Genetic Analysis of Multiplex Rheumatoid Arthritis Families

D Bali1,2, S Gourley1,3, DD Kostyu2, N Goel1,4, I Bruce3, A Bell3, DJ Walker5, K Tran6,DKZhu7, TJ Costello7, CI Amos7 and MF Seldin1,6 1Department of Medicine; 2Department of Immunology, Duke University Medical Center, Durham, NC, USA; 3Department of Rheumatology, The Queen’s University, Belfast, Northern Ireland; 4The University of Texas Medical Branch, Galveston, TX, USA; 5Musculoskeletal Unit, Freeman Hospital, Newcastle upon Tyne, UK; 6Rowe Program in Genetics, Departments of Biological Chemistry and Medicine, University of California, Davis, CA, USA; 7Department of Epidemiology, University of Texas, MD Anderson Cancer Center, Houston, TX, USA

To examine the genetic contribution of HLA and non-HLA genes in the etiopathogenesis of rheumatoid arthritis (RA), 60 Caucasian multiplex families were identiﬁed and DNA analyzed for over 52 markers including DRB1, DQA1 and DQB1 alleles. Many of the markers were chosen because of close proximity to candidate genes suggested by previous studies or models of pathogenesis. Sibling pair analysis (SIBPAL), relative pair analysis (RELPAL) and linkage studies using two different models of inheritance suggested linkage for the MHC and two additional chromosomal regions: chromosome 2 (D2S443 near CD8 and IG␬; 2p13–2p11.1), and chromosome 15 (CYP19–estrogen synthase; 15q15). No support was found for two chromosomal regions, 1p36 and 3q13, recently suggested by other studies. We used transmission disequilibrium testing (TDT), conditional logistic regression, and segregation analysis to study the contributions that the shared epitope and TNF-c have in contributing to risk for RA. These studies provide additional evidence that the association of HLA alleles in RA patients from multiplex families is similar to that observed in sporadic disease, suggest candidate regions for further analysis and ﬁnd additional support for an association of TNF-c alleles with RA susceptibility.

Keywords: rheumatoid arthritis; genetics; MHC

Introduction Additional hypotheses suggesting an important role for DQ alleles or protective effects of DRB1 alleles have also A variety of evidence suggests that susceptibility to rheu- been advanced.9,10 Another recent study strongly sug- matoid arthritis (RA) is in part determined by genetic gests that the TNF gene linked to the MHC is also an predisposition. Twin studies show several-fold higher independent susceptibility gene11 but this has not been levels of concordance in identical compared with non- supported in other studies12,13 that were performed on 1 identical twins and first-degree relatives have an age- different populations and used different methods. standardized relative risk of approximately 8 compared Further studies will be needed to understand the relative 2 with unrelated individuals within similar populations. importance of specific alleles in disease susceptibility and However, only analyses of the MHC region have pro- severity. Estimation of the genetic contribution of HLA vided convincing support for genetic loci that are critical suggests that less than 40% of RA susceptibility can be in the etiopathogenesis of RA. The initial MHC studies attributed to MHC-linked genes.14,15 3 by Stastny showing an association of RA with HLA-DR4 Multiple studies have reported a possible association have subsequently been confirmed in multiple ethnic or linkage with a variety of non-MHC genes and loci, 4 5–7 groups with some exceptions. Further analyses of most notably the T cell receptor B chain. Positive popu- these HLA associations have indicated that the majority lation associations with TCRB alleles have been reported of RA-associated HLA-DRB1 alleles share a common by some groups,16,17 but not others.18,19 These conflicting structural feature at positions 70–74 of the DRB1 chain: results could be due to the confounding effects of popu- Q K R A A (DRB1*0401) or Q R R A A (DRB1*0404, 0405, lation stratification and the discontinuity of linkage dis- 0408, 0101, 1402) or R R R A A (DRB1*1001). This finding equilibrium across the entire TCRB complex.20 A single 8 has given rise to the shared epitope hypothesis. linkage study using the affected sibling (sib) pair method, found increased sib pair sharing for several specific TCRBV regions.21 Other studies have found a significant but weak association of a TCR␣ chain polymorphism Correspondence: Dr MF Seldin, Rowe Chair, Molecular Medicine and with RA.22 Recent studies indicate possible linkage of RA Human Genetics, MS1A Room 4303, University of California, Davis, CA with several other non-MHC loci or genes,23–26 but none 95616-8500, USA This project was supported by US NIH grants AR44422 and have been independently confirmed. The current study AR39162 (CIA and MFS); and NIH/NIAID training Grant No. represents our initial studies of a substantial collection of T32AI07217 (DB and NG) multiplex RA families. Our goals are to ascertain both the Genetic analysis of RA families D Bali et al 29 role of the MHC in familial RA and to identify candidate Evidence for independent effect of TNF-c novel loci that will stimulate additional analyses in other The studies presented above showed linkage to both class data sets. II and class III MHC loci. Since a previous study has suggested linkage of TNF-c that was independent of the shared epitope, we utilized several modeling schemes Results to assess potential independent effects of the alleles of TNF-c vs the shared epitope. First, the effects of TNF-c Linkage studies of candidate genes and chromosomal and the shared epitope were assessed singly using a TDT regions analysis. Using a Fisher exact test, the results showed sig- To identify loci that are linked to RA susceptibility genes, nificant effects of both the shared epitope (31.5 alleles both molecular HLA (DRB1, DQA1 and DQB1) and transmitted vs 11.5 alleles not transmitted, P = 0.002, with highly polymorphic markers (microsatellites) in over 40 an odds ratio of 2.74) and TNF-c (33 TNF-cl alleles trans- chromosomal regions were typed in a total of 190 subjects mitted vs 16 not transmitted, P = 0.02 with an odds ratio from 60 multiplex RA families that included 48 affected of 2.06). sibling pairs. Most of the 40 chromosomal regions were For joint consideration of both TNF-c and the shared selected to include putative candidate genes as shown in epitope, conditional logistic regression was utilized. For Table 1. Each family had at least two affected individuals this analysis, whenever one parent was missing and the who met ACR criteria for RA and included 48 full sibling remaining parent and child were heterozygous, weights pairs concordant for disease. HLA typing indicated that were assigned to the possible outcomes: if both parent for the 142 affected members in these families, 68 had and child were heterozygous for both loci, we assigned one copy of the shared epitope, 35 had two copies of the shared epitope and 39 did not possess a shared epitope. one quarter weight to each of the four possible trans- To assess linkage, a sibling pair analysis was first per- missions. Similarly, if the parent-child pair were hetero- formed using the SIBPAL algorithms. Nominal levels of zygous for only one locus, we assigned one half weight significance were reached for five different chromosomal to each of the uncertain transmissions. Whenever both regions including: the MHC; TCRB; and markers on chro- parents were available or either a parent or child in a mosome 2 (D2S443 near CD8 and IG␬, and D2S273; 2p13– parent-child pair was homozygous, we were able to 2p11.1) and chromosome 15 (CYP19–estrogen synthase; assign unit weight to the observed transmission of an 15q15). For the MHC region, several loci achieved sig- allele to the affected offspring. Joint analysis via con- nificant scores in affected sibling pairs including: DQA1, ditional logistic regression showed a borderline signifi- P = 0.046; DQB1, P = 0.053; TNF-b, P = 0.006; and TNF- cant effect of TNF-c independent from shared epitope (P c, P = 0.007, with approximately 60% sharing of alleles in = 0.056, odds ratio = 2.1) but not of the shared epitope affected pairs (Table 1). Similarly affected sibling pairs (P = 0.50, odds ratio = 1.41). showed significant scores for D2S443 (P = 0.027) and Segregation studies were used as a secondary analysis CYP19 (P = 0.022) (Table 1). For TCRB and D2S273, sig- to assess the joint effects of TNF-c and the shared epitope. nificant scores were observed only in unaffected sibling In this analysis we treated the rheumatoid arthritis obser- pairs (P = 0.017 and P = 0.042, respectively). vations as outcomes, with presence of either the shared A relative pair analysis was also used to assess linkage epitope or TNF-c and the sex of an individual as predic- in this data set that included additional family members tors. Genetic models were fitted under a logistic function (Table 1). This analysis provided stronger evidence in that assumes that the affection risk is a linear function of support of linkage for most of the MHC region markers; the inferred genotype of an individual and the predictors. DRB1, P = 0.002; DQA1, P = 0.044; DQB1, P = 0.001; and To assess the significance of each predictor, a likelihood TNF-b, P = 0.004. Similar levels of significance were ratio test was utilized. This test was defined as the likeli- obtained for two of the non-MHC markers that were sug- hood of disease occurrence under a null model in which = gested by the sibling pair analyses: D2S443, P 0.031; a predictor is eliminated compared to the likelihood = and CYP19, P 0.056. Three additional markers, under an alternate model in which that predictor is D1S1153, D2S114 and D10S521 were also suggested. No included. The results from this segregation analysis linkage was observed for multiple potential candidate showed a trend for effects from both TNF-c and the region genes using either the sibling pair or relative pair shared epitope. Analyses of TNF-c and the shared epi- analysis including those for: CD28, CD80, CD86, FcGR1, tope singly showed significant effects for the TNF-c. The IL2, IL3, IL5, IL10, IL14 and IL5RA (Table 1). estimate of the odds ratio for TNF-c was 1.78 (␹2 = 16.09, Further analysis was also performed using both domi- P = 6 × 10−6) and for the shared epitope the estimated nant and recessive models in which penetrance of RA − odds ratio was 1.63 (␹2 = 17.06, P = 4 × 10 6). From the susceptibility was set to 15% for male carriers and 30% joint analysis, the estimate of the odds ratio was 1.65 for for female carriers. Significant linkage (P Ͻ 0.05) was TNF-c (␹2 = 4.06, P = 0.04) and for the shared epitope, found for both D2S443 and CYP19 (Table 2). For the 1.91 (␹2 = 3.08, P = 0.08). Female sex was associated with recessive model under a homogeneous model in which all of the families are assessed to have the same suscepti- an odds ratio of 1.97. These results are consistent with bility alleles, the maximum LOD scores for D2S443 (LOD independent effects of both TNF-c and shared epitope on = 0.894) and CYP19 (LOD = 0.774) were similar to those risk for rheumatoid arthritis. These results were similar observed for the MHC class II loci: DQA1 (LOD = 0.878), to those obtained using logistical regression. However, in DQB1 (LOD = 0.947), DRB1 (LOD = 0.751), and DR (LOD segregation analysis we used the entire pedigree struc- = 1.14). No evidence of linkage was observed for other ture in the analysis rather than just affected children and markers suggested by the RELPAL algorithm (D1S1153, their parents, who were used for the TDT and conditional D2S114, and D10S521). logistic regression analyses. Genetic analysis of RA families D Bali et al 30 Table 1 Sibling pair and relative pair analyses of multiplex RA families

Locus Sibpal Relpal Genomics

Affecteda Pairs Mean SD T-values P-values P Model 2 Mapb Candidatec Positiond

D1S228 0 10 0.48 0.31 0.34 0.371 TNFR2 1, 35 2 48 0.45 0.28 −1.16 1.000 0.8982 1, 32 CD30 1, 35 D1S233 0 9 0.45 0.27 −0.57 1.000 2 21 0.46 0.34 −0.59 1.000 0.5644 1, 60 IL14 1, 59 D1S255 0 10 0.61 0.25 9.02 0.000 2 48 0.51 0.30 0.24 0.404 0.4557 1, 66 D1S211 0 7 0.37 0.35 −0.95 1.000 2 39 0.49 0.35 −0.26 1.000 0.3688 1, 75 D1S519 0 10 0.49 0.22 −0.07 1.000 2 48 0.56 0.34 1.21 0.116 0.1067 1, 85 D1S216 0 10 0.41 0.29 −1.04 1.000 2 44 0.49 0.33 −0.25 1.000 0.4777 1, 107 D1S236 0 8 0.60 0.33 0.88 0.201 2 45 0.37 0.25 −3.38 1.000 0.9478 1, 132 D1S305 0 10 0.60 0.40 0.68 0.255 CD2 1, 158 2 48 0.49 0.30 −0.16 1.000 0.0547 1, 164 CD1 1, 162 IL6R 1, 164 1, 172 D1S1153 0 10 0.49 0.47 −0.07 1.000 CD48 1, 172 2 48 0.50 0.30 0.02 0.491 0.0106**e 1, 170 FcGR2/3 1, 175 D1S218 0 10 0.46 0.27 −0.49 1.000 2 48 0.45 0.29 −1.17 1.000 0.5049 1, 196 APT1LG1 1, 195 (FAS) AT3 0 10 0.58 0.28 0.52 0.305 2 48 0.44 0.27 −1.50 1.000 0.4577 1, 199 D1S202 0 9 0.37 0.33 −2.20 1.000 2 47 0.48 0.32 −0.50 1.000 0.5917 1, 205 F13B 0 10 0.51 0.33 0.12 0.452 2 48 0.44 0.32 −1.34 1.000 0.5744 1, 216 D1S373 0 10 0.48 0.34 −0.17 1.000 2 48 0.47 0.34 −0.69 1.000 0.3498 1, 220 IL10 1, 218 D1S249 0 10 0.36 0.31 −1.34 1.000 CR1/CR2 1, 228 2 48 0.48 0.33 −0.36 1.000 0.4564 1, 225 CD34 1, 228 D1S103 0 10 0.53 0.30 0.32 0.378 2 48 0.37 0.34 −2.55 1.000 0.9189 1, 270 D2S423 0 10 0.49 0.28 −0.17 1.000 2 48 0.45 0.31 −1.16 1.000 0.5598 2, 22 D2S405 0 10 0.59 0.19 1.47 0.084 2 48 0.48 0.27 −0.62 1.000 0.3649 2, 54 D2S443 0 11 0.42 0.21 −1.27 1.000 TGFA 2, 92 2 48 0.58 0.27 1.97 0.027** 0.0306** 2, 100 CD8 2, 113 IGK 2, 113 D2S114 0 10 0.44 0.37 −0.51 1.000 2 48 0.50 0.32 0.05 0.480 0.0372** 2, 147 IL1RN 2, 134 D2S335 0 10 0.70 0.37 1.74 0.056 2 48 0.38 0.30 −2.81 1.000 0.9453 2, 182 D2S273 0 10 0.63 0.21 1.89 0.042** 2 48 0.47 0.26 −0.73 1.000 0.3019 2, 192 D2S117 0 10 0.56 0.28 0.64 0.267 2 48 0.44 0.30 −1.35 1.000 0.339 2, 201 CTLA4/CD28 2, 203 IL8RB 2, 209 D2S427 0 10 0.51 0.31 0.06 0.478 2 48 0.44 0.25 −1.75 1.000 8773 2, 245 D2S440 0 10 0.54 0.29 0.34 0.369 2 48 0.51 0.27 0.35 0.364 0.3287 2, 255 D3S1304 0 10 0.39 0.29 −1.85 1.000 2 48 0.50 0.29 0.09 0.466 0.4389 3, 16 IL5RA 3, 16

Continued Genetic analysis of RA families D Bali et al 31 Table 1 Continued

Locus Sibpal Relpal Genomics

Affecteda Pairs Mean SD T-values P-values P Model 2 Mapb Candidatec Positiond

D3S1554 0 10 0.55 0.31 0.51 0.309 2 48 0.50 0.26 −0.12 1.000 0.595 3, 33 TGFBR2 3, 36 D3S1266 0 10 0.46 0.37 −0.36 1.000 2 48 0.51 0.31 0.21 0.417 0.2419 3, 46 D3S1767 0 10 0.54 0.34 0.37 0.358 2 48 0.42 0.25 −2.29 1.000 0.9336 3, 69 NKTR 3, 67 D3S1598 0 3 0.46 0.06 −1.00 1.000 2 28 0.44 0.38 −0.79 1.000 0.1032 3, 101 D3S1753 0 10 0.52 0.24 0.27 0.394 2 48 0.52 0.26 0.55 0.291 0.2505 3, 111 D3S2460 0 8 0.44 0.31 −0.56 1.000 2 35 0.47 0.26 −0.78 1.000 0.4247 3, 135 D3S4523 0 5 0.25 0.15 −3.82 1.000 2 29 0.51 0.30 0.14 0.440 0.5825 3, 137 CD80, CD86 3, 140 D4S1559 0 10 0.41 0.26 −0.93 1.000 2 48 0.43 0.26 −1.98 1.000 0.9753 4, 103 NFKB 4, 110 D4S406 0 10 0.41 0.35 −0.72 1.000 EGF 4, 117 2 48 0.40 0.31 −2.13 1.000 0.9992 4, 115 IL2 4, 126 D5S816 0 10 0.53 0.30 0.27 0.395 IL3, IL4 5, 135 2 48 0.49 0.30 −0.32 1.000 0.3583 5, 140 IL5, IL13 5q31 Asthma region DRB1 0 10 0.38 0.20 −1.88 1.000 2 48 0.58 0.33 1.60 0.059 0.002** 6, 44–48 MHC 6, 44–48 DQA1 0 10 0.43 0.26 −0.88 1.000 2 48 0.57 0.27 1.72 0.046** 0.0441 6, 44–48 MHC 6, 44–48 DQB1 0 7 0.41 0.19 −1.21 1.000 2 41 0.58 0.30 1.65 0.053** 0.0009** 6, 44–48 MHC 6, 44–48 TNF-a 0 11 0.33 0.26 −1.80 1.000 2 46 0.57 0.29 1.60 0.058 0.1575 6, 44–48 MHC 6, 44–48 TNF-b 0 10 0.58 0.22 0.90 0.192 2 47 0.57 0.20 2.59 0.006** 0.0039** 6, 44–48 MHC 6, 44–48 TNF-c 0 10 0.49 0.15 0.50 0.313 2 47 0.55 0.15 2.54 0.007** 0.0356** 6, 44–48 MHC 6, 44–48 D7S495 0 10 0.68 0.28 2.05 0.031** 2 48 0.46 0.32 −0.98 1.000 0.4192 7, 147 TCRBN 0 11 0.69 0.27 2.36 0.017** 2 48 0.42 0.29 −1.89 1.000 0.9732 7, 149–157 TCRB 7, 149–157 D7S483 0 10 0.65 0.30 1.58 0.070 2 48 0.39 0.29 −2.54 1.000 0.7751 7, 167 NOS3(163) 7, 163 D10S521 0 10 0.55 0.37 0.43 0.336 2 48 0.50 0.32 0.00 0.499 0.0484** 10, 131 FAS(123) 10, 123 D10S187 0 10 0.69 0.41 1.44 0.091 2 47 0.41 0.30 −2.15 1.000 0.1323 10, 145 D14S125 0 10 0.52 0.19 0.42 0.342 TGFB3 14, 73–76 2 48 0.50 0.29 0.09 0.465 0.4125 14, 70 IL8 14, 124 D14S65 0 10 0.46 0.37 −0.06 1.000 IGH 14, 124 2 48 0.39 0.26 −2.83 1.000 0.9996 14, 115

Continued

Discussion members of several populations. For familial RA the linkage of HLA polymorphisms was previously reported by The current study suggests that susceptibility to RA is Hasstedt27 and Cornelis,23 in contrast to McDermott21 and linked to several chromosomal regions. Not surprisingly, Shiozawa25 who found no signiﬁcant evidence of HLA the most compelling data are those for the HLA region linkage. Further support for a strong role of the MHC in genes that have previously been strongly implicated in the current study was provided by a TDT analysis that association studies comparing affected and unaffected indicated a strong association for the shared epitope as Genetic analysis of RA families D Bali et al 32 Table 1 Continued

Locus Sibpal Relpal Genomics

Affecteda Pairs Mean SD T-values P-values P Model 2 Mapb Candidatec Positiond

CYP19 0 12 0.52 0.40 0.19 0.427 2 48 0.58 0.26 2.06 0.022** 0.0564 15, 43 CYP19 15, 43 D20S112 0 10 0.37 0.24 −1.40 1.000 2 48 0.41 0.30 −1.97 1.000 0.9815 20, 40 D22S315 0 10 0.59 0.27 0.99 0.171 2 48 0.46 0.26 −1.18 1.000 0.1161 22, 16 IGL 22, 13

aThe number of affected individuals in each pair. bThe chromosome followed by the approximate map position in the Genethon genetic map as presented in the Gene Map of the Human Genome (http://www.ncbi.nlm.nih.gov/SCIENCE96/). Some positions were determined by interpolation of additional data contained in the Genome Database (GDB). (http://gdbwww.gdb.org/gdb/gdbtop.html), the Cooperative Human Linkage Consortium (CHLC) (http://www.chlc.org/) or the Centre d’Etude du Polymorphisme Humain (CEPH) (http://www.cephb.fr/cephdb/) databases. cSymbols of candidate genes that are near the genetic markers analyzed. dThe chromosome and approximate map position of the candidate genes as human radiation hybrid mapping (see: Gene Map of the Human Genome, URL given above), or interpolation of additional data contained in CEPH, CHLC, or GDB databases (also see above). eMeets nominal level of signiﬁcance (see text).

well as significant association for the TNF-c polymor- Several non-MHC region loci were also linked to RA phism. Together, these results strongly suggest that the in the current analysis. These must be regarded as only effect of the HLA-DRB1 polymorphisms and the shared suggestive since the results only reach statistical signifi- epitope is similar to that observed in ‘sporadic’ disease. cance when considered alone. It is notable that a similar More controversial is the observation that TNF-c poly- level of significance to that of the MHC loci was achieved morphisms may confer a risk independent of that for two of the non-MHC region loci, D2S443 (on chromo- observed for the shared epitope. The studies by Mul- some 2 near CD8 and IG␬; 2p13–2p11.1), and CYP19 cahy11 strongly suggested that the TNF-c1 allele was an (15q15), using sibling pair analyses and model-dependent independent risk factor. Both a logistical regression linkage analysis. analysis and segregation studies favored an association In contrast to the sib pair analysis, the relative level of with TNF-c1 independent of the shared epitope in the significance for D2S443 and CYP19 compared with MHC current study. Thus, our results using a completely differ- loci was much lower when a relative pair analysis was ent data set than Mulcahy,11 and a different analytic strat- performed. Also, markers in two additional non-MHC egy, provide support for the hypothesis that a second regions, D1S1153 (Chr 1 at 170 cM near CD2, CD1, IL6R, MHC-associated risk factor is important in some popu- CD48 and FcGR2/3; 1p12–1q21) and D10S521 (Chr 10 at lations. Opposing studies have either failed to show any 131 cM near FAS; 10q23–10q24), reached nominal levels independent effect of TNF polymorphisms12,13 or only an of significance using relative pair analysis but did not in association with severity of disease.28,29 The inconsistency the sib pair studies. The differences in the SIBPAL and between these studies may reflect study differences RELPAL results may reflect the sensitivity of the including the populations analyzed and analytic methods RELPAL algorithm to dominant factors segregating in the used as well as the average age of the affected subjects complete family and the inclusion of unaffected individ- and severity of disease. It is of possible importance that uals in the analysis. The LOD score analysis under a both the current study and the study of Mulcahy11 were dominant model failed to confirm effects from D1S1153 primarily composed of families from Ireland. and D10S521 suggesting that these results reflect spuri- It should be noted that it is the TNF-c2, not TNF-c1, ously positive findings from RELPAL. polymorphism which has been shown to be in linkage Finally, it is notable that two other regions, TCRB and disequilibrium with a TNF promoter variant that is asso- D2S273, reached nominal levels of significance in the sib- ciated with increased TNF production in response to ling pair studies, but only in unaffected sib pairs. The some pro-inflammatory stimuli.30 Thus both our results lack of linkage in affected sibling pairs raises the question and those of Mulcahy11 show a seemingly paradoxical of whether these results are the consequence of fully association with a low TNF-producing polymorphism. penetrant susceptibility alleles or are spurious positive This raises the question as to whether the independent findings. effect may be due to another MHC-linked gene(s). Other Association was also examined for those markers sug- MHC region genes linked to RA include HLA-DMA, gested by the linkage results. In contrast to the MHC loci, HLA-DMB and polymorphisms in prolactin microsatel- TDT analyses of these markers did not show an associ- lites.31,32 While the latter was in linkage disequilibrium ation (data not shown). This suggests that these loci, if with DRB1 in the population studied, the DMA effects confirmed in additional linkage studies, are markers for appeared to be independent of DRB1. Additional studies susceptibility genes in these regions that are not in link- examining more MHC region polymorphisms will be age disequilibrium with the true susceptibility genes. For necessary to allow further elucidation of the role of other association-based tests like TDT, markers at very closely non-class II genes. However, these results may in part adjacent chromosomal positions may be necessary for explain the variable effect of the shared epitope and spe- positive tests in most populations. cific DRB1 alleles that are seen in different populations.5,6 Other non-MHC candidate regions including the Genetic analysis of RA families D Bali et al 33 Table 2 Model dependent linkage analysis of multiplex RA family-based association techniques (eg TDT) may be families necessary for uncovering and confirming loci with relatively small genotypic risks.35 This may account for the (a) LOD scores for markers significant at the 0.05 level (dominant lack of linkage in the current study or previous linkage model) studies for candidate genes such as IL10 suggested using association methods.36 Conversely, the association ana- Marker Analysis under Analysis under heterogeneity lytic strategies may require very close proximity between homogeneity the susceptibility gene (ie ¿ 1 cM in non-founder populations) and the marker studied. This is in contrast LODa ⌰b LOD ⌰ Linkedc to linkage methods that are not very sensitive to small chromosomal intervals (ie linkage can be detected for D2S443 0.53 0.2 0.55 0.2 1 Ͼ DQA1 0.92 0.12 0.92 0.12 0.98 10 cM intervals). The current study identifies DQB1 0.90 0.12 0.93 0 0.55 additional loci that should be included in future studies DRB1 0.67 0.2 0.67 0.2 1 of RA to further define this complex genetic disease. DR 1.43 0.1 1.54 0 0.63 TNF-a 0.17 0.28 0.17 0.28 1 TNF-b 1.12 0.1 1.12 0.1 1 Patients and methods TNF-c 1.11 0 1.12 0 1 D7S495 0 0.5 0 0.5 1 Subjects TCRB 0 0.5 0 0.5 1 CYP19 0.51 0.2 0.82 0 0.39 A total of 60 multiplex Caucasian families was identified D1S1153 0 0.5 0.00 0.5 1 in which at least two members met ACR criteria for the D2S114 0 0.5 0.00 0.5 1 diagnosis of RA.37 These families consisted of 47 from D10S521 0 0.5 0.00 0.5 1 Northern Ireland, 10 from Newcastle, United Kingdom, and three from the USA. Of the 142 affected individuals (b) LOD scores for markers significant at the 0.05 level (Recessive from these families who were included in the analysis, model) 122 were rheumatoid factor (RF) positive, 52 had nodular D2S443 0.89 0.18 0.89 0.18 1 disease and 114 had documented erosions on hand films. DQA1 0.88 0.16 0.88 0.16 0.99 We obtained documentation of either a positive RF or DQB1 0.95 0.16 0.99 0 0.47 DRB1 0.75 0.18 1.07 0 0.42 rheumatoid nodules from a total of 133 of the 142 affected DR 1.14 0.16 1.14 0.16 0.95 patients studied. There were 104 females and 38 males TNF-a 0.44 0.24 0.44 0.24 1 and the average age (± SD) of disease onset was 40.3 ± TNF-b 1.10 0.14 1.10 0.14 1 13.7 years of age. In addition, 48 unaffected family mem- TNF-c 1.26 0.08 1.26 0.08 1 bers were also ascertained for determining identity by D7S495 0 0.5 0 0 0.03 descent and linkage analyses. TCRB 0 0.5 0 0.5 0 CYP19 0.77 0.2 0.77 0.2 1 D1S1153 0 0.5 0.00 0.5 1 Sample preparation D2S114 0 0.5 0.00 0.5 1 DNA templates for PCR analyses were prepared either D10S521 0.003 0.42 0.05 0 0.08 from peripheral blood mononuclear cells using modifi- cations of standard techniques38 or prepared from buccal aLOD scores were calculated assuming a genetic model for which cell brushings. Briefly, for buccal cell DNA, after brush- the penetrances were 15% for male carriers of susceptibility and ing the cheek mucosal surfaces, brushes were placed in 30% among female carriers and that the sporadic risk was 0.25% ␮ ° among males and 0.75% among females. For the dominant model 600 l50mm NaOH and heated at 95 C for 10 min. After we assumed a gene frequency of 2% and for the recessive model, vortexing, the brush was discarded and the sample neu- we assumed a gene frequency of 14.1%. bRecombination frequency. tralized with 50 ␮l of 1.0 m Tris/HCl, pH 8.0. Following cLinked proportion under heterogeneity. brief centrifugation (14000 rpm for 1 min), the suspen- sion was transferred to new tubes and stored as stock solutions at 4°C (no loss of DNA in empirical studies over IDDM9 region of chromosome 3 and the proximal region 3 years). For working solutions, stock DNA was sheared of human chromosome 1 suggested in other preliminary and resuspended in T10E1 at a 1–4 dilution and typical studies,23,25 were not confirmed in the current study. PCR reactions utilized 1 ␮l of this working solution. Similarly, these studies did not find a significant linkage for candidate loci suggested in the current study. HLA typing Another putative susceptibility locus at Xq27 recently Oligotyping was performed by PCR amplification using suggested25 was not examined in the current study. Of generic and family or subtype-specific primers and an additional note, another report33 has provided some sup- ELISA, using fluorescein-labeled, sequence-specific oli- port for our initial studies suggesting linkage with CYP19 gonucleotide probes, as previously described.39 Primers at 15q15.26 This study examining 200 sib pairs found mar- and probes target the same sequences used in the 11th ginal linkage between RA and CYP19 (P = 0.04).33 How- and 12th International Histocompatibility Testing Work- ever, the general failure to confirm suggestive results in shops.40,41 different data sets may reflect the relatively low power for linkage methods to detect complex disease loci that DNA marker analyses may confer genotypic risk ratios of less than 4-fold. Much A set of 47 polymorphic microsatellite markers with an larger linkage sets including the current efforts by several average heterozygosity of 0.85 was selected from candi- groups are necessary to provide more definitive results.34 date gene regions as an initial part of a genome-wide Furthermore, the use of case-controlled association or screening effort. Most markers were obtained from Genetic analysis of RA families D Bali et al 34 Research Genetics, Inc (Huntsville, Al, USA) with for- cordant for disease status, the influence of a genetic factor ward-primers already labeled with one of the four fluor- induces an excess sharing of alleles identical by descent, omores (HEX, TET, 6-FAM, JOE phosphormidites). Some while in sib pairs discordant for disease status, genetic of the tetranucleotide repeat markers were kindly pro- effects confer decreased allele sharing. For diseases with vided by Drs L Ballard and R White of the University of low penetrance, such as rheumatoid arthritis, affected sib Utah. Other markers were selected from published pairs are most informative for detecting genetic factors.45 sequences and synthesized commercially (Department of The statistical analysis divides all sibships into all poss- Plant Biology, Duke University, Durham, NC, USA). For ible pairs, but then estimates the appropriate degrees of the latter, forward markers were tagged with a fluor- freedom for the test conditional upon correlations that escent dye using a standard aminolink protocol (ABI may exist among the pairs (SAGE).44 Sib pair analyses manual). Polymerase chain reactions (PCR) were perfor- were restricted to include only full sib pairs, as only a med in 96-well microtiter plates in a 5-␮l reaction volume very limited number of half-sib pairs were available for using 10 ng of sheared human DNA, 0.5 ␮lof10× reac- analysis and their inclusion could have decreased the tion buffer (Perkin-Elmer Cetus, Norwalk, CT, USA), reliability of our results, by increasing the sampling 125 ␮m of each dNTP, 8 ␮m of forward and reverse pri- variability of test statistics. We also used the RELPAL mers, and 0.25 U of Taq DNA Polymerase (Bethesda program44 which can include information from all rela- Research Laboratories, Gaithersburg, MD, USA). tive pairs. For binary traits such as affection with RA, this BIOOVEN III (Biotherm Corporation) or GeneAmp PCR procedure tests the ibd sharing for each class of relative System 9600 (Perkin-Elmer Cetus) were used for PCR, pair among those concordant for disease with those indi- using mostly Weissenbach conditions (92°C for 5 min, viduals discordant for disease. After analyses are conduc- followed by 35 cycles of 40 s at 92°C, 30 s at 55°C, and a ted within each type of relative pair, the results are com- final extension of 2 min at 72°C), with some variations bined using a weighting function that is proportional to

made in the annealing temperatures, MgCl2 concen- the variance within each relative class. This variance can tration, and number of cycles performed to optimize the be estimated from the data or formed by consideration individual marker conditions. of the number of pairs of each type and by assuming no recombination between the marker and disease locus. We Pooling and multiplexing PCR products chose to use the latter method because with our some- In order to maximize efficient typing, a multiplex proto- what limited sample size we were concerned about the col similar to those described previously42 was used for generalizability of our findings, if a parameter had to be many of the markers. Groups of 8–12 markers labeled estimated from each type of relative that was studied. For with three different dyes (FAM, TET, HEX or FAM, JOE, any results that were found to be significant at the 0.05 TAMRA) were pooled together for multiplexing. Internal level from either SIBPAL or RELPAL analysis, we also lane size standards (GeneScan 500-ROX for filter A or applied model-dependent linkage analysis. We assumed GeneScan 500-TAMRA for filter B), 1.25 ␮l of formamide a genetic model for which the penetrance of susceptibility and 0.25 ␮l of loading dye were mixed with one tenth was 15% for male carriers and 30% among female car- volume of the pooled PCR product from each sample. riers. These parameters were chosen according to analy- After denaturing at 93°C for 2 min and quick chilling on ses provided by Hasstedt et al27 which yielded pen- ice, samples were loaded onto 7% polyacrylamide/7 m etrances of 30% at age 50 for homozygous female carriers urea gels (Sooner Scientific, Inc, Garrin, OK, USA) with and 21% for homozygous male carriers for an HLA- 0.1% APC and TEMED. Thirty-six samples were run linked locus. We assumed that the sporadic risk was together on 12-cm gels and electrophoresed in 1 × TBE for 0.25% among males and 0.75% among females. These 3–4 h using a 373A DNA Sequencer (Applied Biosystems, figures were chosen to provide an increased probability Foster City, CA, USA). For semi-automated data collec- of disease occurrence in females, with a higher relative tion and analysis, 672 GeneScan Collection Software probability of disease for females not having inherited (Version 1.1) was used (Applied Biosystems). susceptibility. We fitted two models: for the dominant model we assumed a disease gene frequency of 2% and Allele assignment and analysis for the recessive model, we assumed a disease gene fre- DNA fragment analysis software Genotyper Version 1.1 quency of 14.1%. For females, these frequencies yield (Applied Biosystems) was used for semi-automated allele population prevalences of RA of about 2% and a risk to size assignment in basepairs according to the manufac- siblings of about 10.3% under the dominant model. turer’s manual. The allele size of each PCR product was Under the recessive model, the prevalence is 1.3% and determined in reference to internal lane size standards. the recurrence risk to female siblings is 4.3%. These recur- Allele sizes were rounded off to a discrete whole number rence risks fall within the range of risks provided in the and checked for inheritance and presence of an extra summary manuscript of Rigby et al.14 We used the pack- allele in nuclear families. Marker data were initially pro- age Analyze46 to perform linkage analyses and homogen- cessed using the program PEDSYS43 to identify any eity tests.47 marker inconsistencies. Transmission disequilibrium testing Linkage analysis The TDT test was performed according to published To identify genetic linkages we used both affected sib methods.48 These analyses were restricted to include only pair methods and model-dependent likelihood-based those markers that had been identified through sib pair procedures. We used the SIBPAL module of SAGE44 to linkage analyses as showing a potential genetic linkage. evaluate the null hypothesis that identity by descent (ibd) We first performed an omnibus test that any of the alleles sharing among sib pairs was 0.5 compared with an alter- were excessively transmitted to affected children by per- nate hypothesis of deviation from 0.5. In sib pairs con- forming a marginal homogeneity test in which the mar- Genetic analysis of RA families D Bali et al 35 ginal probabilities for transmission and nontransmission 5 McDaniel DO, Alarcon GS, Pratt PW, Reveille JD. Most African- of all alleles are jointly compared. To assess significance American patients with rheumatoid arthritis do not have the of the results, we permuted the observed transmission rheumatoid antigenic determinant (epitope). Ann Intern Med from parents to children randomly and resampled the 1995; 123: 181–187. data 1000 times to obtain empirical P-values.11 For any 6 Teller K, Budhai L, Zhang M, Haramati N, Keiser HD, Davidson A. HLA-DRB1 and DQB typing of Hispanic American patients loci that were found to show significant results, we sub- with rheumatoid arthritis: the ‘shared epitope’ hypothesis may sequently evaluated evidence for preferential trans- not apply. J Rheumatol 1996; 23: 1363–1368. mission of any alleles using all affected offspring for 7 Boki KA, Panayi GS, Vaughan RW, Drosos AA, Moutsopoulos whom parental data were available, by applying McNe- HM, Lanchbury JS. HLA class II sequence polymorphisms and mar’s test and used Fisher’s exact test to assess signifi- susceptibility to rheumatoid arthritis in Greeks. Arthritis Rheum cance. Fisher’s exact test was used to provide more accur- 1992; 35: 749–755. ate significance levels than can be obtained from 8 Gregersen PK, Silver J, Winchester RJ. The shared epitope application of the normal approximation to the binomial. hypothesis. An approach to understanding the molecular gen- For these analyses, parental alleles were first inferred etics of susceptibility to rheumatoid arthritis. [Review]. Arthritis from all available pedigree members, whenever data Rheum 1987; 30: 1205–1213. were missing. In the analysis, whenever one parent was 9 Zanelli E, Gonzalez-Gay MA, David CS. Could HLA-DRB1 be the protective locus in rheumatoid arthritis? Immunol Today missing and both the remaining parent and affected child 1995; 16: 274–278. were heterozygous for the allele, we scored one half allele 10 Zanelli E, Krco CJ, Baisch JM, Cheng S, David CS. Immune both transmitted and not transmitted as suggested under response of HLA-DQ8 transgenic mice to peptides from the the null hypothesis that there is no preferential trans- third hypervariable region of HLA-DRB1 correlates with predis- mission of either allele. For analyses of the shared epi- position to rheumatoid arthritis. Proc Natl Acad Sci USA 1996; tope, any of the following alleles were classified as con- 93: 1814–1819. taining the shared epitope: DRB1*0401, 0404, 0405, 0408, 11 Mulcahy B, Waldron-Lynch F, McDermott MF et al. Genetic 0101 or 1001. All other alleles were classified as not con- variability in the tumor necrosis factor-lymphotoxin region taining the shared epitope. influences susceptibility to rheumatoid arthritis. Am J Hum Genet 1996; 59: 676–683. 12 Field M, Gallagher G, Eskadale J et al. Tumor necrosis factor Analysis of joint effects of the shared epitope and locus polymorphisms in rheumatoid arthritis. Tissue Antigens TNF-c 1997; 50: 303–307. Both logistic regression and segregation analysis were 13 Hajeer AH, Worthington J, Silman AJ, Ollier WE. Association of performed to assess the independent effects of the shared tumor necrosis factor microsatellite polymorphisms with HLA- epitope and TNF-c. The conditional logistic regression DRB1*04-bearing haplotypes in rheumatoid arthritis patients. analysis was performed using proc logistic in SAS Arthritis Rheum 1996; 39: 1109–1114. (Statistical Analysis Software, Cary, NC, USA). The esti- 14 Rigby AS, Voelm L, Silman AJ. Epistatic modeling in rheuma- mates of the effects of each predictor were assessed by toid arthritis: an application of the Risch theory. Genet Epidemiol forming the antilogit transformation of the coefficient for 1993; 10: 311–320. 49 15 Deighton CM, Walker DJ, Griffiths ID, Roberts DF. The contri- the predictor. As recently summarized by Schaid, appli- bution of HLA to rheumatoid arthritis. Clin Genet 1989; 36: cation of conditional logistic regression provides a TDT 178–182. test, in this case, we used conditional logistic regression 16 Funkhouser SW, Concannon P, Charmley P, Vredevoe DL, so that effects from multiple loci could be modeled. For Hood L. Differences in T cell receptor restriction fragment segregation analysis, the REGD module of the Statistical length polymorphisms in patients with rheumatoid arthritis. Analysis for Genetic Epidemiology (SAGE)44 was used. Arthritis Rheum 1992; 35: 465–471. This approach allowed us to model the risk for RA jointly 17 Mu H, Charmley P, King MC, Criswell LA. Synergy between T associated with TNF-c or the shared epitope, for all indi- cell receptor beta gene polymorphism and HLA-DR4 in suscep- viduals in the pedigree (not just those individuals for tibility to rheumatoid arthritis. Arthritis Rheum 1996; 39: 931–937. whom parental genotype data are available). However, 18 Malhotra U, Concannon P. T cell receptor beta gene polymor- 12 this approach does not condition on family membership phism and rheumatoid arthritis. Autoimmunity 1992; : 75–77. 19 Wallin J, Hillert J, Olerup O, Carlsson B, Strom H. Association and could be influenced by population heterogeneity, if of rheumatoid arthritis with a dominant DR1/Dw4/Dw14 present. sequence motif, but not with T cell receptor beta chain gene alleles or haplotypes. Arthritis Rheum 1991; 34: 1416–1424. 20 Charmley P, Concannon P, Hood L, Rowen L. Frequency and Acknowledgements polymorphism of simple sequence repeats in a contiguous 685- We gratefully acknowledge the expert technical assist- kb DNA sequence containing the human T-cell receptor beta- chain gene complex. Genomics 1995; 29: 760–765. ance of Manai Ghanayem and Jean Ye. 21 McDermott M, Kastner DL, Holloman JD et al. The role of T-cell receptor ␤ chain genes in susceptibility to rheumatoid arthritis. References Arthritis Rheum 1995; 1: 91–95. 22 Cornelis F, Hardwick L, Flipo RM et al. Association of rheuma- 1 Aho K, Koskenvuo M, Tuominen J, Kaprio J. Occurrence of toid arthritis with an amino acid allelic variation of the T cell rheumatoid arthritis in a nationwide series of twins. J Rheumatol receptor. Arthritis Rheum 1997; 40: 1387–1390. 1986; 13: 899–902. 23 Cornelis F, Faure S, Martinez M et al. New susceptibility locus 2 Hasstedt SJ, Clegg DO, Ingles L, Ward RH. HLA-linked rheuma- for rheumatoid arthritis suggested by a genome wide linkage toid arthritis. Am J Hum Genet 1994; 55: 738–746. study. Proc Natl Acad Sci USA 1998; 95: 10746–10750. 3 Stastny P. Association of the B-cell alloantigen DRw4 with rheu- 24 John S, Myerscough A, Marlow A et al. Linkage of cytokine matoid arthritis. N Engl J Med 1978; 298: 869–871. genes to rheumatoid arthritis (RA). Evidence of genetic hetero- 4 Winchester R. The molecular basis of susceptibility to rheuma- geneity. Arthritis Rheum 1997; 40: S330. toid arthritis. Adv Immunol 1978; 56: 389–466. 25 Shiozawa S, Hayashi S, Tsukamoto Y et al. Identification of the Genetic analysis of RA families D Bali et al 36 gene loci that predispose to rheumatoid arthritis. Int Immunol matism Association 1987 revised criteria for the classification of 1998; 10: 1891–1895. rheumatoid arthrtis. Arthritis Rheum 1988; 31: 315–324. 26 Bali D, Gourley IS, Kostyu DD et al. Genetic analysis of mul- 38 Sambrook J, Fritsch EF, Maniatis T. Molecular Cloning: a Labora- tiplex rheumatoid arthritis families. Am J Hum Gen 1996; 59: tory Manual. Cold Spring Harbor Laboratory, NY, 1989. A211. 39 Kostyu DD, Pfohl J, Ward FE, Lee J, Murray A, Amos DB. Rapid 27 Hasstedt SJ, Clegg DO, Ingles L, Ward RH. HLA-linked rheuma- HLA-DR oligotyping by an enzyme-linked immunosorbent toid arthritis. Am J Hum Genet 1994; 55: 738–746. assay performed in microtiter trays. Hum Immunol 1993; 38: 28 Vinasco J, Beraun Y, Nieto A et al. Polymorphism at the TNF 148–158. loci in rheumatoid arthritis. Tissue Antigens 1997; 49: 74–78. 40 Kimura A, Sasazuki T. Eleventh International Histocompat- 29 Mu H, Chen JJ, Jianh Y, King MC, Thomson G, Criswell LA. ibility Workshop reference protocol for the HLA DNA-typing Tumor necrosis factor ␣ microsatellite polymorphism is associa- technique. In: K Tsuji, M Aizawa, T Sasazuki (eds). HLA 1991. ted with rheumatoid arthritis severity through an interaction Oxford Science Publications: Oxford, 1992, pp 397–419. with the HLA-DrB1 shared epitope. Arthritis Rheum 1999; 42: 41 Fernandez-Vina MA, Bignon JD. Primers and oligonucleotide 438–442. probes (SSOP) used for DNA typing of HLA class II alleles. In: 30 Pociot F, Briant L, Jongeneel CV et al. Association of tumor D Charron (ed). HLA—Genetic Diversity of HLA, Functional and necrosis factor (TNF) and class II major histocompatibility com- Medical Implications. EDK: Paris, France, 1997, pp 596–632. plex alleles with the secretion of TNF-alpha and TNF-beta by 42 Reed PW, Davies JL, Copeman JB et al. Chromosome-specific human mononuclear cells: a possible link to insulin-dependent microsatellite sets for fluorescence-based, semi-automated gen- diabetes mellitus. Eur J Immunol 1993; 23: 224–231. ome mapping. Nat Genet 1994; 7: 390–395. 31 Pinet V, Combe B, Avinens O et al. Polymorphism of the HLA- 43 Dyke B. PEDSYS: A Pedigree Data Management System. Depart- DMA and DMB genes in rheumatoid arthritis. Arthritis Rheum ment of Genetics, Southwest Foundation for Biomedical 1997; 40: 854–858. Research: San Antonio, TX, 1997. 32 Brennan P, Hajeer A, Ong KR et al. Allelic markers close to pro- 44 SAGE. Statistical Analysis for Genetic Epidemiology. Release 3.0. lactin are associated with HLA-DRB1 susceptibility alleles Department of Epidemiology and Biostatistics, Rammelkamp among women with rheumatoid arthritis and systemic lupus Center for Education and Research, MetroHealth Campus, Case erythematosus. Arthritis Rheum 1997; 40: 1383–1386. Western Reserve University, Cleveland. 1997. 33 John S, Myercough A, Hajeer A et al. Linkage analysis of candi- 45 Blackwelder WC, Elston RC. A comparison of sib-pair linkage date gene loci in rheumatoid arthriti: evidence suggestive of tests for disease susceptibility loci. Genet Epidemiol 1985; 2: 85– linkage to corticotropin releasing hormone (CRH) and oestrogen 97. synthase (CYP19). Arthritis Rheum 1998; 41: S63 (Abstract). 46 Terwilliger JD. A powerful likelihood method for the analysis 34 Seldin MF, Amos CI, Ward R, Gregersen PK. The genetics revol- of linkage disequilibrium between trait loci and one or more ution and the assault on rheumatoid arthritis. Arthritis Rheum polymorphic marker loci. Am J Hum Genet 1995; 56: 777–787. 1999; 42: 1071–1079. 47 Ott J. Analysis of Human Genetic Linkage. Johns Hopkins Univer- 35 Risch N, Merikangas K. The future of genetic studies of complex sity Press: Baltimore, 1991. human diseases. Science 1996; 273: 1516–1517. 48 Spielman RS, McGinnis RE, Ewens WJ. The trans- 36 Eskdale J, Mc Nicholl J, Wordsworth P et al. Interleukin-10 mission/disequilibrium test detects cosegregation and linkage. microsatellite polymorphisms and IL-10 locus alleles in rheuma- Am J Hum Genet 1994; 54: 559–560, discussion 560–563. toid arthritis susceptibility. Lancet 1998; 352: 1282–1283. 49 Schaid D. Likelihoods and TDT for the case-parents design. 37 Arnett FC, Edworthy SM, Bloch DA et al. The American Rheu- Genet Epidemiol 1999; 16: 250–261.