ARTHRITIS & RHEUMATISMVol. 50, No. 3, March 2004, pp 748–752DOI 10.1002/art.20118© 2004, American College of Rheumatology

Haplotype Analysis in Simplex Families and Novel AnalyticApproaches in a Case–Control Cohort Reveal No Evidence ofAssociation of the CTLA-4 Gene With Rheumatoid Arthritis

Anne Barton, Francine Jury, Stephen Eyre, John Bowes, Anne Hinks,Daniel Ward, and Jane Worthington

Objective. Cytotoxic T lymphocyte–associated an-tigen 4 (CTLA-4) is a negative regulator of T cells andis, therefore, a strong candidate susceptibility gene for Tcell–mediated autoimmune diseases. The association ofCTLA-4 single-nucleotide polymorphisms (SNPs) withrheumatoid arthritis (RA) has been investigated previ-ously, with inconsistent results. Recently, SNPs map-ping to the gene (and not previously investigated in RA)have been associated with both type 1 diabetes mellitusand Graves’ disease. The aim of this study was toinvestigate the association of the CTLA-4 polymorphismwith RA.

Methods. Primer extension methods were used togenotype 5 haplotype-tagging SNPs (htSNPs) (�1722T/C, �1661 A/G, �658 C/T, �319 C/T, and �49 A/G),and the TaqMan 5� allelic discrimination assay wasused to genotype an additional 2 SNPs (CT60 andrs1863800) mapping to the CTLA-4 gene. Association tothe 5 htSNPs was investigated using the transmissiondisequilibrium test in RA simplex families (n � 122).Allele frequencies for the htSNPs were also investigatedin affected sibling pairs (n � 96) and unrelated controls(n � 173). For the SNPs CT60 and rs1863800, unrelatedpatients with RA (n � 759) were compared with controls(n � 755).

Results. No evidence for association to singlemarkers or haplotypes of the 5 htSNPs was detected ineither RA simplex families or the affected sibling–

control cohort. Neither of the 2 SNPs recently associ-ated with Graves’ disease showed evidence for associa-tion in the unrelated patient–control cohort.

Conclusion. No evidence for association ofCTLA-4 with RA was detected using family or case–control methods.

Evidence from animal models, response to treat-ments that deplete T cells, and association with HLAgenes support the hypothesis that rheumatoid arthritis(RA) is a T cell–mediated autoimmune disease (forreview, see ref. 1). Genes involved in the regulation of Tcell responses are, therefore, strong candidate RA sus-ceptibility genes. Cytotoxic T lymphocyte–associated an-tigen 4 (CTLA-4) is a negative regulator of T cells, andpolymorphism within the gene has been associated witha variety of T cell–mediated autoimmune diseases, in-cluding type 1 diabetes mellitus (2), multiple sclerosis(3), Hashimoto thyroiditis, and Graves’ disease (2).Previous investigations of CTLA-4 and RA have yieldedconflicting results, with association reported in somestudies (4–6) but not in others (7–10). All previousstudies have investigated association to 1 or 2 single-nucleotide polymorphisms (SNPs) mapping to exons orthe promoter region, but recent research has shownmany more SNPs mapping to potentially functionaldomains of the gene (11).

The previous apparently conflicting associationsbetween polymorphisms within the CTLA-4 gene andRA could have arisen because the SNPs tested are notfunctional in themselves but are in linkage disequili-brium (LD) with a disease SNP mapping in closeproximity. In this situation, haplotype analysis shouldincrease the power to detect association (compared withsingle-locus methods), because a disease-associated al-lele, if present, should lie on one or more associatedhaplotypes. It has been demonstrated that, rather than

Supported by the Arthritis Research Campaign, UK.Anne Barton, MRCP, Francine Jury, MSc, Stephen Eyre,

MSc, John Bowes, BSc, Anne Hinks, PhD, Daniel Ward, BSc, JaneWorthington, PhD: University of Manchester, Manchester, UK.

Address correspondence and reprint requests to Anne Bar-ton, MRCP, ARC/EU, Stopford Building, Oxford Road, University ofManchester, Manchester M13 9PT, UK. E-mail: ABarton@fs1.ser.man.ac.uk.

Submitted for publication August 4, 2003; accepted in revisedform December 11, 2003.

748

genotyping all polymorphisms, genotyping of 5haplotype-tagging SNPs (htSNPs; the minimum numberof SNPs that must be typed in order to capture the mostfrequently occurring haplotypes) mapping to CTLA-4 issufficient to capture at least 80% of the common hap-lotypes in the European population (11).

In a complex disease such as RA, detection ofweak genetic effects can be complicated by the fact thatgenetic heterogeneity exists (i.e., different genes willcontribute to disease susceptibility in different people).Families containing more than 1 individual with diseaseare likely to have a stronger genetic component todisease susceptibility because, in effect, they concentrategenetic susceptibility factors. Simulation studies haveconfirmed that comparing allele frequencies betweenrelated patients and unrelated controls is more powerfulthan using unrelated patients (12). Studying multicaseRA families, therefore, may facilitate identification ofweak genetic effects.

We used both haplotype analysis in RA simplexfamilies and case–control analysis in affected siblingsand unrelated controls to test for association of theCTLA-4 gene with RA. First, haplotype analysis of the 5CTLA-4 htSNPs in RA simplex families was performed.Second, we tested association of these SNPs in affectedsibling pairs and controls, using a method that adjustsfor the fact that patients are related (13). Recently,association of newly identified SNPs mapping toCTLA-4 has been reported in Graves’ disease, auto-immune hypothyroidism, and type 1 diabetes mellitus(2). Therefore, we also tested 2 of the most highlyassociated SNPs—CT60 and rs1863800—in a large co-hort of unrelated patients with RA and controls.

PATIENTS AND METHODS

Study design. Studies were performed to investigatethe association of 5 CTLA-4 htSNPs and 2 autoimmunedisease–associated SNPs in RA patients living in the UK. Thestudy group comprised 3 cohorts: RA simplex families (af-fected proband plus 1 or both parents), an affected siblingpair–control group (affected siblings and unrelated populationcontrols), and an unrelated patient–control group (unrelatedpatients with RA and unrelated controls from the generalpopulation). Evidence for preferential allele or haplotypetransmission of the 5 htSNPs was investigated in the RAsimplex families. Affected siblings with RA were comparedwith controls for both phenotype and genotype at the 5haplotype-tagging loci, using a method that adjusts for therelatedness of patients. Unrelated RA patients were comparedwith population controls for both phenotype and genotype for2 CTLA-4 SNPs recently associated with autoimmune disease.

Subjects. Patients with RA were identified from theArthritis Research Campaign national repository. All RA

patients satisfied the 1987 American College of Rheumatology(formerly, the American Rheumatism Association) criteria(14) modified for genetic studies of prevalent cases, and allwere Caucasians living in the UK. Subjects were divided into 3cohorts as follows: 1) RA simplex families (n � 122). DNA wasavailable for both parents in 120 families and for 1 parent inthe other 2 families. Eighty percent of the probands werefemale, 68% were rheumatoid factor (RF) positive, and 75%had erosive disease; 2) affected sibling pair–control cohort (96sibpairs [192 individuals] and 173 controls). Of the affectedsiblings, 80% were female, 79% were RF positive, and 82%had erosive disease. The controls were a subset of subjectsused in the unrelated-patient–control study (see below); 3)unrelated RA patient–control cohort (759 patients and 755controls). Of the unrelated patients with RA, 72.6% werefemale, 80.9% were RF positive, and 80.3% had erosivedisease. The median age at disease onset was 43 years (inter-quartile range 32–55 years). Among the patients, 20.4% had 0copies, 47.7% had 1 copy, and 31.9% had 2 copies of sharedepitope (SE) alleles. Control subjects were healthy individualswith no history of inflammatory joint disease; these subjectswere recruited from primary care practices or were blooddonors. Sex information was available for 553 individuals, ofwhom 316 (57.1%) were female; 61.6% had 0 copies, 31.3%had 1 copy, and 7.1% had 2 copies of SE alleles. All patientsand controls were Caucasians living in the UK.

Genotyping methods. Five CTLA-4 htSNPs (�1722T�C [rs733618], �1661 A�G [rs4553808], �658 C�T [CT41],�319 C�T [rs5742909], and exon 1 �49 A�G [rs231775])were genotyped in both the RA simplex families and theaffected sibling–control cohort. A fluorescence-based primerextension method (SNaP-shot; PE Applied Biosystems, War-rington, UK) was used to genotype 4 of the SNPs, and amini-sequencing method (Pyrosequencing AB, Uppsala, Swe-den) was used for the �1722 T�C (rs733618) SNP, accordingto the manufacturer’s instructions (the PCR primers andprobes used in the extension reactions and descriptions of thereaction conditions are available upon request from the au-thors). Genotypes were validated on a subset of the totalcohort using the reverse probe. Two SNPs (CT60 A�G andrs1863800 C�T) were genotyped using a TaqMan 5� allelicdiscrimination assay; the primers and reaction conditions havebeen described previously (2).

Statistical analysis. RA simplex families. Association tothe 5 htSNPs was investigated using single- and multi-locusextensions of the transmission disequilibrium test imple-mented using the TRANSMIT program (http://www.gene.cimr.cam.ac.uk/clayton/software/transmit.txt) in RA sim-plex families.

Case–control analysis using affected sibling pairs. Asso-ciation to the 5 htSNPs was tested in related patients andunrelated controls using a method that accounts for the factthat patients are related. The method is based on the Cochran-Armitage test for trend and is most sensitive when an additivemodel exists (13). The method cannot currently be extended tohaplotype analysis. Therefore, one RA proband was selected atrandom from each sibpair, and estimated haplotypes werereconstructed using the expectation-maximization algorithmimplemented in HelixTree (Golden Helix, Bozeman, MT).Haplotype frequencies were then compared using CLUMPsoftware (http://www.mds.qmw.ac.uk/statgen/dcurtis/

NO EVIDENCE FOR ASSOCIATION OF CTLA-4 WITH RA 749

software.html). P values less than 0.05 were considered signif-icant.

Unrelated patient–control analysis. Association to 2autoimmune disease–associated SNPs was tested using thechi-square test, and stratification analysis was undertaken toinvestigate the effects of sex, severity, age at onset, andcarriage of SE alleles.

RESULTS

Allele frequencies for all of the SNPs were inHardy-Weinberg equilibrium in both the patient andcontrol populations. Furthermore, allele frequencies forthe 5 htSNPs in the control cohort did not differsignificantly from the frequencies of nontransmittedparental alleles in the RA families. The allele andhaplotype frequencies of the 5 htSNPs and the 2 auto-immune disease–associated SNPs closely reflect thosereported previously in a European population (2,11).

RA simplex families. Using single-locus analysis,none of the 5 htSNPs tested showed evidence of prefer-ential allele transmission to affected probands. Analysisof haplotypes defined by the 5 htSNPs revealed noevidence for association using the TRANSMIT program(�2 � 1.14, 7 df, P � 0.99) (Table 1). The frequencies ofthe common haplotypes are similar to those estimatedfrom a large cohort of unrelated controls in the UKpopulation in a previous study (11).

RA-affected siblings and unrelated controls. Us-ing a method by which appropriate adjustment is madefor the fact that RA-affected siblings are compared withunrelated controls, no association to any of the 5 htSNPstested was detected in this independent cohort (Table2). No difference in haplotype frequencies was foundwhen RA patients (the proband from each family) werecompared with controls (P � 0.27 using the T4 option of

CLUMP) (Table 1). In contrast to previous reports(4,5), stratification by HLA subtype also showed noevidence for association to the exon 1 �49 A�G(rs231775) CTLA-4 SNP in either DR3-positive orDRB1*0401-positive patients (P � 0.45 and P � 0.26,respectively). None of the other SNPs showed evidencefor association after stratification by HLA subtype (datanot shown).

Unrelated RA patients and controls. No associa-tion with the SNPs CT60 and rs1863800 by eithergenotype or allele was detected for the patient group asa whole (Table 3). Stratification for the presence of RF,erosions, sex, and carriage of SE alleles revealed noevidence for association (Table 3). SNP CT60 wassignificantly associated with a younger age at diseaseonset (P � 0.02), but this association did not remainsignificant when corrected for the number of stratifica-tions performed (P � 0.18). No association with haplo-types of the 2 SNPs was detected (P � 0.63).

DISCUSSION

Using analysis of 5 htSNPs that define the com-mon haplotypes of CTLA-4 polymorphism in the UKpopulation, we found no evidence for association withRA. Direct testing of a possible functional SNP (CT60)and another SNP (rs1863800) significantly associatedwith other autoimmune diseases also failed to revealevidence for association of the CTLA-4 gene with RA.

Our results could have arisen due to a Type IIerror (false-negative result), but this is unlikely, becauseseveral different approaches designed to maximize thepower to detect association, if one exists, were used, asfollows: haplotype analysis in RA simplex families, com-

Table 1. CTLA-4 haplotypes occurring with a frequency of �1% in RA simplex families and the case–control cohort*

Haplotype

SNP

Haplotype frequency (%)

RA simplex familiesAffected patient–control

cohort†

�1722 �1661 �658 �319 �49 Observed Expected Observed Expected

A T A C C A 89 (36.9) 84 (35.5) 21 (22.8) 53 (33.4)B T A C C G 86 (35.8) 82 (34.7) 35 (38.5) 47 (29.2)C T G C C A 17 (7.2) 19 (8.0) 9 (10.2) 15 (9.2)D T A T C A 16 (6.5) 18 (7.7) 10 (10.8) 11 (6.8)E T G C T A 11 (4.6) 14 (6.1) 6 (6.1) 14 (9.0)F C A C C G 10 (4.3) 8 (3.2) 10 (8.2) 12 (7.8)G T G C C G 6 (2.7) 6 (2.7) 0 (0.0) 1 (0.5)H T G C T A 4 (1.8) 5 (2.2) 2 (1.9) 1 (0.5)

* CTLA-4 � cytotoxic T lymphocyte–associated molecule 4; RA � rheumatoid arthritis; SNP � single-nucleotide polymorphism.† Only 1 patient per family was sampled.

750 BARTON ET AL

parison of allele and genotype frequencies with individ-ual SNPs in affected siblings and controls from the UKpopulation, and direct testing of possible autoimmunedisease causal variances in a large cohort of unrelatedpatients with RA and controls. However, in none ofthese cohorts was association demonstrated either topreviously defined htSNPs or to 2 possible autoimmunedisease causal SNPs. The sample size used had 80%

power to detect the effect of a polymorphism, conferringan odds ratio (OR) of 1.5 at the 5% significance level(assuming an allele frequency of 50% in the controlpopulation). This OR is the same as that conferringsusceptibility to Graves’ disease (2). However, the studywas underpowered to detect a smaller effect, such as theOR of 1.15 associated with type 1 diabetes mellitus (2).

Results of the current study replicate those of 3

Table 2. Genotype frequencies (%) in affected siblings and controls, and results of association test usinga method that accounts for the relatedness of patients*

SNP Patients ControlsHardy-Weinberg

equilibrium, controlsP, by testfor trend†

�1722 (rs733618)CC 159 (86.9) 145 (83.8) 1.0 0.46CT 24 (13.1) 27 (15.6)TT 0 (0.0) 1 (0.6)

�1661 (rs4553808)AA 123 (68.3) 110 (64.7) 0.63 0.25AG 50 (27.8) 52 (30.6)GG 7 (3.9) 8 (4.7)

�658 (CT41)‡CC 147 (77.0) 145 (84.3) 1.0 0.15CT 44 (23.0) 26 (15.1)TT 0 (0.0) 1 (0.6)

�319 (rs5742909)CC 132 (87.4) 122 (80.2) 0.39 0.44CT 18 (11.9) 27 (17.8)TT 1 (0.7) 3 (2.0)

�49 (rs231775)AA 43 (32.6) 59 (37.8) 0.25 0.26AG 55 (41.7) 68 (43.6)GG 34 (25.7) 29 (18.6)

* SNP � single-nucleotide polymorphism.† See ref. 13.‡ See ref. 2.

Table 3. Results of association and stratification analyses in unrelated patient–control cohort*

CT60

P

rs1863800

P1/1 1/2 2/2 1/1 1/2 2/2

All patients 242 (33.7) 336 (46.7) 141 (19.6) 0.65† 266 (35.0) 354 (46.7) 139 (18.3) 0.74†RF� patients 184 (34.4) 248 (46.4) 103 (19.2) 0.83† 200 (35.0) 267 (46.7) 105 (18.3) 0.78†Erosive disease 162 (33.1) 237 (48.4) 91 (18.5) 0.86† 180 (34.5) 249 (47.8) 92 (17.7) 0.96†Patients with age at onset

less than median129 (40.4) 136 (42.7) 54 (16.9) 0.02† 136 (41.0) 143 (43.1) 53 (15.9) 0.10†

All controls 239 (31.7) 370 (49.0) 146 (19.3) 256 (34.2) 363 (48.5) 129 (17.3)Female patients 173 (33.6) 248 (48.2) 94 (18.2) 0.53 187 (34.6) 263 (48.6) 91 (16.8) 0.76Female controls 91 (31.3) 152 (52.2) 48 (16.5) 98 (34.8) 142 (50.3) 42 (14.9)Patients with 0 copies SE 42 (32.3) 63 (48.5) 25 (19.2) 0.78 50 (35.7) 68 (48.6) 22 (15.7) 0.61Controls with 0 copies SE 129 (29.1) 223 (50.3) 91 (20.6) 145 (32.7) 213 (48.1) 85 (19.2)Patients with 1 or 2

copies SE200 (34.0) 273 (46.3) 116 (17.6) 0.78 214 (35.2) 279 (45.9) 115 (18.9) 0.17

Controls with 1 or 2copies SE

110 (35.3) 147 (47.1) 55 (17.6) 97 (35.5) 138 (50.5) 38 (18.9)

* Values are the frequency (%). RF � rheumatoid factor; SE � shared epitope.† Versus all controls.

NO EVIDENCE FOR ASSOCIATION OF CTLA-4 WITH RA 751

previous studies in the UK population, which showed noevidence for association of the exon 1 �49 A�G(rs231775) polymorphism with RA (7,9,10), although in1 of these studies (10) association was detected in RApatients with an associated autoimmune endocrinopa-thy. Using regression analysis, the exon 1 �49 A�G(rs231775) SNP has been excluded as a disease causalvariant for Graves’ disease and type 1 diabetes mellitus,and the evidence points to CT60 as being the most likelyto be an autoimmune disease causal polymorphism (2).The current study is the first to investigate the role ofthis SNP in patients with RA, but the results suggest thatit is unlikely to have a large effect in RA susceptibility.

It is possible that CTLA-4 can act only on certainHLA backgrounds, and previous reports have suggestedassociation with subgroups of RA patients harboringdifferent HLA alleles (4,5). In the current analysis, wefailed to replicate these findings (data not shown).

CTLA-4 was thought to be a strong candidateRA susceptibility gene because it is a negative regulatorof T cells, and pilot studies of a recombinant fusionprotein in humans have shown encouraging preliminaryresults for the treatment of RA (15). However, using themost comprehensive analysis to date of SNPs mappingto the gene, no evidence for association was detected. Itis possible that the lack of association to RA of a genethat has been related to a variety of other T cell–mediated autoimmune diseases may lend support to thehypothesis that T cells are not of primary importance inthe pathogenesis of RA. Indeed, the recent identifica-tion of an RA susceptibility gene that encodes a citrul-linating enzyme, peptidylarginine deiminase 4 (16), andthe success of anti–B cell therapies (17) provide supportfor a crucial role for B cells in disease pathogenesis.

REFERENCES

1. Panayi GS, Lanchbury JS, Kingsley GH. The importance of the Tcell in initiating and maintaining the chronic synovitis of rheuma-toid arthritis. Arthritis Rheum 1992;35:729–35.

2. Ueda H, Howson JM, Esposito L, Heward J, Snook H, Chamber-lain G, et al. Association of the T-cell regulatory gene CTLA4 withsusceptibility to autoimmune disease. Nature 2003;423:506–11.

3. Kantarci OH, Hebrink DD, Achenbach SJ, Atkinson EJ, Walisze-

wska A, Buckle G, et al. CTLA4 is associated with susceptibility tomultiple sclerosis. J Neuroimmunol 2003;134:133–41.

4. Seidl C, Donner H, Fischer B, Usadel KH, Seifried E, KaltwasserJP, et al. CTLA4 codon 17 dimorphism in patients with rheuma-toid arthritis. Tissue Antigens 1998;51:62–6.

5. Gonzalez-Escribano MF, Rodriguez R, Valenzuela A, Garcia A,Garcia-Lozano JR, Nunez-Roldan A. CTLA4 polymorphisms inSpanish patients with rheumatoid arthritis. Tissue Antigens 1999;53:296–300.

6. Yanagawa T, Gomi K, Nakao EI, Inada S. CTLA-4 gene polymor-phism in Japanese patients with rheumatoid arthritis. J Rheumatol2000;27:2740–2.

7. Barton A, Myerscough A, John S, Gonzalez-Gay M, Ollier W,Worthington J. A single nucleotide polymorphism in exon 1 ofcytotoxic T-lymphocyte-associated-4 (CTLA-4) is not associatedwith rheumatoid arthritis. Rheumatology (Oxford) 2000;39:63–6.

8. Matsushita M, Tsuchiya N, Shiota M, Komata T, Matsuta K, ZamaK, et al. Lack of a strong association of CTLA-4 exon 1 polymor-phism with the susceptibility to rheumatoid arthritis and systemiclupus erythematosus in Japanese: an association study using anovel variation screening method. Tissue Antigens 1999;54:578–84.

9. Milicic A, Brown MA, Wordsworth BP. Polymorphism in codon 17of the CTLA-4 gene (�49 A/G) is not associated with susceptibil-ity to rheumatoid arthritis in British Caucasians. Tissue Antigens2001;58:50–4.

10. Vaidya B, Pearce SH, Charlton S, Marshall N, Rowan AD,Griffiths ID, et al. An association between the CTLA4 exon 1polymorphism and early rheumatoid arthritis with autoimmuneendocrinopathies. Rheumatology (Oxford) 2002;41:180–3.

11. Johnson GC, Esposito L, Barratt BJ, Smith AN, Heward J, DiGenova G, et al. Haplotype tagging for the identification ofcommon disease genes. Nat Genet 2001;29:233–7.

12. Risch N, Teng J. The relative power of family-based and case-control designs for linkage disequilibrium studies of complexhuman diseases I: DNA pooling. Genome Res 1998;8:1273–88.

13. Slager SL, Schaid DJ. Evaluation of candidate genes in case-control studies: a statistical method to account for related subjects.Am J Hum Genet 2001;68:1457–62.

14. Arnett FC, Edworthy SM, Bloch DA, McShane DJ, Fries JF,Cooper NS, et al. The American Rheumatism Association 1987revised criteria for the classification of rheumatoid arthritis.Arthritis Rheum 1988;31:315–24.

15. Moreland LW, Alten R, Van den Bosch F, Appelboom T, Leon M,Emery P, et al. Costimulatory blockade in patients with rheuma-toid arthritis: a pilot, dose-finding, double-blind, placebo-con-trolled clinical trial evaluating CTLA-4Ig and LEA29Y eighty-fivedays after the first infusion. Arthritis Rheum 2002;46:1470–9.

16. Suzuki A, Yamada R, Chang X, Tokuhiro S, Sawada T, Suzuki M,et al. Functional haplotypes of PADI4, encoding citrullinatingenzyme peptidylarginine deiminase 4, are associated with rheuma-toid arthritis. Nat Genet 2003;34:395–402.

17. Leandro MJ, Edwards JC, Cambridge G. Clinical outcome in 22patients with rheumatoid arthritis treated with B lymphocytedepletion. Ann Rheum Dis 2002;61:883–8.

752 BARTON ET AL