PADI4 Polymorphisms Confer Risk of Anti-CCP-Positive Rheumatoid Arthritis in Synergy With HLA-DRB1*04 and Smoking

Peptidylarginine deiminases (PADs) catalyze citrullination, a post-translational modification playing a pathogenic role in anti-citrullinated protein antibody (ACPA)-positive rheumatoid arthritis (RA). The interplay between single nucleotide polymorphisms (SNPs) in the PADI genes and known risk factors for ACPA-positive RA, including smoking, HLA-DR4 and -1, and the PTPN22 R620W polymorphism, was investigated. We typed four PADI2 SNPs, four PADI4 SNPs, and the PTPN22 R620W SNP in 445 Danish RA patients and 533 age-matched healthy controls, as well as in 200 North American RA patients and 100 age- and sex-matched controls. The HLA-DRB1 locus was typed in the Danish cohort. Logistic regression analyses, adjusted for age, sex, smoking status, and PTPN22 R620W, revealed increased risk of anti-CCP-positive RA in carriers of rs11203367(T) (OR: 1.22, p=0.03) and reduced risk in carriers of rs2240335(A) in PADI4 (OR: 0.82, p=0.04). rs74058715(T) in PADI4 conferred reduced risk of anti-CCP-negative RA (OR: 0.38, p=0.003). In HLA-DRB1*04-positive individuals, specifically, the risk of anti-CCP-positive RA was increased by carriage of PADI4 rs1748033(T) (OR: 1.54, p=0.007) and decreased by carriage of PADI4 rs74058715(T) (OR: 0.44, p=0.01), and we observed an interaction between these SNPs and HLA-DRB1*04 (p=0.004 and p=0.008, respectively) Thus, PADI4 polymorphisms associate with ACPA-positive RA, particularly in HLA-DRB1*04-positive individuals, and with ACPA-negative RA independently of HLA-DRB1*04.

The five existing PAD isoforms in humans are encoded by the PADI1-4 and -6 genes (15). PAD2 and PAD4 are expressed by immune cells (16)(17)(18) and are present in the synovium of RA patients, where their expression levels correlate with those of inflammatory markers (18,19). Some investigators have found that PAD4 generates autoantigens recognized by ACPAs more efficiently than PAD2 (20), while others have shown that the two PAD isoforms are approximately equally efficient (21).
Studies on Asian populations have shown associations between RA and single nucleotide polymorphisms (SNPs) in PADI2 (22,23) as well as SNPs in PADI4 (22)(23)(24)(25)(26)(27). However, the observed associations with PADI2 SNPs in Asian populations have not been replicated in studies on European or North American cohorts, while a few studies have confirmed the association of the PADI4 SNPs with RA in those populations (28)(29)(30)(31)(32)(33)(34). Notably, one study reported associations between one SNP in PADI4 (rs2240340) and RA in a North American cohort, alone or when combined with a Swedish cohort, but not in the Swedish cohort alone (31). A haplotype of PADI4 has been reported to confer susceptibility to RA in Asians (24), but not in a British study (30). The haplotype consists of the minor alleles of rs11203366, rs11203367, and rs874881 encoding the amino acid substitutions Gly55Ser, Val82Ala, and Gly112Ala, respectively, and rs1748033 which is a synonymous SNP ( Figure 1). Besides, the minor allele (A) of the synonymous SNP rs2240035 has been shown to decrease the risk of RA in Asians and North Americans (26,27,35). Together with rs74058715, which is located in the 5' untranslated region (UTR) of PADI4, rs2240035 has been shown to regulate PAD4 expression levels in neutrophils and monocytes (36).
Only few studies on PADI polymorphism in RA have stratified the patients according to HLA types (28,32,37,38) and, generally, studies addressing the influence of HLA types on development of RA have not taken PADI polymorphism into account. We hypothesized that PADI polymorphisms exert their greatest influence in subjects carrying HLA types that bind citrullinated peptides most strongly, i.e., HLA-DRB1*04.
Our aim was to determine if anti-CCP-positive or anti-CCPnegative RA is associated with SNPs in PADI2 or PADI4 in a Danish and a North American cohort of RA patients and healthy controls. Moreover, we aimed at elucidating the interaction of PADI SNPs with HLA-DRB1*04 or HLA-DRB1*01 in predisposing to RA.

Patients and Controls
The Danish cohort included 445 RA patients clinically identified at 21 rheumatology departments across Denmark and 533 population controls matched for age ( Table 1). The cohort has previously been reported on (39). The RA patients had less than 5 years of disease duration and fulfilled the American College of Rheumatology 1987 classification criteria for RA (40). Healthy controls, frequency-matched by birth year, were randomly selected from the Danish population, and blood samples were collected by general practitioners. Written informed consent was obtained from all study subjects, and the study was approved by the Scientific Ethical Committees for Copenhagen and Frederiksberg (KF 01-039/01), the Danish Data Protection Agency (2001-41-0658), and the Institutional Review Board at Statens Serum Institut (21-00050). The North American cohort included 200 RA patients and 100 age-and sex-matched controls without autoimmune disease selected from the University of Wisconsin (UW) Rheumatology Biorepository described in (35,41) ( Table 1). The RA patients were initially identified by having at least two outpatient visits with RA-associated ICD codes within 24 months (42), or one visit and a positive anti-CCP test. Diagnosis was confirmed by manual review of rheumatology notes. All subjects gave written informed consent, and the study was approved by the Institutional Review Board of the University of Wisconsin-Madison (#2015-0156).
Genotyping DNA was extracted from whole blood samples from the two cohorts as previously described (35,39) and stored at −80°C until use. Samples were genotyped for PADI SNPs and the rs2476601 SNP in PTPN22, encoding R620W, by means of an in-house multiplex SNP assay protocol, as previously described (45). In brief, the method included a polymerase chain reaction (PCR) to amplify the selected SNP sites and an allele-specific primer extension (ASPE) reaction for labeling ASPE-oligonucleotides, followed by hybridization to MagPlex-TAG ™ bead sets (Luminex Corporation, Austin, TX, USA) for analysis on the Luminex platform (Luminex Corporation, Austin, TX, USA). All assay runs included a panel of control samples with known genotypes (Coriell Cell Repository, Camden, NJ, USA) as well as no template negative controls. Low-resolution HLA-DRB1 typing of the Danish cohort had been performed prior to this study, according to the 11 th Histocompatibility Workshop protocol (46).

Statistical Analysis
Hardy-Weinberg equilibrium was calculated with the "hwde"package in R (R Foundation for Statistical Computing, Vienna, Austria). Associations between SNPs and disease were tested using multiple logistic regressions with adjustment for age, sex, smoking status, and carriage of PTPN22 R620W. P-values and odds ratios (ORs) with 95% confidence intervals (95% CI) were calculated for minor allele counts (trend test). Analyses of the Danish and North American cohorts were performed separately and, in order to increase statistical power, repeated after merging

RESULTS
The characteristics of the Danish and North American cohorts included in this study are shown in Table 1. The North American cohort had a median age of 63 years (range 33-95), which was higher than the 53 years (range 19-69) of the Danish cohort, and it contained more women than the Danish cohort, particularly within the anti-CCP-negative RA and healthy control groups. Anti-CCPpositive RA patients were more often smokers than anti-CCPnegative patients in both cohorts, but the Danish cohort contained more ever smokers than the North American cohort in all three study groups. All analyses were adjusted for age, sex, smoking status, and cohort origin.

Associations Between PADI Polymorphisms and RA
We studied associations of the PADI SNPs with anti-CCPpositive and anti-CCP-negative RA separately. Given that carriage of PTPN22 R620W is a known risk factor for anti-CCP-positive RA (14), we adjusted all analyses for this variable too. To increase statistical power, we merged the two cohorts in these analyses but adjusted the analyses for the origin of the cohort. We found that none of the PADI2 SNPs included in this study associated with anti-CCP-positive or anti-CCP-negative RA ( Table 2). Among the PADI4 SNPs examined, the minor allele (T) of rs11203367 [the tag-SNP for the previously reported RA susceptibility haplotype (24,30)], was associated with an increased risk of anti-CCP-positive RA (OR: 1.22, p=0.03) ( Table 2). The minor allele (A) of rs2240335, on the other hand, appeared to protect against both anti-CCP-positive RA (OR: 0.82, p=0.04) and anti-CCP-negative RA (OR: 0.78, p=0.05). Moreover, the minor allele (T) of rs74058715 appeared to protect against anti-CCP-negative RA (OR: 0.38, p=0.003) ( Table 2).
The majority of these reported associations was largely driven by the North American cohort, with exception of the association of rs74058715 and anti-CCP-negative RA (Supplementary Table 1).

Association of PADI Polymorphisms With RA by HLA-Type
The Danish cohort was HLA-typed, so we reanalyzed their data, taking this variable into account. We chose to examine the influence of HLA-DRB1*04 and HLA-DRB1*01 separately, since HLA-DRB1*04 is more strongly associated with anti-CCP-positive RA than HLA-DRB1*01 (12) (subjects were grouped as illustrated in Supplementary Figure 1; of note, subjects carrying both alleles were included in the analysis of HLA-DRB1*04).
Inclusion of HLA-DR type in the analysis did not influence the overall conclusions with respect to SNPs in PADI2 (data not shown). In HLA-DRB1*04-positive subjects, on the other hand, the minor allele (T) of rs74058715 was associated with a significantly reduced risk of anti-CCP-positive RA after adjustment for age, sex, smoking status, PTPN22 R620W, and number of HLA-DRB1*04 alleles (p=0.01; Table 3). This effect can be observed in Figure 2A, which shows the cumulative ORs associated with different combinations of four risk factors (HLA-DRB1*04 alleles, PADI4 alleles, PTPN22 R620W and smoking) compared to the reference group without any of these risk factors. Hence, the ORs for HLA-DRB1*04-positive subjects who carried the minor allele (light blue bars) were generally lower than the ORs for HLA-DRB1*04-positive subjects who did not carry this allele (dark blue bars). By contrast, the minor allele (T) of rs11203367 conferred a borderline significant increase in OR for anti-CCP-positive RA in HLA-DRB1*04-positive subjects (p=0.06, Table 3), as visualized in Figure 2B (light blue bars for carriers versus dark blue bars for non-carriers). Notably, rs1748033(T), which belongs to the same RA susceptibility haplotype as rs11203367(T), associated strongly with anti-CCP-positive RA in HLA-DRB1*04-positive individuals (p=0.007; Table 3), with higher cumulative ORs for carriers of this allele compared to non-carriers ( Figure 2C, light blue bars vs dark blue bars). No clear pattern was observed for HLA-DRB1*01-positive individuals or for subjects who carried neither HLA-DRB1*04 nor HLA-DRB1*01 (Table 3 and gray bars in Figure 2). Neither were any associations between SNPs in PADI and anti-CCPnegative RA revealed after adjustment for HLA type ( Table 3).

Gene-Gene and Gene-Exposure Interactions
In view of the associations between SNPs in PADI4 and anti-CCP-positive RA and the known interaction between the SE and smoking as risk factors for anti-CCP-positive RA, we analyzed the interactions between PADI4 SNPs and other risk factors, i.e., HLA-DRB1*04, HLA-DRB1*01, PTPN22 R620W, and smoking on the risk of anti-CCP-positive or anti-CCP-negative RA.
We observed a strong interaction between the major allele (C) of rs74058715, HLA-DRB1*04 and ever-smoking, that associated with increased risk of anti-CCP-positive RA (p=0.008, Supplementary Table 2). Similarly, the minor allele (T) of rs1748033 and HLA-DRB1*04 interacted synergistically increasing the risk of anti-CCP-positive RA, irrespective of smoking status (p=0.004, Supplementary Table 2). Regarding the risk of anti-CCP-negative RA, we also observed an interaction between the minor allele (T) of rs1748033, HLA-DRB1*04 and ever-smoking (p=0.05, Supplementary Table 2).

DISCUSSION
PAD2 and PAD4 are thought to be involved in the pathogenesis of ACPA-positive RA by catalyzing the formation of citrullinated autoantigens that are recognized by T cells and targeted by ACPAs (8). However, citrullination by PADs may have other potentially pathogenic roles, including regulation of T-cell responses (47,48) and induction of NET formation (3,49). Accordingly, SNPs in PADI2 and PADI4 have been shown to associate with RA, mainly in Asian populations (22)(23)(24)(25)(26)(27).
This study did not show significant associations between PADI2 SNPs and RA in the two mainly Caucasian cohorts examined. We cannot rule out, however, that the minor allele (C) of rs2235912 confers resistance to anti-CCP-positive RA, as suggested by a weak tendency in that direction, which is in agreement with findings from a previous meta-analysis (43).
After merging the Danish and North American cohorts and adjusting the model for the origin of the cohorts, we found that the minor allele (T) of rs11203367 in PADI4 (tag-SNP for a haplotype containing Gly55Ser, Val82Ala, and Gly112Ala) associated with anti-CCP-positive RA with an OR of 1.22. The association was largely driven by the North American cohort, in accordance with a previous study showing association with rs2240340, which is in high LD with the SNP in our study (31). Furthermore, the minor allele (A) of the exon-synonymous SNP rs2240335 was associated with a decreased risk of anti-CCP-positive RA with an OR of 0.82, in agreement with previous results (35).
An especially strong association was observed in HLA-DRB1*04-positive subjects in the Danish cohort between the minor allele (T) of rs1748033 and anti-CCP-positive RA, and a similar tendency was observed for rs11203367, which is also part of RA the susceptibility haplotype. In individuals carrying HLA-DRB1*04, the minor (T) allele of rs74058715 was also associated with reduced risk of anti-CCP-positive RA with an OR of 0.44. This SNP is not part of the RA susceptibility haplotype but is associated with decreased expression of PAD4 in neutrophils and monocytes (36). These results suggest that SNPs in PADI4 alter the cumulative risk of disease conferred by other well-characterized risk factors (HLA-DRB1*04, PTPN22 R620W, and smoking). In accordance, we observed an interaction between the major allele (C) of rs74058715, HLA-DRB1*04, and ever-smoking, as well as between the minor allele (T) of rs1748033 and HLA-DRB1*04 in increasing the risk of anti-CCPpositive RA. Somewhat unexpectedly, we observed that the minor (T) allele of rs74058715 was associated with reduced risk of anti-CCP-negative RA, and the minor allele (A) of rs2240335 showed a similar tendency. Furthermore, we observed an interaction between the minor allele (T) of rs1748033, HLA-DRB1*04, and ever-smoking in conferring risk of anti-CCP-negative RA. Generation of citrullinated autoantigens does not explain a pathogenic role for citrullination in ACPA-negative RA, but PADI polymorphisms may be associated with differences in citrullination patterns that may have an adverse influence on the generation of proinflammatory mediators (50), regulation of T-cell responses (47,48), or NET formation (3,49) potentially leading to tissue damage.
Our study is limited by the relatively low number of subjects included and the lack of HLA-typing of the North American cohort. Furthermore, only low-resolution HLA-typing was performed so the HLA-DRB1*04-positive group may have included individuals carrying HLA-DRB1*0402, which does not contain the SE and has been associated with a protective effect against RA (51).
It is important to note that although HLA-DRB1*04 and HLA-DRB1*01 are often grouped together as SE-containing serotypes, we chose to separate them in our analysis, since the association of HLA-DRB1*04 with RA is considerably stronger than that of HLA-DRB1*01 (12). We found no association between SNPs in PADI and RA in HLA-DRB1*01-positive individuals, and no interaction between PADI4 SNPs and HLA-DRB1*01, which may be due to the low power of the study. On the other hand, cohort stratifications and lack of correction due to the exploratory nature of the study may have led to type I errors. Thus, similar and more robust analyses in larger cohorts are warranted.
Our study suggests that certain SNPs in PADI4 are risk factors for ACPA-positive RA, especially in HLA-DRB1*04-positive individuals. Furthermore, it shows that PADI4 SNPs may also play a role in ACPA-negative RA.

DATA AVAILABILITY STATEMENT
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by the Scientific Ethical Committees for Copenhagen and Frederiksberg (KF 01-039/01), the Danish Data Protection Agency (2001-41-0658), the Institutional Review Board at Statens Serum Institut (21-00050), and the Institutional A B C FIGURE 2 | Influence of PADI4 SNPs on the risk of RA. The influence of the minor alleles of rs74058715 (A), rs11203367 (B), and rs1748033 (C) in PADI4 on the cumulative risk of anti-CCP-positive RA was assessed. Bars represent odds-ratios (ORs) for different combinations of absence or presence of any copy of the minor allele of the three PADI4 SNPs, any HLA-DRB1*04 allele, any PTPN22 R620W, and smoking. Subjects who carried HLA-DRB1*01 in combination with other HLA types than HLA-DRB1*04 were excluded from the analysis. Bars marked with "1" served as the reference condition. All ORs are adjusted for age and sex. ORs and p-values for all comparisons are reported in Supplementary Table 3.
Review Board of the University of Wisconsin-Madison (#2015-0156). The patients/participants provided their written informed consent to participate in this study.

AUTHOR CONTRIBUTIONS
LM, CN, DD, MS, MF, PG, SJ, and CE designed the study. LM and PG carried out the experiments. LM, CE, and SJ carried out the statistical analyses. LM, CN, and CE drafted the manuscript. SJ, MF, and MS provided the sample material and the clinical data for the Danish and North American cohorts, respectively. DD, NØ, and MS revised the manuscript critically. All authors contributed to the article and approved the submitted version.