Objective To examine associations of HLA class I and class II alleles with ankylosing spondylitis (AS) in three cohorts of patients of European, Asian and African ancestry.
Methods HLA-A, HLA-B, HLA-C, HLA-DRB1, HLA-DQB1 and HLA-DPB1 alleles were genotyped in 1948 unrelated white and 67 African-American patients with AS from the Prospective Study of Outcomes in Ankylosing Spondylitis cohort, the North American Spondylitis Consortium and Australo-Anglo-American Spondyloarthritis Consortium, 990 white and 245 African-American Controls and HLA-B alleles in 442 Han Chinese patients with AS and 346 controls from Shanghai and Gansu, China. In addition to the case:control analyses, HLA-B*27-negative patients with AS were analysed separately, and logistic regression and ‘relative predispositional effects’ (RPE) analyses were carried out to control for the major effect of HLA-B*27 on disease susceptibility.
Results Although numerous associations were seen between HLA alleles and AS in whites, among HLA-B*27-negative patients with AS , positive associations were seen with HLA-A*29, B*38, B*49, B*52, DRB1*11 and DPB1*03:01 and negative associations with HLA-B*07, HLA-B*57, HLA-DRB1*15:01, HLA-DQB1*02:01 and HLA -DQB1*06:02. Additional associations with HLA-B*14 and B*40 (B60) were observed via RPE analysis, which excludes the HLA-B*27 alleles. The increased frequency of HLA-B*40:01 and decreased frequency of HLA-B*07 was also seen in Han Chinese and African-Americans with AS. HLA-B*08 was decreased in whites with acute anterior uveitis.
Conclusions These data, analysing the largest number of patients with AS examined to date in three ethnic groups, confirm that other HLA class I and II alleles other than HLA-B*27 to be operative in AS predisposition.
- disease susceptibility
Statistics from Altmetric.com
What is already known about this subject?
Although the association of HLA-B27 with ankylosing spondylitis (AS) is well established, as is the association with B60 (B*40), those of other HLA alleles that have been reported in smaller studies have been inconsistent or unconfirmed.
The role of HLA alleles other than HLA-B27 have not been studied in African-Americans.
In a large recent study imputing HLA alleles in whites from our group, a number of new HLA associations have been reported.
What does this study add?
In this study, the largest to date in whites and Asians and the first to date in African-Americans (other than HLA-B27) to employ direct HLA typing, a number of the HLA-associations we described by imputing HLA alleles in whites are confirmed in whites and in Chinese and African-American patients with AS.
A number of other HLA association not seen in the imputation analysis, including some seen in previous small studies, especially in the largest cohort of HLA-B27-negative patients with AS reported to date, are seen with other HLA-B, DRB1, DQB1 and DPB1 alleles, some reported in earlier smaller studies and other novel associations (though the latter because of possible power or stratification issues will need confirmation).
This study shows a role for HLA-class I alleles other than HLA-B27 and class II HLA genes in predisposition to AS and subsets thereof, independent of linkage with HLA-B27.
How might this impact on clinical practice or future developments?
This study gives the clinician a better understanding of the relative inputs of HLA alleles other than HLA-B27 in predisposition to AS and better informs how to interpret HLA-B27 typing results in blacks.
Studies of the contribution of major histocompatibility complex (MHC)-encoded variants to the heritability of ankylosing spondylitis (AS) suggest that it is responsible for 20.44% of the genetic risk for the disease, with over 114 non-MHC variants identified to date contributing another 7.38%.1 Other MHC region genes have also been implicated in addition to HLA-B*27, including HLA-B*14,2 3 HLA-B*60,4–8 HLA-DRB1, 9–11 HLA-DPB1, 11 12 MICA,13 14 TNF-alpha1,15 TAP1 16 and LMP2, 11 17 although the level of association of these findings has not been definitive and with the exception of HLA-B*60 replication has not been universal.
HLA class I genes (HLA-A, HLA-B and HLA-C) and HLA class-II genes (HLA-DRA1, HLA-DRB1, HLA-DQA1, HLA-DQB1, HLA-DPA1 and HLA-DPB1) encode cell-surface molecules that play an essential role in the immune defence against intracellular infections and in initiating an immune response to invading pathogens, respectively. Linkage disequilibrium (LD) with HLA-B*27 makes it difficult to resolve whether they play an independent role themselves, which would provide additional clues to AS pathogenesis, and potentially serve as biomarkers of diagnosis. Studies of direct HLA typing to date have been small and underpowered, providing inconsistent results.
Recently, we genotyped 7264 MHC single-nucleotide polymorphisms (SNPs) in 9069 AS cases and 13 578 population controls of European descent using the Illumina Immunochip microarray and controlling for the effects of HLA-B*27:02 and B*27:05, identified several other HLA-B allelic associations with AS, including significantly increased frequencies of HLA- B*13:02, B*40:01, B*40:02, B*47:01, B*51:01 and negative associations with HLA-B*07:02 and B*57:01 and association of HLA-A*02:01, HLA-DRB1*0:103 and HLA-DPB1. 18 19 In a Korean study, imputation was used to additionally show that HLA-C*15:02 is associated with AS susceptibility.19
The purpose of this study was, in the largest reported series of patients to date for directly genotyped HLA alleles, to examine the relative contributions of HLA-class I and class II alleles, analysing overall disease associations and conditioning on the presence of HLA-B*27 in three ethnic groups: 1948 whites of European ancestry, 442 Han Chinese and 67 African-Americans, the latter never having been examined for MHC associations (other than HLA-B*27 per se) previously.
White patients with AS in this study came mostly from the Prospective Study of Outcomes in Ankylosing Spondylitis (PSOAS) cohort,20 a multiethnic study conducted at four US academic institutions (The University of Texas Health-McGovern Medical School (UTH-H), Cedars-Sinai Medical Center, the University of California at San Francisco and the National Institutes of Health Clinical Center) and from The North American Spondylitis Consortium (NASC) (n=408).21 The black patients included 53 from the PSOAS study, 7 from NASC and 7 other either from patients followed from the outpatient clinics at UTH or referred by Dr Joel Taurog from the University of Texas Southwestern Medical School. All cases met modified New York criteria for AS.22 Controls were self-identified whites from the USA with no history of rheumatic disease obtained from unaffected non-consanguineous spouses or household or friend controls the Scleroderma Family Registry and DNA repository23 or the NASC21 (unrelated spouses or friends-75%) or from University of Texas-Houston Division Controls from Texas (25%). It should be noted that both the NASC and Scleroderma Family Registry controls and families came from all over the USA, although were enriched from Texas and California (NASC) and from Texas, Maryland, Pennsylvania and Michigan (Scleroderma Family Registry).23 Approximately one-third of the US patients with AS were probands in the NASC study and 2/3 came from the PSOAS study (from which approximately half came from California, 30% from Texas and 20% from the Mid-Atlantic region). All controls were screened by questionnaire for the presence of autoimmune disease or spondyloarthritis and were excluded if having such. For the HLA-B locus analyses, an additional 555 white British and Australian patients with AS from the Australo-Anglo-American Spondylitis Consortium were also analysed. HLA-B locus typing from 442 Han Chinese patients with AS and 346 unrelated Chinese controls were analysed by the same statistical approaches. Though some of the HLA-typing was reported previously by us,24 additional statistical analyses were carried out for the present study on a significantly larger AS cohort heretofore unreported. The patients with AS came from the clinics and hospitals in Shanghai and Jiangsu Province of China and the Chinese controls (who were free of any history of rheumatic disease) were obtained from a study project of Chinese population genetics at Fudan University in Shanghai. All subjects provided written informed consent. The study was approved by the Institutional Review Boards of all participating medical centres.
Single Stranded Conformational Polymorphism (SSCP) typing, of HLA-A, HLA-B and HLA-C alleles, was performed using commercially available kits (Dynal) on genomic DNA extracted from peripheral blood. HLA-DRB1, HLA-DQA1, HLA-DQB1 and HLA-DPB1 typing was performed by standard oligotyping techniques with high resolution HLA-DRB1 typing further achieved by sequence analysis of PCR-amplified HLA-DRB1 exon 2, except for 18 of the African-American patients with AS, where the MHC class II alleles were determined by SSCP typing using commercially available kits (One Lambda). Genomic DNA from the Chinese patients with AS and controls underwent the allele-specific PCR using primers supplied in the SeCore kits and then were followed by sequencing exons 2 and 3 of the HLA-B gene. The HLA SBT uTYPE 6.0 program (Life Technologies) was used in sequencing analysis and assigning HLA-B alleles.
The HLA-B genotyping in HLA-B*40 carriers was confirmed with sequence-based typing using SeCore Kits (Life Technologies, USA). The HLA SBT uTYPE 6.0 program (Life Technologies) was used in sequencing analysis.
We constructed 2×2 tables and tested the proportion of alleles in cases vs controls with adjusted Χ2 test using EPI-INFO (cdc.gov/epiinfo/index.html). Another method of ‘adjusting’ for the effect of HLA-B*27 at the HLA-B locus is to mask the HLA-B*27 alleles and analyse the remaining alleles for association with phenotype by comparing their frequencies (relative predispositional effects analysis (RPE)).25 26 Typically, RPEs are assessed by sequentially removing the alleles with the largest effect among those remaining.25 Such a procedure has no impact on loci other than HLA-B, as we are not removing individuals, but rather alleles. In addition to the univariable associations of each allele with AS, interaction effects between each allele and HLA-B*27 in relation to AS were examined using logistic regression models where the effects of each allele on AS were estimated for HLA-B*27-negative and positive patients separately. For the HLA-B analyses, this study has 94% power to detect an association with an additive OR of 1.5 for an allele with minor allele frequency of 0.1 at nominal significance.
The white patients in the PSOAS and NASC cohorts were strikingly similar in their clinical features, with psoriasis occurring in 10.4%, Crohn’s disease in 5.9%, ulcerative colitis in 3.5% and reactive arthritis in 5.0%. Minor differences between the two cohorts included a slightly lower frequency of HLA-B*27 (85.4%) and uveitis (37%) in the PSOAS cohort compared with 98.5% and 44%, respectively, in the probands from the NASC cohort (the latter enriched for multicase families). The prevalence of HLA-B*27 and uveitis is higher in familial AS.27 Among the 61 blacks with AS studied who had clinical information available, uveitis occurred in 37.7%, psoriasis in 10.2% and inflammatory bowel disease in 11.5%.
Examination of HLA-A, HLA-B and HLA-C allele frequencies in all white patients with AS and controls showed a number of highly significant associations at each locus (table 1). HLA-B*27 occurred in 87.8% of the white patients with AS compared with 7.6% of controls (p<1×10–9, OR=87.7, 95% CI 66.8 to 115.0), while HLA-B*27 homozygosity occurred in 3.5% of patients and 0.1% of controls (p<1×10–9, OR=35.9, 95% CI 4.96 to 258.0) (data not shown—table 1 shows allele frequencies). In addition to HLA-B*27, several other MHC class I alleles, including HLA-A*02, C*01 and C*02 were increased in frequency, whereas HLA-A*01, A*03, B*07, B*08, B*13, B*35, B*40, B*44, B*51, B*57, C*03, C*06 and C*07 were decreased in frequency (table 1). Examination of HLA-DRB1 and HLA -DQB1 allele frequencies in 790 patients and 704 controls demonstrated positive associations with HLA-DRB1*01:01, DRB1*01:03 and DRB1*04:04 as well as with DQB1*03:01 (in LD with DRB1*01:03), DQB1*03:02 (in LD with DRB1*04:04) and DQB1*05:01 (in LD with DRB1*01:01). Examination of HLA-DPB1 alleles in 635 patients and 385 controls revealed an association with HLA-DPB1*03:01. HLA-DRB1*15:01 and DRB1*03:01 and their respective linked alleles DQB1*06:02 and DQB1*02:01 were deceased in frequency in the white patients with AS.
Examining HLA-B*27-negative patients and controls, positive associations were seen with HLA-A*29, B*38, B*49 and B*52 as well as with HLA-DRB1*11 (table 2). On the other hand, HLA-B*07, B*57, DRB1*15:01, DQB1*02:01 and DQB1*06:02 were significantly decreased in frequency. The association were seen with HLA-DPB1*03:01 in the overall group persisted in the HLA-B*27-negative white patients with AS.
Removing the effect of the presence of HLA-B*27 using RPE analysis, new associations emerged with HLA-B*14 and B*40:01 (table 3) and the positive associations with HLA-B*38 and B*52 persisted, as did the negative associations with HLA-B*07 and B*57. Based on findings from HLA-B*27 stratified models for each allele, evidence for a significant interaction effect was observed for B*44 and B*49 (table 3).
Among the Han Chinese patients with AS, HLA-B*27 carriage was observed in 93.0% of the patients with AS and 7.5% of the controls (p<1×10–9, OR=163) (data not shown). Ten (2.3%) of the 442 Chinese patients were homozygous for HLA-B*27 compared with none of the 346 controls (p=0.003). Looking at allele frequencies (table 4), a higher OR was observed for HLA-B*27:04 than HLA-B*27:05, and negative associations were observed for several other HLA-B alleles, including HLA-B*07, B*13, B*15, B*35, B*46 and B*51. A significant interaction effect between B*58 and HLA-B*27 in relation to AS was found (OR=1.24 among HLA-B*27 (-) patients; OR=0.32 among HLA-B*27 (+) patients) (table 4).
Among black subjects, HLA-B*27 occurred in 40 of 67 patients with AS (59.7%) compared with 5 of 245 controls (2.0%) (p<1×10–9, OR=71.1) (table 5). HLA-B*40:01 occurred in 4 of the 67 patients with AS compared with 2 of 245 controls (p=0.03, OR=7.7) (Fisher’s exact). HLA-B*35 and B*42 were decreased in frequency. HLA-B*07 was also decreased in frequency, as seen in whites and Han Chinese, though the results were not significant (p=0.15) unless one applies a one tailed p value by Fisher’s exact testing (p=0.048, OR=0.46). Weak associations encountered with HLA-DRB1*08 and DQB1*03:01. The frequencies of IBD, psoriasis and reactive arthritis were too low for meaningful statistical analyses.
AAU occurred in 380 of 1389 with clinical data available (27.4%) (table 6). HLA-B*27 occurred in 92.9% compared with 85.0% of those without AAU (p=1.4×10–4, OR=2.3). Neither of the two most common HLA-B*27 subtypes seen in whites (B*27:02 and B*27:05) were selectively associated with AAU, independent of their association with AS. Among blacks with AAU, 78.3% were HLA-B*27-positive compared with 42.1% without AAU (p=0.013, OR=4.95). The frequencies of IBD, psoriasis and reactive arthritis were too low for meaningful statistical analyses.
HLA-B*27 is the major genetic association with AS in all major ethnic groups. Association of other MHC Class I and II loci with either AS specifically3–8 10–12 or spondyloarthritis in general2 9 have been suggested, but have been confounded due to LD between HLA-B and the MHC class II loci. In this study, by conducting stratified analyses of HLA-B*27, we have shown associations with other MHC alleles that cannot be attributed to LD with HLA-B*27 in a moderately large number of patients with AS from three ethnic groups. This study also analyses these alleles in the largest collection of HLA-B*27-negative patients with AS reported to date. We confirm by direct HLA typing some of the associations we previously reported by imputation.19
HLA-B60 is a serologically defined specificity that correlates at the DNA level with HLA-B*40:01.28 We were able to confirm the association with HLA-B*40:01 with AS in three ethnic groups. In this study, the largest to date employing direct HLA-typing, we were able to confirm the positive association with HLA-B60 (B*40), especially B*40:01, seen in other studies of whites by direct typing4–7 and by imputed alleles19 and of Taiwanese Chinese.8 We were also able to confirm the negative association with HLA-B*07 and B*57 that we reported previously in whites by imputation19 and in the case of B*07 in Han Chinese by DNA sequencing.24 The association seen with HLA-B*14, although only by RPE analysis, is compatible with what has been observed in French SpA families2 as well as in African blacks.3 The association with HLA-B*38 seen most strikingly in HLA-B*27-negative patients is interesting, given its known association with psoriatic arthritis,29 although this was not seen in the Chinese cohort, where a weak association was described with HLA-B*39 (HLA-B*38 and B*39 being ‘splits’ of the parent specificity HLA-B*16 (http://hla.alleles.org/antigens/broads_splits.html).
We observed an association with HLA-A*02 in the overall cohort that was not seen in the smaller HLA-B*27-negative cohort. This is compatible with the independent association with HLA-A*02:01 that we observed in a much larger patient cohort by imputation,18 although in the current study we were not able to resolve the association to the 4-digit level. Associations with HLA-DRB1*01 and HLA-DRB1*04:04 seen here have previously been described in UK patients with AS and from French spondyloarthritis families,2 10 although the lack of confirmation in HLA-B*27-negatives does not allow us to rule out that this may reflect LD with HLA-B*27. Similarly, the association with HLA-DRB1*01:03 that we observed by imputation19 was again seen here, but was not seen in the HLA-B*27 negatives. HLA-DRB1*01:03 is also strongly associated with inflammatory bowel disease,30 in particular where associated with peripheral spondyloarthritis.31 On the other hand, the decreased frequency of HLA-DRB1*15:01 and its linked allele DQB1*06:02 overall and in the HLA-B*27 negatives confirms what we have observed by imputation analysis in a larger cohort,19 now seen by direct HLA typing. An increased frequency of HLA-B*14 has been observed in French SpA families2 and in African patients with AS3 in other studies.
We were not able to confirm all the associations by direct HLA typing we previously described by imputation.19 In some cases, even with the rather large number of patients studied (1948 whites), given the low ORs (<1.5), there may have not have been adequate power. Alternatively, although the patients and controls were from all over the USA, ancestry informative markers were not examined in the controls and we cannot rule out potential stratification issues. We previously described an association with HLA-B*51 by imputation, seen only in a conditional regression analysis, which we could not confirm here, even by RPE analysis. However, we did find an association with HLA-B*52 in whites, both in the HLA-B*27 negatives and in the overall cohort. HLA-B*51 and B*52 are well-recognised ‘splits’ of the parent specificity HLA-B5 (http://hla.alleles.org/antigens/broads_splits.html) and although HLA-B*51 was decreased overall, it was increased in frequency in the HLA-B*27 negatives and by RPE analysis, although not significantly. HLA-B*51 was significantly reduced in the Chinese cohort overall, but having controlled for the presence of HLA-B*27 by RPE, no association was observed. We were unable to establish an independent association with HLA-B*13, an allele long associated with psoriasis, described previously by imputation in whites.19 However, as with HLA-B*51, HLA-B*13 was reduced in the overall Han Chinese dataset, but having controlled for HLA-B*27, no association was observed. HLA-B*47, observed by imputation as AS-associated,19 only occurred in five patients with AS and hence was too uncommon to establish as an AS association (data not shown).
The association of HLA-DPB1*03:01 with AS seen in whites further extends what was demonstrated previously in smaller cohorts.11 12 This is compatible with the association of SNPs around the HLA-DPB1 locus recently established by imputation.19 We did not examine HLA-DPA1 alleles and so could not confirm the associations with HLA-DPA1*01:02 and DPA1*01:03 reported by Díaz-Peña et al.12 However, we did not observe any association with HLA-DPB1*13:01 that was seen in their cohort.
We could not confirm some other associations described in smaller AS or SpA cohorts elsewhere32–34 perhaps due to clinical heterogeneity. One small recent study of HLA-A, HLA-B, HLA-C and DRB1 alleles in 75 Moroccan patients with AS reported an allele frequency of HLA-B*27 of 32%, with associations also seen with B*57, C*02 and DRB1*15 and negative associations with HLA-B*35 and B*49.32 The reasons for the discrepancies of these data and ours are hard to interpret given the small size of the cohort, which precluded examination of HLA-B*27 negatives. Another recent study of 189 Colombian patients with SpA (including 87 with AS, but also reactive arthritis and undifferentiated SpA) found associations with HLA-B*15 as well as with HLA-DRB1*01 and HLA-DRB1*04 as well as with HLA-B*27, which occurred at an allele frequency of 26.2%.34 Again, the small size of the cohort of patients with AS did not allow examination of HLA-B*27 negatives.
The size of the black AS cohort was small (n=67 patients), which would have restricted the statistical power of our observations. Nevertheless, we were able to confirm the positive association with HLA-B*40:01 and the negative association with B*07, for the first time in this ethnic group. The finding of HLA-B*27 in 60% of the black patients with AS is compatible with what was reported by Khan et al in a smaller cohort several years ago.35
We were able to confirm the association with HLA-DRB1*01:03 as observed by Cortes et al by imputation,19 although not an association of HLA-DRB1*08 with either AS or the uveitis phenotype, as has been reported elsewhere,10 11 in fact HLA-DRB1*08 was actually decreased in those with AAU. We did observe a significant association with the presence of HLA-B*27 and the occurrence of AAU in our patients, as also seen by imputation36 but not the higher frequency of HLA-B*27 homozygosity seen there. Otherwise the most robust association, although negative, was the decreased frequency of HLA-B*08 in those with AAU, which contrasts what was observed by imputation, where an increased frequency of a SNP associated with HLA-B*08, namely rs115937001, was observed.36 The reasons for this was unclear, as the OR of 1.8 in that larger group of 1711 patients with uveitis would suggest that there should have been sufficient power in this cohort of 380 patients with AAU to confirm this.
Thus, in the largest study of direct HLA typing of patients with AS and controls from three ethnic groups (and the only one examining African-American blacks other than HLA-B*27), these data show that the impact of the MHC on AS susceptibility extends beyond HLA-B*27. Many (though not all) of the observations made in a prior larger study of imputed HLA alleles19 are seen here. That positive and negative associations of some of these alleles cross ethnic boundaries suggests an independent role of both MHC class I and II alleles in influencing susceptibility to AS and subsets thereof.
Handling editor Josef S Smolen
Contributors JDR contributed many of the patients, carried out the HLA typing, entered the data, interpreted the results, did many of the statistical analyses, wrote the manuscript and provided the funding for the study. XZ established the Chinese collaboration and oversaw its completion, reviewed the manuscript and provided some of the funding for this study. ML carried out the multivariable and regression statistical analyses and participated in the writing and drafting of the manuscript. MHW contributed a large amount of the patients for this study as well as some of the controls. He also reviewed the manuscript. LY contributed a large amount of the Chinese patients for this study as well as some of the controls and also qc'd the Chinese HLA genotyping and reviewed the manuscript. LSG contributed number of the US patients for this study. She also reviewed the manuscript. Hejian Zou contributed a large amount of the Chinese patients for this study as well as some of the controls. He also reviewed the manuscript. MMW contributed a number of US patients for this study. He also reviewed the manuscript and obtained funding from the NIH Clinical Center to carry on the project there. MLI contributed a number of US patients for this study as well as some of the controls. She also reviewed the manuscript. TJL read all the X-rays for the study to determine who qualified for inclusion. He also reviewed the manuscript. DH contributed a large amount of the Chinese patients for this study as well as some of the controls. He also reviewed the manuscript. MHR oversaw the qc process of the datasets and worked with ML in the statistical analyses. JW oversaw the Chinese segment of this study, carrying out the HLA typing and coordinating the participating centers as well as providing funding for the Chinese portion of this project MAB provided the Australian patients in this study. He also qc'd the HLA typing with the imputed data from our GWAS, confirmed the statistical analyses and assisted JDR in manuscript preparation.
Funding This study was supported by the National Institutes of Health-National Institute of Allergy and Infectious Diseases (NIH-NIAIS) grant UO1 AI09090 (Drs Zhou, Reveille), National Institute of Arthritis, Musculoskeletal and Skin Diseases (NIH-NIAMS) grants R01 AR-46208 and 2P01AR052915-06A1 (Dr Reveille) and by University Clinical Research Grants M01-RR-02558 (The University of Texas Health Science Center at Houston) and M01-RR-000425 (Cedars-Sinai Medical Center) and by a grant from the Spondylitis Association of America. MAB was funded by a National Health and Medical Research Council (Australia) Senior Principal Research Fellowship. LSG is supported by the Rosalind Engelman Research Center at the University of California, San Francisco. MMW is funded by the Intramural Research Program, NIAMS/NIH.
Competing interests None declared.
Patient consent Not required.
Ethics approval The Internal Review Boards at The University of Texas-Health McGovern Medical School, The University of California-San Francisco, Cedars-Sinai Medical Center, The NIH-NIAMS Clinical Center, The Princess Alexandra Hospital (Brisbane), Fudan University reviewed and approved this study.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement We would be happy to share the data published in this manuscript. There are no unpublished data referable to the work included here.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.