Single Nucleotide Polymorphisms in the Tumor Necrosis Factor-Alpha Gene Promoter Region Alter the Risk of Psoriasis Vulgaris and Psoriatic Arthritis: A Meta-Analysis

Background It has been confirmed that tumor necrosis factor-alpha (TNFα), a macrophage-derived pro-inflammatory cytokine, plays an important role in the pathogenesis of psoriasis vulgaris and psoriatic arthritis (PsV&PsA). In contrast, the reported association of TNFα gene promoter region single nucleotide polymorphisms (SNPs) and PsV&PsA has remained controversial. Accordingly, we performed a meta-analysis to provide new evidence that SNPs in the TNFα gene promoter region alter not only the risk of psoriasis vulgaris (PsV) or psoriatic arthritis (PsA) but also of PsV&PsA. Methods Interrelated literature dated to October 2012 was acquired from the PubMed, ScienceDirect, and SpringerLink databases. The number of the genotypes and/or alleles for the TNFα promoter in the PsV and PsA and control subjects was obtained. Odds ratios (ORs) and 95% confidence intervals (CIs) were used to calculate the risk of PsV and/or PsA with TNFα promoter SNPs. Results A total of 26 papers of 2159 for PsV (2129 normal controls) and 2360 for PsA (2997 normal controls) were included in our meta-analysis. The results showed that the variant genotype and allele of TNFα -308A/G was protective in pooled groups of patients with PsV&PsA (OR = 0.682, 0.750; 95% CI, 0.596-0.779, 0.653-0.861). However, the variant genotypes and alleles of TNFα -238A/G and -857T/C had an increased risk of PsV&PsA (OR = 2.493, 2.228, 1.536, 1.486, 95% CI, 1.777-3.498, 1.628-3.049, 1.336-1.767, 1.309-1.685). Moreover, the meta-analysis revealed a significant association between TNFα -238A/G and -857T/C polymorphism and PsA susceptibility (OR = 2.242, 2.052, 1.419, 1.465; 95% CI, 1.710-2.941, 1.614-2.610, 1.214-1.658, 1.277-1.681). In contrast, the variant genotypes and alleles of TNFα -308A/G proved to be protective against PsV (OR = 0.574, 0.650, 95% CI, 0.478-0.690, 0.556-0.759), whereas TNFα -238A/G was found to have a risk association (OR = 2.636, 2.223, 95% CI, 1.523-4.561, 1.317-3.751). Conclusions SNPs in the TNFα gene promoter region alter the risk of PsV and/or PsA.


Introduction
Psoriasis is a common autoimmune disorder that primarily involves the skin (psoriasis vulgaris, PsV). Severe complications, including generalized involvement of the body (erythroderma), extensive pustular lesions, and an associated arthritis known as psoriatic arthritis (PsA), make psoriasis complex and heterogeneous [1]. The worldwide prevalence of PsV varies from 0.91% (United States) to 8.5% (Norway) in adults [2]. However, only 34.4% of patients with PsV also have arthritis, with the onset of skin lesions preceding arthritis in 64.5% of the cases and arthritis appearing after PsV in only 19.35% of cases. In the remaining 16.1% of cases, PsV and PsA began almost simultaneously [3].
Thus, the relationship between the onset of PsV and PsA included in the spondyloarthritis group is closely related, although the exact pathogenesis has yet to be identified. The diseases are multifactorial and result from the interplay between multiple genetic, immunologic, and environmental factors. Under the category of genetic and environmental effects, cytokines, chemokines, adhesion molecules, and growth factors combined with T cells and dendritic cells act in an integrated manner to evolve into unique inflammatory and proliferative processes in psoriasis vulgaris and psoriatic arthritis (PsV & PsA) [1].
Tumor necrosis factor-alpha (TNFa), a macrophage-derived pro-inflammatory cytokine, has been proven to play an important role in the pathogenesis and development of PsV & PsA. By inducing and recruiting multiple cytokines, TNFa activates the inflammatory process of PsV & PsA in the skin, joints, and bowel [4]. The marked clinical response of TNFa blockers in patients has demonstrated the importance of TNFa in both PsV and PsA [5]. The TNFa gene has been localized to the major histocompatibility complex (MHC) region on chromosome 6p21.3, which is 250 kb centromeric from the human leukocyte antigen-B and thought to be a high-priority candidate gene in PsV & PsA [6]. Single nucleotide polymorphisms (SNPs) are common in the TNFa gene, especially in promoter region, and act as markers of disease susceptibility because of their influence on TNFa expression by a cooperative manner [7]. Kaluza et al. found that the psoriasisassociated TNFa promoter allele TNF238.2 showed a significantly decreased transcriptional activity and less tumor necrosis factor alpha production [8]. Although many studies have investigated the relationship between promoter region SNPs in the TNFa gene and the susceptibility to rheumatoid arthritis (RA), ankylosing spondylitis (AS), Crohn's disease, PsV, and PsA [9][10][11][12][13], the association between TNFa promoter SNPs and PsV & PsA risk have produced inconsistent results. Some studies have found a significant association between TNFa promoter SNPs and PsV & PsA, whereas others have failed to confirm these results [14,15]. Two meta-analyses, one of nine studies, the other of ten, were performed in 2006 and 2007, respectively, to evaluate the association between TNFa promoter SNPs and the susceptibility to PsA and PsV [11,13]. However, because of the complexity and heterogeneity of disease, the association between PsV and PsA on pathogenesis is still unclear [16]. Thus, it is necessary to pool the results of these studies and systematically analyze the relationship between TNFa promoter SNPs and PsV & PsA risk. Moreover, new studies about the association between TNFa promoter SNPs and the susceptibility to PsV & PsA have recently been published and provide new evidence that was not included in the previous meta-analyses, particularly regarding the susceptibility alleles TNFa -857T/C, TNFa -1031C/T, and TNFa -863A/C [14,15,[17][18][19][20][21]. Therefore, it is essential to collect the most comprehensive evidence and reassess the association of TNFa promoter SNPs with PsV and/or PsA.

Inclusion and exclusion criteria
For selection in the analysis, each study met all of the following criteria: (1) clinical subtypes of psoriasis, including PsV and/or PsA, according to the Moll and Wright or CASPAR criteria or diagnosed by clinical specialists; (2) a case-control or cohort study; (3) the inclusion the distributions of genotypes and/or alleles within the TNFa promoter provided for both the patients and controls; and (4) the reporting of the basic characters of the participants, including ethnicity, age and gender. The exclusion criteria were as follows: (1) a case report, review or descriptive study; (2) clinical subtypes of psoriasis, including erythroderma or extensive pustular lesions; (3) a lack of normal population as controls; (4) duplicate data in the studies; and (5) an inability to obtain the necessary data.

Quality assessment
The quality assessment of the selected studies was performed independently by two authors using the Newcastle-Ottawa Scale (NOS) [22]. Three parameters of quality, selection, comparability, and exposure assessment were assessed for the case-control studies. We defined ''low-quality studies'' as those with scores of 1-3, ''moderate-quality studies'' as those with scores of 4-6, and ''highquality studies'' as those with scores of 7-9. Any disagreements were resolved by discussion.

Data extraction
The following information was extracted from each study: first author, publication year, country of origin, patient ethnicity and demographic characteristics, subtypes of psoriasis (PsV type I, II, and/or PsA), the number of cases and controls, the number of genotypes and/or alleles for the TNFa promoter in the cases and controls, and consistency tests of genotype frequencies with the Hardy-Weinberg equilibrium model (HWE). Two authors analyzed each paper according to the inclusion and exclusion criteria and independently extracted the information. In some studies, only the novel data were extracted because a portion of the data had already been reported. Disagreements were resolved by discussion.

Statistical analysis
The risks of the variant genotypes and alleles compared with the wild-type for six sites of TNFa promoter SNPs were estimated separately. Different subgroups divided by ethnicity or disease subtype were analyzed. Odds ratios (ORs) and 95% confidence intervals (CIs) were used to calculate the risks of PsV and/or PsA with TNFa promoter SNPs. A p value ,0.05 was considered statistically significant. In the forest plots, OR.1 represented a risk effect and ,1 a protective effect. In our meta-analysis, the heterogeneity among the studies was tested by a x 2 -based Q test and I 2 statistic [23]. When p,0.10 or I 2 .50%, the heterogeneity was considered significant. A fixed-effects model was a better choice to calculate the pooled ORs when the heterogeneity was not significant, otherwise a random-effects model was used because of the heterogeneity among the studies. To explore the source of heterogeneity, a meta-regression analysis was performed when the number of studies was more than ten in a group for which heterogeneity existed [24]. A p value ,0.05 was considered statistically significant. The size of the heterogeneity source (SH) was assessed by percentage. Both Begg's and Egger's tests were performed to check the publication bias [25]; the publication bias was considered significant when p,0.10 in either test. All of the analyses were conducted using Stata software version 11.0 (Stata Corporation, College Station, TX, USA).
Details about the first author, publication year, country of origin, ethnicity, subtype of psoriasis, number of cases and controls, number of genotypes and/or alleles for the TNFa promoter in the cases and controls, consistency tests of genotype frequencies with HWE, and the NOS score are shown in Tables 1 and 2 for each study selected. The publishing date of all the studies ranged from 1997 to 2012. All of the patients were Caucasian, with the exception of five studies that involved Asian patients. Consistency tests of genotype frequencies with HWE were performed in 8 (47.1%) and 9 (64.3%) studies of PsV and PsA, respectively.
Our meta-analysis did not reveal a significant association between TNFa -308A/G, -1031C/T and -863A/C polymorphism and the susceptibility to PsA. Together with the results of heterogeneity and publication bias tests, the values of ORs and 95% CIs for the risk of susceptibility in patients with PsA are shown in Table 3. All of the ORs were calculated using the fixedeffects model, although one pooled OR for the comparison of the TNFa -308A/G allele (A vs G) between PsA and the controls was observed using the random-effects model (Q = 19.20, p = 0.038, I 2 = 47.9%). No pooled OR was performed for the TNFa -488A/ G polymorphism, as only one study was involved.
The overall meta-analysis involving 11 and 12 studies separately indicated that the variant genotypes (AA+AG) and allele (A) of TNFa -238A/G had an associated increase risk of PsV (OR = 2.636, 2.223, 95% CI, 1.523-4.561, 1.317-3.751, p = 0.001, 0.003 [separately], Figure 5 a, b) using the randomeffects model (Q = 45.76, 58.23, p = 0.000, 0.000, I 2 = 78.1%, 81.1% [separately]). To explore the source of heterogeneity, a meta-regression analysis was performed, and the result showed that the concomitant variable of publication year had a statistical significance (p = 0.02, 0.042, SH 54.8%, 37.1% [separately]) while ethnicity, type of PsV and HWE consistency test had no statistical significance (p.0.05). However, we also calculated the pooled ORs in the subgroups divided by ethnicity and type of PsV, and the same risk of the variant genotypes (AA+AG) and allele (A) of TNFa -238A/G was found in all the subgroups with the existence of heterogeneity using the random-effects model, which was not found using the fixed-effects model in subgroups of type II PsV (Figure 5 c, d, e, f).
No pooled OR was calculated for the TNFa -857T/C and -1031C/T polymorphisms, as only two studies were involved.

Publication bias
For the genotypes and alleles of TNFa -308A/G, the publication bias showed statistical significance in the pooled groups of PsV & PsA (Begg's test, p = 0.040, 0.106; Egger's test, p = 0.021, 0.034 [separately]). Similarly, the same publication bias was found in the pooled groups of PsV (Begg's test, p = 0.025, 0.054; Egger's test, p = 0.070, 0.050 [separately]). The funnel plots for publication bias are shown in Figure 6 and suggest the evidence of potential publication bias. In all other groups, neither the Begg's (p.0.1) nor Egger's tests (p.0.1) showed statistical significance, which suggested that no publication bias existed (data not shown).

Discussion
It is well known that TNFa plays a crucial role in autoimmune and infectious diseases as a proinflammatory cytokine; however, the mechanism of altered PsV & PsA risk owing to SNPs in the TNFa gene promoter region remains unknown. A possible explanation is the abnormal expression of the TNFa gene mediated by variations in the TNFa promoter. This premise was supported by studies noting that the TNFa -857C allele was associated with positive response to drug treatment in PsV patients treated with etanercept (Recombinant Human Tumor Necrosis Factor-a Receptora:IgG Fc Fusion Protein) [43]. However, the connection between TNFa transcriptional activity or protein production and TNFa promoter SNPs is unresolved, as the reports are inconsistent with some studies revealing no association between them [44]. Therefore, more complex mechanisms than the abnormal expression of the TNFa gene must exist. Another possible explanation is that TNFa promoter SNPs is also associated with the production of other cytokines that play an important role in the pathologic process of the disease. For example, TNFa -308A allele was associated with higher serum interferon-alpha levels in the patients with juvenile dermatomyositis [45]. Furthermore, because of the complexity of biological systems, a single gene polymorphism may result in further polymorphisms or post-transcriptional regulation under different gene-gene or gene-environment interactions. Accordingly, further studies are needed to assess the functional consequences of TNFa promoter SNPs. In the present study, the variant genotypes and alleles of TNFa promoter SNPs were found to contribute to the same risk of disease susceptibility in all groups or subgroups.  [16]. On the other hand, the diversity of skin lesions, including the presence and severity of a clinical rash, does not affect the clinical manifestation of joint disease, which is found in 30% of patients with skin lesions [4]. However, the common pathogenesis, activation of T lymphocytes and proinflammatory cytokines, is now well recognized in patients with PsV and/or PsA [16]. In addition, the curative effect of TNFa blockers in patients with PsV and/or PsA indicates that, to a certain degree, they share a common immune mechanism [5]. Therefore, both PsV and PsA may have a common onset risk with regard to TNFa promoter SNPs. Our meta-analysis results confirmed this hypothesis through the quantitative synthesis of controversial studies. The TNFa -857T/C variant significantly increased the susceptibility to PsV & PsA (OR = 1.536, 95% CI, 1.336-1.767), without heterogeneity among the groups. For TNFa -238A/G, a greater risk (OR = 2.493, 95% CI, 1.777-3.498) was found using the random-effects model, with heterogeneity primarily regarding the year of publication. However, the TNFa -308A/G variant exhibited a protective effect, with heterogeneity primarily in the disease subtype (OR = 0.682, 95% CI, 0.596-0.779). Further subgroup-analysis found the same risk or protective effect in most of the subgroups without heterogeneity. In a word, a variation in the levels of TNFa production attribute to a functional consequence of these TNFa promoter SNPs, which results to a common onset risk in both PsV and PsA.
Although a meta-analysis completed by Rahman et al. revealed that PsA risk was associated with TNFa -238A/G polymorphism (OR = 2.29; 95% CI 1.48-3.55) and not with TNFa -308A/G polymorphism (OR = 0.92, 95% CI, 0.69 -1.23) [13], more recent studies also paid close attention to other SNPs in the TNFa promoter, such as TNFa -857T/C [17]. Additionally, an increasing number of studies have provided new evidence for PsA susceptibility, although the results are controversial. Our meta-analysis gathered the comprehensive evidence and is the first to reveal that TNFa -857T/C polymorphism is associated with PsA susceptibility using the fixed-effects model (OR = 1.419, 95% CI, 1.214-1.658). Lv, K. et al. have observed in vitro that the TNFa -857T/C variant is a stronger transcriptional activator in response to lipopolysaccharide stimulation in the RAW264.7 cell line [46], indicating that the main mechanism of increased PsA risk can be attributed to the TNFa -857T/C polymorphism. Moreover, the results reflecting the association between TNFa -238A/G or TNFa -308A/G polymorphism and PsA susceptibility are remarkably consistent compared with the previous metaanalysis, although the pooled ORs differ because more studies and the fixed-effects model were used in our meta-analysis.
Furthermore, similar to a previous meta-analysis performed by Chunying Li et al. [11], our meta-analysis results showed that the TNFa -238A/G and TNFa -308A/G polymorphisms separately had the same risks and protective effects with regard to PsV susceptibility. However, in the subgroup-analysis, the results differ because recent studies have provided new evidence for the role of TNFa promoter SNPs in PsV susceptibility [15]. In the Asian subgroup, the TNFa -238A/G variant increased PsV risk by 1.912 (95% CI 1.047-3.491) using the fixed-effects model compared with the value of 1.46 (95% CI 0.66-3.19) revealed previously. However, no association was found between PsV risk and TNFa -238A/G or TNFa -308A/G polymorphism (OR = 1.569, 0.856, 95% CI 1.11-8.71, 0.523-1.400) using the fixed-effects model for the type II PsV subgroup compared to the previous Chunying Li et al. meta-analysis (OR = 3.11, 0.57, 95% CI, 0.800-3.078, 0.37-0.88). The inconsistency in the meta-analysis results indicates that the alterative risk of PsV result from TNFa promoter SNPs may be more complex because of the presence of stratification factors, such as ethnic admixture and the subtype or severity of disease.
Our meta-analysis has several limitations. First, heterogeneity was observed in some groups; although a subgroup analysis was performed, heterogeneity also existed with unclear sources in some subgroups. This finding indicated that the genetic background, combined with linkage disequilibrium patterns, epigenetic modifications, and clinical heterogeneity, made the relationship between the risk of disease susceptibility and TNFa promoter SNPs more complicated. Second, on analyses the risk of psoriasis with TNFa promoter SNPs, patients only with PsV and/or PsA were involved, but not with erythroderma or pustular lesions. Besides, only Caucasian and Asian patients were involved in the subgroup analysis; therefore, the assessment of disease risk is not applicable to other disease subtypes and ethnicities. Third, apart from the limited sample size or quality control of genotyping in some of studies, some subgroup analyses only involved three studies; this small sample size indicated insensitivity with regard to exploring the real association. Finally, as some negative-association studies are often unpublished, it seems unavoidable that publication bias may influence the meta-analysis results. Accordingly, caution should be exercised when interpreting the results of this meta-analysis.

Conclusions
This meta-analysis confirms the previous results that TNFa -238A/G and TNFa -308A/G polymorphisms are associated with PsV or PsA susceptibility. Furthermore, this meta-analysis shows that the TNFa -857T/C variant increases the risk of PsA significantly, whereas TNFa -857T/C and TNFa -238A/G polymorphisms are risk effects in the pooled analysis of PsV & PsA. In contrast, the TNFa -308A/G polymorphism is a protective effect for PsV & PsA. No significant association between TNFa-308A/G, -1031C/T, or -863A/C polymorphism and PsA susceptibility was revealed by this meta-analysis. Thus, further studies are needed to elucidate the precise contribution of those SNPs to the pathogenesis of PsV& PsA by using proteome-wide analysis of SNPs (PWAS) [47]. Additionally, prospective studies on the response to the TNFa blockers in patients with PsV& PsA to their TNFa promoter SNPs are benefit for properly assessing the impact and value of TNFa blockers in patients.