Association of RETN and TNFRSF1B polymorphisms with TNF-α inhibitor response in rheumatoid arthritis patients

Background : Despite the improvement from the introduction of tumor necrosis factor inhibitors (TNFi) in the rheumatoid arthritis (RA), TNFi therapy fails for more than 30% or results in a partial response. Thus, we aimed to explore treatment marker by examining the association of single nucleotide polymorphisms (SNPs) with response to TNFi therapy. Method : Genes associated with RA or RA treatment were reviewed and fourteen SNPs with minor allele frequency ≥ 20% in the East Asian populations were selected and analyzed. Data were collected from 105 RA patients. Our primary endpoint was the disease activity score using 28-joint count after six months of treatment (DAS28- 6month ). The secondary outcomes were the subcomponents of DAS28. Results : A total of 88 patients were included in the final analyses. Among the 14 SNPs analyzed, one SNP showed statistical significance in DAS28- 6month : patients with the GG allele of RETN rs1862513 had a 4.7 times higher chance of low disease activity at 6-months than GC or CC-carriers ( p = 0.033), as indicated by multivariable logistic regression analysis. Rs3397 was marginally significant in univariate analysis (p=0.059), but was significant in the multivariable model (p=0.041). The final model explained 24.5% (Nagelkerke R 2 ) of the variance in DAS28- 6month . Conclusion : Our results demonstrated that, among the genes related to RA, SNPs in RETN and TNFRSF1B were associated with the response of TNFi treatment.

triangle". Genetic susceptibility has been studied in relation to environmental factors, mainly smoking and in part alcohol intake [1][2][3][4]. Indeed, a genetic component may be responsible for up to 60% of susceptibility to RA, suggesting the importance of genetics in RA [5]. On the other hand, studies on responses to treatment in relation to genetic factors are inconsistent and the results vary among different population groups. [1][2][3] Tumor necrosis factor alpha (TNF-α) inhibitors (TNFi) have demonstrated efficacy in RA treatment either as monotherapy or in combination with other disease-modifying antirheumatic drugs. Five TNF-α inhibitors are currently available for RA therapy: etanercept, a fusion protein that was first approved by the US FDA in 1998 to treat RA, and four anti-TNF-α monoclonal antibodies (infliximab, adalimumab, golimumab and certolizumab).
Despite the progress made by the introduction of TNFi, a partial response or treatment failure is observed in more than 30% of patients [6]. Therefore, it is vital to discover prognostic factors associated with TNFi response in order to avoid missing other potentially effective treatments at an early stage of disease.
Pharmacogenomics studies on TNFi show that genetic variations are important in predicting the response to treatment. Genetic variations in the HLA-DRB1 and TNF regions are associated with the response to TNFi in Caucasians, but these associations have failed to reach significance in Korean patients [4][5][6]. Although genes or specific mutations differ among ethnic groups, biological pathways related to immune signaling and inflammation are commonly associated with the response to TNFi in Koreans [6][7][8]. Therefore, genes associated with RA or other autoimmune diseases could affect the therapeutic response to TNFi as they are involved in common inflammatory pathways. In the present study, we examined the association of genetic factors to the TNFi response in RA patients in Korea.

Patients
A total of 105 RA patients from two teaching hospitals who were given TNF-α inhibitors (etanercept, infliximab, adalimumab, or golimumab) from July 2017 to December 2019 were recruited. Patients' data were collected from electronic medical records and included age, sex, age at diagnosis, weight, height, concomitant drugs, comorbidities, and autoantibodies against rheumatoid factor (RF) and anti-cyclic citrullinated peptide (ACPA).
The treatment response was accessed by the DAS28-ESR at 6 months after starting TNFi treatment (DAS28-6m onth ). DAS28-6m onth was ≤3.2 in the low disease activity group and >3.2 of DAS28-6m onth in the moderate-to-high disease activity group [9].

Genotyping
We examined previous genetic studies associated with RA, RA treatment or other autoimmune diseases and selected statistically significant SNPs (Additional file 1). The minor allele frequency from the International HapMap Project was used to capture common SNPs present in > 20% of the East Asian (Han Chinese and Japanese) populations [10].

Statistical analysis
Categorical variables were analyzed by chi-square test, and an independent-samples t-test was used to compare means of continuous variables between patients with DAS28-6m onth ≤3.2 and >3.2. Multivariable linear regression was used to predict independent risk factors associated with TNFi treatment response. Forward selection of variables was used in the regression method using the probability of R 0.05 for entry and 0.10 for removal.
Regression smoothing was carried out using the loess function to validate the regression model. A receiver operating characteristic (ROC) curve was drawn to assess the predictive accuracy of the multivariable model. Statistical significance was considered at p-value of less than 0.05. Statistical analyses were performed using SAS (version 9.4, The SAS Institute, Cary, NC, USA).

Results
Among 105 patients enrolled, a total of 88 patients were included in the analyses.
Seventeen patients were excluded due to the incomplete medical data. Among 88 patients included, 43 patients (48.9%) showed low disease activity (DAS28-6m onth ≤3.2). The mean age of all patients was 44 ± 13 years (range, 20-78), and 81.8% were females. Baseline characteristics of patients were not statistically different between the two groups, except for hypertension, which was more prevalent in patients with moderate-to-high disease activity (p = 0.039). The most prevalent comorbidity was hypertension (15.9%) followed by osteoporosis and hyperlipidemia (both 11.3%). Both RF and ACPA were not associated with treatment response. Methotrexate was the most common concomitant drug prescribed (87.5%), followed by hydroxychloroquine (54.5%) and leflunomide (39.7%) ( Table 1). To determine the possible influence of disease status of patients on their response to TNFi, baseline DAS28 and its subcomponents were examined. Baseline DAS28 was significantly lower in the low disease activity group than in the moderate-to-high disease activity group (p = 0.034) ( Table 2). Among the 14 SNPs genotyped, GG-carriers of rs1862513 were more likely to show low disease activity at 6-months of TNFi treatment than GC or CC-carriers (p = 0.020).
Marginal significance was revealed for Rs3397, with higher disease activity in CC-carriers than in patients with the T allele (p = 0.059). Other 12 SNPs did not reach statistical significance in terms of DAS28-6m onth (Table 3). We performed multivariable logistic regression analysis to determine the independent factors including both genetic and non-genetic variables. Hypertension, baseline DAS28, rs1862513, and rs3397 (all p < 0.1 in univariate analysis), sex, age, and body mass index were also included. The model explained 24.5% (Nagelkerke R 2 ) of the variance in DAS28-6m onth and correctly classified 67.0% of the cases. C-allele carriers of rs1862513 were 4.67-times more likely to exhibit moderate-to-high disease activity than GG carriers. CC carriers of rs3397 were 2.66-times more likely to exhibit moderate-to-high disease activity than T-allele carriers (Table 4).  Figure 1).
To assess the predictive accuracy of TNFi response, a ROC curve based on the multivariable model was drawn. The area under the ROC curve for detecting the response to TNFi was 0.77 (95% CI, 0.67-0.87) (Figure 2).

Discussion
This study aimed to investigate the possible association between genetic variation and response to TNFi treatment. Among the candidate gene polymorphisms examined, two SNPs, rs1862513 and rs3397, were associated with the response to TNFi in our RA patients. Worse treatment outcome was seen in patients with the C allele of rs1862513 and CC carriers of rs3397.
Rs1862513 is located in the promoter region of the RETN gene, which has been investigated by many groups. Previous studies of resistin (RETN) dealt with obesity and insulin resistance in diabetes. [11][12][13] Resistin modulates the release and effect of various chemokines and cytokines, and is a key component associated with metabolic and inflammatory diseases. [14] It is estimated that up to 70% of the variation in circulating resistin levels can be explained by genetic factors, although the mechanism and functional implications of this genetic control are still unknown. [15] In the Framingham Offspring Study, rs1862513 was associated with resistin levels but with high heterogeneity across studies. [16] The C allele was associated with higher resistin levels in a metaanalysis that was mainly driven by the Japanese study. [17] We speculate that the C-allele carriers of rs1862513 have higher resistin levels, which could worsen the outcome following TNFi treatment because an increased level of resistin is linked to enhanced inflammatory and disease activity in RA patients. [18] The TNFRSF1B gene encodes tumor necrosis factor receptor superfamily 1B, which is one of the two receptors that TNF-α binds to and further activates NF-κB, triggering inflammatory pathways. [19] Genetic polymorphisms in TNFRSF1B have been studied mostly in patients with Crohn's disease or tuberculosis. Although the outcomes of these studies were different from that of our study, the CT and TT alleles of rs3397 were associated with lower expression of TNFRSF1B and increased susceptibility to Mycobaterium avium subsp paratuberculosis infection in Crohn's disease patients following TNFi treatment. [20] But, in a systematic review of the TNFi response in patients with inflammatory bowel diseases, rs3397 was a nonsignificant SNP, suggesting that the data on its effect on TNFi response are inconclusive. [21] Also, studies on the role of rs3397 in susceptibility to tuberculosis yielded inconsistent results. [22][23][24] In one study in RA patients, an rs3397 variant was associated with a risk of RA but this study failed to provide evidence regarding the role of rs3397 in the response to TNFi. [25] Thus, most studies of rs3397 showed contradictory and inconclusive results across different populations. From our results, we speculate that the C allele triggers a strong inflammatory response via the TNF-α pathway, which might be related to a weaker response to TNFi treatment. This could be a meaningful evidence for future treatments considering rs3397 in relation to TNFi response in Korean RA patients.
Hypertension was significantly associated with TNFi response in univariate analysis. Since the pathophysiology of increased blood pressure is multifactorial, blood pressure control in inflammatory autoimmune disorders is largely related to chronic inflammation and an immune-mediated mechanism. There is a direct association between inflammation and hypertension, although the underlying mechanism remains undetermined . [26,27] The prevalence of hypertension is high in RA patients and its control is worse than in the general population. [28] In a similar manner, our patients with hypertension had a higher risk of worse outcome following TNFi treatment than patients without hypertension.  this estimates the actual local proportion of patients with moderate-to-high disease activity. The shaded region is a 95% confidence interval of the loess smoother.

Supplementary Files
This is a list of supplementary files associated with the primary manuscript. Click to download.