PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
 
Nat Genet. Author manuscript; available in PMC 2010 August 16.
Published in final edited form as:
Published online 2006 March 5. doi:  10.1038/ng1750
PMCID: PMC2921703
NIHMSID: NIHMS204482

Variation in factor B (BF) and complement component 2 (C2) genes is associated with age-related macular degeneration

Abstract

Age-related macular degeneration (AMD) is the most common form of irreversible blindness in developed countries1,2. Variants in the factor H gene (CFH, also known as HF1), which encodes a major inhibitor of the alternative complement pathway, are associated with the risk for developing AMD38. Here we test the hypothesis that variation in genes encoding other regulatory proteins of the same pathway is associated with AMD. We screened factor B (BF) and complement component 2 (C2) genes, located in the major histo-compatibility complex class III region, for genetic variation in two independent cohorts comprising ~900 individuals with AMD and ~400 matched controls. Haplotype analyses identify a statistically significant common risk haplotype (H1) and two protective haplotypes. The L9H variant of BF and the E318D variant of C2 (H10), as well as a variant in intron 10 of C2 and the R32Q variant of BF (H7), confer a significantly reduced risk of AMD (odds ratio = 0.45 and 0.36, respectively). Combined analysis of the C2 and BF haplotypes and CFH variants shows that variation in the two loci can predict the clinical outcome in 74% of the affected individuals and 56% of the controls. These data expand and refine our understanding of the genetic risk for AMD.

Inflammation has a central role in the pathobiology of AMD914. Dysfunction of the complement pathway is proposed to induce significant damage to macular cells, leading to atrophy, degeneration and the elaboration of choroidal neovascular membranes3,1517. Activation of the alternative pathway is initiated by cleavage of C3b-bound factor B (BF), resulting in the formation of the C3Bb complex (C3 convertase). This complex is stabilized by properdin, whereas its dissociation is accelerated by regulatory proteins, including complement factor H (CFH), the major inhibitor of the alternative complement pathway. As CFH haplotypes are associated with AMD3, we hypothesized that the same may be true for activators of the same pathway, such as complement factor B (BF). BF and complement component 2 (C2), an activator of the classical complement pathway, are paralogous genes located only 500 bp apart on human chromosome 6p21. These two genes, along with genes encoding complement components 4A (C4A) and 4B (C4B), reside in the major histocompatibility complex (MHC) class III region.

We scanned all 18 BF exons in 180 affected individuals and controls from a cohort evaluated at Columbia University. We identified 17 sequence variants; the L9H and R32Q alleles were more frequent in controls than in affected individuals (Table 1). We identified haplotype-tagging SNPs (htSNPs) within BF and C2 (Fig. 1) and genotyped them in 548 affected individuals and 275 controls. Four variants were significantly associated with AMD; the L9H variant in BF, which is in nearly complete linkage disequilibrium (LD) with the E318D variant in C2, is highly protective for AMD (P = 0.00020, odds ratio (OR) = 0.37 (95% confidence interval (c.i.) = 0.18–0.60)). The R32Q allele in BF is in nearly complete LD with the rs547154 SNP in intron 10 of C2 and is also highly protective (P = 6.43 × 10−9, OR = 0.32 (95% c.i. = 0.21–0.48)).

Figure 1
Diagram and haplotype analysis of SNPs in BF and C2. The SNPs used in the study are shown along with the predicted haplotypes, odds ratios (OR), P values and frequencies in the combined analysis of affected individuals (CAS) and controls (CON). The 95% ...
Table 1
Sequence variants in the BF gene detected by DHPLC screening

Genotyping of an independent cohort of 350 cases and 114 controls from the University of Iowa confirmed these findings. For example, the C2 E318D/BF L9H SNP pair was significantly associated with AMD (P = 0.0012). Haplotypes across the C2 and BF loci were analyzed on the basis of data derived from the combined cohorts (Table 2). The common haplotype (H1, Fig. 1) conferred a significant risk for AMD (P = 0.0013). The haplotype tagged by the BF R32Q SNP (H7) was highly protective for AMD (P = 2.1 × 10−7) and the haplotype containing the C2 E318D/BF L9H pair (H10) was also significantly protective (P = 3.4 × 10−6). We obtained even more significant results when using the H1 haplotype as a reference OR = 0.42 (95% c.i. = 0.32–0.58) for H7, and OR = 0.33 (95% c.i. = 0.21–0.52) for H10. Individuals with two protective haplotypes (either homozygous for H7 or H10 or 7/10 compound heterozygotes) were found in 3.4% of the controls but in only 0.77% of the affected individuals (OR = 0.22; 95% c.i. = 0.087–0.56). The odds ratio of individuals with two protective alleles was approximately half that of the individuals with one protective allele, consistent with a codominant model.

Table 2
Association analysis of C2/BF variants in combined Columbia and Iowa cohorts

These associations were statistically significant (Table 2) when the entire AMD cohort was compared with controls, or when major subtypes of AMD, including early AMD and choroidal neovascularization, were analyzed separately. The geographic atrophy (GA) group (133 individuals in two cohorts) deviated from the general trend, similar to our observations with CFH3. Specifically, haplotype H7 demonstrated the strongest protection against the disease (OR = 0.22) when we compared the GA group with controls, in contrast to OR = 0.45 for the remainder of the AMD subjects. Although this deviation may be important in terms of varying etiology of the disease, it did not reach statistical significance, probably owing to the small number of GA cases.

We initially performed combined analyses with CFH by stratifying the subjects according to status at the CFH Y402H allele. Protection conferred by C2 and/or BF was strongest in CFH 402H homozygotes (OR = 0.27), intermediate in 402H/Y heterozygotes (OR = 0.36) and weakest in 402Y homozygotes (OR = 0.44). However, the confidence intervals of all these estimates overlapped. The effect is due to a trend in which the frequency of C2 and/or BF protective alleles is greatest in 402H homozygotes; 40% of these individuals in the control cohort carried at least one protective allele. In contrast, controls that were 402H/Y or 402Y had lower frequencies of C2 and/or BF protection (32% and 26%, respectively). That is, individuals at high risk owing to their CFH genotype who have not developed AMD have a high frequency of protective allele(s) at the C2/BF locus.

To identify the possible combinations of CFH and C2/BF SNPs that are protective for AMD, we analyzed the data with an empirical model and then with a machine-learned model using Exemplar software (Fig. 2). The first model was a hypothesized (hand-built) model, such as one would create by an empirical inspection of the data (Fig. 2a). The model yields four possible combinations of genotypes that protect from AMD (that is, combinations that result in the model being ‘true’). We applied this model to the samples (Fig. 2b) and subjected the resulting distributions to significance testing (Fisher’s exact test; P = 0.00237, P = 4.28 × 10−8 and P = 7.90 × 10−10 for the Iowa, Columbia and combined cohorts, respectively). Subsequently, Exemplar software generated a protective model that provided a ‘best fit’ to the data using a machine-learning method called Genetic Algorithms to test the hypothesis that the machine-learning software can outperform the hand-built model. Models were learned on the Columbia cohort, and the resulting fittest models were applied to the Iowa cohort (for out-of-sample verification) and to the combined sample set. The best-performing model (Fig. 2c,d) describes four possible genotype combinations that protect from AMD (P = 7.49 × 10−5, P = 2.97 × 10−22 and P = 1.69 × 10−23, for the Iowa, Columbia and combined cohorts, respectively). We further validated this method by randomizing the case and control designations and performing 3,000 permutations of the dataset. In summary, combined analysis of these haplotypes with the variation in CFH showed that 56% of unaffected controls harbor at least one protective CFH or C2/BF haplotype, whereas 74% of AMD patients lack any protective haplotype. Approximately 60% of the risk in affected individuals and 65% of the protection of controls are due to the CFH locus, and the remainder (40% and 35%, respectively) to the C2/BF locus. The machine-learned model outperformed the hand-built model, providing more accurate predictions of clinical outcome in cases and controls. An analysis with a classification and regression tree (C&RT) method was used to independently confirm the relative contribution of the CFH and C2/BF loci in AMD. The C&RT models estimate that C2 and BF alleles account for 27–37% of the cases, which is consistent with the 35–40% estimated contribution of the C2/BF locus from the Genetic Algorithm analysis (see Supplementary Methods online for details).

Figure 2
Combined complement gene analyses. Individual SNP analyses suggested several possible combinations of SNPs that protect an individual from developing AMD. To test these, an empirical model was applied first. (a,b) Fraction accounted for by hypothesized ...

BF and C2 are expressed in the neural retina, RPE and choroid. We generated RT-PCR products for the BF and C2 gene from RPE, RPE/choroid complex and neural retina derived from eyes from donors with AMD (ages 67 and 94 years) and without AMD (ages 69 and 82 years) (data not shown). BF protein was present in ocular drusen and Bruch’s membrane and less prominently in the choroidal stroma (Fig. 3a). Ba (a BF-derived peptide) immunoreactivity was less pronounced but distinctly present in patches associated with RPE cells and throughout Bruch’s membrane (Fig. 3b). The distribution of BF was similar to that of C3 (Fig. 3c), both of which are essentially identical to that of CFH and C5b-9 (ref. 3).

Figure 3
Immunolocalization of proteins along the retinal pigment epithelium (RPE)-choroid (Ch) complex in sections from an unfixed eye of a 72-year-old donor with early-stage AMD. (a) BF. (b) Ba (a fragment of the full-length factor B). (c) C3. Anti-BF (red) ...

In summary, these data show that variants in the complement pathway–associated genes C2 and BF are significantly associated with AMD. Protective haplotypes in the C2/BF locus contain nonsynonymous SNPs in the BF gene, an important activator of the alternative complement pathway. Available data confirms the hypothesis that the AMD phenotype may be modulated by abnormal BF activity. Indeed, the BF protein containing glutamine at position 32 (resulting from one of the two BF SNPs tagging a protective haplotype) has been shown to have reduced hemolytic activity as compared with the more frequent Arg32 form18. Notably, the same study did not document a functional effect for the R32W variant, which was not associated with AMD in the current study. On the basis of these data, we suggest that an activator with reduced enzymatic activity may provide a lower risk for chronic complement response that can lead to drusen formation and AMD. This hypothesis is compatible with our previous proposal that insufficient inhibition of the alternative complement cascade owing to variation in CFH results in chronic damage at the retinal pigment epithelium/Bruch’s membrane interface3,9,10. Another BF htSNP, L9H, resides in the signal peptide. Although the functional consequence of this variant has not been demonstrated directly, this variant could conceivably modulate BF secretion.

The genetic and functional data suggest that variation in BF is probably causal for the observed association with AMD. The two haplotype-tagging variants in BF are nonconservative, and one of the two is documented to have a direct functional relevance, whereas the variants in C2 are a conservative change and an intronic SNP. In addition, BF participates directly in the alternative pathway, a pathway that also involves CFH. We cannot rule out a direct role for C2, however, particularly because both C2 and BF regulate the production of C3. C2 and BF have nearly identical modular structures, including serine protease domains and three complement control protein (CCP) modules. Additional support for the involvement of BF in AMD pathogenesis comes from studies of drusen composition. Although the majority of proteins involved in the alternative pathway (CFH, BF, etc.) are found in drusen, their analogs from the classical pathway, such as C2 and C4, are not11,13 (G.S.H., unpublished data). These data further suggest that the C2 SNPs are associated with AMD owing to extensive LD with BF.

Several common functional variants in both C2 and BF have been described1921, but most of these are rare. We analyzed all missense alleles with frequencies >2% in European populations as judged from resequencing data. Moreover, no additional nonsynonymous variants in either gene have been found after complete sequencing of several MHC haplotypes, including examples of our haplotypes H2, H5, and H7 (ref. 22).

Because C2 and BF reside in the MHC locus with many other genes involved in inflammation, it is possible that the associations observed in this study are due to LD with adjacent loci23. Five lines of evidence, however, suggest that the C2/BF locus is the main contributor to the observed association. First, there is only modest LD between C2/BF and adjacent class III loci (Supplementary Fig. 1 online). Second, MHC class II loci and BF haplotypes H7 and H10 do not show strong LD (G.S.H., unpublished data). Third, in a whole-genome scan6, the MHC locus did not demonstrate a statistically significant association with AMD. That analysis included 80 SNPs across the MHC locus but did not include any of the eight SNPs typed in this study. Fourth, estimated recombination rates from HapMap data indicate regions of high recombination on both sides of the C2/BF locus24 (M.D. and B.G., unpublished data). Finally, the single published study on MHC in AMD demonstrates modest protection for the class I locus B*4001 (P = 0.027) and the class II locus DRB1*1301 (P = 0.009)25. Because the protective alleles identified in this study are associated with AMD at a substantially higher statistical significance, it is very unlikely that the C2/BF association is due to LD with these and/or other loci in the MHC.

In conclusion, this study extends and refines the role of the alternative complement pathway in the pathobiology of AMD and further strengthens the proposed model that inflammation, infection or both have a major role in this common ocular disease.

METHODS

Patients

Two independent groups of individuals affected with AMD and age-matched controls of European-American descent over the age of 60 were used in this study. These groups consisted of 350 unrelated individuals with clinically documented AMD (mean age 79.5 ± 7.8 years) and 114 unrelated control individuals (mean age 78.4 ± 7.4 years; matched by age and ethnicity) from the University of Iowa and 548 unrelated individuals with clinically documented AMD (mean age 71.32 ± 8.9 years) and 275 unrelated age- and ethnicity-matched controls (mean age 68.84 ± 8.6 years) from Columbia University. Patients were examined by trained ophthalmologists.

Stereo fundus photographs were graded according to standardized classification systems as described previously3,26,27. Controls did not show any distinguishing signs of macular disease nor did they have a known family history of AMD (stages 0 and 1a). AMD patients were subdivided into phenotypic categories on the basis of the classification of their most severe eye at the time of their recruitment. Genomic DNA was generated from peripheral blood leukocytes using QIAamp DNA Blood Maxi kits (Qiagen).

Studies were conducted under protocols approved by the Institutional Review Boards of Columbia University and the University of Iowa. Informed consent was obtained from all study subjects before participation.

Immunohistochemistry

Posterior poles were processed, sectioned and labeled with antibodies directed against factor B (Quidel), as described previously9. Adjacent sections were incubated with secondary antibody alone to serve as controls. Some immunolabeled specimens were prepared and viewed by confocal laser scanning microscopy, as described9.

Mutation screening and analysis

Coding and adjacent intronic regions of BF and C2 were examined for variants using SSCP analyses, denaturing high performance liquid chromatography (DHPLC) and direct sequencing. Primers for SSCP, DHPLC and DNA sequencing analyses were designed to amplify each exon and its adjacent intronic regions using MacVector. Primer sequences are available upon request. PCR-derived amplicons were screened for sequence variation, as described earlier28,29. All changes detected by SSCP and DHPLC were confirmed by bidirectional sequencing according to standard protocols.

Genotyping

SNPs were discovered through data mining (Ensembl database, dbSNP, Celera Discovery System) and through sequencing. Assays for variants with >10% frequency in test populations were purchased from Applied Biosystems as Validated, Inventoried SNP Assays-On-Demand or were submitted to an Applied Biosystems Assays-By-Design pipeline. The technique used was identical to that described previously3. Briefly, 5 ng of DNA were subjected to 50 cycles on an ABI 9700 384-well thermocycler, and plates were read in an Applied Biosystems 7900 HT Sequence Detection System.

Statistical analysis

Genotypes were tabulated in Microsoft Excel and presented to SPSS for contingency table analysis as described previously3,27. Compliance to Hardy-Weinberg equilibrium was checked using SAS/Genetics (SAS Institute), and all SNPs in both cases and controls survived a cutoff of P <0.05. For haplotype estimation we used snphap (written by David Clayton, Cambridge Institute for Medical Research), SNPEM (written by Nicholas Schork (University of California, San Diego) and M. Daniele Fallin (Johns Hopkins University)) and PHASE version 2.11 (Matthew Stephens, University of Washington, Seattle). The haplotype analysis strategy was, first, to obtain haplotype estimates using the Expectation Maximization (EM) or Gibbs sampling algorithm; second, to identify htSNPs representing a minimal informative set within a region of linkage disequilibrium and third, to assess these for significant association with AMD. All P-values are two-tailed, and X2 values are presented as asymptotic significance. Overall type I error rates (α) were retrospectively calculated using a previously described method30 and are below 2 × 10−3. Significant haplotypes were subjected to permutation testing in both SNPEM and PHASE. The protective SNP model in Figure 2a was presented to Exemplar 2.2 and statistically evaluated by that software for fitness against the three datasets (Iowa, Columbia and combined) in Figure 2b. Generation of the genetic algorithm (GA)-derived model (Fig. 2c) involved Exemplar software. The GA options were set to 1,500 AND/OR models of 15 iterations each, with a model size no larger than 5 (which permits 16 possible genotypes). Further details of the genetic algorithm implementation and significance testing are available in the Supplementary Methods.

A Classification & Regression Tree Analysis was performed with SPSS (v 14.0) with the appropriate module on the Columbia, Iowa and combined data recoded as with (+) or without (−) minor alleles. Models were automatically generated using each of the three data sets that incorporated both CFH and C2/BF loci as contributors to the dependent outcome.

Accession codes

GenBank: complement factor B (properdin, BF): mRNA, AF349679; protein, AAK30167. Complement component 2 (C2): mRNA, AK222537; protein, BAD96257. dbSNP identification numbers for genotyped SNPs are provided in Table 2.

URLs

Resequencing data used for analysis of all missense alleles with frequencies >2% in European populations is available at http://pga.mbt.washington.edu/. The list of SNPs included on the Affymetrix Mapping 100K Array is available at https://www.affymetrix.com/analysis/netaffx/index.affx. snphap is available from http://www-gene.cimr.cam.ac.uk/clayton/software/. PHASE version 2.11 is available at http://www.stat.washington.edu/stephens/software.html. Linkage disequilibrium was assessed using the graphical tools available at http://www.innateimmunity.net. Overall type I error rates (α) were calculated using a previously described method30 implemented at https://innateimmunity.net/IIPGA2/Bioinformatics/multipletestfdrform. Exemplar 2.2 is available at http://www.sapiosciences.com. Further details of the genetic algorithm implementation and significance testing are available at http://www.sapiosciences.com/papers/AMDSupplement.pdf.

Supplementary Material

Supplementary Figure

Supplementary Methods

Acknowledgments

We thank D. Anderson, L. Johnson, D. Bok, and P. Dudley for helpful discussions. We also acknowledge J. Sharp, T. Krezowik, T. Weingeist, C. Boldt, J. Folk, T. Johnson, M. Wilkinson, D. Zumbro, P. Gouras, W. Moscoso, C. McAvoy, S. Thompson, L. Arbisser and A. Arbisser for their assistance in recruiting patients; S. Baruah, R. Wolfe, S. McCormick, J. Donahue, A. Olsh, L. Buckta and M. Busuioc for technical assistance and Vision Share, the Iowa Lions Eye Bank and the Central Florida Lions Eye Banks for their efforts in procuring eyes from human donors. We are especially grateful to those individuals and families who unselfishly donated their time and/or the eyes of their loved ones to this research program. This work was supported in part by the US National Institutes of Health (NIH) (grants EY13435 (R.A.) and EY11515 (G.S.H.)), New York Community Trust (R.T.S.), Wallach Foundation (R.A., G.R.B.), Elyachar Foundation (R.A., G.R.B.), Kaplen Foundation (R.A., G.R.B.), Widgeon Point Charitable Foundation (R.A., J.E.M.), Macula Foundation (R.A.), the International Retina Research Foundation (G.S.H.), the American Macular Degeneration Foundation, Inc. (G.S.H.), the Eye Research Institute (G.S.H.), the Intramural Research Program of the NIH and the National Cancer Institute; by federal funds from the National Cancer Institute of the National Institutes of Health (contract NO1-CO-124000) and by unrestricted grants from Research to Prevent Blindness, Inc., to the University of Iowa Department of Ophthalmology and Visual Sciences and to the Department of Ophthalmology, Columbia University. G.S.H. currently holds an honorary professorship in the School of Medicine, Queen’s University, Belfast. The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products or organizations imply endorsement by the US Government.

Footnotes

Note: Supplementary information is available on the Nature Genetics website.

AUTHOR CONTRIBUTION STATEMENT

The AMD Genetics Clinical Study Group includes Stanley Chang, Lawrence A. Yannuzzi, John C. Merriam and Irene Barbazetto (Department of Ophthalmology, Columbia University, New York); Leonid E. Lerner (F.M. Kirby Center for Molecular Ophthalmology, Scheie Eye Institute, University of Pennsylvania School of Medicine, Philadelphia) and Stephen Russell, Jamal Hoballah, Jill Hageman and Heather Stockman (Department of Ophthalmology and Visual Sciences, Center for Macular Degeneration, University of Iowa, Iowa City, Iowa, USA).

COMPETING INTERESTS STATEMENT

The authors declare competing financial interests (see the Nature Genetics website for details).

Reprints and permissions information is available online at http://npg.nature.com/reprintsandpermissions/

References

1. van Leeuwen R, Klaver CC, Vingerling JR, Hofman A, de Jong PT. Epidemiology of age-related maculopathy: a review. Eur J Epidemiol. 2003;18:845–854. [PubMed]
2. Klein R, Peto T, Bird A, Vannewkirk MR. The epidemiology of age-related macular degeneration. Am J Ophthalmol. 2004;137:486–495. [PubMed]
3. Hageman GS, et al. A common haplotype in the complement regulatory gene factor H (HF1/CFH) predisposes individuals to age-related macular degeneration. Proc Natl Acad Sci USA. 2005;102:7227–7232. [PubMed]
4. Edwards AO, et al. Complement factor H polymorphism and age-related macular degeneration. Science. 2005;308:421–424. [PubMed]
5. Haines JL, et al. Complement factor H variant increases the risk of age-related macular degeneration. Science. 2005;308:419–421. [PubMed]
6. Klein RJ, et al. Complement factor H polymorphism in age-related macular degeneration. Science. 2005;308:385–389. [PMC free article] [PubMed]
7. Zareparsi S, et al. Strong association of the Y402H variant in complement factor H at 1q32 with susceptibility to age-related macular degeneration. Am J Hum Genet. 2005;77:149–153. [PubMed]
8. Conley YP, et al. Candidate gene analysis suggests a role for fatty acid biosynthesis and regulation of the complement system in the etiology of age-related maculopathy. Hum Mol Genet. 2005;14:1991–2002. [PubMed]
9. Anderson DH, Mullins RF, Hageman GS, Johnson LV. A role for local inflammation in the formation of drusen in the aging eye. Am J Ophthalmol. 2002;134:411–431. [PubMed]
10. Hageman GS, et al. An integrated hypothesis that considers drusen as biomarkers of immune-mediated processes at the RPE-Bruch’s membrane interface in aging and age-related macular degeneration. Prog Retin Eye Res. 2001;20:705–732. [PubMed]
11. Mullins RF, Russell SR, Anderson DH, Hageman GS. Drusen associated with aging and age-related macular degeneration contain proteins common to extracellular deposits associated with atherosclerosis, elastosis, amyloidosis, and dense deposit disease. FASEB J. 2000;14:835–846. [PubMed]
12. Johnson LV, Leitner WP, Staples MK, Anderson DH. Complement activation and inflammatory processes in Drusen formation and age related macular degeneration. Exp Eye Res. 2001;73:887–896. [PubMed]
13. Crabb JW, et al. Drusen proteome analysis: an approach to the etiology of age-related macular degeneration. Proc Natl Acad Sci USA. 2002;99:14682–14687. [PubMed]
14. Bok D. Evidence for an inflammatory process in age-related macular degeneration gains new support. Proc Natl Acad Sci USA. 2005;102:7053–7054. [PubMed]
15. Morgan BP, Walport MJ. Complement deficiency and disease. Immunol Today. 1991;12:301–306. [PubMed]
16. Kinoshita T. Biology of complement: the overture. Immunol Today. 1991;12:291–295. [PubMed]
17. Holers VM, Thurman JM. The alternative pathway of complement in disease: opportunities for therapeutic targeting. Mol Immunol. 2004;41:147–152. [PubMed]
18. Lokki ML, Koskimies SA. Allelic differences in hemolytic activity and protein concentration of BF molecules are found in association with particular HLA haplotypes. Immunogenetics. 1991;34:242–246. [PubMed]
19. Davis CA, Forristal J. Partial properdin deficiency. J Lab Clin Med. 1980;96:633–639. [PubMed]
20. Raum D, Glass D, Carpenter CB, Schur PH, Alper CA. Mapping of the structural gene for the second component of complement with respect to the human major histocompatibility complex. Am J Hum Genet. 1979;31:35–41. [PubMed]
21. Alper CA, et al. Immunoglobulin deficiencies and susceptibility to infection among homozygotes and heterozygotes for C2 deficiency. J Clin Immunol. 2003;23:297–305. [PubMed]
22. Stewart CA, et al. Complete MHC haplotype sequencing for common disease gene mapping. Genome Res. 2004;14:1176–1187. [PubMed]
23. Larsen CE, Alper CA. The genetics of HLA-associated disease. Curr Opin Immunol. 2004;16:660–667. [PubMed]
24. Myers S, Bottolo L, Freeman C, McVean G, Donnelly PA. Fine-scale map of recombination rates and hotspots across the human genome. Science. 2005;310:321–324. [PubMed]
25. Goverdhan SV, et al. Association of HLA class I and class II polymorphisms with age-related macular degeneration. Invest Ophthalmol Vis Sci. 2005;46:1726–1734. [PubMed]
26. Bird AC, et al. An international classification and grading system for age-related maculopathy and age-related macular degeneration. The International ARM Epidemiological Study Group. Surv Ophthalmol. 1995;39:367–374. [PubMed]
27. Klaver CC, et al. Incidence and progression rates of age-related maculopathy: the Rotterdam Study. Invest Ophthalmol Vis Sci. 2001;42:2237–2241. [PubMed]
28. Allikmets R, et al. Mutation of the Stargardt disease gene (ABCR) in age-related macular degeneration. Science. 1997;277:1805–1807. [PubMed]
29. Hayashi M, et al. Evaluation of the ARMD1 locus on 1q25–31 in patients with age-related maculopathy: genetic variation in laminin genes and in exon 104 of HEMI-CENTIN-1. Ophthalmic Genet. 2004;25:111–119. [PubMed]
30. Benjamini Y, Hochberg Y. Controlling the false discovery rate—a practical and powerful approach to multiple testing. J R Stat Soc Ser B. 1995;57:289–300.