Genetic polymorphisms associated with psoriasis and development of psoriatic arthritis in patients with psoriasis

Background Psoriasis (PsO) is a chronic inflammatory disease with predominantly cutaneous manifestations. Approximately one third of patients with PsO develop psoriatic arthritis (PsA), whereas the remaining proportion of patients has isolated cutaneous psoriasis (PsC). These two phenotypes share common immunology, but with different heredity that might in part be explained by genetic variables. Methods Using a candidate gene approach, we studied 53 single nucleotide polymorphisms (SNPs) in 37 genes that regulate inflammation. In total, we assessed 480 patients with PsO from DERMBIO, of whom 151 had PsC for 10 years or more (PsC10), 459 patients with PsA from DANBIO, and 795 healthy controls. Using logistic regression analysis, crude and adjusted for age and gender, we assessed associations between genetic variants and PsO, PsC10, and PsA, as well as associations between genetic variants and development of PsA in PsO. Results Eleven polymorphisms in 10 genes were nominally associated with PsO and/or PsC and/or PsA (P < 0.05). After correction for multiple testing with a false discovery rate of 5%, two SNPs remained significant: TNF (rs361525) was associated with PsO, PsC10, and PsA; and IL12B (rs6887695) was associated with PsO. Conclusion Among a cohort of Danish patients with moderate-to-severe psoriasis, two SNPs in the IL12B and TNF genes were associated with susceptibility of psoriasis. None of the SNPs were specifically associated with isolated cutaneous psoriasis or psoriatic arthritis.

Epidemiological studies suggest stronger heritability for PsA than for PsC (9) which indicates that there could be individual risk loci for these two disease entities. Identification of such loci could potentially serve as novel drug targets and possibly improve patient outcomes by facilitating earlier detection of PsA, thereby possibly preventing irreversible joint destruction [7][8][9]. Indeed genetic differences related to the human leukocyte antigen (HLA) class I region of the major histocompatibility complex (MHC) have been well established. Variants of HLA-B, namely HLA-B Ã 27, have been proven to confer risk of PsA [10], while HLA-C Ã 06 has been demonstrated to hold a specific risk for PsC [10][11][12][13][14].

Ethical considerations
The study was conducted in accordance with the Declaration of Helsinki and was approved by

Genotyping
For the patients, DNA was extracted from cryopreserved blood clots by using the Maxwell 16 Blood purification kit (Promega, Madison, Wisconsin, USA), as described by Bank et al. [33]. DNA from the healthy controls was extracted according to the manufacturers' instructions from EDTA-stabilized peripheral blood using either a either PureGene (Qiagen, Hilden, Germany) or Wizard Genomic (Promega, Madison, Wisconsin, USA) DNA purification kit [32].
A candidate gene approach was chosen, with focus on genes involved in the NFκB pathway, TNF-α signaling, the IFN-γ pathway, and the IL-23/IL-17 axis, as described earlier [28][29][30]. In addition, other genes involved in regulation of inflammation including the inflammasome were assessed [34]. In short, relevant genes were identified by searching pathway databases (http://www.genome.jp/kegg/pathway.html and http://www.wikipathways.org/index.php/ WikiPathways) and SNP candidates were identified by searching PubMed with "polymorphism AND Gene name AND (reporter gene OR luciferase OR ELISA OR RT-PCR OR flow cytometry OR EMSA)". The SNPs were selected based on the reported functionality or associated with autoimmune diseases (S1 Table). Additionally, SNPs or genes associated with response to biological treatment of either inflammatory bowel diseases or rheumatoid arthritis were included [28-30].
As a quality control, all SNPs were replicated for 94 randomly selected samples, yielding >99% identical genotypes.

Power analysis
At the 5% significance level, and a minor allele frequencies (MAF) of 0.05, 0.25, and 0.45 there is > 80% power for detecting a dominant effect with an odds ratio (OR) of 1.7, 1.4, and 1.5, respectively, for PsO, an OR of 2.0, 1.7, and 1.9, respectively, for PsC10, an OR of 1.7, 1.4, and 1.5, respectively, for PsA, and for PsA in patients diagnosed with PsO an OR of 1.8, 1.5, and 1.6, respectively ( Table 1). The Genetic Power Calculator [36] was used for power calculations, setting 'prevalence' of PsO to 2%, of PsC10 to 1.4%, of PsA to 0.6%, and of PsA in PsO to 30%, D-prime to 1, type 1 error rate to 0.05, and number of cases and control:case ratio was based on data described in Table 2.

Statistical analysis
Assuming a dominant model, we compared genotype distributions for the following groups: patients with PsO and PsC10 from DERMBIO compared with healthy controls using logistic regression analysis; patients with PsA from DANBIO compared with healthy controls; patients

Polymorphisms associated with the risk of PsO in the general population
The association of SNPs in patients with PsO was compared with healthy controls using a dominant model for association. Nine SNPs were nominally associated with PsO (P < 0.05) (S3 Table). Two SNPs [TNF (rs361525) and IL12B (rs6887695)] withstood correction for multiple testing (Table 3).

Polymorphisms associated with the risk of PsC10 in the general population
Patients were stratified for phenotype and diseased duration, and the genotype distributions of patients with PsC10 were compared with healthy controls. Five SNPs were nominally associated with PsC10 (S3 Table), among which one SNP [TNF (rs361525)] withstood correction for multiple testing (Table 3).

Polymorphisms associated with the risk of PsA in the general population
Genotype distributions for patients with PsA from DANBIO were compared with those from healthy controls using a dominant model for association. Two SNPs were nominally associated with PsA (S3 Table), with one withstanding correction for multiple testing (Table 3).

Polymorphisms associated with the risk of PsA in patients with PsO
Genotype distributions for patients with PsA from DANBIO were compared with patients with PsC10. Only TNF (rs361525) (OR: 0.59, 95% CI: 0.38-0.92, P = 0.019, q = 0.32) showed nominally evidence of association, but this did not withstand correction for multiple testing (S3 Table).

Discussion
This study evaluated 53 SNPs in 37 genes in 480 Danish patients with moderate-to-severe PsO and in 459 patients with PsA. The polymorphisms were primarily chosen as functional polymorphisms targeting the inflammatory signaling pathways. Eleven polymorphisms in 10 genes were nominally associated with PsO and/or PsC and/or PsA (P < 0.05), among which two withstood correction for multiple testing.  [48][49][50][51][52] are all well-known polymorphism associated with susceptibility to PsO. We found that the variant allele of IL23R (rs11209026) was nominally associated with decreased risk of PsO, but this did not withstand correction for multiple testing.
In the current study, the variant allele of TNF (rs361525) was the only variant consistently associated with an increased risk of all phenotypes, thus underlining the importance of TNF-α signaling in the development of psoriasis. Conflicting results for the association of TNF (rs361525) and development of PsA in PsO have previously been reported. While some studies have reported a protective role for the variant allele of TNF (rs361525) in development of PsA in PsO [53,54], other studies have reported increased or no altered risk for development of PsA in PsO [55,56]. In the current study, the variant allele of TNF (rs361525) was associated with an increased risk of PsA and nominally associated with a decreased risk of developing PsA in PsO. The strong association between PsO and TNF (rs361525), and the increased risk of PsC compared to PsA have been suggested to be due to a high LD between TNF (rs361525) and HLA-CW Ã 6 [54]. Another SNP investigated, IL12B (rs3212217), is in close LD with two other polymorphisms: IL12B (rs3212227, r 2 = 0.95) [38- 45,57] and IL12B (rs2082412, r 2 = 0.95) [11,17], both associated with reduced risk of PsO. Differences in effect size between PsC and PsA for IL12B (rs2082412) have been reported [11,17] although the associations could not be replicated by a large GWAS that attempted to confirm previously reported associations [12]. In the current study, no statistically significant difference in effect size between PsA and PsC10 for IL12B (rs3212217) was observed. Thus the current study adds to the evidence that suggest that there is no difference in PsA and PsC10 susceptibility for the variants in IL12B.
In a previous study, the variant allele of PPARG (rs1801282) was associated with decreased risk of PsA [58]. Another study demonstrated the same trend, although the findings were not statistically significant [59]. In the current study, the variant allele of PPARG (rs1801282) was nominally associated with decreased risk of PsA, with odds ratios similar to those previously reported, but the association was not statistically significant after correction for multiple testing. In accordance with previous studies [60], we observed no association between PPARG (rs1801282) and uncomplicated psoriasis, PsO or PsC10. Additional studies investigating association of PPARG (rs1801282) with PsA should be performed in order to clarify the role of PPARG (rs1801282) in PsA. PPARG (rs1801282) encodes a Pro to Ala amino acid substitution in PPARG that leads to reduced activity. PPARG (rs1801282) is a variant of notable interest that has, in addition to PsA, been associated with a lower risk of alcohol-related breast cancer [61,62], but an increased risk of alcohol-related colorectal cancer [63] likely because PPARG inhibits sex hormone synthesis via negative regulation of aromatase [62].
We failed to replicate PTPN22 (rs2476601) as a risk factor of PsA in PsO, probably due to lack of power, since we observed odds ratios similar to those previous reported [12,20].
The main limitation in this and other studies evaluating risk of PsA in PsO is the potential for phenotype misclassification, which inevitably will lead to a decrease in statistical power. In this study, data on PsA, for the DERMBIO cohort, were retrieved from the DERMBIO database, with patient registration performed by dermatologists. Although dermatologist are experienced and trained in assessing PsA, we cannot rule out the possibility that some patients with PsC may have had undiagnosed or undeveloped PsA. This might bias the effect size towards the null hypothesis of no difference. In order to overcome the possibility of undeveloped PsA and reduce the risk of undiagnosed PsA, a cohort of patients with PsC who had the disease for !10 years (PsC10) was established, as proposed by Stuart et al [12]. Similarly, there is risk of potential misclassification in the PsA cohort (e.g. rheumatoid arthritis or osteoarthritis). However, this risk was lowered by only including patients registered in DANBIO, where diagnoses are according to the treating rheumatologist.
In conclusion, this study confirms that two SNPs in the IL12B and TNF genes are associated with susceptibility to psoriasis in Danish patients with moderate-to-severe psoriasis. None of the investigated SNPs were specifically associated with isolated cutaneous psoriasis or psoriatic arthritis.
Supporting information S1