Use of supplementary phenotype to identify additional rheumatoid arthritis loci in a linkage analysis of 342 UK affected sibling pair families

Rheumatoid arthritis (RA) is a chronic inflammatory condition that belongs to the family of diseases referred to as autoimmune. The population prevalence of RA is about 0.01 [1, 2], and the relative risk for siblings of affected persons ranges between 2 and 17 [2–5]. In addition to evidence from familial aggregation, results from twin studies [6–8], segregation analysis [9] and genome-wide association studies [10–12] all indicate evidence of genetic influence in the aetiology of RA. This evidence is further strengthened by the relatively high heritability of RA, estimated to be about 65% [13] from twin studies; and also by reported linkage evidence for regions on chromosomes 2 and 11 [14], and 6 and 16 [15]. Although consistently strong evidence across studies points to association with the HLA locus on chromosome 6, it has however been observed that the RA-HLA region association accounts for only about one-third of the total genetic determinants of susceptibility to RA [16, 17]. On the other hand, none of the other RA-linked loci reported to date has been consistently reported across studies [1, 3, 14, 18, 19]. The implication is that it is likely that a number of other loci involved in the genetic susceptibility to RA exist and are yet to be identified and their individual roles quantified; hence the complete role of genetic factor in the aetiology of this complex disease remains to be clearly understood.

As with many other complex human diseases, the difficulty in achieving a clear understanding of the role of genetics in disease can partly be attributed to limitations in phenotype definition and characterization. RA is a complex disease condition with variation in onset age, severity, and accompanying presence or absence of clinical features such as nodules, erosions, and rheumatoid factor. In addition, phenotypic, clinical and genetic heterogeneity exists between studies. To address the problem of heterogeneity associated with RA phenotype, John et al. [15] and Eyre et al. [20] used stratified analyses based on subgroups defined by age at disease onset, sex and clinical features such as rheumatoid factor (RF) status, erosion and shared epitopes. In addition to the affection status variable, age at RA onset provides the opportunity to investigate genetic influence based on the cumulative hazard of RA by using survival analysis tool in combination with linkage analysis. The premise is that detecting evidence of linkage in regions not previously identified in analysis of affection-based phenotype may be possible in analysis of supplementary phenotype of the same disease. The objective of this study was to use cumulative hazard estimates from proportional hazards model in genome-wide scan for linkage evidence for new and additional genetic loci linked to RA using affected sibling pairs data. We present here the results from the genome-wide linkage analysis of the cumulative hazard to RA in a sample of 728 relatives from 342 affected sibling pair families from the UK.

The RA affected sibling pair families used in this study were part of the UK National Repository of Multicase Rheumatoid Arthritis Families maintained by the Arthritis Research Campaign (http://www.medicine.manchester.ac.uk/epidemiology/research/arc/). Details of family recruitment and assessment protocols have been described elsewhere [1, 20]. Briefly, subjects were examined by trained metrologists using standardized questionnaires and protocols. Radiographs of each subject's hands were reviewed and rheumatoid factor status of each subject was ascertained using a particle agglutination test. All RA affected subjects satisfied the American College of Rheumatology criteria [21, 22]. Access to the UK RA data was requested and approved through the Genetic Analysis Workshop 15 (GAW15). The data used in the present study were from cohorts 1 and 2 families as made available through GAW15 [23]. Cohort 1 comprised of RA affected sibling pairs from the 182 families as described by Mackay et al [1] and cohort 2 was made up of RA affected sibling pairs from the additional 195 families which have been described elsewhere [20]. Available phenotypic and clinical information included RA affection status, age at onset of RA symptoms, erosion status, and RF status. Genotype data on 353 informative microsatellite markers from the ABI Prism Linkage Mapping set, spanning the 22 autosomes at an average inter-marker distance of 10 cM were available on subjects and used in the present study.

We report here genome-wide linkage evidence for RA based on analysis of cumulative hazard estimated from family data on RA affected sibling pairs. The objective of the study was to identify additional evidence of linkage for RA by using supplementary RA phenotype that we envisioned would capture linkage information otherwise undetected by the traditional affection-status phenotype definition used in previous linkage studies of RA. The supplementary phenotype approach was based on application of proportional hazards modelling to estimate cumulative hazard of RA based on affection status and RA onset age for each study subject. 750 affected subjects from 364 families were included in the estimation of cumulative hazard using proportional hazards model. The estimated cumulative hazard was then used as quantitative trait in linkage analysis of genotypes on 353 microsatellite markers spanning the 22 autosomes at an average inter-marker distance of 10 cM.

In this study we have identified three new potential RA loci which have not been reported or detected in previous linkage analyses that used the traditional RA affection status. The new potential RA loci with strong genome-wide suggestive linkage evidence are 9q21.13 with LOD score of 2.7, 15p11.1 with LOD score of 3.2 and 20q13.33 with LOD score of 3.3. In addition to the identified new potential RA loci, we also obtained linkage evidence in chromosomal regions that are consistent with previous findings. The most significant genome-wide linkage evidence was detected on chromosome 6 in the HLA-DRB1 region which is consistent with previous reports [1–3, 14, 18, 28, 29]. The genome-wide suggestive linkage evidence detected in 1q32.1 is also similar to reported level of linkage evidence in previous studies for the region [30]. Among the candidate genes located in locus 1q32.1 are CHI3L1 (chitinase 3-like 1 (cartilage glycoprotein-39)) gene which has been implicated in local autoimmune response that leads to chronic inflammation and joint destruction in RA patients [31] and the IKBKE (inhibitor of kappa light polypeptide gene enhancer in B-cells, kinase epsilon) gene which has been reported to play a role in synovial inflammation, extracellular matrix destruction, and activation of the viral program and innate immune response in RA [32]. Also, the HRH3 (histamine receptor H3) gene located in locus 20q13.33, has a role in neurogenic inflammation [33].

We note that in a previous linkage analysis of a subset of these data (cohort 1 families), Mackay et al [1] reported significant genome-wide evidence on chromosome 6 in the same HLA-DRB1 region on which we have also detected significant evidence, and also identified a region on same chromosome with suggestive evidence plus 9 regions of nominal evidence on other chromosomes. Likewise, in a slightly larger sample from the same UK cohort 1 families, John et al [15] confirmed the linkage evidence on chromosome 6 previously reported by Mackay et al [1], but did not detect any suggestive linkage evidence outside of chromosome 6. However, in another linkage study involving both cohorts 1 and 2 families, Eyre et al [20] failed to detect any significant or suggestive linkage evidence. The fact that previously reported potential RA loci are replicated in the present study underscore the validity of the supplementary RA phenotype approach used here. The identification of new potential RA loci in this study is a demonstration of the importance of the use of supplementary phenotypes for genetic mapping of complex diseases such as RA.

It is worth noting that in this study we did not include clinical covariates such as rheumatoid factor status and erosion status that were included in some of the previous studies because these covariates measure symptoms resulting from RA affection. It is therefore not illogical to assume that an individual's cumulative hazard to RA could not be influenced by such post-RA onset phenotypes or clinical symptoms. We therefore chose not to include in the model variables that we think do not influence an individual's hazard of being affected by RA. Whether all the identified linkage regions in the present study are true positives requires further studies. We declare that the reproduction of linkage evidence at previously reported potential RA loci provides support for the validity of the definition of supplementary RA phenotype and modelling that we have used in the present study. We recognise that supplementary phenotype as defined and treated as quantitative trait in the present study may not be practicable for every complex human disease phenotype. However, for complex diseases for which this approach is applicable, this can lead to identification of new additional genetic loci that could explain part of the missing heritability [34] observed in susceptibility variants identified in association studies of complex diseases. This supplementary phenotype approach can be used to identify new candidate loci or narrow candidate regions in linkage and association studies. It is envisioned that results from both analyses of supplementary disease phenotypes and their corresponding traditional disease phenotype definitions would enhance the efficiency of the "genetic dragnet" at capturing important genetic variants contributing to complex diseases.

This present study demonstrates the potential information gain through the use of supplementary phenotypes in genetic study of complex diseases such as RA to detect new and additional linked loci that are undetected in linkage analysis of the traditional phenotypes. Also, our results provide further evidence of the involvement of multiple loci in the genetic aetiology of RA.