1. Top of page
  2. Abstract
  7. Acknowledgements
  9. Supporting Information


Analyses of families with rheumatoid arthritis (RA) have suggested the presence of a putative susceptibility locus on chromosome 14q21–23. This large population-based genetic association study was undertaken to examine this region.


A 2-stage case–control association study of 950 unrelated Japanese patients with RA and 950 healthy controls was performed using >400 gene-based common single-nucleotide polymorphisms (SNPs).


Multiple SNPs in the PRKCH gene encoding the η isozyme of protein kinase C (PKCη) showed significant single-locus disease associations, the most significant being SNP c.427+8134C>T (odds ratio 0.72, 95% confidence interval 0.62–0.83, P = 5.9 × 10−5). Each RA-associated SNP was consistently mapped to 3 distinct regions of strong linkage disequilibrium (i.e., linkage disequilibrium or haplotype blocks) in the PRKCH gene locus, suggesting that multiple causal variants influence disease susceptibility. Significant SNPs included a novel common missense polymorphism of the PRKCH gene, V374I (rs2230500), which lies within the ATP-binding site that is highly conserved among PKC superfamily members. In circulating lymphocytes, PRKCH messenger RNA was expressed at higher levels in resting T cells (CD4+ or CD8+) than in B cells (CD19+) or monocytes (CD14+) and was significantly down-regulated through immune responses.


Our results provide evidence of the involvement of PRKCH as a susceptibility gene for RA in the Japanese population. Dysregulation of PKCη signal transduction pathway(s) may confer increased risk of RA through aberrant T cell–mediated autoimmune responses.

Rheumatoid arthritis (RA), a common systemic inflammatory disease with autoimmune features, affects ∼1% of the adult population worldwide. The pathogenesis of RA involves an aberrant immune response to self antigens, which induces persistent synovial inflammation and subsequent cartilage destruction and bone erosions. Although the etiology of RA remains largely unknown, numerous epidemiologic studies have provided strong evidence of a major genetic component of RA susceptibility (1, 2). In twin studies, a higher concordance rate for RA in monozygotic than in dizygotic twins has been shown, with heritability approaching 60% (3). Family-based analyses have shown an increased risk of RA in siblings of affected probands compared with the general population (relative risk in siblings [λs] 2–17) (2).

An association between disease risk and the HLA locus has been well established in different ethnic groups. In particular, several HLA–DRB1 alleles (*0401, *0404, *0405, *0408, *0101, *0102, *1001, and *1402) are strongly associated with susceptibility to RA (4). These DRB1 alleles encode a conserved amino acid sequence (QKRAA, QRRAA, or RRRAA) in the third hypervariable region of the molecule, which is commonly called the shared epitope (5). The development of severe disease has also been associated with the HLA–DRB1 shared epitope (6). However, it is estimated that this association accounts for only one-third of the genetic component of RA (1), implying that several other non-HLA genes contribute to RA susceptibility.

To search for gene(s) underlying susceptibility to RA, genome-wide linkage screening in families with RA has been conducted by at least 4 independent research groups focusing on different ethnicities. The European Consortium on Rheumatoid Arthritis Families focused on Europeans (7, 8), Shiozawa et al focused on Japanese (9), the North American Rheumatoid Arthritis Consortium focused on Caucasians living in the US (10, 11), and the Arthritis Research Campaign: UK National Repository of Multicase RA Families focused on Caucasians living in the UK (12, 13). In these studies, linkage to the HLA locus on chromosome 6p was confirmed, and a number of potentially important non-HLA loci were identified. Several loci (e.g., 1p, 1q, 2q, 5q, 13q, 14q, 16p, 18p, and Xq) overlapped ≥2 genome screens (8); however, other than the HLA region, no obvious consensus regarding which chromosomal regions would be most likely to contain RA susceptibility genes was obtained. This is consistent with other complex genetic disorders such as type 1 diabetes mellitus (DM), and could be due to a number of factors, including the limited statistical power of individual studies (false-positive or false-negative findings), intra- and interpopulation differences in the effect size of each gene, and genetic heterogeneity.

One way to circumvent and minimize these problems is to use a meta-analysis, in which data across independent genome scans are combined. Indeed, Fisher et al (14) completed a meta-analysis using the rank-based genome scan meta-analysis method on the early genome scan data obtained in France, the US, the UK, and Japan, and reported linkage signals on 6p (HLA), 6q, 12p, and 16cen (P < 0.01), as well as 8 novel regions with suggestive evidence (P < 0.05). Choi et al (15) also applied the genome scan meta-analysis method to data obtained from European and Caucasian populations, and reported evidence of the involvement of 7 chromosomal regions, including the HLA region. More recently, Etzel et al (16) applied a novel meta-analytical method developed by Loesgen et al (17) to Caucasian-only populations, and reported overwhelming evidence of linkage in the HLA region, strong evidence of 8p and 16p (P < 0.01), and marginal evidence of 1q, 2q, 5q, and 18q (P < 0.05). Taken together, results of these meta-analyses highlight several regions of genetic linkage for RA, especially on chromosomes 1q, 6p, 12p, 16p, and 18q, that can be further studied in fine mapping and candidate gene analyses.

Chromosome 14q is one of the regions for which evidence of linkage to RA was found in the meta-analysis by Fisher et al (14), although the reported significance level was not overwhelming when compared with other possible linkage regions. We noticed that in this region, microsatellite markers showing at least some evidence of significance in each study were tightly clustered in a narrow genomic region of ∼3 Mb. These microsatellite markers were D14S587 (P = 0.037 for Caucasians living in the US by sibpair analysis with the SibPal program) (10), D14S276 (P = 0.03 for Caucasians living in the UK; logarithm of odds [LOD] 0.9 for Japanese) (9, 12), and D14S285 (P = 0.049 for Europeans by single-point analysis) (7). In addition, a recent whole-genome case–control association study in a Japanese population identified another closely located marker, D14S0452i, to be significantly associated with RA (P = 0.001 by Fisher's exact test) (18). These adjacent markers on chromosome 14q23 may be indicative of a common RA susceptibility gene across racial/ethnic groups.

Interestingly, the chromosome 14q21–q23 region has also been implicated repeatedly in susceptibility to systemic lupus erythematosus (SLE). Suggestive evidence of linkage to this region was reported in Caucasian populations in the US and Canada (D14S276; LOD 2.81, P = 0.00016) (19), in Mexican Americans and Caucasians in the US (D14S63; nonparametric linkage [NPL] 1.70, P = 0.04 and D14S258; NPL 2.02, P = 0.02) (20), in the Swedish population (D14S592; LOD 1.15) (21), and in the Finnish population (D14S587; NPL 2.20, P = 0.02) (22), although no single study has obtained significant linkage results. It is well known that different autoimmune diseases share susceptibility loci (23), and there are numerous reports describing concurrence of RA and SLE in individuals (24) or in families (25, 26). In this study, therefore, we hypothesized that a gene predisposing to RA and/or SLE may exist in chromosome 14q21–q23, and we investigated this region in a large-scale association study with a set of >400 single-nucleotide polymorphism (SNP) markers in 950 unrelated Japanese patients with RA and 950 controls. A similar approach was recently used to identify the SEC8L1 gene, located on chromosome 7q31, as a putative RA susceptibility gene in Japanese individuals (27). Our approach was useful in demonstrating an association between RA and multiple variants of the PRKCH gene, which encodes the η isozyme of protein kinase C (PKCη).


  1. Top of page
  2. Abstract
  7. Acknowledgements
  9. Supporting Information


The study was conducted in accordance with the tenets of the Declaration of Helsinki and approved by the Ethics Committee for Human Genome and Gene Research at The University of Tokushima. Written informed consent was obtained from all participants prior to enrollment in the study. A total of 1,900 Japanese subjects, consisting of 950 RA patients (200 men and 750 women; mean ± SD age at enrollment 61.7 ± 12.4 years, mean ± SD age at RA diagnosis 48.7 ± 13.4 years) and 950 healthy controls (205 men and 745 women; mean ± SD age at enrollment 39.1 ± 15.2 years), were recruited through the orthopedic clinic at Tokushima University Hospital and its affiliates, and the rheumatology clinic at Tokushima Kensei Hospital (Tokushima, Japan). Detailed information on patients and control subjects was previously reported (27). All RA patients satisfied the revised criteria of the American College of Rheumatology (formerly, the American Rheumatism Association) (28), with a minimum disease duration of 3 years. Blood specimens were collected from all study subjects. An additional set of 46 DNA samples was obtained from unrelated healthy Japanese volunteers (24 men and 22 women) at The University of Tokushima, and was used for optimization of the SNP genotyping assay and estimating allele frequencies, as previously described (29). Genomic DNA samples were extracted from peripheral blood leukocytes according to standard methods.

SNP selection and genotyping.

For SNP-based fine mapping of the chromosome 14q susceptibility locus, we reviewed 4 RA whole-genome linkage scans published in March 2002 or earlier (7, 9, 10, 12), and searched for the primary target region likely to contain gene(s) predisposing to RA. We chose an ∼20-Mb genomic segment between markers D14S288 and D14S63 on chromosome 14q21–23, centering around the 3 putative RA-associated markers (D14S587, D14S276, and D14S285) (Figure 1). This region also included D14S0452i, which recently was reported to be associated with RA in Japanese subjects (18). According to the human genome data (National Center for Biotechnology Information [NCBI] Build 35.1), the target segment contained a total of 170 known or predicted genes.

thumbnail image

Figure 1. Results of a single-nucleotide polymorphism (SNP) association study on chromosome 14q21–23. Allelic P values (negative log scale) of 378 gene-based SNPs (also referred to as gene-centric evenly spaced common SNPs) from the exploratory sample are plotted against genomic positions. Dashed line indicates a P value of 0.05. SNPs were selected mainly in gene-coding regions rather than in intergenic regions. Locations of 4 putative rheumatoid arthritis–linked microsatellite markers (D14S587, D14S276, D14S285, and D14S0452i) are also shown. A landmark SNP located in the PRKCH gene showing evidence of association in the 2-stage case–control study, rs767755, is indicated.

Download figure to PowerPoint

SNPs used for genotyping were identified from either the NCBI Database of Single-Nucleotide Polymorphisms or the Celera Discovery System database. For this study, we selected gene-based SNPs, also referred to as gene-centric evenly spaced common SNPs (27, 30), using the following criteria: suitability for designing high-throughput TaqMan SNP genotyping assays; mapping to the gene-centric region, which we assigned to an interval between 10 kb upstream from the transcription start site and 10 kb downstream from the final exon; an average distance of 10 kb between adjacent SNPs whenever possible; and common SNPs with a minor allele frequency of >0.15 based on genotyping data from 46 Japanese control individuals, supplied by Applied Biosystems (Foster City, CA) (Scafe CR, Reeve VM: unpublished observations).

A total of 422 high-quality SNPs, with a median inter-SNP distance of 10.3 kb, were selected and used for further association testing. The mean ± SD minor allele frequency was 0.33 ± 0.10. Of 422 SNPs, 406 (96.2%) were mapped within 67 genes (39.4% of the total target genes), and the remaining 16 were mapped in intergenic regions (for further information on the 422 SNPs, see supplementary Table 1, available online at There were significant numbers of allele associations among the SNPs, mostly between 2 or 3 adjacent SNPs, i.e., SNPs in linkage disequilibrium (data not shown). When the SNP data set was analyzed using the method described by Gabriel et al (31), there were 77 genomic regions showing significant linkage disequilibrium (commonly known as a linkage disequilibrium block or a haplotype block).

All but 2 SNPs were genotyped by TaqMan assays in a 384-well plate format, using commercially available reagents (TaqMan Universal polymerase chain reaction [PCR] Master Mix, no AmpErase UNG; Applied Biosystems). Fluorescence was detected using an ABI Prism 7900HT Sequence Detection System (Applied Biosystems). Genotyping accuracy was previously demonstrated by showing 100% concordance with results obtained by direct sequencing and by randomly retyping (30). Two SNPs in the PRKCH gene, rs2230500 and rs2230501, were genotyped by PCR direct sequencing with BigDye Terminator Cycle Sequencing kit version 1.1 and an ABI 3730xl DNA analyzer (Applied Biosystems).

Two-stage association study.

For the SNP association study, we applied a 2-stage case–control analysis strategy in 1,900 subjects by randomly assigning them to 2 independent subsets. The exploratory study consisted of 380 patients and 380 controls, while the validation study consisted of an independent sample set of 570 patients and 570 controls. There were no significant differences in any clinical parameter in either the control or the patient group between these 2 subsets (data not shown). All 422 SNPs were first tested in the exploratory sample, and SNPs exhibiting significant allele associations (P < 0.05) were further tested in the independent validation sample. Significant SNPs (P < 0.05) in the validation sample were tentatively regarded as putative RA-associated SNPs and evaluated further. The standard Bonferroni correction was used for statistical adjustment for multiple comparisons. The overall power of our 2-stage case–control design, which was calculated with CaTS version 0.0.1 (32), was >0.8, assuming a multiplicative genetic model with a genotype relative risk (GRR) of 1.4.

Resequencing and polymorphism screening of the PKCη gene (PRKCH).

To identify novel DNA polymorphism(s)/mutation(s) in the PRKCH gene locus, all coding exons and a 26-kb region in intron 2 were resequenced in 24 randomly chosen individuals (12 RA patients and 12 controls). For direct sequencing, PCR amplicons were purified using ExoSAP-IT (Amersham Biosciences, Piscataway, NJ) and, subsequently, a cycle sequencing reaction was performed as described above. Details of the primers and PCR conditions are available from the authors upon request.

RNA extraction, complementary DNA preparation, and reverse transcription (RT)–PCR analysis.

Synovial specimens were obtained from the knee joints of patients with RA or patients with osteoarthritis (OA) who underwent arthroplasty. Fibroblast-like synoviocytes (FLS) derived from synovial tissue were cultured in Dulbecco's modified Eagle's medium containing 10% fetal calf serum. Total cellular RNA from FLS was extracted using an RNeasy kit (Qiagen, Valencia, CA). Other RNAs from various human tissue samples (Human Total RNA Master Panel II, Human Blood Fractions MTC Panel) were purchased from BD Biosciences (Palo Alto, CA). RNAs were reverse-transcribed with random hexamer primers and Superscript II (Invitrogen, Carlsbad, CA).

For conventional RT-PCR analysis, the following specific primers for human PRKCH and ACTB (β-actin) genes were designed based on established nucleotide sequences (GenBank accession nos. NM_006255 and NM_001101, respectively): for PRKCH, 5′-GGG-AGT-TTC-ACT-GAA-GCT-AC-3′ (forward) and 5′-CAG-CGT-TTA-TGG-ACG-ACA-CA-3′ (reverse); for ACTB, 5′-TGA-CCC-AGA-TCA-TGT-TTG-AGA-3′ (forward) and ACTB 5′-GTG-GTG-GTG-AAG-CTG-TAG-CC-3′ (reverse). RT-PCR was performed using AmpliTaq Gold DNA polymerase (PerkinElmer, Yokohama, Japan), under the following conditions: 95°C for 10 minutes, followed by 30 cycles at 95°C for 30 seconds, 56°C for 10 seconds, and 72°C for 30 seconds, and then a final extension for 5 minutes at 72°C. PCR products were subjected to electrophoresis on a 2% agarose gel, followed by staining with ethidium bromide. Real-time quantitative RT-PCR analysis was performed on an ABI Prism 7900HT sequence detection system, using TaqMan Gene Expression Assay probes (for PRKCH, Hs00178933_m1; for ACTB, Hs99999903_m1). The expression level of the PRKCH transcript was quantified using the threshold cycle (Ct) method, according to the recommendations of the manufacturer (Applied Biosystems), and normalized to the amount of ACTB.

Statistical analysis.

Tests for deviation from Hardy-Weinberg equilibrium were conducted for each SNP locus, using a chi-square goodness-of-fit test. Association tests on individual SNPs were performed with 2 × 2 contingency tables of SNP allele and case–control status with the chi-square test, using SNPAlyze Pro version 5.0 (Dynacom, Chiba, Japan). Odds ratios (ORs) were calculated with 95% confidence intervals (95% CIs). Pairwise linkage disequilibrium between the SNPs was assessed by the 2 most common measures, the absolute Lewontin's coefficient (|D′|) and Pearson's correlation coefficient (r2) (33). Graphical Overview of Linkage Disequilibrium (GOLD) software was used to graphically depict the linkage disequilibrium results (34). The LDMAP program was used to construct a metric linkage disequilibrium map (35), discriminating regions of conserved linkage disequilibrium with additive distances in linkage disequilibrium units. Haplotypes were inferred from SNP genotyping data using Phase version 2.1 (36), and haplotype-tagging SNPs were determined with SNPAlyze Pro version 5.0.


  1. Top of page
  2. Abstract
  7. Acknowledgements
  9. Supporting Information

Chromosome 14q-specific SNP association study.

To map the RA susceptibility genes on chromosome 14q21–23, we first selected 422 SNPs scattered throughout the region, and genotyped all in an exploratory sample (380 RA patients and 380 controls). Among 422 SNPs, 44 were dropped from the analysis either because of observed deviations from Hardy-Weinberg equilibrium (P < 0.05) in control subjects or because of their low allele frequencies (minor allele frequency <0.10), which resulted in insufficient power to detect statistical significance. Consequently, 378 SNPs were tested for a single-locus association between SNP allele frequencies and case–control status, and significant associations (P < 0.05) were detected in 35 SNPs (Figure 1).

These associations were further tested in an independent validation sample (570 RA patients and 570 controls). Three SNPs (rs767755, rs1255784, and rs1255717) located within 2 genes (PRKCH and PPP2R5E) showed significant differences (P < 0.05) in allele frequencies between patients and controls in this stage; however, for rs1255784 and rs1255717, the directions of association in this second cohort were the opposite of the effects seen in the exploratory sample, and results of association tests were not significant when the exploratory and validation sample sets were pooled (data not shown), indicating that their associations were false-positive findings. A consistently reproducible association was observed only for rs767755. When all of the genotyping data were pooled (a total of 950 patients and 950 controls), the overall OR for this SNP was 0.73 (95% CI 0.62–0.86, P = 1.3 × 10−4). The association remained significant after stringent Bonferroni correction for the 378 tests performed in the pooled sample (Pcorr = 0.049). Overall, our association test identified 1 landmark SNP, rs767755, to be associated with RA. This SNP was located in intron 2 of the PRKCH gene, encoding the PKCη isozyme (for full association results, see supplementary Table 1).

RA-associated SNPs in the PRKCH (PKCη) gene region.

The human PRKCH gene consists of 14 exons spanning ∼229 kb of genomic DNA (Figure 2A). To examine the linkage disequilibrium and haplotype structure around rs767755 (landmark SNP), and those across the entire PRKCH gene, we searched for additional SNPs using the public database and, concurrently, by a resequencing strategy. In the public database, 2 ungenotyped SNPs, rs3742633 and rs2209386, were found to be located 16.5 kb upstream and 10.0 kb downstream from rs767755, respectively (Figure 2A). We first resequenced the entire interval between these 2 SNPs, covering an ∼26-kb sequence of intron 2 of the PRKCH gene, in 24 individuals (12 RA patients and 12 controls). Within that region, a total of 63 DNA polymorphisms, including 61 single-base substitutions and 2 DNA insertion/deletion polymorphisms, were identified (see supplementary Table 2, available online at We next searched for PRKCH exonic polymorphisms by sequencing all coding exons and exon–intron boundaries, and identified 5 SNPs: 2 SNPs in the 5′-untranslated region (rs11622327 and rs1951964), 2 synonymous SNPs (V374V in exon 9 [rs2230501] and N558N in exon 12 [rs1088680]), and 1 nonsynonymous SNP (V374I in exon 9 [rs2230500]).

thumbnail image

Figure 2. Human PRKCH gene structure, single-nucleotide polymorphisms (SNPs), and linkage disequilibrium block pattern. A, Genomic organization of the PRKCH gene, consisting of 14 exons (vertical lines), is shown at the top. Horizontal lines represent introns. Genomic regions screened by resequencing included all the coding exons and an ∼26-kb sequence of intron 2 between rs3742633 and rs2209386. Red vertical lines show the locations of SNPs used for linkage disequilibrium analysis or an association study. SNPs rs3742633, rs2209386, and rs767755 (landmark SNP) are indicated. B, Graph of the linkage disequilibrium unit (LDU) map and estimate of the local recombination rate (RR) (generated by the Phase program) between each pair of consecutive SNPs, plotted against their genomic positions. The LDU map identifies regions of strong linkage disequilibrium as plateaus (i.e., linkage disequilibrium or haplotype blocks) and characterizes steps that reflect recombination events. There are 5 distinct linkage disequilibrium blocks, blocks A–E, in the PRKCH gene region. C, Allelic P values (negative log scale) of 43 SNPs in the PRKCH gene region, plotted against genomic positions. Dashed line indicates a P value of 0.05.

Download figure to PowerPoint

We then selected a parsimonious set of 43 SNPs in the PRKCH gene region based on their locations, allele frequencies, and the preliminary results of a linkage disequilibrium analysis using the resequencing data (see Table 1 for a list of the 43 SNPs). We genotyped these SNPs for our entire sample (950 RA patients and 950 controls), and the extent and distribution of linkage disequilibrium within the PRKCH gene were assessed using the 2 most common measures of pairwise linkage disequilibrium, |D′| and r2 (33). We observed ≥5 linkage disequilibrium blocks, based on |D′| values of >0.8. These findings were also supported by the results of r2 analysis, since there were no apparent correlations among SNPs from any of the individual blocks (see supplementary Figure 1, available online at

Table 1. Association analysis of PRKCH SNPs*
SNPLocationPosition 1Position 2AlleleMAF§χ2POR (95% CI)LD block
  • *

    LD = linkage disequilibrium; 5′-UTR = 5′-untranslated region.

  • Numbers refer to the corresponding position starting with +1 at the adenine of initiation codon ATG.

  • Physical position on chromosome 14 (National Center for Biotechnology Information build 35.1).

  • §

    Minor allele frequencies (MAFs) obtained from the pooled (first- and second-stage) population (950 patients and 950 controls) are shown.

  • Odds ratios (ORs) reflect the effects of minor alleles on susceptibility to rheumatoid arthritis. 95% CI = 95% confidence interval.

  • #

    Single-nucleotide polymorphism (SNP) names are given using the human PRKCH cDNA sequence (GenBank accession no. NM_006255) as a reference sequence, according to the MDI/Human Genome Variation Society Mutation Nomenclature Recommendations (

  • **

    Gene-centric evenly spaced common SNPs.

  • ††

    SNPs from the public database.

rs3742633Exon 2+69,14960,927,722AG0.1910.1850.220.64
rs1957907Intron 2+71,41160,929,984GT0.3650.3680.030.87
c.427+2247T>G (this study)#Intron 2+71,43260,930,005TG0.1480.1322.090.15
rs1957910Intron 2+72,48460,931,057AG0.3940.4040.430.51
c.427+6489G>C (this study)#Intron 2+75,67460,934,247GC0.1660.1560.810.37
c.427+8134C>T (this study)#Intron 2+77,31960,935,892CT0.1700.22216.10.0000590.72 (0.62–0.83)A
rs12887002**Intron 2+77,69060,936,263CT0.1690.21312.00.000540.75 (0.63–0.88)A
rs12147491Intron 2+79,17460,937,747GA0.3870.45015.60.0000770.77 (0.68–0.88)A
rs7142290Intron 2+79,96260,938,535CT0.4480.4999.970.00160.81 (0.71–0.92)A
rs7141535Intron 2+80,13260,938,705GA0.2200.2320.820.36A
rs1547148Intron 2+81,37760,939,950AG0.1700.21814.10.000170.73 (0.62–0.87)A
rs7141628Intron 2+81,62960,940,202AT0.2210.2361.090.30A
rs7160675Intron 2+84,87460,943,447TC0.1710.22114.90.000110.73 (0.62–0.85)A
rs7140526Intron 2+85,00460,943,577AT0.1700.21914.70.000130.73 (0.62–0.86)A
rs2352084Intron 2+85,33460,943,907TC0.1700.21914.40.000140.73 (0.62–0.87)A
rs767757Intron 2+85,55460,944,127GT0.4010.45812.70.000360.79 (0.69–0.90)A
rs767755 (landmark SNP)**Intron 2+85,62860,944,201AG0.1700.21914.70.000130.73 (0.62–0.86)A
rs7145570Intron 2+86,00160,944,574TC0.3950.45513.90.00020.78 (0.70–0.90)A
c.427+17895A>G (this study)#Intron 2+87,08060,945,653AG0.1680.1630.140.71A
rs10149839Intron 2+87,73760,946,310AG0.4880.43311.90.000561.25 (1.10–1.43)B
rs912620Intron 2+88,24460,946,817GT0.4770.41216.10.0000621.30 (1.13–1.48)B
rs2882830Intron 2+90,38560,948,958GA0.4760.41713.40.000261.27 (1.12–1.44)B
rs4902055Intron 2+90,67860,949,251AG0.4180.36810.10.00151.24 (1.09–1.39)B
c.427+21986T>C (this study)#Intron 2+91,17160,949,744TC0.2270.1975.210.0221.20 (1.01–1.39)C
rs912621Intron 2+94,76960,953,342AG0.3360.3684.220.0400.87 (0.76–0.99)C
rs2209386††Intron 2+95,66460,954,237GA0.3360.3704.830.0280.86 (0.75–0.99)C
rs11848272**Intron 2+98,38260,956,955TA0.4650.4750.410.52C
rs12435952††Intron 2+114,40360,972,976CT0.0810.0993.740.053C
rs2296273††Intron 4+126,93160,985,504AG0.2770.2830.190.66D
rs2296274††Intron 5+128,35860,986,931AG0.5020.4693.920.0481.14 (0.99–1.28)D
rs2230500Exon 9 (V374I)+135,41960,993,992GA0.2470.20410.00.00161.28 (1.11–1.50)D
rs2230501Exon 9 (V374V)+135,42160,993,994AC0.2470.20410.00.00161.28 (1.11–1.50)D
rs912616††Intron 9+144,65761,003,230GA0.2180.17411.90.000561.33 (1.12–1.56)D
rs912618††Intron 9+144,77161,003,344AG0.2180.17411.90.000571.33 (1.13–1.57)D
rs959728††Intron 9+145,19861,003,771CT0.2890.23415.00.000111.33 (1.15–1.54)D
rs3783789††Intron 9+147,98061,006,553AG0.2290.17914.70.000121.37 (1.17–1.60)D
rs3783786††Intron 9+148,40461,006,977AG0.5100.4862.140.14
rs2181985**Intron 9+152,09061,010,663GA0.3550.3833.310.069E
rs4902064**Intron 10+168,00461,026,577CT0.3360.3521.080.30E
rs12147132**Intron 10+177,29061,035,863AG0.1460.1652.490.11E
rs1088680Exon 12+208,40661,066,979CT0.2720.2434.320.0381.17 (1.01–1.35)E

To define the linkage disequilibrium block structure in a more integrated manner, we constructed a metric linkage disequilibrium units map using the LDMAP program (35), which assigns a location in linkage disequilibrium units to each SNP, identifies regions of strong linkage disequilibrium (i.e., linkage disequilibrium blocks) as plateaus, and characterizes steps reflecting recombination hot spots or recombination events defining the relationships between blocks. In the PRKCH gene locus, we observed 5 distinct linkage disequilibrium blocks (blocks A–E) (Figure 2B), a finding similar to the |D′| results. The linkage disequilibrium block pattern in the linkage disequilibrium units map was consistent with local recombination rates, as estimated by a Phase algorithm (36) (Figure 2B).

When allele frequencies for the 43 PRKCH SNPs were compared between 950 patients and 950 controls, 21 SNPs (11 located within block A, 4 in block B, and 6 in block D) showed a statistically significant association with RA (P < 0.01) (Table 1). As shown in Figure 2C, the distribution of RA-associated SNPs was directly dependent on linkage disequilibrium structure. It is noteworthy that the associations of each SNP in block A (significant P values ranged from 5.9 × 10−5 to 1.6 × 10−3) (Table 1) were attributable to increased frequencies of the major alleles in patients, whereas those for SNPs in blocks B and D were attributable to increased frequencies of the minor alleles in patients (significant P values ranged from 6.2 × 10−5 to 1.5 × 10−3 and from 1.1 × 10−4 to 1.6 × 10−3, respectively), suggesting that multiple causal variants/alleles may be involved.

Logistic regression analysis revealed that the associations between these SNPs and RA remained statistically significant after adjustment for age and sex (data not shown). In addition, allele frequencies for the significant SNPs were compared in subgroups stratified for sex, duration of disease, rheumatoid factor positivity, and the presence of the HLA–DRB1 shared epitope, and no obvious differences were detected (data not shown). Overall, these data imply that there are 3 independent, linkage disequilibrium block–dependent (blocks A, B, and D) associations between PRKCH and RA. Among the significant SNPs, all but 1 were intronic, being located in intron 2, 5, or 9. A significant association result was also obtained for a nonsynonymous SNP found in exon 9 (V374I [rs2230500] [block D]).

Haplotypes generally contain more information than individual SNPs (31, 37), because of their higher levels of genetic heterozygosity (or allele diversity). We therefore investigated the haplotypes in linkage disequilibrium blocks A, B, and D. In each of these 3 linkage disequilibrium blocks, 5 haplotypes that accounted for >98% of haplotypes comprising the respective sites were inferred using the Phase algorithm (36) (designated A1–A5, B1–B5, and D1–D5, respectively) (Table 2). As indicated in Table 2, the minimum numbers of SNPs required to tag these haplotypes (haplotype-tagging SNPs) were 5 for haplotypes A1–A5 (c.427+8134C>T, rs7142290, rs7141535, rs767757, and c.427+17895A>G), 4 for haplotypes B1–B5 (rs10149839, rs912620, rs2882830, and rs4902055), and 3 for haplotypes D1–D5 (rs2296274, rs2230500 [V374I], and rs959728).

Table 2. Haplotype association analysis
Linkage disequilibrium block, haplotypeHaplotype-tagging SNP alleles*Frequencyχ2PPermutation P
  • *

    The order of haplotype-tagging single-nucleotide polymorphisms (SNPs), shown in 5′ to 3′, is (left to right), for linkage disequilibrium block A, c.427+8134C>T, rs7142290, rs7141535, rs767757, c.427+17895A>G; for linkage disequilibrium block B, rs10149839, rs912620, rs2882830, rs4902055; for linkage disequilibrium block D, rs2296274, V374I, rs959728.

Block A      
Block B      
Block D      

Comparisons of haplotype frequencies between patient and control groups revealed 3 common haplotypes, A1, B2, and D3, to be susceptibility haplotypes with increased frequencies in RA patients (P values ranged from 4.6 × 10−5 to 6.7 × 10−4; permutation test–based P values ranged from 6.0 × 10−5 to 8.2 × 10−4), while 3 other common haplotypes, A3, B1, and D1, and a less common haplotype, B5, which were protective haplotypes, had lower frequencies in patients (P values ranged from 1.1 × 10−4 to 2.2 × 10−2; permutation test–based P values ranged from 8.0 × 10−5 to 2.2 × 10−2). However, these differences were no more significant than those observed when the individual SNPs that constituted each linkage disequilibrium block were tested independently.

Genotype combination effects among SNPs residing in different linkage disequilibrium blocks.

Identification of linkage disequilibrium block–dependent associations (blocks A, B, and D) between the PRKCH gene and RA prompted us to examine the genotype combination effects among SNPs residing in individual linkage disequilibrium blocks. First, we selected 1 SNP from each of the 3 linkage disequilibrium blocks for which allele associations with RA were most significant. From block A we selected c.427+8134C>T; from block B, rs912620; and from block D, rs959728. Among the 3 SNPs, all pairwise r2 measures of linkage disequilibrium were confirmed to be <0.1 (supplementary Figure 1). For simplicity, we then designated the putative susceptibility alleles (c.427+8134C, rs912620-T, and rs959728-T) and protective/neutral alleles (c.427+8134T, rs912620-G, and rs959728-C) as allele 1 and allele 0, respectively. At each single SNP site, the presence of the susceptibility allele in both single (0/1) and double (1/1) doses conferred an increased OR relative to the 0/0 genotype (absence of susceptibility allele) (Table 3). The differences were significant for all SNPs in individuals homozygous for the susceptibility allele (1/1 genotype; P values ranged from 2.7 × 10−5 to 2.8 × 10−3); however, in heterozygotes (0/1 genotype), only rs959728 reached statistical significance (P = 0.0026).

Table 3. SNP genotype combination effects on association*
 Frequencyχ2POR (95% CI)
RA PatientsControls
  • *

    For all analyses, n = 948–950 rheumatoid arthritis (RA) patients and 939–948 controls. SNP = single-nucleotide polymorphism; OR = odds ratio; 95% CI = 95% confidence interval.

Single SNP     
  0/00.0250.048  1.00 (referent)
  0/10.2890.3482.860.091.56 (0.93–2.60)
  1/10.6850.6048.910.00282.13 (1.28–3.55)
  0/00.2820.336  1.00 (referent)
  0/10.4830.5041.580.211.14 (0.93–1.41)
  1/10.2350.16017.60.0000271.75 (1.35–2.28)
  0/00.5020.584  1.00 (referent)
  0/10.4170.3639.050.00261.34 (1.11–1.62)
  1/10.0810.0539.350.00221.79 (1.23–2.61)
 c.427+8134C>T and rs912620     
  0/0 and 0/00.0250.043  1.00 (referent)
  0/1 and 0/00.1310.1601.220.271.37 (0.78–2.39)
  0/1 and 0/10.1520.1851.310.621.38 (0.79–2.40)
  1/1 and 0/00.1260.1332.590.111.59 (0.90–2.79)
  1/1 and 0/10.3310.3154.530.0331.77 (1.04–3.01)
  1/1 and 1/10.2280.15610.70.00112.45 (1.42–4.24)
 c.427+8134C>T and rs959728     
  0/0 and 0/00.0190.035  1.00 (referent)
  0/1 and 0/00.1800.2251.560.211.47 (0.80–2.71)
  0/1 and 0/10.0990.1072.720.0991.7 (0.90–3.23)
  0/1 and 1/10.0110.0160.160.691.22 (0.46–3.27)
  1/1 and 0/00.3030.3253.240.0721.72 (0.95–3.12)
  1/1 and 0/10.3130.2428.410.00372.38 (1.31–4.33)
  1/1 and 1/10.0690.03712.40.000433.46 (1.71–7.00)
 rs912620 and rs959728     
  0/0 and 0/00.2070.245  1.00 (referent)
  0/0 and 0/10.0650.0900.610.430.86 (0.59–1.26)
  0/1 and 0/00.2120.2710.300.590.93 (0.71–1.21)
  0/1 and 0/10.2430.1997.230.00721.45 (1.11–1.90)
  0/1 and 1/10.0280.0340.000.980.99 (0.58–1.72)
  1/1 and 0/00.0830.0683.750.0531.46 (0.99–2.13)
  1/1 and 0/10.1090.0749.160.00251.73 (1.21–2.48)
  1/1 and 1/10.0430.01812.60.000392.84 (1.57–5.16)
 c.427+8134C>T and rs912620 and rs959728     
  0/0 and 0/0 and 0/00.0190.034  1.00 (referent)
  0/1 and 0/0 and 0/00.1050.1241.740.191.53 (0.81–2.90)
  0/1 and 0/0 and 0/10.0220.0370.030.871.07 (0.48–2.35)
  0/1 and 0/1 and 0/00.0750.1010.730.391.33 (0.69–2.56)
  0/1 and 0/1 and 0/10.0730.0683.670.0551.92 (0.98–3.75)
  1/1 and 0/0 and 0/00.0820.0872.500.111.69 (0.88–3.26)
  1/1 and 0/0 and 0/10.0380.0451.290.261.52 (0.74–3.16)
  1/1 and 0/1 and 0/00.1370.1691.400.241.45 (0.78–2.71)
  1/1 and 0/1 and 0/10.1700.1268.050.00452.43 (1.30–4.53)
  1/1 and 0/1 and 1/10.0240.0193.680.0552.27 (0.98–5.29)
  1/1 and 1/1 and 0/00.0830.0675.700.0172.23 (1.15–4.34)
  1/1 and 1/1 and 0/10.1040.0728.390.00382.59 (1.35–4.98)
  1/1 and 1/1 and 1/10.0400.01712.30.000444.22 (1.86–9.60)

Using genotyping data for these 3 SNP loci, we calculated the ORs for all observed 2-SNP or 3-SNP combinations. In the 2-SNP-combination analysis, the ORs were generally increased with an increased number of susceptibility alleles (Table 3), and the increase was most significant among individuals homozygous for ≥1 susceptibility SNP allele, depending on the number(s) of second susceptibility SNP alleles. The same tendency was observed in the 3-SNP-combination analysis (Table 3); however, statistical significance was obtained mainly for the effect of the susceptibility allele at rs959728 in individuals who were double-homozygous for susceptibility alleles at c.427+8134C>T and rs912620. The highest OR was observed in individuals homozygous for all susceptibility alleles (1/1-1/1-1/1; OR 4.22, 95% CI 1.86–9.60, P = 4.4 × 10−4). Essentially the same results were obtained with several different 3-SNP sets chosen from blocks A, B, and D (data not shown).

Expression of PRKCH in synovial tissue fibroblasts and peripheral lymphocytes.

As a first step in evaluating the role of PKCη in inflammatory arthritis, we investigated the expression of PRKCH messenger RNA (mRNA) in FLS from patients with RA or OA. As shown in Figure 3A, PRKCH expression was demonstrated in FLS from both RA and OA patients, at levels similar to that in the lung. We next investigated PRKCH expression in peripheral blood fractions (Figure 3B). PRKCH was expressed in resting CD4+ cells (T helper/inducer cells) or CD8+ cells (T suppressor/cytotoxic cells) at higher levels than in resting CD19+ cells (B cells) or CD14+ cells (monocytes). The PRKCH levels in these cells were significantly down-regulated by activation (−85% in CD4+ cells and −84% in CD8+ cells; P < 0.001). Levels of PRKCH mRNA in the mononuclear cell fraction, however, were not significantly affected. These findings indicated that expression of the PRKCH gene in certain cell types is regulated through immune responses.

thumbnail image

Figure 3. PRKCH mRNA expression in fibroblast-like synoviocytes (FLS) and peripheral blood fractions. A, PRKCH and ACTB gene expressions were analyzed by conventional reverse transcriptase–polymerase chain reaction (RT-PCR), using FLS RNA isolated from representative patients with rheumatoid arthritis (RA) or osteoarthritis (OA). Lung total RNA was used as a positive control for PRKCH gene expression. NTC = no template control (without cDNA). B, Levels of PRKCH mRNA in peripheral blood fractions. PRKCH mRNA levels in mononuclear cells (MNCs) (B cells, T cells, and monocytes), CD4+ cells (T helper/inducer cells), CD8+ cells (T suppressor/cytotoxic cells), CD19+ cells (B cells), and CD14+ cells (monocytes) were determined by real-time quantitative RT-PCR and normalized to ACTB levels. Activation was induced by incubation with 2 μl/ml pokeweed mitogen (PWM) and 5 μg/ml concanavalin A (Con A) for 3 days (MNCs), 5 μg/ml Con A for 3–4 days (CD4+ cells), 5 μg/ml phytohemagglutinin for 3 days (CD8+ cells), or 2 μl/ml PWM for 4 days (CD19+ cells). Data are presented as the mean and SEM (n = 3) fold change compared with the PRKCH expression in placenta (set at 1.0). ∗ = P < 0.001 versus resting cells, by Student's t-test. ND = experiments not done.

Download figure to PowerPoint


  1. Top of page
  2. Abstract
  7. Acknowledgements
  9. Supporting Information

The current study was designed to determine whether gene(s) located on chromosome 14q are associated with susceptibility to RA in the Japanese population. Our SNP association scan covers an ∼20-Mb genomic region on chromosome 14q21–23. We focused on the gene-coding region of this area, and chose SNPs yielding priorities in their allele frequencies, rather than relying on linkage disequilibrium information and resources that recently became publicly available (e.g., the International HapMap Project), since a high-density SNP map of the region was not available when we started this study. In addition, to reduce the time and cost of genotyping, we adopted a 2-stage strategy, in which we proceeded to the validation stage of genotyping only if the results from the exploratory stage offered some possibility of overall significance. To test associations, we used the so-called “joint analysis” strategy (32), in which we analyzed the pooled data from both exploratory and validation stages, using a P value of 0.05 with Bonferroni correction for 378 comparisons (the total number of SNP loci used for analysis). A recent report indicated that joint analysis is more efficient and powerful than the standard 2-stage strategy, which considers the second cohort to be a completely independent replication panel and tests statistics for the second stage data alone (32).

Our association study strategy was successful in identifying a significant association for 1 landmark SNP (rs767755), located in intron 2 of the PRKCH gene encoding PKCη. Subsequent analysis of additional SNPs within this gene revealed multiple SNPs scattered in the 3 distinct linkage disequilibrium blocks (blocks A, B, and D) to be significantly associated with RA. Associations were observed in a linkage disequilibrium block–dependent manner. Moreover, the results of genotype combination analysis using SNPs residing in different linkage disequilibrium blocks suggest that 2- or 3-way combinations of “at risk” genotypes increase the risk of RA additively or synergistically. Significant SNPs included a novel missense SNP substituting V for I at residue 374 of the PKCη (V374I). Of note, the V374 residue is located within the conserved ATP-binding site with the consensus sequence GXGXXGX16K, and mutations in this motif reportedly result in abolished functional activities for other PKC isozymes (38, 39). An investigation is now under way in our laboratory to determine whether this variant has functional consequences.

Other significant SNPs were noncoding SNPs, located in intron 2, 5, or 9; these observations are not surprising given the recent reports of similar findings in RA (40) and also in other common diseases with genetic associations, such as type 2 DM (41) and asthma (42). Although the exact causal connection between these intronic SNPs and RA risk is not known, they may alter PKCη expression or function by affecting splicing. In this regard, we searched for potential DNA regulatory sites and protein-binding sites in SNPs but identified no such motifs (TRANSFAC database [43]; data not shown).

One limitation of this genetic association study is that our controls were not matched for age. With respect to age at the time of the interview and subsequent blood sampling, the control subjects were significantly younger than the patients (mean ± SD age 39.1 ± 15.2 years versus 61.7 ± 12.4 years, respectively; P < 0.001). Owing to the late onset of RA, it is possible that at least some of the controls who were included in this study will develop RA in the future. This potential misclassification in the control group due to the inclusion of undiagnosed cases, as well as inappropriate case definition, would bias and weaken any association between PRKCH and RA.

There are, however, several arguments against such bias. First, the mean ± SD age at RA diagnosis in the patients was 48.7 ± 13.4 years, rather similar to the mean age of controls, although the difference was still statistically significant (P < 0.0001), Second, our controls were interviewed extensively about their health and reported no family history of any autoimmune diseases. Therefore, they would have only a low risk of developing RA, and the overall population incidence is estimated to be <0.5–1%. Thus, the sample size of our association study should still provide adequate power to detect an association. Finally, there was no significant difference in PRKCH genotype distribution between our controls and a group of Japanese individuals from the same geographic area who had type 2 DM (n = 711), from whom samples were collected for a different genetic association study (ref.30 and Takata Y: unpublished observations). The mean ± SD age of the type 2 DM group was 62.6 ± 11.4 years, which was not significantly different from that of RA patients in this study. We thus believe our comparison to be acceptable. However, due to the cross-sectional, population-based nature of this study, our findings need to be confirmed in other independent populations and/or by family-based tests of association. In the first instance, the use of populations collected for large prospective cohort studies would be one obvious approach.

The PRKCH gene is an excellent functional candidate for susceptibility to RA, since the disease is believed to be the result of misdirected immune responses by autoreactive T cells, and PKC plays a critical role in signal transduction controlling T cell activation. The PKC gene family consists of ≥11 members, including PRKCH/PKCη, and their gene products (isozymes) are usually categorized into the following 3 groups based on structure and cofactor/activator requirements: conventional (PKCα, PKCβ1, PKCβ2, and PKCγ), novel (PKCδ, PKCε, PKCη, PKCθ, and PKCμ), and atypical (PKCζ and PKCλ/ι). Conventional PKCs are dependent on calcium and diacylglycerol for their functional activity, whereas novel PKCs are calcium independent, and atypical PKCs are calcium and diacylglycerol independent.

Individual PKC isozymes have been shown to exhibit biologically distinct and even opposing cellular functions in different cell types (44), and recent data suggest that at least some isozymes of PKC are involved in critical function(s) of T cells. For example, PKCθ, which is highly expressed in T cells, has been extensively documented to play essential roles in T cell activation, signaling, and gene transcription regulation (45). In addition, PKCβ has been suggested to be a key element in proper T cell migration (46), while PKCζ has been described as maintaining the integrity of the actin cytoskeleton and mediating interleukin-induced T cell proliferation (47). It is also noteworthy that gold sodium thiomalate, which has been used as a therapeutic agent for RA for many years, is now recognized as a PKC inhibitor, suppressing mitogen-induced T cell proliferation (48).

Unfortunately, the physiologic function of PKCη in T cells has not yet been documented, and the pathophysiologic mechanism (or pathway) by which PRKCH gene polymorphisms may influence RA risk remains unknown. However, this study is the first to show that the PRKCH gene is expressed at high levels in helper/inducer (CD4+) or suppressor/cytotoxic (CD8+) T cells, in a manner analogous to that of PKCθ, and that its expression is down-regulated through immune responses. These observations suggest that PKCη functions, at least in part, in signaling pathways unique to T cells, and, therefore, that dysregulation of its gene expression and/or function might be associated with an increased risk of RA.

Another possibility is that PKCη may function as a regulator in both the pathogenic pathway and certain aspects of the cytokine signaling cascade in monocytes and macrophages, although, in our experiments, the PRKCH gene expression level in monocytes (CD14+ cells) was low compared with that in T cells. Interestingly, plasma levels of nitric oxide, a mediator of inflammation, were shown to be elevated in patients with severe RA, and a positive correlation between PKCη and inducible nitric oxide synthase expression was observed in peripheral blood monocyte–derived macrophages from RA patients (49). Although these findings provide evidence supporting the possible involvement of PKCη in immunologic activities of monocytes and macrophages, further investigation is obviously required before firm conclusions can be drawn.

In conclusion, we have identified the PRKCH gene encoding PKCη as a putative candidate gene conferring genetic susceptibility to RA in a Japanese population. Although PKCη certainly has biologic plausibility as an RA gene, replication studies in other independent populations and/or by family-based tests of association are essential for determining whether our observations are consistent. Further investigations into the molecular mechanisms by which PKCη alters RA susceptibility are also required. These studies may ultimately lead to the development of novel therapies that modulate the PKCη pathway in patients with RA and other autoimmune disorders.


  1. Top of page
  2. Abstract
  7. Acknowledgements
  9. Supporting Information

Dr. Itakura had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study design. Drs. Takata, Hamada, and Itakura.

Acquisition of data. Drs. Takata, Hamada, Miyatake, Nakano, and Shinomiya, Mr. Scafe, Mr. Reeve, and Mr. Osabe, and Drs. Moritani, Kunika, and Yasui.

Analysis and interpretation of data. Drs. Takata and Hamada.

Manuscript preparation. Drs. Takata, Inoue, and Itakura.

Statistical analysis. Drs. Takata, Kamatani, Inoue, and Itakura.


  1. Top of page
  2. Abstract
  7. Acknowledgements
  9. Supporting Information

We are grateful to the many patients and volunteers who participated in this study. We greatly appreciate the excellent technical assistance of Dr. Rumi Katashima, Ms Kazue Tsugawa, and Ms Yukiko Sakamoto (The University of Tokushima). We wish to thank Dr. Shunichi Shiozawa (Kobe University) for his efforts in the revision of this manuscript, and the many past and present members of our research group for helpful discussions and assistance.


  1. Top of page
  2. Abstract
  7. Acknowledgements
  9. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  7. Acknowledgements
  9. Supporting Information
jws-art.22262.fig1.doc23K(for full association results, see supplementary Table 1)(see supplementary Figure 1, available online at
jws-art.22262.apx1A.eps2335KSupporting Information file jws-art.22262.apx1A.eps
jws-art.22262.apx2A.eps2327KSupporting Information file jws-art.22262.apx2A.eps
jws-art.22262.apx3A.eps2308KSupporting Information file jws-art.22262.apx3A.eps
jws-art.22262.apx4A.eps2271KSupporting Information file jws-art.22262.apx4A.eps
jws-art.22262.apx5A.eps2326KSupporting Information file jws-art.22262.apx5A.eps
jws-art.22262.apx6A.eps2365KSupporting Information file jws-art.22262.apx6A.eps
jws-art.22262.apx7A.eps2318KSupporting Information file jws-art.22262.apx7A.eps
jws-art.22262.apx8A.eps1876KSupporting Information file jws-art.22262.apx8A.eps
jws-art.22262.apx1B.eps2136KSupporting Information file jws-art.22262.apx1B.eps
jws-art.22262.apx2B.eps1719KSupporting Information file jws-art.22262.apx2B.eps
jws-art.22262.apx1C.eps5485KSupporting Information file jws-art.22262.apx1C.eps

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.