Nature Genetics PDF - Myosin IXB Variant Increases Celiac Disease Risk

LETTERS Myosin IXB variant increases the risk of celiac disease and © 2005 Nature Publishing Group http://www.nature.com/naturegenetics points toward a primar...

LETTERS Myosin IXB variant increases the risk of celiac disease and © 2005 Nature Publishing Group http://www.nature.com/naturegenetics points toward a primary intestinal barrier defect Alienke J Monsuur1, Paul I W de Bakker2, Behrooz Z Alizadeh1, Alexandra Zhernakova1, Marianna R Bevova1, Eric Strengman1, Lude Franke1, Ruben van’t Slot1, Martine J van Belzen1,7, Ineke C M Lavrijsen1, Begoña Diosdado1, Mark J Daly2, Chris J J Mulder3, M Luisa Mearin4, Jos W R Meijer5, Gerrit A Meijer6, Erica van Oort1, Martin C Wapenaar1, Bobby P C Koeleman1 & Cisca Wijmenga1 Celiac disease is probably the best-understood immune-related the gene myosin IXB (MYO9B), which encodes an unconven- disorder. The disease presents in the small intestine and results tional myosin molecule that has a role in actin remodeling of from the interplay between multiple genes and gluten, the epithelial enterocytes2,3. Individuals homozygous with respect triggering environmental factor1. Although HLA class II genes to the at-risk allele have a 2.3-times higher risk of celiac explain 40% of the heritable risk, non-HLA genes accounting disease (P = 1.55
105). This result is suggestive of a primary for most of the familial clustering have not yet been identified. impairment of the intestinal barrier in the etiology of celiac Here we report significant and replicable association disease, which may explain why immunogenic gluten peptides (P = 2.1
106) to a common variant located in intron 28 of are able to pass through the epithelial barrier. a b 0.000001 19-kb block rs12996130 rs2305756 rs1979261 rs1060367 rs7254359 rs2279008 rs2305767 rs2279004 rs2279003 rs2305765 rs2305764 rs1043963 rs962917 rs891201 rs891205 0.00001 0.0001 96 91 56 17 14 91 95 70 72 91 14 20 0 2 80 98 94 97 95 70 63 99 8 1 21 22 79 35 99 90 97 91 64 93 P 0.001 1 12 10 34 35 10 95 90 60 64 93 7 0 68 15 0 31 95 74 60 28 0 7 37 20 25 31 91 68 83 0.01 5 9 9 22 24 46 60 6 1 44 22 25 21 28 0 3 43 34 6 11 0.1 4 2 1 15 31 2 4 15 8 11 17 29 1 18 14 12 MYO9B Figure 1 Fine-mapping study in and surrounding MYO9B. (a) The association data are plotted in blue (circles) on a logarithmic scale for each of the 15 SNPs across a 191-kb region (17,030,936–17,222,116 bp) encompassing MYO9B and tested in 216 individuals with Marsh III celiac disease and 216 controls (set 1). Subsequent follow-up studies in a second cohort of 247 affected individuals and 470 controls (set 2) of the six SNPs with P o 0.05 are plotted in green (squares). The association data from the combined cohorts (463 affected individuals and 686 controls) are plotted in red (triangles). The spacing between SNPs reflects the distances between them. (b) Pairwise LD between the 15 SNPs, given by the D’ statistics computed with the genotype data from the 216 controls. The overall LD structure of these data is very similar to that of the data on a European population in the HapMap database (Centre d’Etude du Polymorphisme Humain; Utah residents with ancestry from northern and western Europe), where all associated SNPs fall into a single 22-kb haplotype block (defined as previously described24). The darker shade of red indicates the higher D’. 1Complex Genetics Section, DBG-Department of Medical Genetics, University Medical Center Utrecht, Universiteitsweg 100, 3584 CG Utrecht, the Netherlands. 2Broad Institute of Harvard and MIT, 320 Charles St., Cambridge, Massachusetts 02141, USA; and Center for Human Genetics Research, Massachusetts General Hospital, 185 Cambridge Street, Boston, Massachusetts 02114, USA. 3Department of Gastroenterology, VU University Medical Center, P.O. Box 7057, 1007 MB Amsterdam, The Netherlands. 4Department of Pediatrics, Leiden University Medical Center, P.O. Box 9600, 2300 RC Leiden, and VU University Medical Center, P.O. Box 7057, 1007 MB Amsterdam, The Netherlands. 5Department of Pathology, Rijnstate Hospital, P.O. Box 9555, 6800 TA Arnhem, The Netherlands. 6Department of Pathology, VU University Medical Center, P.O. Box 7057, 1007 MB Amsterdam, The Netherlands. 7Present address: Laboratory of DNA Diagnostics, Leiden University Medical Center, P.O. Box 9503, 2300 RA Leiden, The Netherlands. Correspondence should be addressed to C.W. ([email protected]). Received 29 June; accepted 27 September; published online 13 November 2005; doi:10.1038/ng1680 NATURE GENETICS VOLUME 37 [ NUMBER 12 [ DECEMBER 2005 1341  LETTERS Table 1 Names, chromosomal locations and P values for the 15 SNPs tested in and around MYO9B SNP rs2305756 rs1979261 rs1060367 rs7254359 rs12996130 rs2279008 rs2305767 rs2279004 rs962917 rs2279003 rs2305765 rs2305764 rs891201 rs891205 rs1043963 Position (bp) 17,030,936 17,054,518 17,076,202 17,100,566 17,140,752 17,144,303 17,155,296 17,157,679 17,163,247 17,167,031 17,173,992 17,174,833 17,199,099 17,215,586 17,222,116 P Set 1 0.7836 0.9885 0.4085 0.4198 0.4741 0.3114 0.0032* 0.9049 0.0186 0.0069 0.0089 0.0121 0.4832 0.6614 0.0205 Set 2 0.0412 0.0022* 0.0151 0.0005* 0.0003* 0.1233 Set 1 and set 2 0.0003* 0.000049* 0.0003* 0.0000097* 0.0000021* 0.0095 *P values still significant after correcting for multiple testing. P values were calculated in Haploview using the w2 test. The locations of the SNPs in MYO9B are given in Supplementary Figure 2. © 2005 Nature Publishing Group http://www.nature.com/naturegenetics With a prevalence close to 1% (ref. 4), celiac disease is the most correcting for six independent tests (nominal Bonferroni P o 0.05/6 ¼ common food intolerance in general western populations. In indivi- 8
103; Fig. 1 and Table 1). Combining the cohorts (set 1 and set 2) duals with celiac disease, ingestion of gluten leads to inflammation strengthened the association considerably and showed highly signifi- and tissue remodeling of the intestinal mucosa, resulting in malnutri- cant association for all five SNPs in MYO9B (Fig. 1 and Table 1). We tion and severe complications. Long regarded as a gastrointestinal observed the smallest P value for rs2305764 (P ¼ 2.1
106) located disorder of childhood, the disease is now considered to be a chronic in intron 28. Our data showed strong linkage disequilibrium (LD) systemic autoimmune disease5 and is more often diagnosed in adults between the five associated SNPs (Fig. 1b), which are located in a than in children6. Celiac disease has a strong heritable component, single, 19-kb haplotype block spanning exons 15–27 of MYO9B although the inheritance is complex and multifactorial1. Gluten is the (Supplementary Fig. 2 online). key environmental risk factor, and HLA class II genes are associated to To rule out the possibility that our initial study (comprising only celiac disease7. HLA-DQ2 has a key role in the disease by presenting five microsatellite markers11) had overlooked an association signal gluten peptides to CD4+ cells in the lamina propria8. Much of the elsewhere in the CELIAC4 locus, we carried out a comprehensive research on celiac disease has focused on the activation and regulation fine-mapping study using 359 tag SNPs with minor allele frequencies of gluten-reactive T cells. But the etiological steps preceding this T-cell 4 0.02 and r2 4 0.7, and covering the lod-1.5 region (99% activation are still poorly understood; for example, why are the confidence interval (c.i.); 15,385,880–21,075,237 bp) on set 2 supple- antigenic gluten peptides resistant to further breakdown in the mented with the cases from set 1 (totaling 463 independent cases and intestinal lumen, and how do the gluten peptides pass through the 470 controls). After carrying out quality checks on the raw data, we epithelial barrier? The integrity of the intestinal barrier is impaired in excluded 60 poorly performing or monomorphic tag SNPs (Supple- active celiac disease9,10, which implies that the epithelial cell barrier mentary Table 3 online), yielding 299 tag SNPs for further analysis, of has a role in the early pathogenesis of the disease. which 19 were located in the 191-kb region covered previously. Three Genome-wide screens have been done to identify non-HLA genes SNPs were overlapping in the two screens (rs7254359, rs2279008 and involved in celiac disease but, so far, no gene has been positionally rs2305767; Supplementary Table 2). After excluding tag SNPs that cloned. We previously obtained strong evidence for linkage (multiple showed deviation from Hardy-Weinberg equilibrium, we were able to maximum lod score 4.43, nominal P ¼ 6.2
106) to chromosome analyze 291 tag SNPs successfully. These allowed us to capture most of 19p13.1 in celiac disease (CELIAC4) in a cohort of affected sibling the untyped genetic variation present in the Centre d’Etude du pairs of European descent from the Netherlands11. This chromosomal Polymorphisme Humain population of the HapMap Project (Supple- location was also suggested by meta and pooled analyses of European mentary Table 3). We observed a single, strong peak of association in celiac disease data that did not include the Dutch cohort12. In our ori- MYO9B (Fig. 2, Table 2 and Supplementary Table 3). The most- ginal study11, further association analysis was done with five micro- associated tag SNP (rs1457092, P ¼ 7.8
105), located between satellite markers spanning the 3.5-Mb lod-1 interval in a cohort of rs962917 and rs2279003, remained significant after stringent Bonfer- 216 case-control pairs, resulting in weak association to D19S899 (P ¼ roni correction for 291 tag SNPs (Pc o 0.05/291 ¼ 1.72
104) and 1.3
103), located in intron 1 of the gene myosin IXB (MYO9B). is in strong LD with the five associated SNPs identified in the random Here we focused on the region surrounding D19S899 and, more screen described above (rs2305767, rs962917, rs2279003, rs2305765 specifically, on MYO9B using the same 216 cases and 216 controls (set 1; Supplementary Fig. 1 and Supplementary Table 1 online). 0.0001 Initially, we typed 15 SNPs with an average spacing of 13 kb across rs1457092 MYO9B. Using a conservative Bonferroni correction, a single SNP 0.001 (rs2305767) located in intron 14 of MYO9B showed significant P association (P ¼ 0.0032; Fig. 1a, Table 1 and Supplementary 0.01 Table 2 online), confirming our initial observation. Another five 0.1 SNPs proximal to rs2305767 also showed significant association without correction (P o 0.05; Table 1); these SNPs allowed us to 1 15,523,544 21,004,252 refine the region of association, as they were located in the 3¢ part of bp MYO9B, with one SNP (rs1043963) located 36.4 kb downstream of the gene. A replication study of the six SNPs with P o 0.05 was done in a Figure 2 Comprehensive fine-mapping study of the CELIAC4 region on second, fully independent, cohort of 247 unrelated individuals with 19p13.1. P values are plotted for all 291 tag SNPs within the 99% c.i. Marsh III celiac disease who met the same strict diagnostic criteria11, (15,385,880–21,075,237 bp, in chromosomal order) in a cohort of 463 independent individuals with Marsh III celiac disease and 470 controls. and 470 Dutch blood bank donors (set 2). We observed significant P values were calculated in Haploview using the w2 test. The P value cutoff association (P o 0.05) for all five SNPs in MYO9B but not for the after Bonferroni correction was 1.72
104. Only SNP rs1457092, located SNP downstream of the gene (rs1043963, P ¼ 0.123). Three SNPs in intron 20 of MYO9B and indicated with an arrow, was significant after (rs962917, rs2305765 and rs2305764) were still significant after correction for multiple testing. 1342 VOLUME 37 [ NUMBER 12 [ DECEMBER 2005 NATURE GENETICS  LETTERS Table 2 Names, chromosomal locations and P values for the 11 tag SNPs from the comprehensive screen located in MYO9B (17,072,928– 17,185,703 bp) SNP rs7246865 rs4808571 rs7254359 rs1870068 rs10409461 rs11673417 rs2279009 rs2279008 rs7259292 rs2305767 rs1457092 Position (bp) 17,080,105 17,087,229 17,100,566 17,119,828 17,129,784 17,131,499 17,143,899 17,144,303 17,153,014 17,155,296 17,165,236 P 0.3022 0.4125 0.9009 0.4977 0.7905 0.6813 0.4592 0.1617 0.7726 0.0008 0.000078* *P value was still significant after correcting for multiple testing. Details are shown in Supplementary Table 3. and rs2305764; Supplementary Fig. 3 online). One of the three 95% c.i. ¼ 1.23–2.13; P ¼ 5.3
104), whereas individuals homo- © 2005 Nature Publishing Group http://www.nature.com/naturegenetics overlapping SNPs (rs2305767 located in intron 14) was already zygous with respect to the A allele have a risk of developing celiac significant in the initial cohort (P ¼ 3.2
103) and showed even disease that increases to 2.27 (95% c.i. ¼ 1.56–3.30; P ¼ 1.55
105), stronger association in the combined cohorts (set 1 and set 2, with population-attributable risks of 25% and 23%, respectively. comprising 463 cases and 686 controls; P ¼ 3
104; Table 1). These results suggest that the 3¢ part of MYO9B is associated with Haplotype analysis with these six SNPs showed that in this 19-kb celiac disease. region, only four haplotypes account for 497% of all the observed MYO9B is a good candidate gene for celiac disease because of the haplotypes; these can be captured by typing only three SNPs function of its encoded protein, and it may be one of the long-sought (rs2305767, rs1457092 and rs2305764). The AAACTA haplotype was factors involved in the early mucosal events preceding the well-under- present in 39% of cases and 30% of controls (odds ratio (OR) ¼ 1.56; stood inflammatory response. MYO9B encodes a single motor protein2 95% c.i. ¼ 1.27–1.93, P ¼ 1.8
105; Table 3). The GGCTCG haplo- belonging to the class IX myosin molecules, which are unique in type is the most low-risk haplotype and is very unlikely to encompass comparison with other classes because they contain a Rho-GTPase- the unknown causal disease variant. We were able to narrow the activating domain within their tails. This GTPase activity converts haplotype to a single SNP, as we observed that the presence of the A active Rho-GTP into inactive Rho-GDP, thereby downregulating Rho- allele of SNP rs2305764 constantly predisposes to a higher risk of celiac dependent signaling pathways3. Rho-family GTPases are involved in disease, independent of the variation in the other five SNPs. This remodeling of the cytoskeleton and tight junction assembly, both of observation was confirmed by logistic regression analysis using a con- which result in enhanced epithelial paracellular permeability13,14. It is ditional forward selection procedure showing that only SNP rs2305764 therefore tempting to speculate that a genetic variant in the 3¢ part of was left in the regression model (data not shown) and, similarly, that MYO9B leads to an impaired interaction with RhoA, thereby perturb- none of the haplotypes showed stronger association. Therefore, the ing tight junction gate and fence function. Hence, a subtle, underlying rs2305764 SNP located in intron 28 of MYO9B can alone completely intestinal barrier abnormality may be involved in the etiology of celiac explain the association observed at MYO9B and can be considered a disease, which is in line with the recent observation of intestinal marker for celiac disease risk. Sequence analysis of the entire MYO9B permeability in individuals with celiac disease with normal histol- coding region in 16 affected individuals homozygous with respect to ogy15,16. As a consequence, immunogenic gluten peptides can enter the the at-risk haplotype identified 22 variants, 2 of which were already deeper mucosal layer more easily. Notably, this is the site at which the included in our study. Of the 20 new SNP variants, 2 were considered HLA-DQ2–mediated antigen presentation to the CD4+ cells initiates relevant (i.e., nonsynonymous SNPs) and were typed in our entire the inflammatory response. So far, MYO9B is the only non-HLA gene cohort of 463 independent cases and 686 independent controls. identified for celiac disease by positional cloning. It will be interesting Unfortunately, neither of these two SNPs turned out to be more to determine the effect of this gene in other populations with celiac strongly associated than rs2305764, which agrees with the observation disease. The identification of MYO9B as a susceptibility gene in celiac that these two SNPs are not in LD with the at-risk haplotype disease is a notable finding that may open new avenues for studying the (Supplementary Table 4 and Supplementary Fig. 4 online). early events of celiac disease pathogenesis, a process that has not yet The genotype frequency data for rs2305764 showed that the risk received much attention, but which might prove important in devel- of celiac disease rose in proportion to the number of copies of the A oping alternative treatments to the strict gluten-free diet currently used. allele (frequency of 46.5% in affected individuals compared with 37.9% in controls), implying a codominant allele-dose effect. Indivi- METHODS duals heterozygous with respect to the A allele have a modest but Subjects and controls. DNA, isolated from whole blood, was available from significantly higher risk of developing celiac disease (OR ¼ 1.66; two independent cohorts of Dutch individuals with celiac disease (set 1, 216; set Table 3 The prevalence of MYO9B haplotypes reconstructed from selected SNPs and their association to celiac disease MYO9B haplotypes rs2305767 rs962917 rs1457092 rs2279003 rs2305765 rs2305764 Cases (%)a Controls (%)a ORb 95% c.i.b Pc G G C T C G 308 (37.3) 565 (45.5) 1.00d – – A G C C C G 128 (15.5) 216 (17.4) 1.09 0.84–1.41 0.53 A A A C T A 319 (38.7) 375 (30.2) 1.56 1.27–1.93 0.000018 A G C C T A 46 (5.6) 67 (5.4) 1.26 0.84–1.88 0.26 Rare haplotypes 24 (2.9) 20 (1.6) 2.21 1.24–4.05 0.009 aNumber represents the frequency of haplotypes estimated using an expectation maximization algorithm embedded log linear model21,22. bOR represents maximum likelihood estimate of odds ratio, and the corresponding 95% c.i. was approximated using Woolf’s method. cP values were calculated using a w2 test. dThis haplotype was taken as the reference. NATURE GENETICS VOLUME 37 [ NUMBER 12 [ DECEMBER 2005 1343  LETTERS 2, 247). All the affected individuals were diagnosed in accordance with the The HapMap database and the Haploview program are available at http:// revised ESPGHAN criteria17. More than 90% of the affected individuals were www.hapmap.org/. HLA-DQ2-positive (Supplementary Table 1). The initial biopsy specimens of the individuals were retrieved; all showed a Marsh III lesion upon reevaluation Accession codes. GenBank: Homo sapiens chromosome 19 complete sequence, by one of two experienced pathologists (G.A.M. and J.W.R.M.). Both cohorts NC_000019; H. sapiens MYO9B mRNA, NM_004145. Ensembl: transcript included children and adults. There were also two cohorts of controls available. MYO9B, ENST00000319396. Set 1 controls (n ¼ 216) comprised random hospital controls. Set 2 controls Note: Supplementary information is available on the Nature Genetics website. (n ¼ 470) were random blood bank donors. All cases and controls were from The Netherlands and of European descent, and at least three of their four ACKNOWLEDGMENTS grandparents were also born in The Netherlands. The comprehensive screen We thank the affected individuals, their physicians and the Dutch Celiac included all individuals in set 2 supplemented with the cases from set 1. Disease Foundation for participating in this study; C. Gouw, A. Bardoel and © 2005 Nature Publishing Group http://www.nature.com/naturegenetics Supplementary Table 1 lists the baseline demographic parameters of the H. van Someren for practical assistance and J. Senior for critically reading the cohorts. All individuals gave their informed consent. This study was approved manuscript. This study was supported by grants from The Netherlands by the Medical Ethical Committee of the University Medical Center Utrecht. Organization for Scientific Research, the Dutch Digestive Disease Foundation and the Celiac Disease Consortium. Random screen: SNP selection and genotyping. We selected 15 random SNPs from Applied Biosystems, covering MYO9B and its surroundings. These SNPs COMPETING INTERESTS STATEMENT were obtained as Assays on Demand or Assays by Design (Applied Biosystems) The authors declare that they have no competing financial interests. and initially typed on set 1 (216 individuals with Marsh III celiac disease and 216 controls) using a 7900 Taqman (Applied Biosystems). In the follow-up Published online at http://www.nature.com/naturegenetics/ Reprints and permissions information is available online at http://npg.nature.com/ study, six SNPs were selected for further typing on set 2 (247 individuals with reprintsandpermissions/ Marsh III celiac disease and 470 controls). Comprehensive screen: tag SNP selection and genotyping. For the compre- 1. Van Heel, D.A., Hunt, K., Greco, L. & Wijmenga, C. Genetics in coeliac disease. Best hensive screen we took the 99% c.i. (1.5-lod) of the CELIAC4 linkage peak on Pract. Res. Clin. Gastroenterol. 19, 323–339 (2005). chromosome 19, ranging from 15,385,880 to 21,075,237 bp (National Center 2. Post, P.L. et al. Myosin-IXb is a single-headed and processive motor. J. Biol. Chem. 277, 11679–11683 (2002). for Biotechnology Information build 34). SNPs were selected by downloading 3. Muller, R.T., Honnert, U., Reinhard, J. & Bahler, M. The rat myosin myr 5 is a GTPase- all the SNPs typed in the Centre d’Etude du Polymorphisme Humain (Utah activating protein for Rho in vivo: essential role of arginine 1695. Mol. Biol. Cell 8, residents with ancestry from northern and western Europe) population in this 2039–2053 (1997). region from the HapMap database18. From these SNPs, the program Tagger19 4. Dube, C. et al. The prevalence of celiac disease in average-risk and at-risk Western European populations: a systematic review. Gastroenterology 128, S57–S67 (2005). was used to select tag SNPs so that all SNPs with a minor allele frequency Z2% 5. Rewers, M. Epidemiology of celiac disease: what are the prevalence, incidence, and were captured with r2 Z 0.7 (excluding SNPs with low Illumina quality design progression of celiac disease? Gastroenterology 128, S47–S51 (2005). scores). A final set of 359 tag SNPs was obtained for genotype analysis. 6. Green, P.H. The many faces of celiac disease: clinical presentation of celiac disease in We genotyped SNPs using the GoldenGate assay on an Illumina BeadStation the adult population. Gastroenterology 128, S74–S78 (2005). 7. Falchuk, Z.M. & Strober, W. HL-A antigens and adult coeliac disease. Lancet 2, 1310 500 GX (Illumina). All tag SNPs were examined for their resulting quality; all (1972). those that were not polymorphic in our population, had a low signal or had 8. Arentz-Hansen, H. et al. The intestinal T cell response to alpha-gliadin in adult celiac clusters that were too wide were excluded (Supplementary Table 3). We also disease is focused on a single deamidated glutamine targeted by tissue transglutami- carried out an evaluation of the tagging efficiency for two reasons. First, several nase. J. Exp. Med. 191, 603–612 (2000). 9. Schulzke, J.D., Schulzke, I., Fromm, M. & Riecken, E.O. Epithelial barrier and ion tag SNPs were dropped because of low-quality or bad clusters, because they were transport in coeliac sprue: electrical measurements on intestinal aspiration biopsy not in Hardy-Weinberg equilibrium in the controls, or because they were not specimens. Gut 37, 777–782 (1995). polymorphic in our population. Second, in the meantime, HapMap had 10. Van Elburg, R.M., Uil, J.J., Mulder, C.J. & Heymans, H.S. Intestinal permeability in expanded the number of SNPs typed in the region. We evaluated the tagging patients with coeliac disease and relatives of patients with coeliac disease. Gut 34, 354–357 (1993). efficiency of the 291 working tag SNPs using the following parameters (Supple- 11. Van Belzen, M.J. et al. A major non-HLA locus in celiac disease maps to chromosome mentary Table 3): aggressive tagging was used to get the most information; all 19. Gastroenterology 125, 1032–1041 (2003). the SNPs in the region were used; lod score threshold was set to 1; number of 12. Babron, M.C. et al. Meta and pooled analysis of European coeliac disease data. Eur. J. iterations was set to 1,000; maximum number of tries was set to 100,000. Hum. Genet. 11, 828–834 (2003). 13. Bruewer, M., Hopkins, A.M., Hobert, M.E., Nusrat, A. & Madara, J.L. RhoA, Rac1, and Statistical analysis. We calculated association w2 and two-tailed P values using Cdc42 exert distinct effects on epithelial barrier via selective structural and biochem- ical modulation of junctional proteins and F-actin. Am. J. Physiol. Cell Physiol. 287, the Haploview program20, for each stage of the study (set 1, set 2, set 1 and set C327–C335 (2004). 2, and the comprehensive screen). SNPs that were not in Hardy-Weinberg 14. Matter, K. & Balda, M.S. Signalling to and from tight junctions. Nat. Rev. Mol. Cell equilibrium (P o 0.001) in the controls were excluded from further analysis. Biol. 4, 225–236 (2003). We used multiple logistic regression analysis to estimate allelic and genotypic 15. Smecuol, E. et al. Permeability, zonulin production, and enteropathy in dermatitis herpetiformis. Clin. Gastroenterol. Hepatol. 3, 335–341 (2005). OR and the corresponding 95% c.i. for SNP rs2306764. To obtain genotype 16. Uil, J.J. et al. Follow-up of treated coeliac patients: sugar absorption test and intestinal information for the combined cohorts on the six SNPs comprising this 19-kb biopsies compared. Eur. J. Gastroenterol. Hepatol. 8, 219–223 (1996). region and showing P o 103 (rs2305767, rs962917, rs1457092, rs2279003, 17. United European Gastroenterology. When is a coeliac a coeliac? Report of a working rs2305765 and rs2305764), the associated tag SNP (rs1457092) was also typed group of the United European Gastroenterology Week in Amsterdam, 2001. Eur. J. Gastroenterol. Hepatol. 13, 1123–1128 (2001). for the 216 controls from set 1. As parental information was missing, an ex- 18. The International HapMap Consortium. The International HapMap Project. Nature pectation maximization algorithm was used to estimate haplotype frequencies 426, 789–796 (2003). in a multiple locus system21,22. This function recast the regression model as a 19. De Bakker, P.I. et al. Efficiency and power in genetic association studies. Nat. Genet., generalized linear mixed model with random effects, which is fitted as part of advance online publication 23 October 2005 (doi:10.1038/ng1669). 20. Barrett, J.C., Fry, B., Maller, J. & Daly, M.J. Haploview: analysis and visualization of LD the maximization step and allows testing for LD and disease association21,22. and haplotype maps. Bioinformatics 21, 263–265 (2005). The OR and the corresponding 95% c.i. were calculated for all haplotypes. 21. Chiano, M.N. & Clayton, D.G. Fine genetic mapping using haplotype analysis and the Population attributable risk was calculated for rs2306764 using the corresponding missing data problem. Ann. Hum. Genet. 62, 55–60 (1998). genotype frequencies of this marker as explained previously23. All the analyses 22. Bitti, P.P. et al. Association between the ancestral haplotype HLA A30B18DR3 and multiple sclerosis in central Sardinia. Genet. Epidemiol. 20, 271–283 (2001). were done using STATA statistical software, version 8.0 for MS Windows. 23. Greenland, S. & Rothman, R.K. Measures of effect and measures of association. in Modern Epidemiology 2nd edn. 47–67 (Lippincott-Raven, Philadelphia, 1998). URLs. The website for Applied Biosystems is http://myscience.appliedbiosystems. 24. Gabriel, S.B. et al. The structure of haplotype blocks in the human genome. Science com/. The program Tagger is available at http://www.broad.mit.edu/mpg/tagger/. 296, 2225–2229 (2002). 1344 VOLUME 37 [ NUMBER 12 [ DECEMBER 2005 NATURE GENETICS