Methotrexate Normalizes Up-Regulated Folate Pathway Genes in Rheumatoid Arthritis

Authors


Abstract

Objective

The folate antagonist methotrexate (MTX) is an anchor drug in the treatment of rheumatoid arthritis (RA), but its mechanism of action with regard to the impact on folate metabolism remains elusive. The aim of the present study was to investigate the cellular pharmacologic impact of MTX on peripheral blood cells, by comparing MTX-treated RA patients to MTX-naive RA patients and healthy controls.

Methods

Gene expression microarray data were used to investigate the expression of 17 folate pathway genes by peripheral blood cells from a cohort of 25 RA patients treated with MTX, 10 MTX-naive RA patients starting treatment with MTX, and 15 healthy controls (test cohort). Multiplex real-time polymerase chain reaction was used to validate the results in an independent cohort, consisting of 151 RA patients treated with MTX, 28 MTX-naive RA patients starting treatment with MTX, and 24 healthy controls (validation cohort).

Results

Multiple folate metabolism–related genes were consistently and significantly altered between the 3 groups in both cohorts. Concurrent with evidence of an immune-activation gene signature in MTX-naive RA patients, significant up-regulation of the folate-metabolizing enzymes γ-glutamyl hydrolase and dihydrofolate reductase, as well as the MTX/folate efflux transporters ABCC2 and ABCC5, was observed in the MTX-naive RA group compared to healthy controls. Strikingly, MTX treatment of RA patients normalized these differential gene expression levels to the levels observed in healthy controls.

Conclusion

These results suggest that under inflammatory conditions, basal folate metabolism in the peripheral blood cells of RA patients is markedly up-regulated, and treatment with MTX restores folate metabolism to normal levels. Identification of this novel gene signature provides insight into the mechanism of action of MTX, thus paving the way for development of novel folate metabolism–targeted therapies.

Rheumatoid arthritis (RA) is a systemic autoimmune disease that is characterized by chronic inflammation of the joints. RA is heterogeneous in terms of its clinical presentation, pathologic processes, and response to therapy ([1, 2]). The folic acid antagonist methotrexate (MTX) is by far the most widely used disease-modifying antirheumatic drug (DMARD) in the treatment of RA, either as a single agent or in combination with other DMARDs. Failure to respond to DMARD-based therapy may prompt the use of biologic agents, such as anti–tumor necrosis factor α monoclonal antibodies, in combination with MTX ([3, 4]).

Beyond its application in the treatment of RA, MTX is also commonly prescribed in the treatment of malignant diseases ([5, 6]). The antiinflammatory effects elicited by low-dose MTX treatment of RA may involve mechanisms that are distinct from those observed with high-dose MTX treatment of various cancers. However, with respect to the cellular pharmacologic effects of MTX, there are overlapping mechanisms of action in RA and cancer. MTX suppresses proliferation of malignant cells by blocking the de novo biosynthesis of purines and pyrimidines ([5]). Moreover, MTX inhibits the homocysteine and methionine cycle involved in methylation of DNA, RNA, and proteins. MTX-induced inhibition of 5-aminoimidazole-4-carboxamide ribonucleotide formyltransferase (AICARTF/ATIC) promotes release of the endogenous antiinflammatory mediator adenosine ([5, 7, 8]). MTX has also been reported to achieve additional pharmacologic effects, e.g., generation of reactive oxygen species through JNK activation ([9, 10]), induction of T cell apoptosis ([11]), and inhibition of NF-κB ([12]), for which the contribution of its therapeutic efficacy is not fully defined.

As shown in Figure 1, treatment with MTX primarily interferes with folate homeostasis, a process that is tightly controlled at multiple levels, including the following: 1) routes of cellular uptake and efflux of MTX/folate; 2) intracellular metabolism and retention of MTX/folate; and 3) expression of key regulatory folate-dependent enzymes (for review, see refs.[5] and[13]). Oral MTX is taken up in the upper small intestine via the proton-coupled folate transporter (PCFT/SLC46A1), which functions optimally at the acidic pH of the duodenum and jejunum, both of which show predominant expression of PCFT ([5, 14]). Cellular uptake of MTX by immune cells proceeds via the reduced folate carrier (RFC/SLC19A1), although a third transport route, involving a receptor-mediated process via folate receptor β (FRβ), which is expressed predominantly in macrophages, can be operative in activated macrophages in inflamed synovial tissue ([15, 16]). Upon entry into the cell, MTX is metabolically activated by the enzyme folylpolyglutamate synthetase (FPGS) to a polyanionic polyglutamate form that is highly retained within cells ([17]). For deactivation, polyglutamylation can be reversed by γ-glutamyl hydrolase (GGH), which cleaves off the polyglutamyl chain attached to folates and polyglutamatable antifolates such as MTX ([18]).

Figure 1.

Key components of methotrexate (MTX)/folate metabolism, involved in the uptake and extrusion, metabolism, and targeting of MTX/folate. The influx transporters comprise reduced folate carrier (RFC), proton-coupled folate transporter (PCFT), and folate receptor α and β (FRα/β) isoforms. The metabolizing enzymes comprise folylpolyglutamate synthetase (FPGS) and γ-glutamyl hydrolase (GGH) (compartmentalized in lysosomes). The folate-dependent enzymes involved in folate cycling include dihydrofolate reductase (DHFR), thymidylate synthase (TYMS), 5-aminoimidazole-4-carboxamide ribonucleotide formyltransferase (AICARTF/ATIC), glycinamide ribonucleotide transformylase (GART), methylenetetrahydrofolate reductase (MTHFR), and methionine synthase (MS). The MTX efflux transporters, members of the ATP-binding cassette (ABC) transporters, include ABCC1–5 and ABCG2. Inhibition of folate-dependent enzymes by MTX is depicted with the long broken lines. Proteins and enzymes are shown in boldface. PG = polyglutamate; OP = organic phosphate; H+ = hydrogen; Pi = phosphates; DHF = dihydrofolate; THF = tetrahydrofolate; CH3-THF = 5-methyl-tetrahydrofolate; CH2-THF = 5,10-methylene-tetrahydrofolate; CHO-THF = 10-formyl-tetrahydrofolate. Adapted, with permission, from refs.[5] and[13].

MTX blocks the key enzyme dihydrofolate reductase (DHFR), whereas MTX polyglutamates (MTXGlu) potently inhibit thymidylate synthase (TYMS), both of which are enzymes that are crucial for the de novo biosynthesis of purines and pyrimidines required for DNA replication and cellular proliferation. MTXGlu and dihydrofolate polyglutamates that accumulate after DHFR inhibition can also exert an inhibitory effect on other folate-dependent enzymes downstream of DHFR, such as glycinamide ribonucleotide formyltransferase (GART) and methylenetetrahydrofolate reductase (MTHFR) ([19, 20]), which also function in de novo purine biosynthesis and methylation of DNA, RNA, and proteins, respectively. Finally, folate/MTX can be extruded from cells, mainly as the nonpolyglutamated form, via ATP-binding cassette (ABC) transporters, including ABCC1 and ABCG2 ([5, 21]).

Despite the key role of MTX in current RA therapies, its mechanism of action remains elusive, and a recurrent theme is the lack of a clinical response and the risk of toxicity in a considerable fraction of patients. Many studies have investigated genetic variants of folate pathway genes and the levels of MTXGlu in red blood cells (RBCs) as predictive markers for therapy response or toxicity, and these markers could be used to decipher the molecular basis underlying the loss of MTX efficacy in RA ([22-26]).

Although these studies offer promising avenues toward the discovery of markers that could predict and explain the mechanisms underlying loss of MTX function, the mode of action of MTX in RA remains unclear. As a step toward this end and in order to better understand the impact of MTX treatment on RA, we compared expression levels of multiple MTX/folate pathway genes in RA patients prior to and during treatment with MTX. Furthermore, expression profiling in these 2 RA groups was compared with that in healthy controls.

PATIENTS AND METHODS

Patients and controls

This study consisted of 2 independently collected cohorts, a test cohort and a validation cohort. The test cohort included 25 RA patients who were receiving treatment with MTX, 10 RA patients who had not been treated with MTX or other DMARDs and were about to start MTX treatment (MTX-naive), and 15 healthy controls, as previously reported ([27]). The validation cohort consisted of 151 RA patients who were receiving treatment with MTX, 28 MTX-naive RA patients who were about to start MTX treatment, and 24 healthy controls. The characteristics of the patients and controls in both cohorts are summarized in Table 1. All patients were diagnosed as having RA according to the American College of Rheumatology 1987 revised criteria for RA ([28]), and each subject was randomly selected for the study. All RA patients and healthy individuals gave their informed consent to participate, and the study protocol was approved by medical ethics committees of the Academic Medical Center, the VU University Medical Center, and the Jan van Breemen Research Institute|Reade (Amsterdam, The Netherlands).

Table 1. Characteristics of the patients and healthy control subjects in the test cohort and validation cohort*
 Test cohortValidation cohort
Healthy controls (n = 15)MTX-naive RA patients (n = 10)RA patients receiving MTX treatment (n = 25)Healthy controls (n = 24)MTX-naive RA patients (n = 28)RA patients receiving MTX treatment (n = 151)
  1. Except where indicated otherwise, values are the mean (range). RA = rheumatoid arthritis; CRP = C-reactive protein; ND = not determined; ESR = erythrocyte sedimentation rate; RF = rheumatoid factor; ACPAs = anti–citrullinated protein antibodies; NA = not applicable; DAS28 = Disease Activity Score in 28 joints.
  2. aMissing in 2 patients receiving methotrexate (MTX) treatment.
  3. bMissing in 1 patient without MTX treatment.
  4. cMissing in 39 patients receiving MTX treatment.
Age, years43 (27–63)49 (25–67)49 (23–63)35 (23–63)52 (21–83)55 (29–79)
Sex, % female608068506175
Disease characteristic      
CRP, mg/literNDND19 (3–76)ND18 (1–57)17 (0–131)
ESR, mm/hourND41 (13–70)27 (2–82)ND28 (5–82)24 (2–82)
Positive for RF, no. (%)ND5 (50)10 (43)aND13 (48)b107 (71)
Positive for ACPAs, no. (%)ND9 (90)20 (86)aND17 (63)b82 (73)c
Disease duration, monthsNA7 (1–12)113 (8–417)NA6 (1–18)118 (0–516)
Erosions, no. (%)NA1 (10)24 (96)NA5 (18)108 (72)
DAS28NA5.1 (3.1–7.4)5.5 (3.4–7.2)NA5.3 (3.4–7.6)5.1 (1.8–8.3)
Medication      
Dose of MTX, mg/weekNA021 (7.5–30)NA021 (5–30)
Prednisone at <20 mg/day, no. (%)NA06 (24)NA042 (28)

Blood sampling for RNA isolation

Samples of peripheral blood (2.5 ml) for RNA isolation were obtained from all patients and healthy controls and directly collected in PAXgene Blood RNA isolation tubes (PreAnalytiX) to avoid manipulation and activation of the cells ([29]). Incubation at room temperature for 2 hours on a roller bank ensured complete cell lysis and RNA stabilization, after which tubes were stored at −20°C. Total RNA was isolated using a PAXgene RNA isolation kit according to the manufacturer's instructions, including a DNase treatment (Qiagen) to remove genomic DNA. The quantity and quality of the RNA were determined using a NanoDrop Technologies spectrophotometer.

Gene expression microarray data

For gene expression analysis ([27]) of peripheral blood samples from the test cohort, data were retrieved from the Stanford Microarray Database ([30]) (http://smd.stanford.edu/cgi-bin/search/basicSearch.pl), and expression levels of specific genes were extracted and analyzed. A subanalysis in the test cohort involved microarray expression analysis of genes in the MTX/folate pathway ([13]), in particular those involved in cellular uptake of MTX/folate (FR, PCFT, and RFC) and efflux of MTX/folate (ABCC1–ABCC5 and ABCG2), the metabolism of MTX/folate (FPGS and GGH), and intracellular targeting of MTX (DHFR, TYMS, MTHFR, ATIC, and GART).

Complementary DNA (cDNA) synthesis and real-time polymerase chain reaction (PCR) analysis

RNA (0.25 μg) was reverse-transcribed into cDNA with a RevertAid H Minus First Strand cDNA Synthesis Kit (MBI Fermentas) according to the manufacturer's instructions. The BioMark 96.96 dynamic array (Fluidigm) used for gene expression analysis was also used for real-time PCR analysis of 17 folate/MTX pathway genes, performed by an outsourcing company (ServiceXS). Before using the BioMark array, Specific Target Amplification of the cDNA (14 cycles) was performed. The reaction was diluted 5-fold before being loaded onto the BioMark array. PCR and imaging were performed on a BioMark instrument, after which Ct values were extracted using BioMark Real-Time PCR analysis software. Results of the TaqMan gene expression assays (Applied Biosystems) for the 17 selected folate-related genes are shown in Supplementary Table 1 (available on the Arthritis & Rheumatism web site at http://onlinelibrary.wiley.com/doi/10.1002/art.38094/abstract). Expression levels were log2 transformed, and data are presented as the median, relative to the values for GAPDH.

Statistical analysis

Following filtering of the array data and normalization of the data from the Stanford Microarray Database, 2 different data analysis approaches were used. The first, an unbiased approach, included genome-wide expression analysis using Significance Analysis of Microarrays (SAM) ([31]). Differentially expressed genes between patient groups and/or healthy controls were considered to be significant at q values of less than 5%. Subsequently, for interpretation of pathways represented by those genes, we applied gene ontology analyses using PANTHER (details available at http://pantherdb.org). PANTHER compares a gene list of interest with a reference list (NCBI list of Homo sapiens genes) to identify significantly overrepresented groups of genes (those with P values less than 0.05 after Bonferroni correction for multiple comparisons).

The second, a biased approach, included subanalysis of single MTX/folate pathway genes from the microarray data (test cohort) and real-time PCR data (validation cohort). These data were analyzed using GraphPad Prism 5.01 software. Statistical analyses for 2-group comparisons were performed using Student's t-test or Mann-Whitney U test, where appropriate. The Bonferroni-Holm method was used to correct for multiple testing. P values less than 0.05 were considered significant.

RESULTS

Gene expression profiling in MTX-naive and MTX-treated RA patients in the test cohort

As an initial, unbiased exploration of the impact of MTX treatment, gene expression profiles in a test cohort of 10 MTX-naive RA patients, 25 RA patients treated with MTX, and 15 healthy controls were analyzed ([27]). We first compared expression profiles between MTX-treated and MTX-naive RA patients to examine whether MTX induced specific alterations in the expression levels of individual or related genes. SAM analysis of the peripheral blood of MTX-treated and MTX-naive RA patients revealed 4,311 differentially expressed genes, of which 1,968 genes were up-regulated and 2,343 genes were down-regulated in the MTX-treated group compared to the MTX-naive group.

PANTHER analysis of the differentially expressed genes identified several pathways and biologic processes that were associated with the up- or down-regulated genes in the MTX-treated group compared to the MTX-naive group of RA patients (Table 2), e.g., an up-regulated cysteine biosynthesis pathway (P = 0.0432), which is associated with folate metabolism, and a down-regulated folate biosynthesis pathway (P = 0.00756).

Table 2. PANTHER analysis of significantly differentially expressed genes in pathways in the methotrexate (MTX)–treated group compared to the MTX-naive group of rheumatoid arthritis patients*
PathwayP
  1. FGF = fibroblast growth factor; PI3K = phosphatidylinositol 3-kinase; EGF = epidermal growth factor; IGF = insulin-like growth factor; PDGF = platelet-derived growth factor; TCA = tricarboxylic acid; VEGF = vascular endothelial growth factor.
Up-regulated genes in MTX-treated vs. MTX-naive 
Ras pathway6.99 × 10−5
Unclassified9.62 × 10−4
FGF signaling pathway4.16 × 10−3
Parkinson's disease5.69 × 10−3
Ornithine degradation8.03 × 10−3
Apoptosis signaling pathway9.80 × 10−3
Angiogenesis1.25 × 10−2
PI3K pathway1.49 × 10−2
Alzheimer's disease–amyloid secretase pathway1.49 × 10−2
EGF receptor signaling pathway1.88 × 10−2
Insulin/IGF pathway–MAP kinase kinase/ MAP kinase cascade2.07 × 10−2
Toll receptor signaling pathway2.17 × 10−2
PDGF signaling pathway2.70 × 10−2
Xanthine and guanine salvage pathway2.94 × 10−2
p53 pathway3.10 × 10−2
B cell activation3.13 × 10−2
Huntington's disease3.75 × 10−2
TCA cycle4.05 × 10−2
Cysteine biosynthesis4.32 × 10−2
Alzheimer's disease–presenilin pathway4.75 × 10−2
Down-regulated genes in MTX-treated vs. MTX-naive 
Unclassified7.03 × 10−4
Inflammation mediated by chemokine and cytokine signaling pathway1.42 × 10−3
p53 pathway by glucose deprivation1.51 × 10−3
Integrin signaling pathway5.14 × 10−3
Folate biosynthesis7.56 × 10−3
Acetate utilization1.70 × 10−2
VEGF signaling pathway2.86 × 10−2
Formyltetrahydroformate biosynthesis3.67 × 10−2

Gene expression profiling of folate pathway–related genes in MTX-naive and MTX-treated RA patients in the test and validation cohorts

The results obtained from the PANTHER analysis encouraged us to pursue further in-depth analysis of the genes that regulate cellular folate metabolism (Figures 1-3). As an initial evaluation in the test cohort, data on expression levels of individual folate pathway–related genes were selected from the microarray data set, and these genes were divided into 4 groups based on their known function: folate influx systems (FRα and FRβ), folate-metabolizing enzymes (FPGS and GGH), folate-dependent enzymes (DHFR, GART, and TYMS), and folate efflux transporters (ABCC1–5 and ABCG2) (Figure 1). The effect of MTX treatment on folate metabolism (comparing MTX-treated RA patients with MTX-naive RA patients) was studied both in the test cohort and in an independent validation cohort. Detailed analysis showed that expression levels of these folate pathway genes were, overall, lower in the MTX-treated group compared to the MTX-naive group (Figures 2A–D), as had been anticipated from the pathway analysis.

Figure 2.

Gene expression analysis of the test cohort. Gene expression of folate/MTX influx transporters (A), metabolizing enzymes (B), folate-dependent enzymes (C), and efflux transporters (D) was assessed in the MTX-naive group (MTX−) (n = 10) and MTX-treated group (MTX+) (n = 25) of rheumatoid arthritis patients compared to healthy controls (HCs) (n = 15). The original microarray template did not include probes for RFC, PCFT, MTHFR, and ATIC, and therefore these genes are not presented. Results are the log2-transformed relative expression data (RQ), relative to the values for GAPDH. Data are presented in box plots, where the boxes represent the 25th to 75th percentiles, the lines within the boxes represent the median, and the lines outside the boxes represent the 10th and 90th percentiles. ∗ = P < 0.05; ∗∗ = P < 0.01; ∗∗∗ = P < 0.001. See Figure 1 for other definitions.

Figure 3.

Gene expression analysis of the validation cohort. Gene expression of the folate/MTX influx transporters (A), metabolizing enzymes (B), folate-dependent enzymes (C), and efflux transporters (D) was assessed in the MTX-naive group (MTX−) (n = 28) and the MTX-treated group (MTX+) (n = 151) of rheumatoid arthritis patients compared to healthy controls (HCs) (n = 24). Results are the log2-transformed relative expression data (RQ), relative to the values for GAPDH. Data are presented in box plots, where the boxes represent the 25th to 75th percentiles, the lines within the boxes represent the median, and the lines outside the boxes represent the 10th and 90th percentiles. ∗ = P < 0.05; ∗∗ = P < 0.01; ∗∗∗ = P < 0.001, after Bonferroni-Holm correction for multiple testing. See Figure 1 for other definitions.

Thereafter, we studied these genes individually, supplemented with genes encoding the influx transporters RFC1 and PCFT and genes encoding the folate-dependent enzymes MTHFR and ATIC (not represented on the microarray), in an independent validation cohort using multiplex real-time PCR. These results revealed a pattern similar to that in the test cohort, even after stringent Bonferroni-Holm correction for multiple testing. This analysis in the validation cohort confirmed that levels of the multidrug efflux transporter genes ABCC1, ABCC3, and ABCC4 were significantly decreased in the MTX-treated group compared to the MTX-naive group (P = 0.0369, P = 0.0022, and P = 0.0004, respectively) (Figure 3D). In addition, the expression of GGH, which catalyzes the hydrolysis of folate/MTXGlu tails, was significantly lower in the MTX-treated group compared to that in the MTX-naive group (P = 0.0004) (Figure 3B). A similar profile was observed for the folate-dependent enzymes DHFR, TYMS, and ATIC (P = 0.0024, P < 0.0001, and P = 0.0040, respectively) (Figure 3C). Furthermore, gene expression levels of the folate influx receptor FRβ were significantly lower in the MTX-treated group compared to the MTX-naive group (P = 0.0124), but significance was lost after Bonferroni-Holm correction (Figure 3A).

No significant differences were observed in the expression levels of various folate influx systems (RFC1 and PCFT), the metabolizing enzyme FPGS, the folate-dependent enzyme MTHFR, and the efflux transporter ABCG2 between the 2 groups of RA patients (Figures 3A–D). One notable difference was observed in the expression levels of GART, which were significantly increased in the MTX-treated group compared to the MTX-naive group in the test cohort (P < 0.0001) (Figure 2C), but were significantly decreased in the MTX-treated group compared to the MTX-naive group in the validation cohort (P < 0.0001) (Figure 3C). This apparent discrepancy was explained by the fact that only one GART isoform was detected on the microarray, whereas both isoforms were captured by multiplex real-time PCR. Overall, the differential expression of folate-related genes showed that MTX is capable of suppressing the expression of these genes.

Folate/MTX pathway gene expression profiling in MTX-naive RA patients and healthy controls in the test and validation cohorts

Next, folate/MTX pathway gene expression profiling studies were extended to the MTX-naive RA patients and healthy controls. First, SAM analysis performed on microarray data from the test cohort revealed 3,150 genes that were differentially expressed between the MTX-naive group and healthy controls, of which 1,553 were up-regulated and 1,597 were down-regulated. PANTHER analysis of these differentially expressed genes revealed that several biologic processes involved in immune responses were significantly up-regulated in the MTX-naive group compared to healthy controls; these included processes such as the immune system response, response to stimuli, cellular process, (primary) metabolic process, protein metabolic process, angiogenesis, signal transduction, and cell surface receptor–linked signal transduction (Table 3). This profile, together with the immune responses that were previously observed by our group ([27]), point toward the possibility that an immune-activation gene signature is acquired in MTX-naive RA patients.

Table 3. PANTHER analysis of significantly differentially expressed genes in biologic processes in methotrexate (MTX)–naive rheumatoid arthritis patients compared to healthy controls
Biologic processP
Up-regulated genes in MTX-naive vs. healthy controls 
Unclassified2.48 × 10−11
Immune system process6.73 × 10−6
Response to stimulus1.00 × 10−5
Metabolic process8.41 × 10−5
Cellular process4.89 × 10−4
Mesoderm development6.21 × 10−4
Primary metabolic process9.55 × 10−4
Immune response1.33 × 10−3
Developmental process2.85 × 10−3
Protein metabolic process4.09 × 10−3
Skeletal system development2.29 × 10−2
Angiogenesis3.03 × 10−2
Signal transduction3.33 × 10−2
Cell surface receptor–linked signal transduction3.75 × 10−2
Down-regulated genes in MTX-naive vs. healthy controls 
Metabolic process3.57 × 10−8
Primary metabolic process5.64 × 10−8
Protein metabolic process2.94 × 10−6
Translation2.09 × 10−3
Unclassified1.69 × 10−2

Moreover, in the test cohort, the overall picture emerged that folate pathway–related gene expression levels within the MTX-naive group were elevated compared to those in the healthy controls (Figures 2A–D). A similar pattern was observed in the validation cohort, even after stringent Bonferroni-Holm correction for multiple testing (Figures 3A–D).

These analyses showed that gene expression levels of the folate-metabolizing enzyme GGH (P < 0.0001), folate-dependent enzymes DHFR (P < 0.0001) and GART (P = 0.006), and efflux transporters ABCC2 and ABCC5 (both P < 0.0001) were consistently and significantly higher in the MTX-naive group compared to healthy controls (Figures 3B–D). Gene expression levels of the folate-dependent enzyme ATIC and efflux transporters ABCC3 and ABCC4 were also significantly higher in the MTX-naive group compared to healthy controls (P = 0.0204, P = 0.0179, and P = 0.0176, respectively), but not after Bonferroni-Holm correction (Figures 3C and D). Gene expression analysis of the influx transporters (RFC, PCFT, and FRβ) and metabolizing enzyme FPGS revealed no significant differences between the MTX-naive group and healthy controls (Figures 3A and B). For FPGS, this differed from the findings observed in the test cohort, in which the MTX-naive RA patients had FPGS expression levels that were significantly higher than those in healthy controls (Figure 2B); however, this could be accounted for by different isoforms being measured in the microarray as compared to the multiplex real-time PCR.

Collectively, the differential expression of critical folate/MTX pathway–related genes was consistent with the elevated expression of an immune-activation gene signature. These findings suggest that immune system activation in RA patients may account for the differences in expression levels in folate/MTX pathway genes between the MTX-naive group and healthy controls.

Folate/MTX pathway gene expression profiling in MTX-treated RA patients and healthy controls in the validation cohort

Although the up-regulated expression of folate/MTX pathway genes in the MTX-treated group appeared to be dampened by the pharmacologic action of MTX, we examined to what extent the levels were restored to the levels in healthy controls. With MTX treatment, the levels of FRβ, GART, ATIC, ABCC1, ABCC3, and ABCC4 were restored to the values in healthy controls (Figure 3), whereas expression levels of GGH (P < 0.001), DHFR (P = 0.0026), ABCC2 (P < 0.001), and ABCC5 (P < 0.001) remained significantly higher in the MTX-treated group compared to healthy controls; however, these elevated levels did not exceed the expression levels in the MTX-naive group.

Finally, both cohorts were tested for possible confounding factors that could have influenced the results. No significant differences in positivity for anti–citrullinated protein antibodies, C-reactive protein levels, and the erythrocyte sedimentation rate were found between the groups, and therefore these could be excluded as confounding factors. We also analyzed age, sex, rheumatoid factor (RF) titer, prednisone use, disease duration, and presence of erosions; only disease duration and presence of bone erosions were potential confounding factors. In the validation cohort, a clear difference in disease duration was observed between the groups, in that the disease duration was <19 months in the MTX-naive patients compared to a maximum of 516 months in the MTX-treated patients. We therefore compared patients with a disease duration of <19 months in both groups. With regard to erosions, we compared patients without erosions in both groups.

Reanalyses including these 2 factors did not change the results, thus establishing that the gene expression patterns in the validation cohort (depicted in Figure 3) were independent of these possible confounding factors. Confounding by indication was largely overcome, because patients in the MTX-naive group were facing the start of treatment and were at the same stage of disease as that in patients in the MTX-treated group at the start of treatment. Moreover, a comparison between patients undergoing short-term treatment with MTX and those who had been treated with MTX over the long term revealed no differences.

Finally, we examined whether alterations in abundance of specific immune cell subsets (CD14+ monocytes, CD20+ B cells, and CD3+ T cells) ([32]) could account for the alterations in the gene expression profiles between the MTX-naive group of RA patients and healthy controls. Monocyte, T cell, and B cell counts were, on average, increased 1.37-fold, 0.83-fold, and 0.9-fold, respectively, in the MTX-naive group compared to healthy controls, but the percentage of monocytes as a fraction of total leukocytes was only marginally higher in the MTX-naive group (mean ± SD 6.9 ± 1.4% versus 6.1 ± 1.6% in healthy controls) (see Supplementary Table 2, available on the Arthritis & Rheumatism web site at http://onlinelibrary.wiley.com/doi/10.1002/art.38094/abstract), whereas the fold change differences in expression of folate-related genes ranged up to 3.6-fold.

We also analyzed the correlation of expression levels of immune cell subset–specific genes and expression of the most significantly differentially expressed folate-related genes (i.e., GGH, DHFR, and ABCC3) between the MTX-naive and MTX-treated groups of RA patients and healthy controls. Consistently, no correlation between the groups was found in the levels of monocyte gene expression and the differentially expressed folate-related genes. With regard to expression of other immune cell subset–specific transcript markers (CD3 and CD20), a random distribution of some correlations between the groups was observed (results shown in Supplementary Table 2, available on the Arthritis & Rheumatism web site at http://onlinelibrary.wiley.com/doi/10.1002/art.38094/abstract). Taken together, these results underscore the notion that changes in immune cell subsets are not associated with the expression profiles of folate-related genes.

Collectively, in these 2 independent cohorts, a consistent profile of increased expression of folate/MTX pathway genes was observed in MTX-naive RA patients, and these levels exceeded those in healthy controls. MTX treatment led to stabilization and suppression of the gene expression levels in RA patients, when compared to the levels in RA patients who had not received treatment with MTX.

DISCUSSION

Microarray techniques have been successfully applied to identify gene expression signatures and relevant pathways involved in the pathophysiology of RA ([27, 33]). In the current study, we quantified expression levels of folate/MTX pathway–related genes in the peripheral blood of RA patients (prior to and during MTX treatment) and healthy control subjects in order to address 2 questions: 1) is the expression of these genes altered during chronic inflammation?; and 2) what impact does MTX treatment have on the expression of these genes? Since these expression profiles included the majority of folate pathway–related genes, it could generate an integrated picture of dynamic alterations in folate metabolism as a function of chronic inflammation and/or MTX treatment.

An important finding of the current study was that in blood cells of MTX-naive RA patients, expression levels of multiple folate/MTX pathway–related genes were significantly increased, whereas in MTX-treated patients, these expression levels were either retained or largely suppressed to the levels observed in healthy controls. These findings were unrelated to the many possible confounding factors tested, including age, sex, co-medication, confounding by indication, RF seropositivity, disease duration, presence of bone erosions, and alterations in immune cell–type abundance, although an influence of residual confounding factors cannot be ruled out.

The finding of up-regulated expression of folate/MTX pathway–related genes under inflammatory conditions is compatible with evidence of an immune-activation program that did emerge from PANTHER pathway analysis of the microarray data. In this context, immune activation, with accompanying proliferation and lymphopoiesis, requires (folate-dependent) purine and pyrimidine de novo biosynthesis for increased nucleotide generation, occurring concomitant with up-regulated expression of the folate-dependent enzymes DHFR, GART, and ATIC. MTX could harbor the capacity to offset the imbalance in homeostasis caused by the immune activation in this chronic inflammatory disease. Consistently, part of the therapeutic effects of MTX may stem from suppression of proliferation and induction of apoptosis in activated immune cells ([11]).

In accordance with the mode of action of MTX, pathway analysis revealed down-regulation of folate biosynthesis and up-regulation of cysteine biosynthesis in the MTX-treated patients compared to the MTX-naive patients. The latter may be explained by the observation that MTX diminishes the intracellular folate pool, which gives rise to accumulation of potentially toxic levels of homocysteine. By converting homocysteine to cysteine, the cell has an escape route to prevent induction of cell death ([34]).

Although pathway analysis involves multiple levels of complexity, and the results should be interpreted with care, we and other investigators have successfully applied pathway analysis tools, such as PANTHER analysis, to reveal biologically relevant processes, characterized by differentially expressed gene signatures ([27, 35]). Herein, we applied PANTHER analysis to select folate metabolism as a regulated process, which was further validated and extended by quantitative PCR analysis of the originally identified genes and additional biologically related genes in an independent cohort.

Evaluation of alterations in expression of functionally related genes may provide additional information about the dynamic features of folate metabolism and homeostasis. When we examined the group of folate/MTX influx transporters, we found that expression of RFC and PCFT was largely unchanged, and therefore the functional capacity of these transporters may not be a limiting factor in MTX uptake. FRβ is expressed particularly on activated macrophages, and although FRβ has a relatively low affinity for MTX ([15, 16]), its reduced expression in MTX-treated patients compared to MTX-naive patients may be a reflection of the MTX-induced suppression of macrophage activation.

Initially, one may have anticipated that expression of FPGS, which has been implicated in the activation of MTX to polyglutamate forms, could have played a critical role in folate/MTX homeostasis in the blood cells of RA patients. However, no significant differences in expression were found between the groups. Nevertheless, it should be mentioned that FPGS messenger RNA (mRNA) expression levels may not necessarily be indicative of FPGS catalytic activity, as FPGS pre-mRNA is subject to aberrant splicing, particularly under conditions of antifolate pressure ([17]). Remarkably, a prominent alteration observed was the increased GGH expression in MTX-naive RA patients as compared to that in healthy controls. In response to MTX treatment, GGH expression was suppressed to levels that were in the intermediate range between those in MTX-naive patients and those in healthy controls. Conceivably, GGH may serve as a potential marker of the onset of inflammatory processes and could be used to demonstrate a successful antiinflammatory response upon MTX treatment. In this context, GGH may regulate the increased need for folates by regulating cleavage of intracellularly stored folylpolyglutamate pools ([5, 18, 36]). MTX may suppress this need and facilitate restoration of cellular folate homeostasis. Given the notion that GGH is compartmentalized in the lysosome, and folylpolyglutamates reside in the cytoplasm and mitochondria ([36]), this implies that intercompartmental cross-talk may take place in folate homeostasis and the response to MTX.

ABC drug–efflux transporters have dual functions. From a pharmacologic perspective, they may extrude a variety of cytotoxic drugs, including MTX, and thereby confer drug resistance ([5, 13, 37]), whereas physiologically, one important function is to secrete inflammatory mediators, e.g., leukotrienes ([38]). As such, increased expression of ABC transporters in MTX-naive RA patients as compared to that in healthy controls may be compatible with an inflammatory response. In this respect, it is of interest to note that upon MTX treatment, expression levels of ABC transporters do not further increase, which would suggest that they have no role in resistance development, but rather, the levels decrease, which is a possible indication of a genuine antiinflammatory effect.

Given its cross-sectional approach, the present study was not designed to directly assess MTX-response predictions. Moreover, on a cautionary note, it is advised that (clinical) MTX-response parameters should not be overinterpreted in cross-sectional studies when compared to longitudinal studies ([39]). However, over the past decade, multiple efforts have been made to build MTX-response prediction models based on genetic variants (single-nucleotide polymorphisms [SNPs]) of folate/MTX–related genes ([22, 24, 40]) or MTXGlu in RBCs ([26, 41, 42]). Although these studies showed promising results, genetic associations were not always consistent ([23]) or could not be replicated in various international cohorts ([43]). For most SNPs, their impact on the functional activity of gene products is still unknown, and it remains to be determined whether they contribute individually to sustaining folate homeostasis. Regarding MTXGlu formation, measurements in RBCs provide valuable information and are well accepted. However, RBCs do not represent the most ideal cell type, since they do not have a nucleus and intracellular organelles (and therefore lack lysosomal GGH activity) to control folate homeostasis, as white blood cells do. Therefore, a longitudinal cohort study combining analysis of gene expression profiles, SNPs, and RBC MTXGlu would be most revealing in identifying relevant parameters for MTX-response prediction studies.

Understanding the MTX response is relevant not only in a monotherapy or DMARD-combination setting. It is also a crucial factor in a treatment setting of combination therapy with biologic agents such as infliximab, adalimumab, or etanercept ([44, 45]), to achieve less radiologic progression of damage in the joints ([46]). Based on our observations of a stabilization/decrease in folate/MTX pathway–related gene expression levels in the MTX-treated group compared to the MTX-naive group, the results of the present study suggest that RA patients may have a sustained response to MTX therapy, even though they may be clinically classified as poor responders eligible for biologic treatment. This would imply that clinical nonresponsiveness to MTX may be associated with mechanisms downstream of the folate/MTX pathway ([9, 13]). Moreover, one cannot exclude the possibility that for a small fraction of RA patients with a poor response to MTX, classic mechanisms of acquired resistance to MTX may still apply ([5, 13]).

Finally, identification of differential expression profiles of folate-related genes in RA patients may open new avenues for therapeutic interventions beyond drugs such as MTX. In fact, from a clinical oncology perspective, many MTX analogs were designed to overcome MTX resistance ([5]). These next-generation antifolate drugs have emerged from clinical application with known efficacy and safety/toxicity profiles. For example, these novel antifolate agents harbor properties of more efficient cellular uptake (via RFC, FR, or PCFT), polyglutamylation by FPGS, impaired efflux via ABC transporters, and targeting of key enzymes in folate metabolism ([5, 47, 48]). Expanding our knowledge of the differential expression levels of folate transporters in various blood-cell types may also contribute to their selective targeting. For example, FRβ expression on synovial macrophages has been exploited for selective delivery of antifolate drugs for which FRβ has a high affinity ([16, 49]). Given the response profile for MTX, the present study results (as shown in Figure 3) could suggest that the folate-dependent enzyme GART (involved in de novo purine biosynthesis) may be an attractive drug target. Currently, several inhibitors of this enzyme are being evaluated in a (pre)clinical oncology setting ([5, 50]). Whether these novel GART inhibitors could elicit potential antiinflammatory responses remains to be established.

In conclusion, the current study identified remarkable differences in expression profiles of folate/MTX pathway–related genes in the peripheral blood cells of MTX-treated RA patients compared to MTX-naive RA patients and healthy controls. The increased immune-activation status known to be present in patients with RA paralleled the elevated expression levels of these genes in MTX-naive RA patients compared to healthy controls. The impact of MTX treatment on RA was manifested in the stabilization and suppression of the folate/MTX pathway–related gene expression levels. This novel gene signature could provide additional insights into the mechanism of action of MTX, and thus pave the way to facilitate further exploration of MTX with the aim of continually improving its pharmacologic properties and efficacy.

AUTHOR CONTRIBUTIONS

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Verweij had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Blits, Jansen, Assaraf, Lems, Nurmohamed, van Schaardenburg, Verweij.

Acquisition of data. Blits, Jansen, Nurmohamed, Voskuyl, Wolbink.

Analysis and interpretation of data. Blits, Jansen, Assaraf, van de Wiel, Nurmohamed, Wolbink, Vosslamber, Verweij.

Acknowledgments

We extend our gratitude to the rheumatologists who allowed inclusion of their patients in the present study. We also thank Michiel Pegtel for his support.

Ancillary