Higher baseline global leukocyte DNA methylation is associated with MTX non-response in early RA patients

Four hundred ninety-six subjects were eligible from the Treatment in the Rotterdam Early Arthritis Cohort (tREACH, ISRCTN26791028), a multicenter, stratified single-blind, randomized controlled trial of eRA patients, as previously described [8]. In brief, included patients were diagnosed with RA based on the American College of Rheumatology (ACR) 1987 classification criteria for RA [9] and were categorized in high, intermediate, or low probability groups for persistent disease, according to the Visser prediction model [10]. All patients who received MTX mono or combination (MTX + corticosteroids and MTX + SSZ + HCQ + corticosteroids) therapy were enrolled in this study (n = 336). An escalating dose of MTX was prescribed in the first 3 weeks from 10 mg (week 1) up to 17.5 mg (week 2) and 25 mg (week 3). Additionally, all patients received weekly 10 mg folic acid, at least 24 h after MTX administration, as recommended [2]. Whole blood leukocytes were collected at baseline (T0) and after 3 months of MTX therapy (T3) and stored at − 80 °C. This study was approved by the medical ethics committee of the Erasmus University Medical Center: MEC-2006-252. Medical ethics committees at each participating center approved the study protocol, and written informed consent was obtained for all patients.

LINE-1 global DNA methylation was determined in DNA from leukocytes, isolated using the Sequenom EpiTYPER® assay (Agena Bioscience™) as previously described [11, 12]. Briefly, 500 ng of purified genomic DNA was treated with sodium bisulfite to distinguish methylated from non-methylation cytosines using the EZ DNA MethylationTM Kit (Zymo Research®) according to the manufacturer’s instructions. Converted DNA concentrations were quantified using a ND-1000 NanoDrop Spectrophotometer (NanoDrop Technologies Inc.) using the RNA-40 default settings, as recommended (Zymo Research®). Bisulfite-converted DNA was stored no longer than 1 month at − 80 °C or until the experiment was performed. A LINE-1 bisulfite-targeted PCR was performed on the C-1000 Touch Thermal Cycler™ (Bio-Rad) using the following primers: 5′aggaagagagGTGTGAGGTGTTAGTGTGTTTTGTT-3′ and 3′cagtaatacgactcactatagggaggaaggctATATCCCACACCTAACTCAAAAAAT-′5. The PCR was followed by Shrimp Alkaline Phosphatase treatment, RNA transcription, and Sequenom analysis as previously described [12]. A mixture consisting of 100% enzymatically methylated DNA and 0% methylated DNA, due to a genetic knockout for methyltransferases, resulted in 50% DNA methylation and was used as a positive control during all steps. Milli-Q water was used as a negative control. DNA methylation was quantified using a Matrix-assisted Laser Desorption/Ionization-Time Of Flight (MALDI-TOF) MassARRAY® (Sequenom) analyzer according to the manufacturer’s instructions. Methylation percentage was calculated using the following formula: % Methylation = (area methylate peak)/(area unmethylated peak + area methylated peak) × 100. All samples were measured in triplet and samples with a variation coefficient (CV) of > 10% were checked for outliers by means of Dixon’s Q test. Outliers were removed, and if CV still exceeded 10% for the remaining duplicate, the sample was excluded. Twelve CpG sites were present within the LINE-1 PCR fragment of which CpG 6.7, CpG 8.9, and CpG 11.12 were combined, since these sites could not be separated. CpG 4 could not be analyzed because of a silent signal, and CpG 10 could not be analyzed due to a low mass fragment. Finally, the following seven CpG sites (CpG1, CpG2, CpG3, CpG5, CpG6.7, CpG8.9, and CpG11.12) were analyzed for differences in methylation [12].

We performed paired analysis to assess the change in global DNA (hydroxy)methylation over the first 3 months using paired-sample t tests. Associations between global DNA (hydroxy)methylation at baseline, at 3 months and over 3 months (Δ(hydroxy)methylation) with response (ΔDAS28-ESR) were first analyzed in a univariate linear regression model, after which the associations were adjusted for confounders. Baseline DAS28 score, baseline erythrocyte folate levels, BMI, age, sex, and smoking status (current versus former + never) are known to be associated with MTX response and DNA methylation and were therefore tested as confounders [13‐16]. The presence of anti-citrullinated protein antibodies (ACPA) has previously been related to decreased MTX response [17]. ACPA positivity was therefore tested as potential covariate. Confounders and covariates were only considered important when the effect size (beta coefficient, B) changed with > 10% upon adjustment. In addition, the relation between baseline global DNA methylation and MTX response was assessed dichotomously (non-responders versus moderate/good responders), according to the EUropean League Against Rheumatism (EULAR) response criteria at 3 months [18]. Associations between global DNA methylation and response were assessed in a crude logistic regression model and in a model adjusted for baseline DAS28, baseline erythrocyte folate, and BMI. Results are expressed in odds ratios (OR) with 95% confidence interval (CI). Incomplete cases were excluded prior to the analysis. For the correlation analysis, distributions of the variables were tested for normality using the Shapiro-Wilkinson test, where p > 0.05 was considered normally distributed. The correlation between baseline erythrocyte folate and global DNA methylation was tested using Spearman’s correlation due to the skewed distribution of erythrocyte folate, and the correlation between global DNA methylation determined by LC-ESI-MS/MS and LINE-1 was tested using Pearson’s correlation test (normally distributed variables). All statistical analyses were conducted using R Studio Software (Version 1.1.423; RStudio Team 2015), and p values< 0.05 were considered significant. Models tested for both methylation and hydroxymethylation were corrected for multiple comparisons using the Bonferroni correction, where p < 0.025 (0.05/2 = 0.025) was considered significant. LINE-1 analysis was corrected using the Bonferroni correction for the 7 CpGs that were tested simultaneously; hence, p < 0.007 (0.05/7 = 0.007) was considered significant.

Genomic DNA was available and isolated from leukocytes of 265 treatment-naive early RA patients and from 275 subjects at T3. A minimum of 600 ng was required for reliable (hydroxy)methylation measurements. Nine (T0) and five (T3) extracted DNA samples did not reach up to this minimum and were therefore excluded. Global DNA (hydroxy)methylation was successfully quantified in 294 patients, comprising 249 (T0) and 257 (T3) samples. Baseline characteristics of these 294 subjects are summarized in Table 1. The mean age was 53.4 ± 14.2 years, and 70.4% was female. Mean DAS28 at baseline was 4.7 ± 1.2 and decreased to 3.0 ± 1.2 over the first 3 months (Table 1). All patients were treatment naive at baseline and received MTX mono- or combination therapy for at least 3 months (Table 1).

Mean global DNA methylation at baseline was 4.41 ± 0.13% and did not change significantly over the first 3 months of therapy (p = 0.454) (Additional file 1: Table S1). Global DNA hydroxymethylation increased significantly with 0.0008% over the first 3 months (p = 0.013; Additional file 1: Table S1).

Baseline global DNA methylation was associated with ΔDAS28 over the first 3 months when assessed in a crude univariate linear regression model (B = 1.36, p = 0.044). One percent difference in global DNA methylation at baseline corresponds to 1.41 difference in ΔDAS28 between patients, when adjusted for baseline DAS28, baseline erythrocyte folate, and BMI (B = 1.41, p = 0.013; Table 2).

In addition, we stratified subjects accordingly: non- and moderate/good responders according to the EULAR response criteria at 3 months.

Higher baseline global DNA methylation was associated with EULAR non-response in both a crude logistic model (OR = 0.027, 95% CI = 0.002–0.377) and when adjusted for baseline DAS28, baseline erythrocyte folate, and BMI (OR = 0.010, 95% CI = 0.001–0.188; Fig. 1).

Baseline global DNA hydroxymethylation was not significantly associated with ΔDAS28 in a crude univariate model (B = 19.56, p = 0.288), nor when adjusted for baseline DAS28, baseline erythrocyte folate, BMI, age, and sex (B = 6.90, p = 0.664; Table 2) and was therefore not further assessed between non-responders and moderate/good responders.

As folate is related to DNA methylation through one-carbon metabolism, we examined the correlation between baseline global DNA methylation and erythrocyte folate. We did not observe a correlation between baseline global DNA methylation and baseline erythrocyte folate concentrations (R = 0.084, p = 0.24).

Global DNA methylation at 3 months of therapy was not associated with ΔDAS28 (B = 0.40, p = 0.471), nor was global DNA hydroxymethylation at 3 months (B = 12.52, p = 0.517; Table 2). In addition, differences between DNA (hydroxy)methylation at baseline and after 3 months of therapy were not associated with changes in DAS28 (Δmethylation B = − 0.68, p = 0.182, Δhydroxymethylation B = − 1.55, p = 0.925; Table 3).

LINE-1 global DNA methylation was determined in DNA isolated from leukocytes of 120 patients and was successfully quantified in 104 subjects. Seventy-eight individuals had no missing data in any of the variables needed in the analysis and were therefore used. LINE-1 methylation in CpG2 was not significantly associated with ΔDAS28 in a crude univariate model (B = 0.09, p = 0.242). However, it was associated with ΔDAS28 when adjusted for baseline DAS28, baseline erythrocyte folate, BMI, and smoking status (B = 0.16, p = 0.026; Table 4). Methylation at the other 6 CpG sites within LINE-1 was not associated with DAS28 nor was mean LINE-1 methylation (Additional file 1: Table S2).

Furthermore, we examined the correlation between global DNA methylation obtained with LC-ESI-MS/MS and global DNA methylation in CpG2 assessed with the LINE-1 method. Here, we found a significantly positive correlation, although not strong (R = 0.34, p = 0.00061; Additional file 1: Fig. S1).

To assess whether the association between baseline global DNA methylation and disease activity was specific for MTX response and not due to combination therapy, subjects were stratified by therapy and univariate linear regression analyses were performed. The effect size was 1.8-fold higher in the MTX monotherapy group (B = 2.06, p = 0.074) compared to the triple therapy group (B = 1.12, p = 0.173), although the associations were not significant (Table 5).