To identify genomic regions linked with determinants of age at symptom onset, disease activity, and functional impairment in ankylosing spondylitis (AS).
To identify genomic regions linked with determinants of age at symptom onset, disease activity, and functional impairment in ankylosing spondylitis (AS).
A whole genome linkage scan was performed in 188 affected sibling pair families with 454 affected individuals. Traits assessed were age at symptom onset, disease activity assessed by the Bath Ankylosing Spondylitis Disease Activity Index (BASDAI), and functional impairment assessed by the Bath Ankylosing Spondylitis Functional Index (BASFI). Parametric and nonparametric quantitative linkage analysis was performed using parameters defined in a previous segregation study.
Heritabilities of the traits studied in this data set were as follows: BASDAI 0.49 (P = 0.0001, 95% confidence interval [95% CI] 0.23–0.75), BASFI 0.76 (P = 10−7, 95% CI 0.49–1.0), and age at symptom onset 0.33 (P = 0.005, 95% CI 0.04–0.62). No linkage was observed between the major histocompatibility complex (MHC) and any of the traits studied (logarithm of odds [LOD] score <1.0). “Significant” linkage (LOD score 4.0) was observed between a region on chromosome 18p and the BASDAI. Age at symptom onset showed “suggestive” linkage to chromosome 11p (LOD score 3.3). Maximum linkage with the BASFI was seen at chromosome 2q (LOD score 2.9).
In contrast to the genetic determinants of susceptibility to AS, clinical manifestations of the disease measured by the BASDAI, BASFI, and age at symptom onset are largely determined by a small number of genes not encoded within the MHC.
Ankylosing spondylitis (AS) is a common chronic arthritis with a prevalence of 1–3/1,000 in white populations (1). It affects predominantly the spine and sacroiliac joints, causing pain and stiffness, with initial erosion and subsequent fusion of involved joints. Susceptibility to the disease is probably oligogenic (2), the gene HLA–B27 being almost essential but not sufficient for disease recurrence in families, with several other loci having been implicated in genome-wide linkage studies (3).
It is apparent, however, that the clinical manifestations of the disease vary considerably, both in the extent and severity of axial disease and in the occurrence of other disease features, such as iritis and peripheral arthritis. Very little is known about the risk factors affecting clinical manifestations of AS. Using self-administered questionnaires that assess either current disease activity (Bath AS Disease Activity Index [BASDAI] ) or functional impairment due to AS (Bath AS Functional Index [BASFI] ), we demonstrated that these traits are highly heritable (h2 [residual multifactorial familial component] 0.51 and 0.68, respectively) (6). Complex segregation analysis suggested that these measures are probably determined by the effects of a single major gene, with additional polygenic influences. In that study, age at symptom onset was not shown to be significantly heritable. The aim of the current study was to identify genomic regions determining the BASDAI, BASFI, and age at symptom onset.
We performed a genome-wide linkage analysis on 188 affected sibling pair families with 454 affected individuals. AS was defined according to the modified New York criteria (7). Affected family members were contacted by telephone and the diagnosis confirmed by semistructured interview conducted by one of the authors (JH, LB, or MAB). All patients were then sent a questionnaire to complete, including the BASDAI and BASFI instruments and date of onset of AS symptoms. More than 90% of the individuals contacted about participating in the study completed the questionnaire. Most, but not all, of these family members were included in our previous segregation study (6); 43 additional families not enrolled in the segregation study were included because we were able to obtain trait data on family members, and 28 families included in the segregation study had not been genotyped and were therefore excluded from the present study.
Genotype information was obtained from a previously reported whole genome scan in these families, in which 367–505 microsatellite markers were assessed (3). Nonparametric quantitative linkage analysis was performed, with subsequent parametric multipoint analysis of regions achieving “nominal” evidence of linkage (see below), using parameter estimates from our previous segregation study (6). The parameters estimated in that study apply generally to familial AS in British Caucasians, and therefore can be applied to the families studied here.
All distances quoted are from the p-telomere. Genome-wide significance thresholds used for autosomal analysis were for “nominal” linkage (marker P < 0.05) (logarithm of odds [LOD] score 0.6–2.1), “suggestive” linkage (likely to occur once by chance per whole genome screen) (LOD score 2.2–3.5), and “significant” linkage (likely to occur once by chance per 20 whole genome screens) (LOD score ≥3.6) (8). For X chromosome analysis, the threshold for “nominal” linkage is 1.18, that for “suggestive” linkage is 3.1, and that for “significant” linkage is 4.6. Multipoint nonparametric quantitative linkage analysis of the autosomes was performed using a modified Haseman-Elston algorithm (9) as implemented by the program Merlin (10). For the BASDAI and BASFI, parametric 2-point and multipoint quantitative linkage analyses were performed using the program VITESSE (11). Parameters for dominant, codominant, and recessive models were taken from segregation study estimates (6). Parametric assessment of age at symptom onset was not possible since our previous segregation study had not demonstrated a major gene effect (6). Trait heritability was assessed using the program SOLAR (12). Nonparametric linkage analysis of susceptibility to AS was performed using GeneHunterPlus (13) for the autosomes and ASPEX (14) for the X chromosome.
Mean ± SEM values for the BASDAI, BASFI, and age at symptom onset are shown in Table 1. These values were similar to the averages from the Bath AS database (mean BASDAI 4.1, BASFI 4.0, age at symptom onset 25.2 years ), indicating the lack of a major ascertainment bias with regard to these traits. No significant correlation between disease duration and age at symptom onset was seen.
|BASDAI (n = 383)||3.9 ± 0.1|
|BASFI (n = 382)||3.8 ± 0.1|
|Age at symptom onset (n = 343)||22.6 ± 0.4|
Heritabilities of the traits studied in this data set were as follows: disease activity by BASDAI 0.49 (P = 0.0001, 95% confidence interval [95% CI] 0.23–0.75), functional impairment by BASFI 0.76 (P = 10−7, 95% CI 0.49–1.0), and age at symptom onset 0.33 (P = 0.005, 95% CI 0.04–0.62). Although the heritability of the BASDAI was virtually the same as that previously reported (0.51) (2), the heritability estimates for BASFI and age at symptom onset were higher than previously reported (0.68 and 0.18, respectively) (2), with age at symptom onset being significantly heritable in this data set.
Nonparametric linkage findings with P values less than 0.05 are listed in Table 2. Among all 3 phenotypes studied, nominal linkage was achieved in 37 regions. Of these, only 5 were linked to more than 1 trait. On chromosome 3p, linkage was observed with age at symptom onset, BASDAI, and BASFI, with peak LOD scores of 1.3, 0.8, and 0.9 at 30.4, 2.5, and 30.4 cM, respectively, from the p-telomere. Although the maximum point of linkage differed slightly between traits, all 3 linked regions overlapped. On chromosome 11, nominal linkage was observed with age at symptom onset and BASDAI at 84–89 cM from the p-telomere, with maximum LOD scores of 0.8 and 0.9, respectively (Figure 1). A further locus at 137 cM was linked with both age at symptom onset and BASFI, with maximum LOD scores of 3.3 and 1.1, respectively. On chromosome 16, nominal linkage was observed with age at symptom onset and BASFI, peaking at 88 and 86 cM, and with maximum LOD scores of 1.6 and 0.9, respectively (Figure 2). On chromosome 18, linkage was observed with age at symptom onset, BASDAI, and BASFI, with peaks between 16 and 18 cM from the p-telomere, and with maximum LOD scores of 0.6, 0.9, and 1.1, respectively.
|Chromosome||Distance from p-telomere, cM||Age at symptom onset||BASDAI||BASFI|
Regions for which nonparametric linkage was achieved with P values of less than 0.05 were examined by parametric quantitative linkage analysis. Linkages with LOD scores of ≥2.2 in this analysis are shown in Table 3. Five regions demonstrated at least suggestive linkage with either the BASDAI or the BASFI. Suggestive linkage with the BASFI was seen on chromosome 2q (maximum LOD score 2.9 at 237.7 cM [multipoint codominant parametric analysis]). With the BASDAI, marker D13S175 at 7.4 cM achieved a LOD score of 2.4 (2-point recessive model), but no significant linkage was observed with other traits or in nonparametric analysis. On chromosome 16, strong linkage with the BASDAI was observed (maximum LOD score 2.9 at D16S3136 at 60 cM [multipoint dominant model]). Significant linkage was observed on chromosome 18 for the BASDAI (maximum LOD score 4.0 at marker D18S452, 17.7 cM [multipoint recessive model]). There was also nominal linkage of the BASFI to this region (maximum LOD score 1.7 also at D18S452 [recessive model]), and nominal linkage of age at symptom onset (maximum nonparametric LOD score 0.6 at the same position). On chromosome 20q, strong linkage was also observed with the BASDAI (maximum LOD score 3.3 at D20S119, 61 cM [multipoint recessive model]), but not with either the BASFI or age at symptom onset.
|Chromosome||Distance from p-telomere, cM||BASDAI||BASFI||Analysis model|
Moderate linkage was observed between the X chromosome and age at symptom onset (maximum LOD score 1.5 at 138 cM [P = 0.004]), but no linkage of the X chromosome with the BASDAI, BASFI, or disease susceptibility was seen.
A region on chromosome 16q that we have previously demonstrated to be strongly linked with susceptibility to AS (3) also exhibited moderate linkage with the BASFI and age at symptom onset (Figure 2). In the families in this study, the maximum linkage for susceptibility to disease was found at 109 cM (LOD score 5.1), with nominal linkage extending from 75 cM to the q-telomere. Maximum linkage with the BASFI was found at 82 cM (maximum LOD score 1.7, dominant model), with nominal linkage extending from 74 to 92 cM. For age at symptom onset, the maximum LOD score was 1.6 (P = 0.004) at 88 cM, with nominal linkage extending from 78 to 118 cM.
No significant linkage was observed between the region encoding the major histocompatibility complex (MHC) and any of the traits examined (LOD scores <1.0) (Figure 3). In contrast, in the families studied, disease susceptibility showed strong linkage with the MHC (maximum LOD score 15.7).
We have identified a small number of regions for which there is at least moderate evidence of linkage with the 3 traits studied. In particular, we observed linkage (likely to occur by chance less than once per whole genome screen) with the BASDAI on chromosomes 11q, 13p, 16p, 18p, and 20q, with age at symptom onset on chromosome 11q, and with the BASFI on chromosome 2q. Five regions, on chromosomes 3p, 11p, 11q, 16p, and 18p, were linked to more than 1 of the phenotypes studied, making it less likely that these were false-positive findings. While some of the data will inevitably be found to represent false-positive results, they support the findings of our segregation study, with major gene effects identified for both the BASDAI and the BASFI. In our segregation study, restricted models (major gene only, with no residual genetic effect) were not different in likelihood from models consisting of a major gene with residual genetic effects primarily determining the variance of the trait. Thus, the finding of a small number of loci with moderate linkage to the BASDAI and BASFI is consistent with the segregation results.
In the current study, significant heritability of age at symptom onset was identified, in contrast to the findings in our segregation study. The observed heritabilities of age at symptom onset in our segregation study and the current study are not, however, statistically significantly different (heritability observed in segregation study 0.18, in current study 0.33 [95% CI 0.04–0.62]). The difference between the two in statistical significance is likely due to the increased sample size in the current study and minor differences between the families studied.
There was little overlap observed between regions linked with susceptibility to AS in this data set and those linked with clinical characteristics of AS. None of the regions for which suggestive or greater linkage was achieved for either susceptibility or clinical characteristics was linked (LOD score ≥2.2) to both. Most striking was the lack of linkage of the MHC with clinical characteristics of disease. This region is very strongly linked with susceptibility to AS (LOD score 15.7 in these families), and the results of association studies (16, 17) have also suggested effects of HLA–DRB1 genes on clinical manifestations of AS. In the present study, all of the subjects with AS were B27 positive, and therefore we can only study the effect of B27 heterozygosity and homozygosity with regard to disease severity. There is disagreement between previous studies as to the effect of B27 homozygosity on disease severity, with some studies suggesting only that B27 homozygosity may be associated with more peripheral arthritis (18) and others suggesting more severe disease in B27 homozygotes (19, 20). The modest power of this study may not be sufficient to detect such effects. However, the finding of strong linkage elsewhere in the genome indicates that it is unlikely that MHC-encoded genes are major determinants of disease severity or age at symptom onset in AS, and certainly play a far less significant role than they do in determining disease susceptibility.
One region that did show moderate evidence of linkage with the traits investigated in the current study as well as with disease susceptibility was a centromeric region on chromosome 16, which showed nominal evidence of linkage at 78–118 cM for age at symptom onset and 74–92 cM for the BASFI. In this data set, the linkage of this region with disease susceptibility was extremely strong, with a peak LOD score of 5.1 and nominal evidence of linkage extending from 75 cM to the q-telomere. Thus, the linkage with the BASFI and age at symptom onset overlaps almost completely with the area linked to susceptibility, suggesting a common genetic effect. This adds further evidence that this is a true-positive linkage, although ultimate proof of this will require identification of disease-associated genetic variants.
Polymorphisms of the NOD2 gene are major susceptibility factors for Crohn's disease (21, 22). Although it has been conclusively demonstrated that NOD2 polymorphisms are not associated with susceptibility to AS (23–27), we have recently demonstrated association of the Pro268Ser NOD2 variant with the BASDAI in patients with AS associated with inflammatory bowel disease (27). In the nonparametric analysis in the present study, all 3 quantitative trait loci studied showed at least nominal evidence of linkage at this locus. In the parametric analysis, marker D16S3136 showed suggestive linkage with the BASDAI (LOD score 2.9). This marker lies <25 kb from the gene NOD2. This finding supports the hypothesis that genetic variation in NOD2 affects the severity of AS.
The results of this study highlight the difference in the genetic control of susceptibility to AS and of its clinical manifestations. No linkage with the MHC was observed, whereas other regions not linked to AS susceptibility were found to be strongly linked with its clinical manifestations, supporting the theory that the pathophysiologic mechanisms involved in susceptibility to and clinical manifestations of AS are distinct processes. Identification of the genes involved in controlling the rate of progression of AS may indicate novel targets for therapies that would be effective in preventing the process of ankylosis, and thus long-term disability, in AS.
The authors wish to thank the members of the National Ankylosing Spondylitis Society (UK) and the study families for their support of and participation in this project.