Volume 3 Supplement 7
Association of KCNB1 to rheumatoid arthritis via interaction with HLA-DRB1
© Xiao et al; licensee BioMed Central Ltd. 2009
Published: 15 December 2009
With the rapid development of large-scale high-throughput genotyping technology, genome-wide association studies have become a popular approach to mapping genes underlying common human disorders. Some genes are discovered, but many more have not been. Because these genes were not initially identified, it is reasonable to assume that their main effect is weak. We propose a method to accommodate such a situation. It is applied to the Genetic Analysis Workshop 16 Problem 1 case-control data in which shared-epitope alleles of HLA-DRB1 show very strong association with rheumatoid arthritis. Because some previous functional studies have reported association of gene KCNB1 to rheumatoid arthritis, we evaluate whether the gene KCNB1 contributes to the genetics of rheumatoid arthritis in this data set. Fifteen single-nucleotide polymorphisms from this gene were chosen. The association of KCNB1 gene to rheumatoid arthritis seems to be moderate.
Conventional methods such as linear regression and logistic regression are widely used in genetic association tests in the presence of gene-gene or gene-environment interaction. In recent years several new methods have been developed. Based on Tukey's non-additivity model, Chatterjee et al.  proposed a model that makes use of "generalized association parameters" to map additional genes while allowing for gene-gene and gene-environment interactions. This method may have higher power to detect genetic variants than some existing methods. One drawback of their work is that it does not accommodate the situation in which the main effect of the second gene is bounded by constant times its interaction effect with the first gene. To overcome this drawback, Wang  proposed a method and derived the asymptotic distribution of the likelihood-ratio test statistic. This method allows the main effect of the second gene to be bounded but does not assume it to be zero. However, the method proposed by Wang is for a continuous phenotype. In this report, we generalized that method to case-control study so it can be applied to the Genetic Analysis Workshop 16 (GAW16) Problem 1 data.
We choose to study genes HLA-DRB1 using its shared-epitope alleles and KCNB1. The relationship between HLA-DRB1 and increased risk for rheumatoid arthritis has been studied for 20 years, and several high risk alleles of HLA-DRB1 are treated as shared-epitope alleles [3, 4], which are used widely in association study for rheumatoid arthritis . There are also some studies covering regions including genes PTPN22 and TRAF-C5, and at least two alleles of PTPN22 were associated with increased risk for rheumatoid arthritis . In a genome-wide study, TRAF-C5 was found to be a risk locus for rheumatoid arthritis . Unlike PTPN22 and TRAF-C5, which are supported by statistical evidence, there is some functional evidence supporting the active role of KCNB1 in rheumatoid arthritis. The functional channel in human T lymphocytes is composed of four identical KCNB1 sub-units, and several peptide inhibitor of KCNB1 have been developed as therapy for autoimmune diseases  such as type 1 diabetes mellitus and rheumatoid arthritis. An expression study  found that there is a downregulation of potassium channels including KCNB1 in autoimmune diseases.
Because the association signal of KCNB1 is not very strong (it has not been discovered by other genome-wide association studies), we conjecture that its effect may mainly be manifested through its interaction with the HLA-DRB1 gene. The goal of the current analysis is to investigate whether the gene KCNB1, a previously reported gene associated with rheumatoid arthritis, is associated with rheumatoid arthritis in this GAW16 dataset. For this purpose, we use two statistical techniques, including one particularly developed for this study, which may be helpful in mapping genes that have weak main effect.
The GAW16 Problem 1 data comes from the North American Rheumatoid Arthritis Consortium (NARAC) and includes 868 cases and 1194 controls. Single-nucleotide polymorphisms (SNPs) of all individuals were genotyped using 550 k Illumina chip (n = 545,080). A genome-wide association scan based on the trend test was performed to assess the association of all SNPs to rheumatoid arthritis.
The regular logistic regression of two-locus disease model with interaction is logit(π (x)|SE, SNP) = α0 + α1 × SE + α2 × SNP + α3 × SE × SNP, where SE is the number of shared-epitope alleles of HLA-DRB1 and SNP is the number of a chosen allele at an SNP of KCNB1, and the coefficients α2 and α3 measure the additional contribution of locus 2 over that of locus 1. A traditional test of both effects with two degrees of freedom is used to assess the association of the second locus, and the hypotheses to be tested are H0: α2 = α3 = 0 vs. H1: α2 ≠ 0 or α3 ≠ 0. This test treats the coefficients α2 and α3 as unrelated to each other because one of two coefficients can be 0 regardless of the values of the other one. However, this test may not be effective because the values of α2 and α3 could depend on whether there is association due to locus 2.
The situation of interest in this study is that main effect |α2|, is constrained by the interaction |α3| effect. The constraint is α3 ≠ 0, |α2| ≤ M × |α3|, with M a pre-specified constant, and then the hypotheses to test in this study are H0: α2 = α3 = 0 vs. H1 : α3 ≠ 0, |α2| ≤ M × |α3|. Based on this constraint, two extreme situations need to be considered. When M is set to be 0, α2 has to be 0 as well, and then the test measures interaction effect only (H1 : α2 = 0, α3 ≠ 0) with 1 df; when M is very large, α2 is not affected by α3, and then the test measures both effects (H1 : α2 ≠ 0 or α3 ≠ 0) with 2 df.
An appropriate value for M should achieve a balance between measuring interaction only and both effects. Based on some simulation results in linear regression , when M is in the range of [0.1, 0.3], the asymptotic quantiles of statistic based on this test seem to be rather different from those of the interaction-only test and the both-effects test. We set M to 0.4 in this analysis because the value of M should reflect the moderate association of KCNB1 and it should not be too small. Because the distribution of the likelihood-ratio statistic of this proposed test is unknown for case-control study, and in order to compare p-values, the estimation of the p-values need to be based on the same procedure. Ten thousand permutations are used to estimate the p-values in three tests for each SNP.
We also use principal-component analysis to test multiple SNPs in KCNB1 jointly. The principal components were obtained via the statistical package R (version 2.6.2) based on the correlation coefficient matrix.
The result of a genome-wide association scan is consistent with previous studies , with the strongest signal coming from HLA region on 6p21. For genes HLA-DRB1 and KCNB1, a p-value < 10-100 for HLA-DRB1 and the range of p-values for KCNB1 from 10-4 to 10-7 were set as thresholds for significance.
p-Values in three association tests based on 10,000 permutations
First principal component
The association between HLA-DRB1 gene and rheumatoid arthritis is very strong. Except for some known genes, a genome-wide scan found no other genes that show obvious association. One possibility is that other genes have a weak main effect, making them hard to detect. We propose a method for case-control study that allows the main effect of the second gene to be weak relative to its interaction effect with the first gene. Using this model, we studied the association between KCNB1 and rheumatoid arthritis.
The results of three tests used in this study support our assumption that the effect of KCNB1 may mainly be manifested through its interaction with the HLA-DRB1 gene.
The interaction-effect only test (M = 0) and the proposed test (M = 0.4) perform better than the both-effects test; for six SNPs, the proposed test even performs better than interaction-effect-only test.
The M value reflects the strength of main effect in the proposed model. M values of 0.1 and 0.4 represent small effect and moderate effect, respectively. Considering the moderate main effect of KCNB1, we set M value to be 0.4. The M value should be determined by researchers, every gene has different M value. A wrongly chosen M value will reduce the power in the test, so we suggest that a rough range of M values should be determined by some prior information, and then a number of interations be used to get the appropriate M value. Computing burden is a big concern in case-control study. When the number of permutations is larger than 10,000, comparisons cannot be made, so a more efficient algorithm for permutation needs to be developed. This method is more applicable in linear regression because the asymptotic distribution of the likelihood-ratio test statistic has been derived .
Previous studies have reported association of rheumatoid arthritis to KCNB1 gene. Based on our analysis, the association strength between KCNB1 and rheumatoid arthritis seems to be moderate in the GAW16 Problem 1 data.
We used two methods, including one developed by us, to investigate the association between KCNB1 gene and rheumatoid arthritis. Based on results, the strength of the association is moderate. This association needs further confirmation.
List of abbreviations used
Genetic Analysis Workshop 16
The Genetic Analysis Workshops are supported by NIH grant R01 GM031575 from the National Institute of General Medical Sciences. We thank Drs. Deborah Dawson, Trudy Burns, and Jian Huang for their valuable comments and suggestions. We also thank two anonymous reviewers for their valuable comments.
This article has been published as part of BMC Proceedings Volume 3 Supplement 7, 2009: Genetic Analysis Workshop 16. The full contents of the supplement are available online at http://www.biomedcentral.com/1753-6561/3?issue=S7.
- Chatterjee N, Kalaylioglu Z, Moslehi R, Peters U, Wacholder S: Powerful multilocus tests of genetic association in the presence of gene-gene and gene-environment interactions. Am J Hum Genet. 2006, 79: 1002-1016. 10.1086/509704.PubMed CentralView ArticlePubMedGoogle Scholar
- Wang K: Genetic association tests in the presence of epistasis or gene-environment interaction. Genet Epidemiol. 2008, 32: 606-614. 10.1002/gepi.20336.View ArticlePubMedGoogle Scholar
- Gregersen PK, Silver J, Winchester RJ: The shared epitope hypothesis. An approach to understanding the molecular genetics of susceptibility to rheumatoid arthritis. Arthritis Rheum. 1987, 30: 1205-1213. 10.1002/art.1780301102.View ArticlePubMedGoogle Scholar
- du Montcel ST, Michou L, Petit-Teixeira E, Osorio J, Lemaire I, Lasbleiz S, Pierlot C, Quillet P, Bardin T, Prum B, Cornelis F, Clerget-Darpoux F: New classification of HLA-DRB1 alleles supports the shared epitope hypothesis of rheumatoid arthritis susceptibility. Arthritis Rheum. 2005, 52: 1063-1068. 10.1002/art.20989.View ArticlePubMedGoogle Scholar
- Feitsma AL, Helm-van Mil van der AH, Huizinga TW, de Vries RR, Toes RE: Protection against rheumatoid arthritis by HLA: nature and nurture. Ann Rheum Dis. 2008, 67 (suppl 3): 61-63. 10.1136/ard.2008.098509.View ArticleGoogle Scholar
- Carlton VE, Hu X, Chokkalingam AP, Schrodi SJ, Brandon R, Alexander HC, Chang M, Catanese JJ, Leong DU, Ardlie KG, Kastner DL, Seldin MF, Criswell LA, Gregersen PK, Beasley E, Thomson G, Amos CI, Begovich AB: PTPN22 genetic variation: evidence for multiple variants associated with rheumatoid arthritis. Am J Hum Genet. 2005, 77: 567-581. 10.1086/468189.PubMed CentralView ArticlePubMedGoogle Scholar
- Plenge RM, Seielstad M, Padyukov L, Lee AT, Remmers EF, Ding B, Liew A, Khalili H, Chandrasekaran A, Davies LR, Li W, Tan AK, Bonnard C, Ong RT, Thalamuthu A, Pettersson S, Liu C, Tian C, Chen WV, Carulli JP, Beckman EM, Altshuler D, Alfredsson L, Criswell LA, Amos CI, Seldin MF, Kastner DL, Klareskog L, Gregersen PK: TRAF1-C5 as a risk locus for rheumatoid arthritis - a genomewide study. N Engl J Med. 2007, 357: 1199-1209. 10.1056/NEJMoa073491.PubMed CentralView ArticlePubMedGoogle Scholar
- Frolov RV, Berim IG, Singh S: Inhibition of delayed rectifier potassium channels and induction of arrhythmia: a novel effect of celecoxib and the mechanism underlying it. J Biol Chem. 2008, 283: 1518-1524. 10.1074/jbc.M708100200.View ArticlePubMedGoogle Scholar
- Qin D, Huang B, Deng L, El-Adawi H, Ganguly K, Sowers JR, El-Sherif N: Downregulation of K+ channel genes expression in type I diabetic cardiomyopathy. Biochem Biophys Res Commun. 2001, 283: 549-553. 10.1006/bbrc.2001.4825.View ArticlePubMedGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.