1410996 showed unfavorable odds of disease in a recessive manner. That is, we did not see a statistically significant difference in these two for the heterozygous genotype, whereas we did for homozygous mutants (ORs=3.7 and 6.6, respectively). Haplotype analysis of this gene allowed us to assess the effect of all the coding variants together. The presence of a risk allele from rs1061170 in the high-risk haplotype (GCGN) clearly confers a significantly increased risk for AMD.
A large deletion 3′ to CFH is associated with protection from AMD. The CFHR1 and CFHR3 genes deleted in this region are closely related to CFH and their proteins are abundant in serum, but they have as yet no defined role in the complement cascade and no clear function. Rarely did the deletion segregate with the high-risk C allele for Y402H in our sample set, and so its protective effect could not be independently assessed against the risk exerted by a mutated CFH gene. Unfortunately, we were not able to distinguish the copy number for the deletion, which could change the haplotype estimations, and thus our risk assessments. Future plans are to develop a method of distinguishing heterozygous (+/−) from homozygous (+/+) non-deleted individuals.
CFB is an alternative complement cascade protein. Activation of the alternative pathway is initiated by cleavage of C3b-bound factor B (CFB), resulting in the formation of the C3Bb complex (C3 convertase). Of the two SNPs we studied in CFB
, only rs4151667 showed a statistically significant difference between cases and controls. This SNP gives rise to a missense variant that results in amino acid 9 changing from Leu to His in the signal sequence region of the protein. There appears to be some interaction between one SNP in CFB
and Y402H in CFH
. The unadjusted OR for rs2072633 is not statically significant alone (P
=0.63), but after adjusting for Y402H it becomes significant (P
=0.01). This suggests CFB effectively cancels the aberrant activity of a mutated CFH
gene, thus modifying risk in a favorable direction; however, it does not reduce risk further beyond a wild-type 402Y phenotype. Also of note, because C2 and CFB are closely related in structure and function, rs4151667 from CFB
and rs9332739 from C2
show the same frequency of common and rare alleles between cases and controls, resulting in very consistent protective effect between them. In fact, as we reported previously, the C2
genes are located only 500
bp apart on chromosome 6p21 in the major histocompatibility complex class III region, and are evolutionarily and structurally closely related.11
Previous genetic studies have shown that mutations in C2
, as in CFB
, can compromise the efficiency of the complement cascade. The variants in C2
show a protective effect in our study (), but it is unclear whether the association is in any way causal. Studies of drusen composition have found deposits of proteins from the alternative complement pathway, but their analogs from the classical pathway, including C2
, have not been found. Analysis of haplotypes across this region shows two that are protective, both of which appear to be due to the variants from CFB
C3 is the most abundant complement component, synthesized predominantly in the liver but to a lesser extent in other cells and tissues. Significant C3 transcripts are detectable in the neural retina, choroid, retinal pigment epithelium, and cultured retinal pigment epithelium cells. In contrast to the other complement cascade proteins showing a protective effect, variants in C3 increase the risk for AMD. Experimental evidence of functional differences between wild-type versus mutated alleles in C3 is not conclusive; however, this gene is mutated in many other diseases affecting epithelial tissues that are very similar in structure to choriocapillaris, Bruch's membrane, and retinal pigment epithelium. It is unclear how a mutated form of this gene would lead to sequestered complement protein deposits in some kidney disorders, but it is strikingly similar to the drusen deposits seen in the dry form of AMD. Functional studies are needed to further characterize this important protein and how it contributes to the incremental risk associated with its variants.
Polymorphism at 10q26 spanning ARMS2
and the promoter region of HTRA1
, as well as the estimated haplotypes from these variants, shows very strong association with AMD case status. Our sample showed a slightly higher risk from rs11200638 in HTRA1
, yet much attention and focus remain on rs10490924 in ARMS2
; Fritsche et al
extensively sequenced several individuals with the ARMS2
risk haplotype and found non-sense-mediated decay of ARMS2
resulting from a complex variant in the 3′ UTR (c.*
372_815delins54) and proposed this as the etiologic variant. On the other hand, a non-sense allele in ARMS2
(R38X) does not appear to confer significant risk (suggesting that a loss of function is not responsible for the ARMS2
and data not shown). Thus, confirmation that the c.*
372_815delins54 is the functional variant remains elusive.19
The APOE gene has been found to be associated with AMD in some but not all studies.24, 25
As can be seen in the data available at our FTP site, there were a total of 132 E4 allele-bearing subjects in the present study. A two-sided test for association yielded a P
-value of 0.097. It could be argued that a one-sided P
-value is more applicable; however, that value was also not significant at P
=0.054. Yet, there were only seven homozygous E2 individuals and five homozygous E4 individuals in our study. An additional nine individuals were E2/E4. Thus, we conclude that we did not have power comparable to that of Schmidt et al
who examined 1260 research subjects to come to the conclusion of association.
Recently, Yang et al27
reported the TLR3
SNP, rs3775291, to be very protective against geographic atrophy but not other AMD categories. This gene, which encodes a viral sensor that supports innate immunity and host defense,27
is a plausible candidate for AMD etiology. Edwards et al28
found a borderline association with rs3775291, in some cohorts and negative results in others, and we and others have failed to replicate this association, nor did it contribute to our multistate modeling.29
Surprisingly, we did not see our risk profiles change appreciably when we adjusted for smoking status, where most other studies show smoking to be a risk factor.30
There were 46 never smoker individuals in our research study, of which 32 were cases and 14 were controls. Although we saw no statistically significant relationship, this may be on account of our age choice of samples: There were no never smoker controls below the age of 60 years with whom to compare the cases. Among the cases below the age of 60 years, 14 were past smokers, 18 were current smokers, and 7 were never smokers. Therefore, our sample size, ascertainment, and paucity of younger controls make it inappropriate to come to a conclusion concerning smoking and AMD. Other reasons that we did not see a higher risk in smokers are that we did not stratify by AMD disease categories as some researchers have done, nor did we adjust for smoking in the haplotype analysis. We were limited to three categories of smoker by the AREDS self-reported questionnaire.
Our multistate modeling, using two independent analytic approaches on the same data set, concluded that we could account for ~80% of the cases and ~50% of the controls on the basis of the CFH locus, the chromosome 10 gene associated SNPs, and one C2 intronic SNP on chromosome 6. We realize, however, that the size of this cohort and number of controls may be insufficient to make a definitive interaction assessment. To this caution, we add the confounding effects of population stratification and a small but consequential effect of missing genotypes. In addition, new statistical methods for assessing interactions are emerging; thus, our current assessment is no more than a best guess estimate. On the other hand, we believe it sufficiently meritorious and diligent to report in this paper at this time.
Identifying genetic risk factors for AMD is important for understanding the etiology of the disease. Currently, AMD is not preventable and its causes are not fully understood although a number of risk factors and a few protective factors have been identified. A better understanding of the pathology and genetic predisposition of AMD is vital for both the affected individual and the larger public's health. In the coming years, we will have a greatly increased population over the age of 65 years with greater longevity, and therefore AMD disease will continue rising as an urgent public health issue. Vision loss resulting from AMD will have a considerable impact on the health-care system, social services, on families and communities, and quality of life. Significant efforts and progress have been made in learning the genetic causes of AMD, yet further research is needed to bring forth prevention and treatment strategies.