Molecular Vision 2014; 20:105-116
Received 08 October 2013 | Accepted 11 January 2014 | Published 14 January 2014
Alejandra V. Contreras,1 Juan Carlos Zenteno,3,4 Juan Carlos Fernández-López,1 Ulises Rodríguez-Corona,1 Ramcés Falfán-Valencia,5 Leticia Sebastian,1 Fabiola Morales,1 Daniel Ochoa-Contreras,2 Alessandra Carnevale,1 Irma Silva-Zolezzi6
The last two authors contributed equally to this work.
1Instituto Nacional de Medicina Genómica. Mexico City, Mexico; 2Asociación Para Evitar la Ceguera en México, Hospital “Dr. Luis Sánchez Bulnes,” Mexico City, Mexico; 3Department of Genetics and Research Unit, Institute of Ophthalmology Conde de Valenciana, Mexico City, Mexico; 4Biochemistry Department, Faculty of Medicine, UNAM, Mexico City, Mexico; 5Laboratorio HLA, Instituto Nacional de Enfermedades Respiratorias Ismael Cosío Villegas, Mexico City, Mexico; 6Nutrition and Health Department, Nestlé Research Center, Lausanne, Switzerland
Correspondence to: Alejandra V. Contreras. Periférico Sur 4809, Arenal Tepepan, Tlalpan, 14610 Mexico City, Mexico; Phone: +525553501974; FAX: +525553501974; email: firstname.lastname@example.org
Purpose: To evaluate the contribution of genetic variants of complement factor H (CFH), complement component 2 and 3 (C2 and C3), complement factor B (CFB), and age-related maculopathy susceptibility 2 (ARMS2) to age-related macular degeneration (AMD) risk in the Mexican Mestizo population.
Methods: Analysis included 282 unrelated Mexican patients with advanced AMD, 205 healthy controls, and 280 population controls. Stereoscopic fundus images were graded on the Clinical Age-Related Maculopathy System (CARMS). We designed a resequencing strategy using primers with M13 adaptor for the 23 exons of the CFH gene in a subgroup of 96 individuals clinically evaluated: 48 AMD cases and 48 age- and sex-matched healthy controls. Single nucleotide polymorphisms (SNPs) in C3 (Arg80Gly and Pro292Leu), C2 (rs547154), CFB (Leu9His), and ARMS2 (Ala69Ser) were genotyped in all patients, healthy and population controls using TaqMan assay.
Results: All evaluated individuals were Mexican Mestizos, and their genetic ancestry was validated using 224 ancestry informative markers and calculating Fst values. The CFH resequencing revealed 19 SNPs and a common variant in the intron 2 splice acceptor site; three CFH haplotypes inferred from individual genotypes, showed significant differences between cases and controls. The risk alleles in C3 (rs1047286, odds ratio [OR]=2.48, 95% confidence interval [CI]=1.64–3.75, p=1.59E-05; rs2230199, OR=2.15, 95% CI=1.48–3.13, p=6.28E-05) and in ARMS2 (rs10490924, OR=3.09, 95% CI=2.48–3.86, p=5.42E-23) were strongly associated with risk of AMD. The protective effect of alleles in C2 (rs547154) and CFB (rs4151667) showed a trend but was not significantly associated after correction for multiple testing.
Conclusions: Our results show that ARMS2 and C3 are major contributors to advanced AMD in Mexican patients, while the contributions of CFH, C2, and CFB are minor to those of other populations, reveling significant ethnic differences in minor allele frequencies. We provide evidence that two specific common haplotypes in the CFH gene predispose individuals to AMD, while another may confer reduced risk of disease in this admixed population.
Age-related macular degeneration (AMD) is a late-onset disease recognized as a serious condition that leads to vision loss in a growing number of elderly persons [1,2]. Early AMD is characterized by the presence of medium-size drusen or retinal pigment abnormalities; while advanced or late AMD can be either non-neovascular (dry, atrophic, or nonexudative) or neovascular (wet or exudative). The advanced non-neovascular form is characterized by drusen and geographic atrophy extending to the center of the macula, whereas advanced neovascular AMD is characterized by choroidal neovascularization and its sequelae . In the U.S. population, more than 8 million have AMD, and the prevalence of the advanced form is projected to increase by more than 50% by 2020 . In the United States, the highest prevalence of most AMD lesions was found in persons aged 60 years and older . Prevalence of advanced AMD in self-identified Latinos from Los Angeles, California, increased from 0% in those 40–49 years of age to 8.5% in those 80 or older, and that of early AMD from 6.2% to 29.7% .
One of the earliest genome-wide association studies (GWASs) examined AMD and demonstrated that a common variant in the complement factor H (CFH) gene was strongly associated with the disease . CFH resequencing revealed a polymorphism in linkage disequilibrium (LD) with the risk allele representing an amino acid change, Tyr402His (rs1061170) . Later, several groups replicated the finding , and new CFH variants were reported [9,10]. In addition, other polymorphisms in CFH besides Tyr402His appear to increase the risk for developing AMD in Asians [11,12]. Multivariate models have shown that variables such as ethnic group are significantly associated with the incidence of AMD , indicating that the molecular basis of the disease is polygenic, involving multiple genes, and with different levels of severity by ethnicity [14,15]. In fact, racial differences between Latinos, Asian Americans, and Caucasians have been proposed to explain the differences in the risk of developing nonexudative and exudative AMD among these ethnic groups . Thus, the AMD genetic risk and the frequency of associated genetic polymorphisms differ across ethnicities . For instance, while white and black groups show the highest frequency for the Tyr402His polymorphism, Hispanics and Chinese have the lowest frequencies of this variant . In addition, it has been suggested that although the Tyr402His variant appears to play a causal role in the etiology of AMD, this polymorphism could be simply in LD with additional nearby alleles that might show even stronger association .
However, additionally AMD susceptibility loci have been identified in several gene regions, including ARMS2/HTRA1 (Ala69Ser, rs10490924) , as well as others related to complement pathway genes such as factor B (CFB) (Leu9His, rs4151667), C2 (rs547154), and C3 (Pro292Leu, rs1047286 and Arg80Gly, rs2230199) [20-22]. Similar to the differences in frequency reported for Tyr402His across populations, the risk allele frequencies of Ala69Ser diverge greatly between European and Asian populations from the HapMap International Project, with 45% in Asian populations compared to 21% in Caucasians (HapMap). The Mexican population is mainly composed of Mestizos, who are individuals with a genetic background consisting of Amerindian, European, and, to a lesser extent, African ancestries. As a result of admixture, the resulting linkage disequilibrium (LD) patterns in this locus may affect the frequency and effect of the variants previously associated with AMD, and may enable the identification of additional AMD-related variants [23,24].
Therefore, we aimed to analyze the genetic variation in CFH, and evaluate the role of known polymorphisms in ARMS, CFB, C2, and C3 in AMD risk in a Mexican population. For this purpose, all 23 exons of CFH were resequenced in 96 Mexican subjects (48 AMD cases and 48 controls), and genotypes of seven previously AMD-associated single nucleotide polymorphisms (SNPs; rs10490924, rs4151667, rs547154, rs1047286, rs2230199, rs1410996, and rs1061170) were investigated in 754 unrelated Mexican Mestizos: 273 patients with advanced AMD, 201 healthy controls, and 280 population controls. The results of our study strengthen the understanding of the genetic contribution of the CFH, ARMS, CFB, C2, and C3 genes to susceptibility of AMD in Mexicans or other Latin American populations.
The protocol adhered to the tenets of the Declaration of Helsinki. This study was approved by the Scientific, Ethics, and Biosafety Review boards from the Instituto Nacional de Medicina Genómica (INMEGEN, Mexico). Informed consent was obtained from all individuals.
Patients with AMD and healthy controls were unrelated men and women aged >60 years enrolled at the Institute of Ophthalmology Conde de Valenciana and Hospital Dr. Luis Sánchez Bulnes de la Asociación Para Evitar la Ceguera in Mexico City. A total of 282 unrelated Mestizo Mexican patients with either of the two forms of advanced AMD (geographic atrophy or neovascular) and 205 healthy, age-matched controls were recruited and examined by an ophthalmologist. Stereoscopic fundus images of the macular region were obtained from each patient with AMD and healthy control for evaluation and diagnosis; thus, the groups of cases and healthy controls were equally rigorously clinically evaluated. The images were then graded based on the Clinical Age-Related Maculopathy System (CARMS), which categorizes patients into five mutually exclusive groups based on assessment of drusen, retinal pigment epithelial irregularities, geographic atrophy, retinal pigment epithelial detachment, and choroidal neovascularization . In addition, we analyzed DNA samples from 280 population controls, individuals from six different regions in Mexico from the Mexican genome diversity project (MGDP) , and 156 Mexican Amerindians from two ethnic groups, Mayas and Zapotecos. The population controls were recruited with the aim of analyzing genetic population structure within Mexico and were not assessed for AMD status.
Genomic DNA was extracted from peripheral-blood leukocytes using the QIAamp DNA Blood Maxi Kit (Qiagen, Hilden, Germany). We selected seven SNPs that have been strongly associated with AMD in several ethnic groups, including two in the CFH gene (rs1061170 [Tyr402His] and rs1410996), two in C3 (rs2230199 [Arg80Gly] and rs1047286 [Pro292Leu]), one in C2 (rs547154 [IVS10]), one in CFB (rs4151667 [Leu9His]), and one in ARMS2 (rs10490924 [Ala69Ser]). We performed genotyping with the use of an allelic discrimination assay with TaqMan SNP genotyping probes (Applied Biosystems, Foster City, CA), following the manufacturer’s protocol.
The entire CFH gene was resequenced in a subgroup of 96 individuals clinically evaluated: 48 AMD cases and 48 age- and sex-matched healthy controls. The subset of 48 cases was AMD Grade 5 according to the CARMS system, and were selected based on age (75±8.0) and sex (same number of men and women), to minimize the age and gender differences between them and the healthy control group used in the same analysis. All 23 exons of CFH were amplified in a GeneAmp PCR System 9700 (Applied Biosystems). The amplifications (600 bp on average) were performed with a proofreading DNA polymerase, Pfu Ultra II Fusion HS DNA Polymerase (Stratagene, La Jolla, CA), in a total volume of 25 μl PCR. Final concentrations were the following: 0.3 mM deoxynucleotide triphosphate mixture (Takara Bio, Shiga, Japan), 1 mM Mg2SO4, 0.3 mM primer mix (10 uM each), 1 U Pfu Ultra II Fusion HS DNA Polymerase, and 50 ng DNA template. The PCR products were cleaned up and then sequenced using the BigDye Terminator V3.1 and M13 universal primers on a 3730×l DNA Analyzer (Applied Biosystems). Sequences were analyzed using Module SeqMan V7.0 (Lasergene software V7.0, DNASTAR, Madison, WI). The reference gene sequences were obtained from GenBank, accession number NG_007259, version: NG_007259.1 RefSeqGene, 102,494 bp. Primers pair sequences used for the 23 exons are shown in Appendix 1; 16 had been previously reported .
Linkage disequilibrium analysis was done with Haploview 4.0 . Haplotypes were inferred from individual genotypes using the software PHASEv2.1 [28,29]. Conditions used for the haplotype inference procedure using 23 biallelic SNPs consisted of 100 iterations and one thinning interval step through the Markov chain.
To verify the distribution of the ancestral components between the case and control samples, we used 224 ancestry informative markers (AIMs) , which were SNPs with allele frequency differences (δ)≥0.4 for pairwise comparisons between the HapMap and Zapotecos groups. These AIMs were identified in the MGDP  and are publicly available upon request. Thus, we genotyped this set in our sample of cases and healthy controls using a customized microarray by Illumina (GoldenGate, San Diego, CA). The principal component analysis (PCA) and Fst were performed with EIGENSOFT 4.2 (Harvard, MA) [30,31].
PLINK association analysis toolset was used in data analysis . Hardy–Weinberg equilibrium was checked for all SNPs evaluated in the population and healthy controls. Logistic regression analyses and further adjustment for appropriate covariates, such as age and sex, were applied to evaluate whether the SNPs were associated with AMD. The significance values were adjusted for multiple testing, and Bonferroni adjusted p values were reported.
Data on disease status, sex, and age are provided in Table 1. In our sample, 32.3% of cases had geographic atrophy (CARMS 4) while 63.7% had choroidal neovascularization (CARMS 5). All evaluated individuals were Mexican Mestizos, and their genetic ancestry was validated using a PCA and calculating the Fst values. The variation observed between our sample groups represents the distribution of ancestral components previously reported in the Mexican population , thus indicating that our study was not biased by a population stratification effect (Appendix 2 and Appendix 3).
The CFH coding region (23 exons and the flanking sequences) was resequenced in 96 individuals (48 AMD cases and 48 healthy controls). This assay revealed 19 SNPs that have been observed previously, and a common variant in the intron 2 splice acceptor site, IVS2–18insTT. Two novel synonymous coding variants were found, one in the controls’ exon 5 (Asp194Asp, minor allele frequency=1.06%), and the other one in cases’ exon 17 (Gln816Gln, minor allele frequency=2.2%). The frequency distributions of the minor allele for all variants and their respective disease odds ratios (ORs) are provided in Appendix 4. Eleven SNPs showed significant allele frequency differences between the cases and controls; eight of them were more frequent in the cases than in the controls and had ORs ≥2.84, including Tyr402His. Interestingly, SNPs rs551397 in intron 1, Val62Ile (rs800292) in exon 2, and Ala473Ala (rs2274700) in exon 11 were observed more frequently in the controls than in the cases (OR=0.45, 95% confidence interval [CI]=0.22–0.93, p=3.17E-02; OR=0.45, 95% CI=0.22–0.93, p=3.17E-02; and OR=0.47, 95% CI=0.23–0.94, p=3.49E-02, respectively), suggesting that they confer a protective effect against AMD. However, after adjusting for multiple testing, these three suggestive protective variants were not significant.
LD analysis performed on 96 Mexican individuals showed extensive LD across an extended region of CFH (Figure 1). Five SNPs located in exons 7–9 were in virtually complete LD, and the same was observed for variants in exon 14 and exon 19 (r2=0.82) and SNPs in exon 7 and intron 16 (r2=0.82). Haplotype construction was done considering all polymorphisms found with direct sequencing. The haplotypes were inferred from the genotypes of 96 individuals, and the eight haplotypes that had a frequency greater than 4% are shown in Table 2. CFH haplotypes 6, 7, and 8 showed differences between the cases and the controls, 0.0% versus 4.2%, 9.4% versus 2.1% and 11.5% versus 0%, respectively. Haplotype 6, which was not present in the cases, comprised the three protective SNPs (rs551397, rs800292 and rs2274700). CFH haplotype 7 included the six AMD risk SNPs that conserved significant p values after adjustment for multiple testing (rs1061147, rs482934, rs12029785, rs1061170, rs4658046, and rs375046) and the four SNPs identified in exon 21. Haplotype 8 was not present in the controls and comprised the six risk alleles.
Seven SNPs previously associated with AMD in the CFH, C3, C2, CFB, and ARMS2 genes were analyzed in 273 unrelated patients with advanced AMD (geographic atrophy or choroidal neovascularization) and 201 healthy controls. Nine cases and four healthy controls were discarded for genotyping because of quality control of call rate <70% SNPs per individual. Moreover, we adopted an intraethnic experimental design by analyzing 280 population controls from six different regions in Mexico, northern, southern, central and both coastal, western-Pacific and east, with ancestral component differences previously described .
All 754 samples were genotyped to determine the allele frequencies in three groups for the seven SNPs. The fact that we observed few differences between these two control groups indicates that there would be little bias due to use of both samples as a control group and justifies our combining two control groups to form a single group of 481 reference subjects. Screening of the seven SNPs revealed that the risk alleles in CFH (rs1410996, p=6.81E-09; rs1061170, p=1.02E-06), in C3 (rs1047286, p=1.59E-05; rs2230199, p=6.28E-05) and in ARMS2 (rs10490924, p=5.42E-23) were strongly associated with AMD (Table 3). Previously identified protective alleles in C2 (rs547154, p=3.12E-02, OR=0.54, 95% CI=0.30–0.94) and CFB (rs4151667, p=5.06E-02, OR=0.34, 95% CI=0.12–1.00) were also found associated with reduced risk of AMD in this group. However, after adjusting for multiple testing, these two suggestive protective SNPs were not significant (Table 3).
To analyze the genetic structure at the CFH locus and identify additional susceptibility alleles for AMD, we resequenced all 23 exons of CFH in 96 Mexicans, 48 cases and 48 controls. We estimate that more than 99% of polymorphic sites with a minimal allele frequency of 1% were detected, in the 192 haploid genomes (96 individuals) evaluated in our study, and about 95% of all polymorphic sites with a similar frequency in the cases and the healthy controls . We did not find novel alleles with potential functional effect in the resequenced region; however, we could not exclude the possibility of genetic contributions of unidentified polymorphisms in intron or regulatory regions to AMD.
We identified eight CFH haplotypes with frequencies higher than 4% in at least one of the groups (Table 2). Within these, haplotype 6 including three alleles that showed a suggestive protective effect for AMD development, T in rs551397 (intron 1), A in rs800292 (Val62Ile, exon 2), and A in rs2274700 (Ala473Ala, exon 11), was more frequent in the controls (4.2% versus 0%) than in the cases (p value=0.043). Interestingly, this is consistent with the study by Ng et al., who showed these three alleles are part of haplotypes that conferred a reduced risk of AMD in a Chinese population . In addition, one common extended haplotype in the CFH gene has been described as associated with lower risk if atypical hemolytic uremic syndrome (aHUS), membranoproliferative glomerulonephritis type II or dense deposit disease (MPGN2/DDD), and AMD [8,35]. This CFH haplotype carries the Val62Ile variant within the short consensus repeat 1 (SCR1) domain in the N-terminal region that is essential for factor H (fH) regulatory activities . Tortajada et al. showed that the fH-Ile62 variant exhibits increased binding to C3b compared with fH-Val62 and is a more efficient cofactor for factor I in the proteolytic inactivation of C3b . These findings provide evidence of CFH haplotypes that may have a protective effect for AMD; however, these results need further validation in an independent or larger sample size.
Haplotypes 7 and 8 were more frequent in the cases than in the controls (p value=0.030 and 0.001, respectively). Both haplotypes include two risk SNPs previously reported, Ala307Ala (rs1061147) in exon 7 and Tyr402His (rs1061170) in exon 9 [8,18], and the rs375046 in intron 16. Tyr402His has shown a functional effect, because it lies within SCR7 domain in fH, which binds heparin and C-reactive protein . We found that rs375046 is in high LD with rs1061170 and rs1061147 (r2=0.82), and shows association with AMD risk, while in a Chinese population the frequency distribution of these variants is different and they are not associated .
In our population, the LD between Ala473Ala (rs2274700) and Gln672Gln (rs3753396) was less than that reported in a Chinese population, where both SNPs were significantly associated with exudative ADM . The SNP Gln672Gln was not associated with AMD risk in our study and neither in Caucasians . The frequencies pattern of this SNP and Ala473Ala suggests that the LD in our population is similar to that of Caucasians. Therefore, our results highlight the importance of considering LD differences observed in the CFH region and the frequencies patterns of genetic variants between populations to interpret the results of association studies and determine genetic susceptibility to AMD. In this context, haplotype and conditional analysis of individual variants should be performed to better understand the interaction between the different variants.
The AMD-associated variants in CFH, C2, C3, CFB, and ARMS2 have been established in populations worldwide [10,12,19,20,22,37,38]. However, genetic risk and the frequency of associated genetic polymorphisms differ across ethnicities . Thus, when we compared the frequency of the seven genotyped SNPs in our control population group with two Mexican Amerindian groups (Zapotecos and Mayas) and individuals from three continental groups included in the HapMap International Project, Caucasians, Asians (Chinese and Japanese population), and individuals from a Nigerian region, we observed significant differences for rs1061170 (Tyr402His) in CFH between Mexican Mestizos and those populations (Appendix 5). In our population, the Caucasian ancestral contribution for this SNP seems to be important, although its frequency remains not as high as in Caucasians. The association for CFHrs1061170 has been reported mainly in Caucasians [39,40]. Our results showed that in Zapotecos, an Mexican Amerindian group, the Tyr402His frequency was as low as that reported for Japanese and Chinese, populations in which this variant is not associated with AMD [11,12,41] and where this allele may confer risk on just a few individuals in the population, and thus have a much lower contribution to overall genetic risk of AMD than in Caucasians .
Replication of genetic association studies in different populations helps in determining the effect size of the alleles associated with the disease and provides evidence of the underlying biologic mechanisms in the disease etiology [42,43]. Therefore, our analysis of these five genes provides additional support of their involvement in AMD in an admixed population. An important consequence of using population controls for which detailed phenotyping for the trait of interest is not available relates to the potential of misclassification bias: A proportion of the controls is likely to have AMD, and others will develop it in the future . However, the effect that this has on power should be modest unless the extent of misclassification bias is substantial; for example, if 5% of controls met the definition of cases at the same age, the loss of power would be approximately the same as that due to a decrease in the sample size by 10% . In Mexico, a report showed that the prevalence of AMD in a reference hospital at Mexico City was 3.3% . Including controls from the same geographic area (population controls) is considered an effective approach and has been previously used in the analyses of multiple disease phenotypes, including AMD [44,47].
The SNPs showing strong associations with AMD (rs1410996, rs1061170, rs1047286, rs2230199, and rs10490924) using the combined groups of healthy plus population controls as reference were associated only when healthy controls were used. The SNPs in C2 and CFB (rs547154 and rs4151667) were not as strongly associated as those in CFH, C3, and ARMS, but the power to detect the association is limited because their frequencies are low in the Mexican population (6% and 3%, respectively); however, the patients with AMD had an allele frequency lower than that of the general population (3% versus 6% and 0.7% versus 3%, respectively). The previous observation indicates that these alleles may confer a protective effect in this population, but due to their low frequency, a larger sample size is needed to have the power to detect their association and confirm the protective effect previously described in other populations [20,37].
A previous study in Mexicans also reported significant association of CFH Tyr402His (OR=3.8, p=1.0E-05) and ARMS2 Ala69Ser (OR=3.04, p=4.0E-07) . The SNP evaluated in the ARMS2 gene (rs10490924) showed a strong association with AMD in our sample (OR=3.09, 95% CI=2.48–3.86, p value=5.42E-23). Thus, as a result of logistic regression analysis to assess the conditioning effect of SNPs, we observed that for ARMS Ala69Ser (rs10490924) the BETA coefficient was still significant (p=5.23E-08) after entering CFHrs1410996, CFH Tyr402His (rs1061170), sex and age as covariates, suggesting that it does indeed have an effect independent of all these covariates (Appendix 6). This polymorphism had been reported to show high association with AMD [19,49,50]; Fritsche et al. proposed a functional role of ARMS2 in mitochondrial homeostasis, although the precise mode of action remains to be elucidated .
In summary, our data show no evidence of the presence of novel alleles associated with AMD in the coding region of the CFH gene, although unidentified polymorphisms in the intron or regulatory regions may exist. In addition, we confirmed the association of CFH, C3, and ARMS2 with the risk of developing AMD in Mexicans, and revealed significant ethnic differences in the allele frequency of SNPs, such as rs1061170 in CFH, rs2230199 in C3, and rs10490924 in ARMS2. The SNPs in C2 and CFB (rs547154 and rs4151667) may confer a protective effect in our Mexican Mestizo population, but due to their low frequency, a larger sample size is needed to confirm this observation. Moreover, our results provide evidence that at least two specific common haplotypes in CFH predispose individuals to AMD, while another one may confer reduced risk of disease in this admixed population. In addition, other genetic factors not yet discovered may be responsible for AMD risk and protection in our Mexican Mestizo population. Gene–gene interactions between CFH, C2, and CFB and biologic and functional studies could be analyzed to identify the role of these three genes in conferring protection against AMD in admixed and parental populations.
Appendix 1. List of primer sequences used for Complement Factor H gene resequencing.
Appendix 2. Principal component analysis for the three samples groups used.
Appendix 3. Fst values between Age-Macular Degeneration cases, healthy controls and population controls.
Appendix 4. SNPs identified in the Complement Factor H gene re-sequencing in Mexicans, 48 Age-Macular Degeneration patients and 48 healthy controls.
Appendix 5. Comparison of allele frequencies of SNPs in CFH, C2, CFB, C3 and ARMS2 genes between Mexican Mestizos and Mexican Amerindians and HapMap populations.
Appendix 6. Genetic effect of rs10490924 to Age-Macular Degeneration risk in Mexicans.
This work was supported by Consejo Nacional de Ciencia y Tecnología grant SALUD-2008-C01–87887, and by Instituto Nacional de Medicina Genómica grant 06/2007. The authors declare no conflict of interest.