Ankylosing spondylitis (AS) is a common chronic inflammatory disorder primarily affecting the spine, although the peripheral joints and the attachments of ligaments and tendons to joints (the entheses) are also frequently involved, as well as the eyes, and, less commonly, the heart and lungs (1, 2). AS can occur alone or in the setting of other types of spondylarthritis, including reactive arthritis, psoriatic arthritis, and inflammatory bowel disease (IBD). The prevalence of AS in European Caucasoid populations ranges from 0.2% to 0.9% (2, 3). Disease susceptibility is clearly attributable to genetic factors, because the sibling recurrence risk ratio is 82 (4), and twin-based studies estimate that disease heritability exceeds 90% (5). HLA–B27, which is encoded in the class I major histocompatibility complex (MHC) region, confers the greatest known susceptibility to AS, and up to 95% of patients of European ancestry are HLA–B27 positive (1, 5). In addition to HLA–B27, other MHC genes have also been implicated in AS, including HLA–DRB1 (6), tumor necrosis factor α (TNFα) (1), transporter-associated protein 1 (7), and low molecular weight protein 1 (8).
Although the MHC is likely the primary mediator of genetic susceptibility to AS, it explains <50% of the total genetic risk for AS (6, 7). Genome-wide linkage scans have implicated numerous non-MHC genomic regions, including 1p, 2p, 2q, 3p, 9q, 10q, 11p, 19q, and 16q (9, 10). Thus, delineation of the molecular etiology of AS is contingent on the identification and characterization of these non-MHC susceptibility genes.
Phenotypic heterogeneity can confound the analysis of the genotypic contribution to any disease. The modified New York criteria for the classification of AS rely not only on historical findings (inflammatory back pain) and physical findings (limitation of spinal motion and chest expansion) but also on the presence of radiographic sacroiliitis (11). In this study we adhered strictly to these criteria by insisting on the availability of sacroiliac radiographs (or appropriate recorded evidence documenting the presence of unequivocal radiographic sacroiliitis) as a prerequisite for inclusion, in order to focus on the phenotype of spondylitis, either alone or in the setting of other spondylarthropathies.
PATIENTS AND METHODS
- Top of page
- PATIENTS AND METHODS
The families were recruited by the Spondylitis Association of America (SAA) and investigators from the North American Spondylitis Consortium (NASC). Only families with 2 or more siblings meeting the modified 1984 New York criteria for AS (11) and for whom radiographs were available for review were included in the present analysis. The diagnoses were reevaluated by rheumatologists with experience in assessing patients with AS (DTY or JDR for patients enrolled by the SAA, and the collaborating rheumatologists at the other sites), and the radiographs were read by 3 different reviewers: a radiologist (OCW), a rheumatologist (JDR), and the referring study rheumatologist; discrepancies were resolved by a fourth reviewer (MW), who read the radiographs in a blinded manner. In the few instances in which radiographs were not available (<10% of the patients), either magnetic resonance imaging (MRI) or computed tomography (CT) scans or radiologist reports using wording consistent for grade II, III, or IV sacroiliitis as required by the modified New York Criteria (11) were used. In cases in which only MRI or CT scans were available, the diagnosis of spondylitis was confirmed by both the patient's radiologist and the NASC radiologist (OCW).
Informed consent was obtained from each family member. The Committee for the Protection of Human Subjects at the University of Texas Health Science Center at Houston approved this study, as did the institutional review boards at each enrolling site. A questionnaire about various clinical features of AS was administered to each affected family member. Blood samples were obtained, and genomic DNA was extracted by standard methods. HLA–B27 and other HLA–B typing was carried out by single-stranded conformational polymorphism analysis using commercially available kits (Dynal, New Hyde Park, NY), and class II (DRB1, DQA1, DQB1, DPB1) high resolution typing was performed by standard oligotyping techniques using primers and probes recommended at the Thirteenth International Histocompatibility Workshop (Victoria, British Columbia, Canada; May, 2002).
All 400 markers in ABI PRISM linkage map MD-10 (Applied Biosystems, Foster City, CA) were successfully amplified and typed using multiplex polymerase chain reaction in 5-μl reaction volumes. A preliminary analysis of the data for 171 sibpairs suggested, in addition to markers in the MHC, linkage to markers on 6q and 11q (data not shown). Hence, 16 markers not included on the ABI PRISM panel on chromosomes 6p (HLA–DRB1, DQA1, DQB1, and DPB1), 6q (D6S1654, D6S1577, D6S305, D6S1599, and D6S1719), and 11q (D11S4147, D11S917, D11S4090, D11S4127, D11S4094, D11S912, and D11S4126) were subsequently used for fine mapping analyses.
Genotyping was achieved using ABI 3100 sequencers for electrophoresis and GeneMap software for allele calling (Applied Biosystems). On average, amplification was successful in 92.5% of the individuals, and they were typed at all 400 markers; >90% of the individuals were typed at 316 markers (79%), and >80% were typed at 386 markers (96.5%). Specified relationships in each pedigree were verified with the Relative program (12). Mendelian inconsistencies and likely genotyping errors were identified using PedCheck software (13). After excluding 3 pedigrees with apparent nonpaternity, the estimated genotyping error was <1%. Inconsistent genotypes were removed from further analysis. The average distance between 2 adjacent markers was 8.85 cM.
Two-point and multipoint nonparametric linkage (NPL) analyses were conducted using the NPL-all (nonparametric linkage score using all affected relatives) statistic implemented in GeneHunter version 2.1 (14) as well as in GeneHunter Plus (15). One-parameter Allele-Sharing Model (ASM) LOD scores were calculated based on the distribution of test statistics under the null hypothesis and conditional on the data, using the ASM computer program (15). The estimation of the relative contribution of each locus was similar to that described by Laval et al (10), and the probabilities of sharing n alleles (zn) were estimated using GeneHunter under the assumption of no dominant variance (14). Family-based tests of linkage disequilibrium were conducted using the transmission disequilibrium test (TDT) (16) and the pedigree disequilibrium test (17), which were implemented using UNPHASED software (18).
- Top of page
- PATIENTS AND METHODS
In total, 244 sibpairs from 180 pedigrees who were concordant for AS were included in this study. Also included were 602 family members, 424 of whom had AS. Sixty-three families had affected sibpairs and both parents available, 25 had affected sibpairs and 1 parent available, and 92 had affected siblings only (Table 1). One hundred fifty-two families had 1 affected sibpair, 1 family had 1 affected sibpair in 2 different generations, 24 families had 3 affected siblings, and 3 families had 4 affected siblings.
Table 1. Composition of families studied
|Affected sibpairs, both parents||63|
|Affected sibpairs, 1 parent||25|
|Affected sibpairs, no parents||92|
|Affected parent/offspring pairs only||9|
|One affected offspring, both parents*||30|
| Total number of families||219|
|One affected sibpair||152|
|One affected sibpair, 2 different generations||1|
|Three affected siblings†||24|
|Four affected siblings‡||3|
| Total number of affected sibpairs||244|
Spondylitis occurred more frequently in men than in women in these families (Table 2). Ninety-four percent of the patients were Caucasian, with the rest being of either mixed Caucasian/American Indian or Caucasian/eastern Asian heritage. The frequency of uveitis in the multiplex families (40%) was similar to what is typically observed in patients with AS (2) (Table 2). Inflammatory bowel disease and psoriasis occurred less frequently (10% and 11%, respectively). HLA–B27 was found in 97% of the patients with AS, with the affected siblings in the two B27-negative families sharing B39 in one family and B52 in another family. Further detailed studies of HLA alleles are under way.
Table 2. Demographic, clinical, and genetic characteristics of patients with ankylosing spondylitis included in this genome-wide scan*
|Reactive arthritis, excluding hips and shoulders||14/393||3.6|
An additional 39 pedigrees without affected sibpairs studied containing 129 family members (48 with AS) were included; 2 families had affected sibpairs, but the other sibling was not available for study, and 37 were “trio” families (1 affected proband, with parents available) (Table 1). Nine families had parent/offspring pairs with AS, and 28 were simplex families in which the diagnosis of AS (or other spondylarthropathies) could not be confirmed in any other family member. In one of these 28 families, clinical information was not available in the affected relative (other than enough to confirm the diagnosis of AS). Of the remaining 27 AS probands, 44% were male, 44% had a history of uveitis, and all were HLA–B27–positive. A history of peripheral arthritis was given by 55% of patients, and 15% had a history of heel pain. None had psoriasis or IBD. These characteristics were not significantly different from those in the multiplex families (data not shown).
Table 3 summarizes the results of 2-point linkage and TDT analyses and provides the physical and genetic locations for each marker where putative linkage was observed, using P < 0.05 as the threshold. The results of multipoint linkage analysis are shown in Figure 1; the NPL score and the LOD score were estimated using the ASM program. The highest ASM LOD scores (2-point and multipoint) were observed in the regions on chromosomes 6p, 6q, and 11q, while relatively weaker ASM LOD scores were found on chromosomes 1, 3, 4, 5, 16, 17 and, 19 (Table 3 and Figure 1).
Table 3. Two-point linkage and TDT analyses of chromosomal markers and AS susceptibility*
|Chromosome||Marker||Position, cM||NPL||ASM LOD||P, 2-point analysis||P, TDT analysis|
|6p (MHC)||HLA–B||44.92||6.378||12.37||1.8 × 10−11||0|
|6p (MHC)||DRB1||45.90||8.720||20.49||6.8 × 10−20||0.0003|
|6p (MHC)||DQA1||45.98||6.793||14.88||9.0 × 10−13||0.002|
|6p (MHC)||DQB1||46.05||7.623||18.37||1.3 × 10−15||0.001|
|6p (MHC)||DPB1||46.65||5.251||10.72||2.5 × 10−8||0.019|
|6p (MHC)||D6S1610||53.81||4.377||7.232||2.7 × 10−6||0.157|
|11q||D11S4090||105.74||2.159||1.732||0.013||6.2 × 10−5|
Figure 1. Multipoint linkage analysis. Broken line represents the nonparametric linkage (NPL) score; solid line represents the ASM-derived logarithm of odds (LOD) score.
Download figure to PowerPoint
Linkage of the MHC region was supported by both 2-point and multipoint analyses, as well as by TDT analysis. The region covers almost the entire p arm of the chromosome, spanning 71 cM for 2-point analysis and 100 cM for multipoint analysis. The strongest linkage was for the HLA loci, including HLA–B, DRB1, DQA1, DQB1, and, as has not been noted previously, DPB1, with ASM LOD scores of >10. Of note was the finding that the strongest peak (45.90 cM) in the MHC was at the HLA–DRB1 locus (NPL score 8.720, ASM LOD score 20.490, P = 6.8 × 10−20 for 2-point and NPL score 9.582, ASM LOD score 20.00, P = 1.14 × 10−23 for multipoint analyses). Of note, the HLA–B, DRB1, DQA1, DQB1, and DPB1 markers are not part of the ABI marker panel used but were examined to better characterize the MHC contribution to AS.
A second region on chromosome 6 was found to show high ASM LOD scores at the q arm of the chromosome, with the strongest evidence of linkage at 154.10 cM (D6S441; NPL score 2.457, ASM LOD score 2.433, P = 0.005) in the 2-point analysis. This is supported by a 39.13-cM region (135.58–174.71 cM) in multipoint analysis, with the greatest significance at 166.39 cM (NPL score 2.532, ASM LOD score 1.762, P = 4.2 × 10−3). This region was not observed in 2 previous genome-wide screens, described by Brown et al (9) and Laval et al (10) (hereinafter referred to as the Oxford study). Linkage in this region was confirmed by typing an additional 5 markers that were not included in the ABI panels (see Patients and Methods). On chromosome 11, a 17.3-cM region (105.74–123 cM) with a high ASM LOD score was found, which is separated by 2 single markers (D11S908 and D11S4127) into 2 regions in 2-point analysis. The strongest evidence for linkage was for D11S4094 at 123 cM (NPL score 2.235, ASM LOD score 1.939). The TDT analysis showed that the strongest association was with D11S4090 at 105.74 cM (P = 6.2 × 10−5). Of note, these 2 markers (D11S4090 and D6S4094) were not part of the ABI marker panel (see Patients and Methods) but were used when preliminary results in 171 sibpairs indicated linkage to this region (data not shown). This association maintained significance even after applying Bonferroni correction. Linkage in this region is further supported by multipoint analysis in which the putative linkage spans 22.19 cM continuously from 101.68 cM to 123.87 cM, with the strongest peak at 112.33 cM (NPL score 2.122, ASM LOD score 1.324, P = 0.014).
Several other regions showed putative linkage with AS (Table 4). These included a region on chromosome 1q encompassing D1S238 (202.73 cM), a region on chromosome 4p between positions 28.91 cM and 43.59 cM, and a region on chromosome 5q between 188.57 cM and 195.49 cM. These were seen on both 2-point and multipoint analyses and were not seen in previous scans (9, 10), although linkage nearby on chromosomes 1q and 5q was seen in the 2 genome-wide scans described previously (9, 10). In contrast, other regions with possible or putative linkage were seen on chromosome 3p, spanning 25.58 cM (between 76.76 cM and 102.34 cM, on both 2-point and multipoint analysis), on chromosome 10q (between positions 102.26 cM and 126.95 cM, on multipoint analysis only), on chromosome 16q (between 88.68 cM and 100.39 cM), and on chromosome 17p (in a 21.45-cM region encompassing D17S938, D17S1852, and D17S921, which for both chromosomes was observed on 2-point and multipoint analyses as well as by TDT analysis), and at D19S931 (49.75 cM, on 2-point analysis alone); these regions were also observed to have marginal evidence for linkage in the Oxford study (9, 10). The same results were seen again in GeneHunter Plus, with no new linkages found and previous linkages confirmed. Of note, the region seen on chromosome 16q was not overlapped with the region implicated in susceptibility to psoriasis (19).
Table 4. Similarities between the present genome-wide scan and the oxford scans (9, 10)*
|Marker||Current study||Oxford study|
|Distance from p telomer, cM||Single-point analysis||Multipoint analysis||Distance from p telomer, cM||Screen 1||Screen 2||Screens 1 and 2|
|NPL||P||ASM LOD||NPL||P||ASM LOD||ASM LOD||P||ASM LOD||P||ASM LOD||P|
|D6S470||18.22||1.22||0.1010||0.97||3.22||0.0004||3.15||17.7||0.1||0.26||3.1||7.9 × 10−5||2.2||0.00066|
|D6S289||29.93||3.11||0.00062||3.22||4.54||1.21 × 10−6||5.58||29.55||0.8||0.028||1.9||0.0017||2.5||0.00033|
|D6S422||35.66||3.38||0.00022||4.90||5.50||5.63 × 10−9||8.41||35.7||0.9||0.02||2.9||0.00012||3.6||2.3 × 10−5|
|D6S276||40.69||3.83||3.4 × 10−5||7.07||7.21||3.55 × 10−14||13.58||44.8||1.8||0.0022||4.8||1.3 × 10−6||6.5||<1 × 10−6|
|D6S1610||53.81||4.38||2.7 × 10−6||7.23||7.30||1.74 × 10−14||13.62||53.9||1.5||0.0042||NA||NA||1.5||0.0042|
An estimation of the contribution of the regions implicated or possibly implicated in AS susceptibility suggested that 45% of the overall risk for AS is provided by the MHC (Table 5) (λs = 7.2). Overall, the individual contributions of non-MHC regions were small (λs < 2.0).
Table 5. Estimation of locus-specific contribution to AS susceptibility*
|Marker||Chr||Position, cM||z0||z1||z2||λ||Contribution, %|
- Top of page
- PATIENTS AND METHODS
Besides confirming linkage of the MHC region with AS, we also found evidence for linkage at 10 non-MHC regions, on chromosomes 1q, 3p, 4p, 5q, 6q, 10q, 11q, 16q, 17p, and 19q. The highest ASM LOD scores were observed in the regions on chromosome 6p (MHC), 6q (where the highest ASM LOD score [2.433] was encountered at D6S441), and 11q (where a significant association on TDT analysis was found for 1 marker, D11S4090, even after applying Bonferroni correction), while the putative linkages found in chromosomes 1q, 3q, 4p, 5q, 10q, 16q, 17p, and 19q were marginal. Earlier, Laval et al (10) identified suggestive or significant linkages with AS on chromosomes 1 (66.6 cM), 2, 5, 6 (44.8 cM, MHC region), 9 (160.2 cM), 10 (137.6 cM), 16 (109.1 cM, 122.1 cM), and 19 (53.2 cM, 66 cM). Except for the MHC region at 6p, none of the non-MHC regions that showed high ASM LOD scores overlap in both studies. Interestingly, the regions that show weak putative linkage on chromosomes 3, 10, 16, 17, and 19 in this study also showed weak linkage in the study by Laval et al (10). These loci have not been implicated, however, in genome-wide scans from other rheumatic diseases, although a susceptibility region for insulin-dependent diabetes mellitus is found at chromosome 6q25 (20), a region that was identified in the present scan. Other potential candidate genes in this region include TNFα-induced protein 3 (6q23), SNF2 histone linker PHD RING helicase (6q24.2), katanin p60 (ATPase-containing) subunit A1 (6q24.3), protein-L-isoaspartate (D-aspartate) O-methyltransferase (6q24–25), and NADPH oxidase 3 (6q25.1–26). The region of chromosome 11q identified in the present study also contains a number of potential candidate genes for AS susceptibility, including the matrix metalloproteinase family of genes (11q21–22), the caspase family of genes (11q22), interleukin-18 (IL-18; 11q22–23), IL-10α receptor (11q23), the CD3 gene family (11q23), and Thy-1 cell surface antigen (11q23). Studies are currently under way to better map and confirm the linkage to both of these regions.
AS is just one member of the spondylarthropathy (SpA) group, and we decided to specifically focus on the phenotype of spondylitis in the setting of SpA (as classified by the modified New York criteria ), in order to reduce phenotypic heterogeneity, which is likely genetically more heterogeneous. Greater phenotypic heterogeneity might reduce statistical power, which we sought to avoid.
The highest ASM LOD score for the MHC genes analyzed in this study was for HLA–DRB1. This might lead some to postulate that HLA–DRB1 forms a separate locus for AS predisposition. Other studies have suggested this (6). However, the tight linkage disequilibrium known to exist between MHC genes makes it impossible to assign a definite and dominant role for any 1 MHC locus in this study. Of note, however, this is the first study to examine the HLA–DPB1 locus in AS susceptibility. Overall, the MHC contributed less than half the susceptibility to AS, as has been reported elsewhere (5).
The fact that close agreement between our scan and the 2 from the Oxford study (9, 10) was not seen in many non-MHC genes is not entirely surprising. Both our study and the Oxford study confirm linkage with the MHC, the dominant region in susceptibility to AS. Moreover, as noted above, evidence for linkage for certain non-MHC regions was seen by both our group and investigators for the Oxford study of investigators (specifically on chromosomes 3, 5, 10, 16, 17, and 19). The best explanation for the discrepancies is that the contribution of each non-MHC gene to the overall susceptibility to AS is likely small (λs < 2.0), and the number of sibpairs examined by our group (n = 244) and by Laval et al (n = 255) (10) may not, when examined individually, provide adequate power to discern such “small effect” genes. Further evidence for this is provided by the identification of significant association of AS in our families with the ANKH gene encoded on chromosome 5p (21), despite the fact that this region was not identified by the ABI PRISM linkage map MD-10 marker panel used in this study. Alternatively, the marker density could be increased, using microsatellite markers spaced much more closely. This is part of the principle used here in the fine mapping of candidate regions on chromosomes 6q and 11q as well as in the MHC. However, doing this throughout the genome would markedly increase the time and expense associated with genome-wide scanning to the point of unfeasibility. Ultimately, a combined collaborative approach between the different family collections that is under way will probably be necessary to comprehensively dissect the genetic basis of AS.
Another possible explanation for the differences in results between the scans of the Oxford group (9, 10) and the present study is clinical heterogeneity between the 2 family collections. However, the frequencies of “other features” of SpA in the patients with AS were not significantly different between patients in the current study and those in the Oxford study (63% male, 9% with IBD, 16% with psoriasis, and 44% with uveitis in the Oxford study  versus 65%, 10%, 12%, and 35%, respectively, in the present study). In both studies, the modified New York criteria for AS (11) were used as a basis for inclusion; these criteria require the presence of at least 1 clinical criterion (inflammatory low back pain, limitation of motion of the lumbar spine, or limitation of chest expansion) and radiographic sacroiliitis. However, in the Oxford study (9, 10), the referring physicians determined whether the patient met these criteria, whereas in the present study inclusion required the presence of either available pelvic radiographs (>90% of patients) or appropriately worded reports of such by a radiologist (<10% of patients). Whether this difference in the application of the New York criteria contributed to the discrepancies between patients in the current study and those in the Oxford study cannot be ascertained. Moreover, the inclusion of only HLA–B27–positive AS probands in the Oxford study makes the presence of a significant number of false positives unlikely. Alternatively, differences in disease severity between the 2 cohorts (which the Oxford group has shown to be at least in part genetically determined ) stemming from differences in referral patterns to rheumatologists between North America and the UK could also cause differences in genetic data.
One concern about these data is whether the genetic basis of familial and nonfamilial AS is the same. Although we observed no psoriasis or IBD in patients with nonfamilial AS, there were no differences in other disease features (uveitis, heel pain) in the trio families, and there were no statistically significant differences between patients with familial versus nonfamilial AS. Moreover, these clinical features occur at similar frequencies in patients with sporadic AS (19). Nevertheless, because these families were referred to as multiplex families (in which familiality could not be confirmed), genetically we cannot rule out the possibility that they may not be truly representative of sporadic AS cases. Similar studies in additional families that are being conducted in our cohort are under way to confirm the linkages reported here. Further fine mapping of the regions identified in this study is also being performed.
The relevance of identifying the genes involved in susceptibility to AS is clear. First, identifying individuals at highest risk for developing AS might allow earlier diagnosis, in which case interventions (such as TNF blockers) could be used before the loss of spinal motion or joint destruction could occur. Moreover, elucidation of these genes would allow potential targets for new therapies. Finally, in those individuals at highest risk for developing AS, the eventual discovery of the triggering event(s) might allow actual disease prevention.
- Top of page
- PATIENTS AND METHODS
We acknowledge the contributions of the other NASC members, including Allen Sawitzke, MD (University of Utah Health Science Center, Salt Lake City) and Frank B. Vasey, MD (University of South Florida, St. Petersburg); the coordinating assistance of Laura Diekman (University of Texas–Houston), Parthajeet R. Chowdhuri, Erin Dowling, and Erin Skrok (Cedars-Sinai Medical Center, Los Angeles, CA), David Ritter (MetroHealth Hospital, Cleveland OH), Karen Adams (University of Pennsylvania, Philadelphia), Vicki Lapp (University of Toronto, Toronto, Ontario, Canada), Peter Majewski (University of Minnesota, Minneapolis), Cathy Mallon (University of Alberta, Edmonton, Alberta, Canada), and Trudy Doyle (Oregon Health and Science University, Portland); the statistical advice and support of Joshua Akey, PhD (Fred Hutchinson Cancer Center, Seattle, Washington); and the technical assistance of Rui Jin and Binh Vu (University of Texas–Houston), Elizabeth Dickman, and Joanna Watson (Center for Genome Information, University of Cincinnati, Cincinnati, OH).