ARTHRITIS & RHEUMATISM Vol. 46, No. 4, April 2002, pp 946–952 DOI 10.1002/art.10149 © 2002, American College of Rheumatology Genome Scan for Quantity of Hand Osteoarthritis The Framingham Study S. Demissie,1 L. A. Cupples,1 R. Myers,2 P. Aliabadi,3 D. Levy,4 and D. T. Felson5 Objective. To search for markers linked to quantity of radiographic hand osteoarthritis (OA) in the Framingham Heart Study. Methods. The sample included 684 original cohort members and 793 offspring in 296 pedigrees. Radiographic OA features evaluated included the Kellgren/Lawrence (K/L) score, osteophytes, and joint space narrowing (0–3 scale). Four quantitative phenotypes were computed from these measurements: sum of K/L scores across hand joints, sum of osteophyte scores, sum of joint space narrowing scores, and proportion of affected joints. Prior to linkage analysis, these phenotypes were adjusted for age using a linear regression analysis from which standardized residuals were computed. The regression analysis was performed separately for each sex and each generation. The variance component model (SOLAR) was then applied to the normalized scores of the residuals. Results. The average age was 62 years for the original cohort and 54 years for the offspring. Fifty percent of the original cohort and 30% of their offspring had at least 1 affected joint (K/L score >2). Heritability ranged from 28% (proportion of joints affected with OA) to 34% (sum of K/L scores). Eight chromosomal regions indicated suggestive linkage (multipoint logarithm of odds [LOD] score >1.5) for at least 1 phenotype; LOD scores were highest for joint space narrowing, with a multipoint LOD score ⴝ 2.96 on chromosome 1p at D1S1665. Chromosomes 7, 9, 13, and 19 indicated consistent LOD score elevation for multiple OA phenotypes. Conclusion. There are several chromosomes that may harbor OA susceptibility genes. Further investigation of these regions using larger studies and finer maps will be important to confirm linkage. Osteoarthritis (OA) is the most common form of arthritis and is among the leading causes of disability throughout the world. As part of the complex etiology of this disease, there is an increased recognition that a genetic component plays an important role (1,2). Estimates of heritability of OA have ranged from 10% to 60% and may vary by the joint affected (2–4). Although it is clear that risk of OA is affected by inheritance, possible mutations or polymorphisms that may increase the risk of disease are numerous. It is conceivable that OA risk would be affected by abnormalities in such widely varying molecules as minor forms of collagens, enzymes involved in synthesis or degradation of cartilage or other tissues in or around the joints, hormone or growth factors or their receptors, and even important structural molecules outside the cartilage that give joints their shape. It is also possible that particular structural abnormalities such as ligamentous laxity play a role in increasing risk for disease (5). All of the possible structural and functional susceptibilities make it hard to make an educated guess about associations of particular polymorphisms or mutations with OA. With a disease as complex in etiology as OA, an approach to searching for sites within the genome that accounts for variability in individual risk of OA will provide insights to the most likely genes or cluster of genes that affect disease risk. In addition, recent studies (6) increasingly suggest that susceptibility linkages may be joint specific. Supported by NIH grant AR-20613 and by NIH/National Heart, Lung, and Blood Institute contract N01-HC-38038. 1 S. Demissie, PhD, MPH, L. A. Cupples, PhD: Boston University School of Public Health, Boston, Massachusetts; 2R. Myers, PhD: Boston University School of Medicine, Boston, Massachusetts; 3 P. Aliabadi, MD: Brigham and Women’s Hospital, Boston, Massachusetts; 4D. Levy, MD: National Heart, Lung, and Blood Institute, Framingham Heart Study, Framingham, Massachusetts; 5D. T. Felson, MD, MPH: Clinical Epidemiology Research and Training Unit and Arthritis Center, Boston University School of Medicine, Boston, Massachusetts. Address correspondence and reprint requests to David T. Felson, MD, MPH, 715 Albany Street, A203, Boston University School of Medicine, Boston, MA 02118. E-mail: email@example.com. Submitted for publication July 13, 2001; accepted in revised form November 2, 2001. 946 GENOME SCAN FOR HAND OA Linkage studies, those in which the genome is explored to test whether sharing of particular anonymous markers correlates with sharing of an OA phenotype, have yielded varying results. Wright et al (7), in a study of 66 sibpairs, reported that nodal hand OA was linked to chromosome 2q. Finnish investigators also reported a linkage of OA in the distal interphalangeal (DIP) joints with chromosome 2q (8). Results of another study of sibpairs with end-stage hip or knee OA, which was defined as having undergone total hip or knee replacement, also showed a linkage of hip disease with chromosome 2q (9). While all suggested linkage with 2q, the linkage sites on 2q differed substantially across studies. These linkage studies were limited by small size and by a focus on patients undergoing surgery for severe disease. There have been no large-scale linkage studies of OA from general populations. The recent focus on hip OA has led to a neglect of disease at sites where the disease is more prevalent, such as the hands and knees. There also have been no linkage studies reported from the US white population. Here, we report the results of a whole genome scan evaluating radiographic hand OA in the Framingham Heart Study (FHS) cohort and offspring, a population-based group of multiple nuclear and extended families in which both parents (the cohort) and their children (the offspring) were studied at adult ages. Because most of our middle-aged and elderly subjects had some evidence of radiographic OA, we focused our inquiry into genetic linkage on the amount of disease, summarizing the quantity of OA occurrence across hand joints. While we were also interested in knee OA, heritabilities of knee OA were not high enough to yield reliable linkage information. SUBJECTS AND METHODS The FHS is a population-based, multigenerational cohort study that began in 1948 and includes ⬎1,300 pedigrees (10). The original cohort (parents in the families studied here) was evaluated twice for the occurrence of hand OA (1967–70 and 1992–93). For the purposes of this study, the first radiograph of hands in a subject was used (11). As part of a study investigating the heritability of OA (1993–94), we evaluated the Framingham offspring who had at least 1 parent included in the earlier OA studies, using the same techniques. Hand OA was characterized radiographically using the standard Kellgren/Lawrence (K/L) grading scheme (12). Radiographs were also read for osteophytes (OST) and for joint space narrowing (JSN) using the Framingham OA atlas (13), with each scored on a 0–3 scale. To ensure that radiographs in offspring and cohort subjects would be read according to the same standards, 1 musculoskeletal radiologist (PA) read all 947 hand radiographs, with films from offspring and cohort mixed. For intraobserver agreement of ordinal-scale K/L scores, the weighted kappa for all hand joints was 0.77; for the OST score, it was 0.75; and for the JSN score, it was 0.74. All P values were ⬍0.001. When we defined joints as affected or not affected based on a K/L score of ⱖ2, the kappa value was 0.82 and the P value was ⬍0.001. DNA was extracted from whole-blood or buffy coat specimens using a standard protocol (14,15). DNA aliquots from the largest 330 families were sent to the Mammalian Genotyping Service Laboratory at the Marshfield Clinic (Marshfield, WI), where a 10-cM genome scan was performed (average heterozygosity 0.77). For quality control, 1) the family relationships were verified on the basis of all available markers by using the sib_kin program of the ASPEX package (ftp://lahmed.stanford.edu/pub/aspex/index.html); and 2) Mendelian inconsistencies were identified and eliminated by using the GENTEST program in PEDSYS (http://www.sfbr.org/sfbr/ public/software/pedsys/pedsys.html). Of these 330 families, a genome-wide scan was carried out in the 296 extended pedigrees who had information on the OA traits derived from ⬃1,477 subjects with hand measurements. Hand OA summary variables. Information from the DIP joints, proximal interphalangeal (PIP) joints, metacarpophalangeal (MCP) joints, and the carpometacarpal joint at the base of each thumb was used, because these are the joints in the hand most commonly affected by OA. There are 10 DIP joints (we treated the IP joint of the thumb as a DIP), 8 PIP joints, 10 MCP joints, and 2 thumb base joints; thus, a person could have as many as 30 joints affected. Measurements on the 30 joints were obtained for K/L, OST, and JSN. The first hand radiograph of cohort members was of the right hand only; thus, OA scores for these readings were doubled, as previously described (11). Based on these measurements, we defined 4 quantitative traits: the K/L-sum (sum of K/L scores), OST-sum (sum of OST scores), JSN-sum (sum of JSN scores), and the proportion of affected joints (K/L-proportion) (a joint was considered affected when its K/L score was ⱖ2). For the linkage analysis, each of the 4 measurements was adjusted for age using a linear regression analysis from which standardized residuals were computed. The regression analysis was performed for men and women separately within the original cohort and the offspring samples. Normalized scores were then computed based on the ranks of standardized residuals for the combined sample (i.e., the ranks of the adjusted scores). Statistical analysis. We used the computer program SOLAR (16) for linkage analysis. All linkage analyses were performed using the normalized deviates derived from the standardized residuals. SOLAR uses a variance-components approach that decomposes the total phenotypic variance into quantitative trait locus (QTL; unobservable trait affecting a major locus), residual polygenic, and residual nongenetic (or random environmental) components. Linkage is assessed by comparing the likelihood of a restricted model (a model where the QTL variance is set to 0) with a full model in which the variance due to the QTL is estimated. In the full model, the QTL variance is estimated based on the correlation between the proportion of genes shared that are identical by descent at a single marker (for 2-point linkage analysis) or multiple 948 DEMISSIE ET AL Table 1. Description of the subjects* Table 2. Heritability estimates Study variable (no. in original cohort/ no. of offspring) Original cohort Offspring Hand measurement n Heritability ⫾ SE P* Sex, % men (684/793) Age at time of hand radiograph, years (684/793) % of subjects with hand OA (680/792)† Sum of K/L scores (0–120) (680/792) % of joints with OA (680/792) Sum of osteophyte scores (0–90) (681/792) Sum of joint space narrowing scores (0–90) (681/792) 44.0 62.1 ⫾ 10.6 49.6 53.6 ⫾ 9.2 1,472 0.34 ⫾ 0.056 ⬍0.0001 50.0 11.6 ⫾ 15.5 12.7 ⫾ 19 8.4 ⫾ 11.2 30.4 4.8 ⫾ 9.7 5 ⫾ 12 3.6 ⫾ 7.1 Sum of Kellgren/Lawrence scores % of joints with osteoarthritis Sum of osteophyte scores Sum of joint space narrowing scores 1,472 0.28 ⫾ 0.057 ⬍0.0001 1,473 1,473 0.33 ⫾ 0.055 0.30 ⫾ 0.054 ⬍0.0001 ⬍0.0001 4.2 ⫾ 8.1 1.4 ⫾ 4.4 * Significance of the mean heritability. * Except where indicated otherwise, values are the mean ⫾ SD. OA ⫽ osteoarthritis; K/L ⫽ Kellgren/Lawrence. † Those with at least 1 affected joint. markers on a chromosome (for multipoint linkage analysis) and the QTL. To calculate multipoint linkage values, SOLAR uses the Fulker and Cherny regression approach (17). Decisions regarding linkage or no linkage are made using the logarithm of odds (LOD) score, which is computed as the log10 of the ratio of the likelihoods for the restricted and full models. Linkage is considered to be significant genome-wide if the LOD score exceeds 3.3, and to represent suggestive linkage if it exceeds 1.5. The cutoff we used for suggestive linkage (LOD score ⫽ 1.5, for a 10-cM genome map) is ⬃20% less than LOD score ⫽ 1.9 (for a dense map) (18). In this article we have reported results from both 2-point and multipoint linkage analyses. Since multipoint linkage analysis is more powerful and less prone to false-positive results, the results and conclusions regarding linkage are based on the multipoint LOD scores only. RESULTS Table 1 presents statistics on the subjects included in the linkage analyses. The sample in this study included 684 participants in the original cohort and 793 participants in the offspring study. The average age at which the hand measurements were obtained was 62 years for the original cohort members and 54 years for the offspring. The average unadjusted hand OA scores, constituting a sum of the scores of all joints in the hands, among the original cohort members were 11.6, 8.4, and 4.2 on the K/L, OST, and JSN scales, respectively. The corresponding values for the offspring were 4.8, 3.6, and 1.4. Fifty percent of the original cohort members and 30% of their offspring had at least 1 affected joint. The average proportion of joints with OA was 12.7% among the cohort members and 5% among the offspring. In Table 2, the estimated heritability values of the various hand OA phenotypes are presented. These estimates (⫾SE) ranged between 0.28 ⫾ 0.06 for the proportion of joints affected with OA and 0.34 ⫾ 0.06 for the sum of K/L scores. Linkage results are presented in Figures 1 and 2. In the 2-point linkage analyses, there were 11 chromosomes (1–3, 7, 9, 11, 14, 16–19) with a maximum LOD score ⬎1.5 for at least 1 of the 4 hand OA phenotypes (Figure 1A). In the multipoint linkage, there were 8 chromosomes (1, 2, 7, 9, 11–13, 19) with maximum LOD scores ⬎1.5 (Figure 1B) for at least 1 of the 4 hand phenotypes. The LOD score was highest on chromosome 7 (LOD score ⫽ 2.86 for JSN-sum) in the 2-point analysis and on chromosome 1 (LOD score ⫽ 2.96 for JSN-sum) in the multipoint linkage analysis. Figures 2a–h show the results of the multipoint linkage analyses for chromosomes with a maximum LOD score ⬎1.5. For chromosome 1, the peak LOD score was obtained at 102 cM (D1S1665; LOD score ⫽ 2.96 for JSN-sum); chromosome 2 at 48 cM (D2S405; LOD score ⫽ 2.23 for JSN-sum); chromosome 7 at 50 cM (D7S817; LOD score ⫽ 2.32 for JSN-sum); chromosome 9 at 76 cM (D9S1122; LOD score ⫽ 2.29 for JSN-sum); chromosome 11 at 77 cM (⬃D11S2371; LOD score ⫽ 1.60 for JSN-sum); chromosome 12 at 166 cM (D12S392; LOD score ⫽ 1.66 for JSN-sum); chromosome 13 at 36 cM (between D13S894 and D13S325; LOD score ⫽ 1.61 for K/L-sum); and for chromosome 19 at 52 cM (D19S433; LOD score ⫽ 1.82 for JSN-sum) and at 68 cM (D19S178; LOD score ⫽ 1.83 for K/Lsum). DISCUSSION Our linkage of hand OA, a heritable OA phenotype, in Framingham subjects suggests that there are several chromosomes that may harbor OA susceptibility genes. The ones that met the criteria for suggestive linkage (multipoint LOD score ⬎1.5) include chromosomes 1, 2, 7, 9, 11–13, and 19. For those chromosomes GENOME SCAN FOR HAND OA 949 Figure 1. A, Maximum 2-point logarithm of odds (LOD) scores by chromosome for hand osteoarthritis (OA). B, Maximum multipoint LOD scores by chromosome for hand OA. KL-Sum ⫽ sum of Kellgren/Lawrence scores; KL-Prpn ⫽ proportion of affected joints; JSN ⫽ sum of joint space narrowing scores; OST ⫽ sum of osteophyte scores. with suggestive linkage, LOD scores were highest for the JSN phenotype (Figures 2a–h). The strengths of our study are considerable. First, unlike other linkage studies that focus on subjects with joint replacement or subjects with symptoms, we had uniform, comprehensively obtained radiographs on all 950 DEMISSIE ET AL Figure 2. Multipoint LOD scores by chromosome for hand OA. See Figure 1 for definitions. GENOME SCAN FOR HAND OA subjects; all were read by 1 expert reader (PA) whose readings have high reproducibility. Further, parents and children were both studied at adult ages when OA was sufficiently prevalent to be easily characterized. We were able to evaluate several different related phenotypes, and the data on linkage using these phenotypes provide possibly valuable information suggesting different linkages with different elements of disease and help explore the consistency of linkage across phenotype definitions. To our knowledge, our study constitutes the first linkage study of hand OA in the US white population. Depending on the phenotype definitions, various studies have suggested that ⬃10–60% of the variation in OA is attributed to genetic factors (2–4). Our ageadjusted heritability estimates for the 4 hand OA phenotypes (ranging from 28% to 34%) were in the lower range of those reported by others (2,4). Part of this difference may be due to our inclusion of both men and women, since we (3) and others (8) have reported higher heritabilities in women. Also, our study includes both middle-aged and elderly subjects, the latter group being at highest risk of disease; our relatively low heritability results may suggest that those with early disease who have been the focus of other studies may have higher disease heritability, a finding similar to that in studies of other chronic diseases. This is one of the main reasons we chose to study the first hand radiographs obtained from our subjects rather than followup films obtained more than 20 years later. Our heritability results may be more generalizable to OA in the overall community than are results of studies of much younger subjects. A few linkage studies have indicated chromosomes 2q, 7p, and 11q as potential regions for markers that are linked to OA susceptibility genes. Those indicating linkage to 2q showed suggestive linkage of severe end-stage OA to 2q31 (D2S202 and D2S72 regions) (9), suggestive linkage to DIP OA near the locus IL1R1 on 2q12-q14 (8), and a significant association between nodal OA and 2q23-32 and 2q33-35 (7). In our study, however, none of these regions on 2q showed linkage to hand OA; the peak LOD score for chromosome 2 was observed on 2p at 48 cM (D2S405). Otherwise, our results indicated suggestive linkage to regions on 7p (⬃D7S817), ⬃10–30 cM proximal to 7p15-p21, the regions indicated in the Finnish study (8); and to 11q (D11S2371), ⬃10–20 cM proximal to markers D11S901, D11S1358, and D11S35, reported by Chapman et al (19). Other regions that indicated suggestive linkage, with multipoint LOD scores ⬎1.5, were on chromosomes 1, 9, 12, 13, and 19. In fact, the highest LOD score 951 in the genome-wide scan was obtained on 1p (LOD ⫽ 2.96, ⬃D1S1665 region). Although the LOD scores in our study were generally highest for the JSN phenotype, on chromosomes 7, 9, 13, and 19 there were consistent LOD score elevations for multiple hand phenotypes (Figures 2c, d, g, and h). In addition, chromosomes 7 and 19 each showed suggestive linkage evidence with multipoint LOD scores ⬎1.5 for ⬎1 phenotype (chromosome 7 for JSN-sum and K/L-proportion and chromosome 19 for JSN-sum and K/L-proportion). The high LOD scores for JSN may indicate that the chromosomal sites represent genes influencing cartilage loss rather than genes whose products are involved with bony response to disease, which might be reflected if LOD scores were highest for osteophytes. While this does not necessarily indicate that JSN should be the phenotype of interest for genetic studies, it suggests that the JSN phenotype may be separable in some circumstances from the phenotype for other radiographic features. It is important to note that the heritability estimates from this study suggest a strong genetic involvement for all 4 OA phenotypes. A number of candidate genes lie within the regions that showed suggestive linkage to OA. For example, the region of high LOD score on 2p is near the gene for tissue inhibitor of metalloproteinases 3, a matrix inhibitor of active stromelysin. On chromosome 19, our marker is close to the transforming growth factor ␤1 gene, and, on chromosome 11, it lies close to the ␣ estrogen receptor gene. It is not surprising that we did not see overlapping results (regions) among the aforementioned studies since they vary in sample sizes, analysis strategies, and, perhaps most important, in OA phenotype definitions. Moreover, the results from various studies, including ours, constitute only suggestive evidence of linkage (possibly due to insufficient statistical power). Only Chapman et al (19) found significant linkage (LOD score ⫽ 3.15) between D11S1358 and D11S35. Therefore, it will be important to undertake further investigation of several chromosomal regions using finer maps and larger studies to confirm linkage. The higher LOD scores on 1p and 2p (our study), the suggestive linkage evidence for 7p, 11q, and 2q (other studies), and the consistent signal of linkage to several hand OA phenotypes for regions on chromosomes 9, 13, and 19 (our study), in particular, warrant further investigation. The study has a number of potential limitations. First, while we were also interested in evaluating linkage to knee OA, the low heritabilities for this trait prevented further analyses. A simulation analysis of power in our 952 DEMISSIE ET AL sample for knee data indicated that we had 80% power to detect a locus when it accounts for 25% heritability or more. Our heritability estimates for knee OA were generally lower than this estimate. Second, no hip OA data were available in this population. Third, although there was very high intraobserver reliability in the assessment of OA, the presence of measurement errors or misclassification in the phenotype definitions could not be ruled out. For example, the use of only the right hand OA data for the cohort members could have resulted in phenotype misclassification and might have underestimated the presence of linkage. Finally, it is possible that the linkage findings in this study represent false-positive signals. It is also possible that true linkages were not detected because of insufficient power. While our sample size is typical of many family studies, it is clear that only QTLs with large effects can be detected through linkage analyses. Our study only has power of ⱖ80% to detect a QTL that accounts for ⬎25% of the total variation in the trait, a threshold that may not be reached for a complex trait such as OA. In summary, we report the results of a genome scan suggesting linkage of multiple chromosomal sites that may harbor genes for hand OA. While we cannot confirm previous reports of linkage to chromosome 2q, we can corroborate sites of potential interest on 7p and 11q and report additional promising sites on several other chromosomes. REFERENCES 1. Felson DT, Zhang Y. An update on the epidemiology of knee and hip osteoarthritis with a view to prevention. Arthritis Rheum 1998;41:1343–55. 2. Spector TD, Cicuttini F, Baker J, Loughlin J, Hart D. Genetic influences on osteoarthritis in women: a twin study. BMJ 1996; 312:940–3. 3. Myers RH, Couropmitree NN, Chaisson CE, Hannan MT, Zhang Y, McAlindon T, et al. Relative heritability of osteoarthritis (OA) in different hand and knee joint groups: the Framingham study [abstract]. Arthritis Rheum 1996;39 Suppl 9:S170. 4. Hirsch R, Lethbridge-Cejku M, Hanson R, Scott WW Jr, Reichle R, Plato CC, et al. Familial aggregation of osteoarthritis: data from the Baltimore Longitudinal Study on Aging. Arthritis Rheum 1998;41:1227–32. 5. Jimenez SA, Dharmavaram RM. Genetic aspects of familial osteoarthritis. Ann Rheum Dis 1994;53:789–97. 6. Loughlin J, Mustafa Z, Irven C, Smith A, Carr AJ, Sykes B, et al. Stratification analysis of an osteoarthritis genome screen: suggestive linkage to chromosomes 4, 6, and 16. Am J Hum Genet 1999;65:1795–8. 7. Wright GD, Hughes AE, Regan M, Doherty M. Association of two loci on chromosome 2q with nodal osteoarthritis. Ann Rheum Dis 1996;55:317–9. 8. Leppavuori J, Kujala U, Kinnunen J, Kaprio J, Nissila M, Heliovaara M, et al. Genome scan for predisposing loci for distal interphalangeal joint osteoarthritis: evidence for a locus on 2q. Am J Hum Genet 1999;65:1060–7. 9. Loughlin J, Sinsheimer JS, Mustafa Z, Carr AJ, Clipsham K, Bloomfield VA, et al. Association analysis of the vitamin D receptor gene, the type I collagen gene COL1A1, and the estrogen receptor gene in idiopathic osteoarthritis. J Rheumatol 2000;27: 779–84. 10. Dawber TR, Kannel WB, Lyell LP. An approach to longitudinal studies in a community: the Framingham study. Ann N Y Acad Sci 1963;107:539–56. 11. Felson DT, Couropmitree NN, Chaisson CE, Hannan MT, Zhang Y, McAlindon TE, et al. Evidence for a Mendelian gene in a segregation analysis of generalized radiographic osteoarthritis: the Framingham Study. Arthritis Rheum 1998;41:1064–71. 12. Kellgren JH, Lawrence J. Radiological assessment of osteoarthritis. Ann Rheum Dis 1957;16:494–502. 13. Felson DT, McAlindon TE, Anderson JJ, Naimark A, Weissman BW, Aliabadi P, et al. Defining radiographic osteoarthritis for the whole knee. Osteoarthritis Cartilage 1997;5:241–50. 14. Gross-Bellard M, Oudet P, Chambon P. Isolation of high-molecular-weight DNA from mammalian cells. Eur J Biochem 1973;36: 32–8. 15. Miller SA, Dykes DD, Polesky HF. A simple salting out procedure for extracting DNA from human nucleated cells. Nucleic Acids Res 1988;16:1215. 16. Almasy L, Blangero J. Multipoint quantitative-trait linkage analysis in general pedigrees. Am J Hum Genet 1998;62:1198–211. 17. Fulker DW, Cherny SS. An improved multipoint sib-pair analysis of quantitative traits. Behav Genet 1996;26:527–32. 18. Lander E, Kruglyak L. Genetic dissection of complex traits: guidelines for interpreting and reporting linkage results. Nat Genet 1995;11:241–7. 19. Chapman K, Mustafa Z, Irven C, Carr AJ, Clipsham K, Smith A, et al. Osteoarthritis-susceptibility locus on chromosome 11q, detected by linkage. Am J Hum Genet 1999;65:167–74.