ARTHRITIS & RHEUMATISMVol. 50, No. 4, April 2004, pp 1117–1121DOI 10.1002/art.20169© 2004, American College of Rheumatology

A Functional Haplotype of the PADI4 Gene Associated WithRheumatoid Arthritis in a Japanese Population Is Not

Associated in a United Kingdom Population

Anne Barton, John Bowes, Stephen Eyre, Kristian Spreckley, Anne Hinks, Sally John,and Jane Worthington

Objective. In the era of postgenomic research,linkage- and association-based strategies are beginningto reveal novel complex disease genes. Using such anapproach, a functional haplotype of the peptidylargin-ine deiminase 4 gene (PADI4) has recently been identi-fied as a gene conferring susceptibility to rheumatoidarthritis (RA) in a Japanese population. In the presentstudy, we investigated the association of single-nucleo-tide polymorphisms (SNPs) in the PADI4 gene with RAin a UK population.

Methods. Association with 4 exonic SNPs(padi4_89*G/A, padi4_90*T/C, padi4_92*G/C, andpadi4_104*T/C), mapping to the PADI4 gene and defin-ing a haplotype previously reported to be associatedwith RA, was investigated. Genotyping was performedusing 5� allelic discrimination assays. Estimated haplo-types were generated using the expectation-maximization algorithm, and frequencies of the SNPsand haplotypes were compared between unrelated Cau-casian RA patients from the UK (n � 839) and popu-lation controls (n � 481).

Results. Allele frequencies for the 4 SNPs in theUK population were similar to those reported in theJapanese control population, but none of these wasassociated with RA. As in the Japanese population, theSNPs in the UK population defined 2 major haplotypes,but neither was associated with RA (P � 0.79).

Conclusion. A PADI4 susceptibility haplotype as-sociated with RA in a Japanese population is notassociated with RA in a UK population. Other genesinvolved in the citrullinating pathway remain strongcandidate RA-susceptibility genes and require furtherinvestigation.

It is generally accepted that rheumatoid arthritis(RA) is a complex disease in which carriage of riskalleles at a number of genetic loci predisposes anindividual to develop the disease following exposure toan as-yet-unidentified environmental stimulus. In viewof the number of people with this chronic condition(�1% of the adult population) (1) and the resultinghealth burden, the potential benefits of understandingthe genetic basis of RA are great. The importance of theHLA locus has been well established through linkageand association studies and it is thought to account forabout one-third of the genetic component of suscepti-bility (2). Many other potential candidate genes havebeen investigated, but results have been inconsistent (forreview, see ref. 3). Previously, these candidates werelargely selected on the basis of existing knowledge aboutthe biology and pathology of RA. Recently, for the firsttime, information obtained from linkage-based ap-proaches has led to the identification of an RA-susceptibility gene in a Japanese population (4).

Peptidylarginine deiminase 4 (PADI4)(NM_012387) is one member of a family of relatedgenes, the PADI genes, which encode enzymes that areinvolved in the posttranslational conversion of arginineresidues to citrulline. Previous studies have shown thatanticitrullinating autoantibodies are specific for RA andarise early in the disease course, suggesting that theymay have a pathogenic role (for review, see ref. 5). Thegene maps to 1p36 and linkage to this locus have

Supported by the UK Arthritis Research Campaign. Dr.Barton was the recipient of a Wellcome Advanced Fellowship. Dr.John’s work was supported by the UK Medical Research Council.

Anne Barton, MRCP, John Bowes, BSc, Stephen Eyre, MSc,Kristian Spreckley, BSc, Anne Hinks, PhD, Sally John, PhD, JaneWorthington, PhD: University of Manchester, Manchester, UK.

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

Submitted for publication October 8, 2003; accepted in re-vised form January 10, 2004.

1117

previously been described in European and Japanesepopulations (6,7). Recently, these linkage data werefollowed up in a Japanese population using a fine-mapping approach, and association with a number ofsingle-nucleotide polymorphism (SNP) markers map-ping to the region was identified (4).

Further genetic studies refined the associationwith the PADI4 gene and identified novel SNPs mappingto the gene. The SNPs formed a haplotype that wasassociated with susceptibility to RA, and functionalstudies demonstrated that this haplotype affected stabil-ity of the messenger RNA (mRNA) transcripts and wasassociated with levels of antibodies to citrullinated pep-tide (anti-CCP) in the sera of patients with RA. It ispostulated that presence of the susceptibility haplotyperesults in greater transcript stability, which, in turn, leadsto more citrullination of epitopes. These citrullinatedepitopes are then thought to become a target for RA-specific autoantibodies (anti-CCP antibodies).

Although there is strong functional evidence tosupport a role for polymorphism in the PADI4 gene inthe etiology of RA in a Japanese population, it is notclear whether these findings can be extrapolated to otherpopulations. The aim of the current study, therefore, wasto investigate whether the same haplotype associatedwith susceptibility in a Japanese population is associatedwith RA in a UK Caucasian population.

PATIENTS AND METHODS

Study design. Association studies were performed inUK patients with RA to investigate the association of 4 PADI4SNPs that can be used to define a susceptibility haplotype thathas been associated with RA in a Japanese population. Unre-lated RA patients were compared with population controls forboth the phenotype and the genotype of 4 PADI4 SNPs.Haplotypes were estimated and the frequency of the suscepti-bility haplotype was compared between patients and controls.

Subjects. RA patients were obtained from 2 mainsources: the Arthritis Research Campaign National Repositoryof RA patients and families, and local clinics. For patientsobtained through the National Repository, 1 affected patientper family was selected for investigation. All RA patientssatisfied the 1987 American College of Rheumatology (for-merly, the American Rheumatism Association) criteria for RA(8) modified for genetic studies of prevalent cases (9).

Control subjects with no history of inflammatory ar-thritis were recruited from among healthy blood donors andfrom general practice registers. All of the patients and controlsubjects had a UK Caucasoid ethnic origin, were recruited withethics committee approval, and provided their informed con-sent.

Statistical power. Sample sizes were calculated basedon published allele frequencies (minor allele frequencies of33–40%) so that each of the association studies had 80%

power to detect a gene conferring a genotypic relative risk of1.5 at the 5% significance level (4). For the stratificationanalyses, the study had 80% power to detect a genotypicrelative risk of 1.5–2.0 at the 5% significance level.

Genotyping methods. Four PADI4 SNPs(padi4_89*G/A, padi4_90*T/C, padi4_92*G/C [rs874881], andpadi4_104*T/C [rs1748033]) were genotyped using a TaqMan5� allelic discrimination assay according to the manufacturer’sinstructions, except that a 5-�l rather than a 25-�l reactionvolume was used (see the Applied Biosystems Web site athttp://www.appliedbiosystems.com). Briefly, each polymerasechain reaction (PCR) contained 15 ng DNA, 2.5 �l TaqManmaster mix, and 0.125 �l of 40� assay mix. The PCR wasperformed using 384-well plates on an ABI 9700 thermal cycler(Applied Biosystems, Warrington, UK) (reaction conditions50°C for 2 minutes, 95°C for 10 minutes followed by 40 cyclesof 95°C for 15 seconds and 60°C for 1 minute). The TaqMan7700 was used to perform plate reading with the allelicdiscrimination option.

The SNPs were selected for investigation because theywere all exonic. Two of the SNPs, padi4_92*G/C (rs874881)and padi4_104*T/C (rs1748033), had been associated with RAusing single-point analysis in a Japanese population, 3 of theSNPs involved nonconservative amino acid substitutions, andall 4 SNPs together defined a haplotype associated with RA ina Japanese population (4).

Statistical analysis. Single-point analysis. Associationof RA with each of the 4 PADI4 SNPs was tested using thechi-square test implemented with Stata software (CollegeStation, TX). RA is a heterogeneous disease with variation inthe presence of various clinical features such as erosions,rheumatoid factor (RF), and age at onset, as well as genotypicheterogeneity in terms of carriage of HLA–DR4 susceptibilityalleles. When such heterogeneity exists, stratification analysiscan increase the power to detect association by creating morehomogeneous subsets of cases despite the loss of power due toreductions in sample size (10). Stratification analysis wasundertaken to investigate the effects of sex, disease severity,age at onset, and carriage of shared epitope (SE) alleles.

Haplotype analysis. Pairwise linkage disequilibriummeasures were investigated and haplotypes constructed usingthe expectation-maximization (EM) algorithm implemented inHelixTree software (Golden Helix, Bozeman, MT). Themarker chain Monte Carlo (MCMC) algorithm, implementedin PHASE (11), was also used to estimate haplotypes to ensurethat the frequencies were similar using 2 different approaches.Haplotype frequencies were compared between patients andcontrols using CLUMP software (12).

RESULTS

RA-susceptibility cohort. DNA was availablefrom 839 RA patients and 481 population control sub-jects. Of the RA patients, 68.1% demonstrated erosionsand 75.6% were RF positive. The median age at onset ofRA was 43 years (interquartile range 32–55 years) and73.6% of the RA patients were female. Among thepatients, 19.6% had no copies, 49.2% had 1 copy, and31.2% had 2 copies of the SE alleles. Among the

1118 BARTON ET AL

population-based controls, sex information was availablefrom 366 individuals, of whom 205 (56.0%) were female,and 63.7% had no copies, 29.9% had 1 copy, and 6.4%had 2 copies of the shared epitope alleles.

Association study. Allele frequencies for all theSNPs were in Hardy-Weinberg equilibrium in both thepatient and the control populations. The allele andhaplotype frequencies of the 4 SNPs (Tables 1 and 2)closely reflect those reported previously in a Japanesepopulation (4).

Single-point association with susceptibility. Noassociation was detected with the SNPs padi4_89*G/A,padi4_90*T/C, padi4_92*G/C (rs874881), orpadi4_104*T/C (rs1748033), by either genotype or allele,in the RA patient group as a whole (Table 1). Stratifyingthe patients by the presence of RF, presence of erosions,sex, and carriage of SE alleles revealed no evidence ofassociation with RA (Tables 1 and 3).

Haplotype analysis. Full genotype informationfor all 4 SNPs, from which the haplotypes could beestimated, was available from 666 patients and 383controls. The 4 SNPs were in strong linkage disequili-brium, with r values between 0.75 and 1.00. Estimatedhaplotypes for the 4 SNPs were generated using the EMalgorithm, and the frequencies were checked by compar-ing them with estimated frequencies generated using theMCMC algorithm. The 2 programs yielded very similarhaplotype frequency estimates (data not shown). As inthe Japanese population (4), 2 major haplotypes ac-counted for more than 85% of the total chromosomes inboth the UK RA patients and UK controls (Table 2).Haplotype frequencies were not statistically differentbetween the UK RA patients and UK population con-trols (P � 0.79) (Table 2).

Table 1. Comparison of genotype frequencies for 4 exonic PADI4 single-nucleotide polymorphisms (SNPs) in RA patients and controls*

SNP, allele All patients All controls

Patients Age at onset of RA

RF positive Erosions �median age �median age

SNP891/1 243 (33.3) 164 (36.9) 181 (32.9) 170 (34.1) 100 (30.7) 143 (35.5)1/2 341 (46.8) 208 (46.8) 256 (46.5) 237 (47.5) 159 (48.8) 182 (45.2)2/2 145 (19.9) 72 (16.2) 113 (20.5) 92 (18.4) 67 (20.6) 78 (19.4)P 0.218 0.163 0.54 0.119 0.49

SNP901/1 262 (33.4) 165 (36.8) 193 (32.4) 179 (34.4) 120 (32.8) 142 (34.0)1/2 374 (47.7) 212 (47.3) 281 (47.2) 247 (47.4) 180 (49.2) 194 (46.4)2/2 148 (18.9) 71 (15.8) 121 (20.3) 95 (18.2) 66 (18.0) 82 (19.6)P 0.296 0.12 0.544 0.437 0.321

SNP921/1 237 (30.7) 152 (32.7) 179 (30.8) 161 (31.5) 115 (29.9) 122 (31.2)1/2 382 (49.5) 231 (49.7) 285 (49.1) 253 (49.5) 188 (48.8) 194 (49.6)2/2 153 (19.8) 82 (17.6) 117 (20.1) 97 (19.0) 82 (21.3) 75 (19.2)P 0.578 0.56 0.84 0.519 0.809

SNP1041/1 356 (42.4) 227 (47.2) 252 (40.1) 237 (42.5) 160 (40.1) 196 (44.5)1/2 388 (46.2) 205 (42.6) 301 (47.9) 256 (46.0) 199 (49.9) 189 (43.0)2/2 95 (11.3) 49 (10.2) 76 (12.1) 64 (11.5) 40 (10.0) 55 (12.5)P 0.243 0.057 0.317 0.082 0.485

* Except where indicated otherwise, values are the number (%). Rheumatoid arthritis (RA) patients were stratified by presence of rheumatoid factor(RF), erosions, and age at onset of RA below or above the median age, and genotype frequencies were compared with controls. The common allelewas always referred to as allele 1 and the rare allele as allele 2. All P values are versus controls. SNP89 � padi4_89*G/A; SNP90 � padi4_90*T/C;SNP92 � padi4_92*G/C (rs874881); SNP104 � padi4_104*T/C (rs1748033).

Table 2. Comparison of estimated haplotype frequencies betweenpatients with RA and unrelated controls

Haplotype

SNP*Haplotype frequency,

no. (%)

89 90 92 104 Patients Controls

1 1 1 1 1 372 (55.9)† 224 (58.6)2 2 2 2 2 215 (32.3)† 113 (29.6)3 2 2 2 1 58 (8.7) 36 (9.4)4 1 1 2 1 15 (2.2) 8 (2.1)

* The common allele was always referred to as allele 1 and the rareallele as allele 2. See Table 1 for definitions.† P � 0.79 versus controls, calculated using the T1 option in CLUMP(12).

PADI4 NOT ASSOCIATED WITH RA IN A UK POPULATION 1119

DISCUSSION

A previous study in a Japanese population dem-onstrated an association of RA with a haplotype of thePADI4 gene defined by exonic SNPs (4). Two of theseexonic SNPs were associated with RA using single-pointtests and the haplotype was shown to affect mRNAstability and to be related to levels of anti-CCP antibod-ies. In the present study, which is the first to attempt toreplicate this association with the PADI4 gene in anindependent population, we have shown no associationof RA either with the exonic SNPs or with the JapaneseRA-susceptibility haplotype in our UK population.

There are several possible explanations as to whythe association was not replicated. First, the findingsmay have arisen due to a Type 2 error (false negative),but this is unlikely because the study had 80% power todetect the relative risk for the individual SNPs reportedin the Japanese study (odds ratios 1.66–2.00) at the 5%significance level. Furthermore, genotype frequenciesdid not differ from Hardy-Weinberg expectations in thecontrol population, and allele and haplotype frequenciesin our UK population are similar to those reportedpreviously (4).

Second, previous studies of antibodies to citrulli-nated peptides (anti-CCP antibodies) in patients withRA suggest that they may be markers of disease severity

and, therefore, it is possible that the haplotype isassociated with disease severity, rather than with suscep-tibility (13–15). If the Japanese patients had a moresevere form of RA, it may explain why an associationwith RA was not detected in the UK cohort. However,the prevalence of seropositivity in our UK RA cohortwas similar to that in the Japanese RA patients investi-gated (4). Furthermore, stratification based on markersof disease severity, such as the presence of RF anderosions, showed no association with either the SNPsindividually or the haplotype. It should be noted that thepatients in our cohort had mainly longstanding RA, withthe majority having erosive and/or seropositive disease.Thus, it is not an ideal group in which to attempt todifferentiate severity from susceptibility factors. Thisquestion would be better addressed by investigating aninception cohort of patients with inflammatory arthritisfollowed up prospectively. Anti-CCP antibody levelshave not been tested in our UK cohort and, therefore,one limitation of the current study is that we were unableto examine the reported relationship between the hap-lotype and anti-CCP antibody levels (4).

A third possibility is that the gene plays a role insusceptibility to RA only in the presence of environmen-tal factors to which the Japanese, but not the UKpopulation, are exposed. Differences in diet and other

Table 3. Comparison of genotype frequencies for 4 exonic PADI4 SNPs in patients with RA and controls stratified by sex and carriage of SEalleles*

Female Male 0 copies SE 1/2 copies SE

Patients Controls Patients Controls Patients Controls Patients Controls

SNP891/1 171 (32.6) 69 (36.7) 66 (34.6) 59 (38.8) 46 (33.3) 106 (39.1) 194 (33.4) 53 (33.5)1/2 243 (46.3) 91 (48.4) 92 (48.2) 64 (42.1) 66 (47.8) 119 (43.9) 268 (46.1) 80 (50.6)2/2 111 (21.1) 28 (14.9) 33 (17.3) 29 (19.1) 26 (18.8) 46 (17.0) 119 (20.5) 25 (15.8)P 0.165 0.533 0.519 0.383

SNP901/1 193 (33.3) 69 (35.6) 65 (33.5) 58 (37.7) 52 (32.9) 106 (38.7) 207 (33.7) 54 (33.8)1/2 275 (47.5) 96 (49.5) 93 (47.9) 66 (42.9) 78 (49.4) 123 (44.9) 288 (46.9) 81 (50.6)2/2 111 (19.2) 29 (14.9) 36 (18.6) 30 (19.5) 28 (17.7) 45 (16.4) 119 (19.4) 25 (15.6)P 0.414 0.624 0.485 0.514

SNP921/1 179 (31.0) 63 (31.8) 54 (29.2) 51 (32.1) 46 (29.7) 98 (33.9) 188 (31.0) 51 (31.5)1/2 278 (48.2) 102 (51.5) 98 (53.0) 73 (45.9) 80 (51.6) 136 (47.1) 296 (48.8) 85 (52.5)2/2 120 (20.8) 33 (16.7) 33 (17.8) 35 (22.0) 29 (18.7) 55 (19.0) 122 (20.1) 26 (16.0)P 0.441 0.398 0.609 0.484

SNP1041/1 258 (41.7) 100 (48.8) 91 (44.0) 74 (46.0) 78 (44.6) 148 (50.5) 274 (42.1) 70 (41.4)1/2 292 (47.2) 88 (42.9) 91 (44.0) 66 (41.0) 82 (46.9) 119 (40.6) 299 (45.9) 77 (45.6)2/2 69 (11.1) 17 (8.3) 25 (12.1) 21 (13.0) 15 (8.6) 26 (8.9) 78 (12.0) 22 (13.0)P 0.166 0.846 0.405 0.934

* Except where indicated otherwise, values are the number (%). The common allele was always referred to as allele 1 and the rare allele as allele2. All P values are versus controls. SE � shared epitope (see Table 1 for other definitions).

1120 BARTON ET AL

lifestyle factors are known to exist between the 2 ethnicgroups, but investigation of possible gene–environmentinteraction would require exploration in a cohort inwhom a detailed history of environmental exposures hadbeen recorded.

Alternatively, it is possible that the PADI4 SNPstested in the current study are carried on a singleextended haplotype that carries the disease mutationand is in linkage disequilibrium with the haplotype in theJapanese, but not in the UK population. It is possible tospeculate that recombination within the gene may occurin the UK population such that a functional mutationup- or downstream of the exonic SNPs investigated is notcarried on the same haplotype. Given the fact that thesusceptibility haplotype in the Japanese population af-fects mRNA transcript stability, one possible hypothesisis that a disease mutation may lie in part of the 3�-untranslated region that has not so far been screened. Incontrast, for the 4 SNPs examined, the Japanese and UKpopulations appear remarkably alike: the same haplo-types exist and both the individual SNP and haplotypefrequencies are very similar (4). Therefore, it is unlikelythat the 2 populations differ with respect to linkagedisequilibrium across the gene. Linkage studies in awhole-genome scan of UK RA families have revealed noevidence for linkage with the 1p36 region (16), althoughit should be noted that failure to detect linkage does notexclude the possibility of a disease gene mapping to aregion (17).

Finally, it is possible that the process of citrulli-nation is important in RA pathogenesis, but that differ-ent genes in the pathway are associated with RA indifferent populations. Whether other members of thePADI family are important in this respect remains to bedetermined.

In summary, association of RA with a PADI4RA-susceptibility haplotype identified in a Japanesepopulation has not been replicated in a UK population.However, other genes involved in the citrullinatingpathway remain strong candidate RA-susceptibilitygenes and require further investigation.

REFERENCES

1. Symmons D, Turner G, Webb R, Asten P, Barrett E, Lunt M, etal. The prevalence of rheumatoid arthritis in the United Kingdom:

new estimates for a new century. Rheumatology (Oxford) 2002;41:793–800.

2. Deighton CM, Walker DJ, Griffiths ID, Roberts DF. The contri-bution of HLA to rheumatoid arthritis. Clin Genet 1989;36:178–82.

3. Barton A, Ollier W. Genetic approaches to the investigation ofrheumatoid arthritis. Curr Opin Rheumatol 2002;14:260–9.

4. 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.

5. Zhou Z, Menard HA. Autoantigenic posttranslational modifica-tions of proteins: does it apply to rheumatoid arthritis? Curr OpinRheumatol 2002;14:250–3.

6. Cornelis F, Faure S, Martinez M, Prud’homme JF, Fritz P, Dib C,et al. New susceptibility locus for rheumatoid arthritis suggested bya genome-wide linkage study. Proc Natl Acad Sci U S A 1998;95:10746–50.

7. Shiozawa S, Hayashi S, Tsukamoto Y, Goko H, Kawasaki H,Wada T, et al. Identification of the gene loci that predispose torheumatoid arthritis. Int Immunol 1998;10:1891–5.

8. 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.

9. MacGregor AJ, Bamder S, Silman AJ. A comparison of theperformance of different methods of disease classification forrheumatoid arthritis: results from an analysis from a nationwidetwin study. J Rheumatol 1994;21:1420–6.

10. Leal SM, Ott J. Effects of stratification in the analysis of affected-sib-pair data: benefits and costs. Am J Hum Genet 2000;66:567–75.

11. Stephens M, Smith NJ, Donnelly P. A new statistical method forhaplotype reconstruction from population data. Am J Hum Genet2001;68:978–89.

12. Sham PC, Curtis D. Monte Carlo tests for associations betweendisease and alleles at highly polymorphic loci. Ann Hum Genet1995;59:97–105.

13. Bas S, Genevay S, Meyer O, Gabay C. Anti-cyclic citrullinatedpeptide antibodies, IgM and IgA rheumatoid factors in thediagnosis and prognosis of rheumatoid arthritis. Rheumatology(Oxford) 2003;42:677–80.

14. Meyer O, Labarre C, Dougados M, Goupille P, Cantagrel A,Dubois A, et al. Anticitrullinated protein/peptide antibody assaysin early rheumatoid arthritis for predicting five year radiographicdamage. Ann Rheum Dis 2003;62:120–6.

15. Visser H, le Cessie S, Vos K, Breedveld FC, Hazes JM. How todiagnose rheumatoid arthritis early: a prediction model for persis-tent (erosive) arthritis. Arthritis Rheum 2002;46:357–65.

16. MacKay K, Eyre S, Myerscough A, Milicic A, Barton A, Laval S,et al. Whole genome linkage analysis of rheumatoid arthritissusceptibility loci in 252 United Kingdom affected sibling pairs.Arthritis Rheum 2002;46:632–9.

17. Van Heel DA, McGovern DP, Cardon LR, Dechairo BM, LenchNJ, Carey AH, et al. Fine mapping of the IBD1 locus did notidentify Crohn disease-associated NOD2 variants: implications forcomplex disease genetics. Am J Med Genet 2002;111:253–9.

PADI4 NOT ASSOCIATED WITH RA IN A UK POPULATION 1121