The expression of autoimmunity in mice deficient in programmed death 1 (PD-1) suggests that PD-1 is a candidate gene involved in the development of human autoimmune diseases such as systemic lupus erythematosus (SLE) and rheumatoid arthritis (RA). We therefore tested the potential association between PD-1 and the development of SLE and RA by conducting case–control genetic-association studies.
Ninety-eight SLE patients, 84 RA patients, and sex-matched control subjects for each disease group were recruited and genotyped for a single-nucleotide polymorphism, C+872T, in the human PD-1 gene. The significance of the association of the PD-1 gene with SLE or with RA was analyzed by statistical tests for the difference in genotype distribution between disease and control groups.
The human PD-1 gene was found to be significantly associated with disease development in RA patients, but not SLE patients. The risk of RA development appeared to be significantly increased by carriage of the T allele (odds ratio 3.32, P < 0.0001) or the C/T genotype (odds ratio 3.52, P < 0.00005).
The PD-1 gene is significantly associated with RA susceptibility, suggesting the possibility that PD-1 may contribute to the pathogenesis of RA.
Systemic lupus erythematosus (SLE) and rheumatoid arthritis (RA) are two human autoimmune diseases that are mediated by damaging immune overreactivity to self antigens. Although the pathogenesis of SLE and RA remains unclear, it is possible that dysregulated lymphocyte activation initiates the breakdown of tolerance and predisposes the patient to the development of these autoimmune diseases, because lymphocyte activation appears to be governed by immunostimulatory and immunoinhibitory signals that are delivered through lymphocyte surface receptors. More solid evidence to support this hypothesis comes from the fact that autoimmune diseases can develop in mice that have either an overexpression of a stimulatory receptor or a deficiency of an inhibitory receptor. However, further efforts toward establishing a correlation between each immunoregulatory receptor and human autoimmune diseases are still needed in order to gain more insight into the disease pathogenesis and to develop better therapeutic strategies.
Programmed death 1 (PD-1), which was originally identified in a T cell line undergoing activation-induced cell death, is a CD28 family member that contains a cytoplasmic immunoreceptor tyrosine-based inhibitory motif and is expressed on the surface of activated T cells and B cells (1–5). As an immunoinhibitory receptor, PD-1 has been shown to inhibit lymphocyte activation and cytokine production after interacting with its ligands PDL-1 (B7-H1) and PDL-2 (B7-DC) (6–10). The immunoinhibitory function of PD-1 was further supported by the observation that mice deficient in PD-1 expression developed autoimmune diseases, despite having distinct phenotypes on different genetic backgrounds. C57BL/6 mice with PD-1 deficiency had an increased incidence of progressive glomerulonephritis, and, more impressively, all of these mice exhibited progressive arthritis, with synovial cell proliferation, lymphocyte infiltration, and pannus formation (11). These phenotypes resembled the clinical manifestations of SLE and RA in humans. In contrast, PD-1–deficient BALB/c mice developed autoimmune cardiomyopathy, with IgG deposition in the heart (12). These data, together with the observation of the wide tissue expression of the PD-1 ligands, suggest that interaction between PD-1 and its ligands plays an important role in maintaining peripheral tolerance.
Based on the solid evidence for the immunoinhibitory function of PD-1 in mice, we hypothesized that PD-1 is associated with the development of SLE and RA in humans. We therefore attempted to test this hypothesis by identifying a single-nucleotide polymorphism (SNP) in the human PD-1 gene that we could use for conducting genetic-association studies. Directly validating candidate SNPs, which were predicted by nucleotide mismatches in documented human PD-1 complementary DNA (cDNA) sequences, we discovered a SNP located at nucleotide position +872 (counting from the first base of the PD-1 cDNA sequence) in Chinese subjects (4). This SNP was then used in case–control genetic-association studies to determine whether the PD-1 gene is associated with the development of RA and SLE.
PATIENTS AND METHODS
Patient and control populations.
Case–control genetic-association studies were conducted to test the association between the PD-1 gene and the development of SLE and RA. A total of 98 SLE patients (93% women) and 94 RA patients (69% women) were recruited into this study from Kaohsiung Medical University Hospital in Kaohsiung and from Cathay General Hospital in Taipei. The diagnoses of SLE and RA were established using the classification criteria of the American College of Rheumatology (formerly, the American Rheumatism Association) (13,14). One hundred healthy blood donors (95% women) with no family history of an autoimmune disease were recruited into the control group for the SLE patients. The control group for the RA patients consisted of 135 individuals (69% women) who were also randomly selected from healthy blood donors with no family history of an autoimmune disease. Each control group was matched for sex with the corresponding disease group. All patients and controls were Chinese subjects residing in Taiwan.
Genomic DNA extraction.
Leukocytes were derived from EDTA-treated peripheral whole blood obtained from patients and healthy blood donors. Genomic DNA was extracted from the leukocytes using a GFX Genomic Blood DNA Purification kit (Amersham Pharmacia Biotech, Uppsala, Sweden) according to the manufacturer's protocol. Genomic DNA was stored at a concentration of 100 ng/μl in Tris–EDTA buffer.
Genotyping by nested polymerase chain reaction (PCR) and restriction analysis.
To amplify the PD-1 gene region encompassing the C+872T SNP, nested PCR was performed using purified genomic DNA as the template and 2 sets of PD-1 gene–specific primers. Primers were designed based on the exon 5 sequence in the published human PD-1 cDNA sequence (GenBank accession no. U64863). The sequences of outer primers for the first-round PCR were 5′-AGACGGAGTATGCCACCATTGTC-3′ and 5′-AAATGCGCTGACCCGGGCTCAT-3′. The sequences of inner primers for the second-round PCR were 5′-TAGCGGAATGGGCACCTCATC-3′ and 5′-AGTGTCCATGCTCAGGCCTCA-3′.
For the first-round PCR, the 30-μl reaction sample was set up to contain 1 μl of genomic DNA, 1× PCR buffer, 100 pmoles of each outer primer, 2.5 mM of each dNTP, 2.5 mM MgCl2, and 0.5 units of Taq polymerase (Boehringer Mannheim, Mannheim, Germany), and the DNA fragment was amplified for 30 cycles (1 minute at 94°C, 1 minute at 50°C, and 1 minute at 72°C). For the second-round PCR, the DNA fragment was amplified for 25 cycles (1 minute at 94°C, 1 minute at 50°C, and 1 minute at 72°C) in a 30-μl reaction sample that contained 3 μl of the first-round PCR product, 1× PCR buffer, 100 pmoles of each inner primer, 2.5 mM of each dNTP, 2.5 mM MgCl2, and 0.5 units of Taq polymerase.
An 89-bp fragment was amplified from genomic DNA after nested PCR. Three microliters of the nested PCR product was used for overnight restriction enzyme digestion with Pvu II, which digested DNA amplified from the T allele, but not from the C allele, into 48-bp and 41-bp fragments. After Pvu II digestion, DNA fragments were separated on 10% polyacrylamide gels and were visualized by ethidium bromide staining. Therefore, 3 different genotypes, T/T, T/C, and C/C, were determined by evaluating the ability of Pvu II to digest nested PCR products, as shown in Figure 1. In addition, the presence of the C+872T SNP was confirmed by automatic sequencing of some nested PCR products, using an ABI 377 Genetic Analyzer (Mission Biotech, Taiwan).
The genotype (C/C, C/T, and T/T) and allele (C or T) frequencies and the allele carriage frequencies (the percentage of individuals carrying at least 1 copy of the T or C allele) were determined by direct counting. The chi-square test with 2 degrees of freedom was used to determine the significance of the difference in genotype distributions. The chi-square test with Yates' correction and Fisher's exact test for ≥1 cell with <5 counts were used to test the significance of differences in 2 × 2 contingency tables. P values for T allele positivity were obtained from the comparison between individuals with at least 1 T allele (T/T genotype plus T/C genotype) and those with the C/C genotype; P values for C allele positivity were obtained in a similar way. Two-sided P values were calculated; P values less than 0.05 were considered significant. For statistically significant results, an odds ratio (OR) and a Cornfield's 95% confidence interval (95% CI) were calculated.
Detection of a SNP in the human PD-1 gene and determination of genotypes.
To identify a PD-1 SNP in the Chinese population in Taiwan, we focused on validating 3 candidate SNPs that were recognized by nucleotide mismatches in 3 previously documented human PD-1 cDNA sequences (National Center for Biotechnology Information [NCBI] accession nos. L27440, U64863, and NM_005018). Nested PCR was performed to amplify the genomic regions flanking each candidate SNP, and the presence of a SNP was then determined by restriction analysis, using specific restriction enzyme recognition sites (preexisting or created by PCR primers) in amplified genomic fragments. In a pilot study, we therefore discovered a PD-1 SNP located in exon 5 of the PD-1 gene at nucleotide position +872 (counting from the first base of the published cDNA sequence) (4). Direct nucleotide sequencing of the amplified PCR products verified that in these Chinese subjects, this SNP contained the C or T allele, and it was therefore named C+872T SNP. This SNP was found to be present in Europeans as well, and has been documented in the NCBI database (SNP ID: rs2227981) (15).
Since the Pvu II restriction enzyme can recognize the T allele, but not the C allele, of the C+872T SNP, each person can be genotyped to have C/C, C/T, or T/T alleles by analyzing the patterns of Pvu II digestion of nested PCR products (see Figure 1 for examples). Although the nucleotide variation of this SNP does not change the encoded amino acid, this SNP still can be a useful genetic marker in linkage association studies to test whether the PD-1 gene is associated with the occurrence of SLE or RA. Thus, if a positive association is found, the causative genomic regions within or adjacent to the PD-1 gene can be identified.
Lack of association between the PD-1 gene and SLE susceptibility.
To determine whether there was an association between the PD-1 gene and SLE susceptibility, we tested this association in a genetic case–control study after genotyping 98 SLE patients and 100 sex-matched healthy control subjects for the C+872T SNP. The genotype and allele frequencies and the allele carriage frequencies in both groups were calculated by direct counting, as shown in Table 1.
Table 1. Frequency distribution of the PD-1 C+872T polymorphism in healthy subjects and in patients with SLE*
No. (%) in healthy subjects (n = 100)
No. (%) in SLE patients (n = 98)
PD-1 = programmed death 1; SLE = systemic lupus erythematosus.
We found that both the SLE and control populations appeared to be in Hardy-Weinberg equilibrium. More importantly, we found no significant difference in the genotype distribution or allele frequency between the SLE patient and control groups by genotype and allele tests (see Table 1 for P values). In fact, the genotype distribution in the SLE patient group was almost indistinguishable from that in the healthy control group. These data strongly suggest that there is no correlation between the PD-1 gene and SLE susceptibility in this Chinese population.
Association between the PD-1 gene and RA susceptibility.
Despite the lack of statistical association between the PD-1 gene and SLE susceptibility, the arthritic phenotype in PD-1–deficient mice prompted us to hypothesize that the PD-1 gene is associated with RA susceptibility. To test for this disease association, we conducted another case–control study and genotyped 84 RA patients and 135 sex-matched healthy control subjects for the C+872T SNP. Genotype distributions and allele frequencies in the RA patient and control groups were then calculated and compared for statistical analyses.
As shown in Table 2, the RA population, but not the control population, appeared to deviate from Hardy-Weinberg equilibrium. Statistical analysis of the genotype distribution revealed a significant difference between the RA patient and control groups, as did the analysis for the difference in allele frequencies (Table 2). When the allele carriage frequency in the RA patient group was compared with that in the control group, we also discovered a significant association between the T allele, but not the C allele, and RA susceptibility (OR 3.32, P < 0.0001, with >95% power at the 5% significance level). It therefore appears that the T allele influences RA susceptibility in a dominant mode.
Table 2. Frequency distribution of the PD-1 C+872T polymorphism in healthy subjects and in patients with RA*
No. (%) in healthy subjects (n = 135)
No. (%) in RA patients (n = 84)
OR (95% CI)
PD-1 = programmed death 1; RA = rheumatoid arthritis; OR = odds ratio; 95% CI = 95% confidence interval.
In addition, the carriage of the C/T genotype was overrepresented in RA patients, showing a statistically significant difference compared with the C/C genotype (OR 3.52, P < 0.00005, with 95% power at the 1% significance level). When compared with the C/C genotype, the T/T genotype also tended to increase the risk of RA susceptibility (OR 1.8), although this was not statistically significant. However, this direct comparison between the C/C and T/T genotypes may not be appropriate, since the number of subjects carrying the T/T genotype is inversely related to that of the C/T genotype, and the T/T genotype is therefore underrepresented in a fixed disease population of a case–control study, especially if 1 T allele can affect disease susceptibility and the frequency of the T/T genotype is relatively low. Taken together, these results strongly suggest that the PD-1 gene is associated with RA susceptibility and that the disease risk is increased by carriage of the T allele or the C/T genotype.
Lymphocyte activation is strictly regulated by positive and negative signals that are delivered through various immunoregulatory receptors (16). A defect in the negative signals from immunoinhibitory receptors may reduce the threshold of autoreactive lymphocyte activation and lead to the development of autoimmune diseases. This has been evidenced by the expression of the autoimmune phenotype or lymphocyte hyperactivity in genetically manipulated mice with a deficiency in an immunoinhibitory receptor (Fcγ receptor IIB [FcγRIIB], CD22, CTLA-4, or PD-1) and in mice deficient in an inhibitory signaling molecule (Lyn or src homology 2 domain–containing phosphatase 1) (17–21). Thus, negative signals initiated by immunoinhibitory receptors appear to play an important role in the maintenance of peripheral tolerance in mouse.
In humans, the contribution of the immunoinhibitory receptors and their signal molecules to the development of autoimmune diseases remains unclear. Nevertheless, the immunoinhibitory receptors FcγRIIB, CD22, and CTLA-4 have been used as candidate genes in SNP case–control association studies to test their possible association with the human autoimmune diseases SLE and RA. It was found that FcγRIIB and CD22, but not CTLA-4, are associated with SLE susceptibility, whereas neither CD22 nor CTLA-4 appears to be linked to RA susceptibility (22–28).
Our case–control genetic-association study using the C+872T SNP in the PD-1 gene is the first to show evidence of an association of the PD-1 gene with susceptibility to RA, but not to SLE, in the Chinese population. In contrast to our findings, Prokunina and coworkers (15) recently found that the PD-1 gene is associated with SLE susceptibility in European and Mexican multicase families, but not in African American multicase families. This discrepancy may be explained by the influence of the genetic background on susceptibility to autoimmunity, similar to the differential expression of autoimmune phenotypes in different strains of mice deficient in FcγRIIB or PD-1 (11, 12, 17). Yet, further studies of ethnic differences in the expression and function of PD-1 may provide better insight into the role of the PD-1 protein in initiating autoimmune diseases in different human populations.
The molecular mechanisms that account for the association between RA susceptibility and the T allele and the C/T genotype of the C+872T SNP remain to be clarified. One possible mechanism is that the C+872T SNP may be associated with an alteration in the level of expression of the PD-1 gene, possibly as a result of linkage disequilibrium with other PD-1 gene polymorphisms, which differentially control PD-1 gene transcription. This possibility is supported by the identification of several SNPs, including C+872T, in the promoter, introns, exon 5, and 3′-untranslated region of the PD-1 gene in the European population and the finding that a SNP in intron 4 alters the runt-related transcription factor 1 binding site (15). Since there was a significant association between RA and the T allele of the C+872T SNP, we may, by this hypothesis, expect a dose-dependent effect of the T allele on RA susceptibility, but this is inconsistent with the observed lack of a significant association between the T/T genotype and RA susceptibility. However, the absence of an association between RA and the T/T genotype in the present study may be due to the occurrence of a genotype distribution that is skewed toward the C/T genotype in the disease group in a case–control study, because carriage of 1 T allele may be able to dominantly influence susceptibility to RA. An evaluation of the correlation between different genotypes and PD-1 expression levels may help in the assessment of this hypothesis.
Another possible explanation for the significant association of the C+872T SNP with RA susceptibility is the connection between this SNP and the functional change in the PD-1 protein via linkage disequilibrium with other nucleotide polymorphisms that alter the sequence and structure of the PD-1 protein. This possibility may be tested by evaluating the ability of PD-1 protein to transduce inhibitory signals in response to ligand stimulation in individuals with different genotypes. It is possible that finding a correlation between genotypes and modifications in the function of the PD-1 protein may provide an explanation for the significant association between RA susceptibility and the C/T genotype, but not T/T genotype, that was observed in this study. However, we still cannot exclude the possibility that another genuine RA-susceptibility gene is located adjacent to the PD-1 gene in the region of chromosome 2q37.3 and affects the significant association between the PD-1 gene and RA susceptibility, particularly if the expression and function of PD-1 are not correlated with different C+872T SNP genotypes.
In conclusion, we have shown experimental evidence that a genetic locus around an immunoinhibitory receptor gene, PD-1, is linked to RA susceptibility in the Chinese population. These data may provide a starting point for further investigations of whether the defect in PD-1 expression and function plays an important role in the pathogenesis of RA.
We are grateful to the SLE and RA patients and the healthy blood donors who provided blood samples for this research. We also thank Yun-Mei Hsiao and Yanfeng Lu for technical assistance and for review of the manuscript.