Methotrexate pharmacogenetics: The first step toward individualized therapy in rheumatoid arthritis



Rheumatoid arthritis (RA) affects 0.5–1% of the population worldwide. Over one-half the patients with RA are being treated with methotrexate (MTX), making this the most commonly used disease-modifying antirheumatic drug (DMARD). Since the first reports of the efficacy of low-dose MTX in RA (1), several studies have established its central role in the treatment of RA, both as monotherapy and in combination with other DMARDs (2–8). Despite the advent of the newer biologic therapies, MTX retains this central role because it is relatively inexpensive, there is broad experience with its use, and it is widely used in combination regimens with other DMARDs (4–7).

Although well proven, the efficacy of MTX in RA is variable. The response to MTX according to the American College of Rheumatology (ACR) criteria for 20% improvement varies from 46% to 65% (9, 10). There are no useful and reliable clinical or molecular markers of response to therapy. Serum levels of MTX are not useful in determining drug efficacy, since the drug is eliminated from the circulation within 24 hours of administration, which is much shorter than the standard weekly dosing interval in RA (11). While circulating intracellular levels of MTX polyglutamates (MTXPGs) in erythrocytes and polymorphonuclear cells do correlate with clinical efficacy, the assay for MTXPGs is technically difficult and is not available in most clinical facilities (12). Levels of various cytokines and other mediators of inflammation, such as tumor necrosis factor α (TNFα), interleukin-10 (IL-10) (13), matrix metalloproteinase 3 (MMP-3), IL-6 (14), and tissue inhibitor of metalloproteinases 1 (TIMP-1) (15), may correlate with the efficacy of MTX, but rapid clinical assays to measure them are not readily available. Serum levels of chemokines such as RANTES and growth-related oncogene α (GROα) may predict the effect of MTX on radiographic erosions, but are not easily measurable (16). Hence, although sophisticated cytokine and enzyme assays may assess response to treatment, they are often expensive and are not available in most clinical settings, thus limiting their utility in the clinical management of individual patients.

Toxicity is another factor that limits the use of MTX. Approximately 10–30% of patients with RA discontinue MTX because of toxicity (17). Several toxicities have been reported, including nodulosis (8%) (18), hypersensitivity pneumonitis (2–5%) (19), central nervous system toxicity (1–35%) (20), postdosing reactions (10%) (21), gastrointestinal (GI) symptoms such as nausea, vomiting, abdominal pain, and diarrhea (60%) (22, 23), hepatitis with elevated transaminase levels (20–58%) (24), hematologic abnormalities (1–2%) (25), rash (1–2%) (10), alopecia (5%) (10), and osteopathy (rare) (26). Although it has been suggested that serum levels of MTX may predict GI toxicity and myelosuppression (27), their predictive ability is low. Hence, in current clinical practice, toxicity monitoring is performed routinely with blood counts and liver function tests. Not surprisingly therefore, MTX has the highest monitoring costs among the commonly used DMARDs, although the drug itself is inexpensive (28).

The significant variability in clinical response and unpredictable toxicity in patients taking MTX, coupled with the presence of effective, but expensive, alternative therapies, has led to approaches for screening of patients prior to the initiation of MTX. This may help decrease the morbidity associated with side effects, reduce the need for laboratory tests to monitor toxicity, and help select patients who are more likely to respond to the drug. The rapidly growing field of pharmacogenetics may make such screening methods feasible and available for use in clinical practice.

We review the concepts of pharmacogenetics as they apply to MTX, highlight major, relevant publications, and describe the potential implications of this field for future research and clinical care. As is evident from the contents of this review, pharmacogenetics is a rapidly advancing area of research and holds promise that in the near future, therapy can be tailored to a given patient's genetic profile.

Pharmacogenetics, an emerging field

Pharmacogenetics is the study of genetic polymorphisms in drug-metabolizing enzymes and the translation of inherited differences to differences in drug effects (29). Genes are described as “polymorphic” when allelic variants exist in the population, one or more of which alters the activity of the encoded protein compared with the wild-type sequence. Typically, the polymorphism leads to reduced activity of the encoded protein. Although the focus of pharmacogenetics is the study of drug-metabolizing enzymes, polymorphisms in drug transporters (such as P-glycoprotein) and drug targets (such as receptors) have also received attention. Pharmacogenetic studies may soon make it possible to select medications and dosages precisely for individual patients (30).

Until recently, clinically relevant polymorphisms in drug metabolism were only discovered if they occurred in families or led to obvious phenotypic differences among individuals in the population. Major advances in the field of genetics have changed this. With molecular sequencing and high-throughput technology, large numbers of genetic polymorphisms, such as single-nucleotide polymorphisms (SNPs), can now be detected accurately and rapidly. Used in the context of clinical studies, such technology may allow the rapid identification of SNPs in drug-metabolizing enzymes that may have clinically important consequences, such as variability in drug response among individuals (31).

Cellular pathway of methotrexate

MTX, a folate analog, is a competitive inhibitor of the enzyme dihydrofolate reductase (DHFR). It enters cells through an active transport mechanism that is mediated by the reduced folate carrier 1 (RFC-1), also called SLC19A1 (Figure 1). MTX efflux from the cell is accomplished by members of the ATP–binding cassette (ABC) family of transporters. The ABC family includes 48 proteins classified into 7 distinct subfamilies ABC A–G (32). Current evidence suggests that, of these transporters, ABCC 1–4 and ABCG2 play a major role in the efflux of MTX from the cell (33–36).

Figure 1.

Cellular pathway of methotrexate. Genes shown in italics have been targets of pharmacogenetic analyses in published studies. RFC-1 = reduced folate carrier 1; ABCB1 and ABCC1–4 = adenosine triphosphate–binding cassette (ABC) transporters; GGH = γ-glutamyl hydrolase; FPGS = folylpolyglutamyl synthase; MTX-PG = methotrexate polyglutamate; TYMS = thymidylate synthase; dTMP = deoxythymidine monophosphate; dUMP = deoxyuridine monophosphate; DHFR = dihydrofolate reductase; FH2 = dihydrofolate; FH4 = tetrahydrofolate; MTR = methyl tetrahydrofolate reductase; 5,10-CH2-THF = 5,10-methylene tetrahydrofolate; MTHFR = methylenetetrahydrofolate reductase; 5-CH3-THF = 5-methyl-tetrahydrofolate; AICAR = aminoimidazole carboxamide ribonucleotide; ATIC = AICAR transformylase; FAICAR = 10-formyl-AICAR; IMP = inosine monophosphate; AMP = adenosine monophosphate; ADP = adenosine diphosphate; ATP = adenosine triphosphate; ADA = adenosine deaminase.

Once inside the cell, MTX is converted into a polyglutamate form by the enzyme folylpolyglutamyl synthase (FPGS). This process can be reversed by the enzyme γ-glutamyl hydrolase (GGH), which by catalyzing the removal of the γ-linked polyglutamates, facilitates MTX efflux from the cell. The polyglutamate form of MTX, which can have up to 7 glutamic acid moieties (MTXPG2–7), has several important functions. It retains MTX within the cell (37), and it inhibits DHFR, which reduces dihydrofolate (DHF) to tetrahydrofolate (THF). THF is important for the production of biologically active folate cofactors, such as 5-methyl-THF, which is required for the generation of methionine from homocysteine and for the synthesis of polyamines (38). The polyglutamated form of MTX also inhibits thymidylate synthase (TYMS), which converts deoxyuridylate to deoxythymidylate in the de novo pyrimidine biosynthetic pathway (39). The enzyme methylenetetrahydrofolate reductase (MTHFR) in the folic acid pathway is not directly inhibited by MTX. However, because of the effects of MTX on the intracellular folate pool, this enzyme is influenced by MTX.

Recent evidence suggests that the effects of MTX on purine synthesis may be critical to its antiinflammatory properties. MTXPGs inhibit the enzyme aminoimidazole carboxamide ribonucleotide (AICAR) transformylase, which causes intracellular accumulation of AICAR. AICAR and its metabolites can then inhibit two enzymes involved in adenosine metabolism, adenosine deaminase and AMP deaminase, leading to increased intracellular concentrations of adenosine and adenine nucleotides. Subsequent dephosphorylation of these nucleotides results in increased concentrations of adenosine in the extracellular space. Adenosine is a potent antiinflammatory agent, and some of the antiproliferative effects of MTX are thought to be mediated through this mechanism (40).

This review is divided into gene polymorphisms that influence MTX transport across the cell membrane, those that influence enzymes in the cellular pathway of MTX, and the effects of gene–gene interactions.

MTX transporter pharmacogenetics


RFC-1 is important for the transport of MTX into the cell. Defective transport can occur as a result of polymorphisms that either inactivate the enzyme or change the function of transcription factors, leading to loss of RFC-1 gene expression (41, 42). A G80A polymorphism leading to the substitution of arginine for histidine at codon 27 in the first transmembrane domain (TMD1) of the RFC-1 protein (43) and a 61-bp repeat polymorphism in the RFC-1 promoter associated with increased transcriptional activity of the gene have been recently discovered (44). It remains controversial whether the RFC G80A SNP affects the uptake and intracellular levels of MTX. One study showed no significant differences in MTX uptake rates between leukemic blast cells with the arginine-27 RFC proteins versus those with the histidine-27 RFC proteins (45). However, in another study, children with acute lymphoblastic leukemia homozygous for the RFC G80A variant had higher MTX plasma levels (P = 0.004) than other genotype groups (46). In yet another study, RA patients taking MTX who had the RFC 80A/A genotype had higher MTXPG levels compared with those who had the RFC 80G/G or G/A genotypes (P = 0.007) (47).

The G80A SNP has further been studied in association with response to MTX in 105 patients with RA (48). In that study, patients within the top 25th percentile of MTX responders were identified using a visual analog scale (VAS) to measure the physicians' assessment of patients' response to MTX. Patients homozygous for the RFC SNP 80A/A had a greater response to MTX compared with patients with the wild-type allele (80G/G). Patients homozygous for the A allele were 3 times more likely to be within the top 25th percentile of MTX responders (95% confidence interval [95% CI] 1.3–8.4; P < 0.01) compared with those with the wild-type G allele, suggesting that this SNP is associated with an increased response to the drug (48).


P-glycoprotein, the product of the ABCB1 gene (a member of the ABC family of transporters) is a membrane transporter implicated in the disposition and bioavailability of several drugs including MTX (Figure 1). Recently, SNPs in the ABCB1 gene have been identified and correlated with P-glycoprotein expression (49). One of them is the C3435T polymorphism in exon 26 of the gene (50, 51). Like RFC-1, there is considerable controversy regarding whether genetic variations in ABCB1 and/or P-glycoprotein expression influence MTX efflux from the cell. Although there are no data demonstrating that the ABCB1 C3435T SNP directly affects the cellular transport of MTX, some studies indicate that higher P-glycoprotein expression may be a marker of MTX resistance (52), whereas other studies contradict this finding (53, 54).

In a recent study, 92 patients with RA and 97 healthy controls were genotyped for the C3435T polymorphism (55). Patients were classified as those with active disease (n = 62) despite 6 months of therapy with MTX (7.5–15 mg/week) and prednisone (5–10 mg daily) and those with inactive disease (n = 30) after 6 months of the same treatment. There was no difference in the distribution of the ABCB1 genotypes among the RA patients and controls. Patients with the 3435CC and 3435CT genotypes had a greater risk of having active RA compared with patients with the 3435TT genotype (odds ratio [OR] 2.89 [95% CI 0.87–9.7], P < 0.05). The 3435T allele also seemed to confer a protective effect, with patients homozygous for this allele having a less severe form of RA that was more likely to respond to MTX and prednisone (55).

Although genetic variations in other members of the ABC family have not been studied in the context of MTX response in RA, SNPs in these transporters are quite common. In our study of 95 patients with RA and 190 controls, we established that SNPs in the ABCB1 (1236C>T and 3435C>T), ABCC1 (4002G>A, IVS14+115C>T, and IVS18-30C>G), and ABCC2 (–24C>T, 1249G>A, 4488C>T, 4544G>A, IVS23+56T>C, and IVS31+74C>T) transporters occur quite frequently. Moreover, allele frequencies of the ABCB1 3435C>T SNP were significantly different (P < 0.001) between European Americans and African Americans in our cohort of RA patients, suggesting that racial differences need to be considered when interpreting pharmacogenetic studies (56).

Thus, genetic variations in transporters such as RFC-1 and ABCB1 may affect the response to MTX in RA, although this remains a controversial subject (Table 1). Genetic variations in other transporters occur frequently and need to be studied further for their effects on drug response. Drugs such as sulfinpyrazone and probenecid are inhibitors of some of the ABC transporters. Future strategies to improve MTX efficacy may include coadministration of such inhibitors with MTX in patients with an unfavorable transporter genotype encoding increased drug efflux.

Table 1. Pharmacogenetics of MTX transporters*
GenePolymorphismAmino acid substitution in enzymeBiochemical effectsClinical effectsReference
  • *

    MTX = methotrexate; RFC-1 = reduced folate carrier 1; ABCB1 = ATP binding casette transporter B1.

RFC-1G80AHistidine to arginine at codon 27May affect transcriptional activity of RFC1 gene and MTX entry into cellMay affect response to MTX48, 72
ABCB1C3435TNo amino acid substitutionMay affect MTX entry into cellMay affect response to MTX55

MTX cellular pharmacogenetics

Genes in the MTX cellular pathway that have been studied so far to assess response to MTX in RA are MTHFR and TYMS.


This is the best studied of the genes in the MTX cellular pathway. MTHFR is important in the generation of 5-methyl-THF, which is the methyl donor for the methylation of homocysteine to methionine by methionine synthase (MS). About a dozen SNPs in the MTHFR gene have been described; two nonsynonymous SNPs have been extensively studied. The C677T polymorphism, which was first described in 1995, causes an alanine-to-valine substitution at codon 222 of the MTHFR gene. It results in a thermolabile variant of MTHFR with decreased enzyme activity and subsequent increased plasma homocysteine levels (57). The homozygous C677T variant, with ∼30% of the wild-type activity, has a prevalence of ∼8–10% in the general population. Heterozygotes have ∼60% activity and form ∼40% of the population.

Another polymorphism in the MTHFR gene, A1298C, which leads to a glutamine-to-alanine substitution at codon 222, was described in 1998 (58, 59). Homozygotes and heterozygotes for A1298C have reduced activity of the MTHFR enzyme, although they do not have a thermolabile variant of MTHFR. The homozygous genotype with ∼60% of enzyme activity in lymphocytes has been observed in ∼10% of the Canadian population (worldwide prevalence unknown) (59). The A1298C polymorphism, by itself, does not influence plasma homocysteine levels. However, individuals heterozygous for the C677T and A1298C polymorphisms have significantly decreased activity of the MTHFR enzyme and elevated plasma homocysteine levels comparable to those in individuals homozygous for the C677T polymorphism (58).

Increased plasma homocysteine levels, accentuated further in individuals with the C677T and A1298C SNPs in the MTHFR gene, may play a role in MTX toxicity. A number of studies have examined this association (Table 2). In one study, 105 patients with RA, 35 of whom were treated with MTX, 34 with sulfasalazine (SSZ), and 36 with MTX plus SSZ, were recruited, and plasma homocysteine levels were measured in all patients (60). All patients were genotyped for the C677T SNP. The two treatment groups receiving MTX showed a greater increase in plasma homocysteine levels as compared with the group receiving SSZ alone, with the patients taking MTX plus SSZ showing the greatest increase. Heterozygotes and homozygotes for the C677T SNP had higher plasma homocysteine levels than did patients in whom the SNP was absent. Although a statistically significant increase in plasma homocysteine levels (17%; P < 0.05) was found in patients experiencing a GI adverse event (compared with patients without an adverse event), there was no direct association between the MTHFR genotype and GI events. Thus, patients with RA who are taking MTX may have increased plasma homocysteine levels, which are further exacerbated by the presence of the C677T SNP and this may be important in mediating the GI toxicity from MTX (60).

Table 2. Pharmacogenetics of MTHFR*
GenePolymorphismAmino acid substitution in enzymeBiochemical effectsClinical effects (ref.)
  • *

    MTHFR = methylenetetrahydrofolate reductase; GI = gastrointestinal; MTX = methotrexate; RA = rheumatoid arthritis.

MTHFRC677TAlanine to valineThermolabile MTHFR with decreased activity; increased plasma homocysteineMay increase the following: GI toxicity (60); hepatic and GI toxicity, alopecia, stomatitis, and rash (62, 63); headache, lethargy (74). No effect on toxicity (62); no effect on efficacy or toxicity (71)
MTHFRA1298CGlutamine to alanineMay decrease MTHFR activity and increase plasma homocysteineMay affect MTX efficacy (63); may increase susceptibility to RA and decrease MTX toxicity (62). No effect on efficacy or toxicity (71)

In another study, 236 patients with RA who where taking MTX were genotyped for the C677T SNP, and MTX toxicity and disease activity were assessed in all patients (61). Nineteen patients (8%) were homozygous, 95 patients (40%) were heterozygous for the C677T polymorphism, and in 122 patients (52%), the SNP was absent. Patients with the SNP (homozygous or heterozygous) had an increased risk (relative risk [RR] 2.01 [95% CI 1.09–3.70]) of discontinuing MTX therapy because of adverse events, such as GI symptoms, hair loss, and hepatotoxicity. The strongest association of the polymorphism was with an increased risk of elevated liver enzyme levels, specifically, alanine aminotransferase (RR 2.38 [95% CI 1.06–5.34]). Folate supplementation did not influence the predictive power of the genotype on MTX toxicity. The C677T genotype had no influence on the efficacy of MTX in this study. This study suggests that the C677T polymorphism may be a marker for MTX toxicity, particularly, hepatotoxicity (61).

In a more recent cross-sectional study, 93 patients with RA who were taking MTX and 377 healthy subjects were genotyped for the C677T and A1298C SNPs (62). Serum folate and plasma homocysteine levels were measured in all patients. Disease activity was assessed using standard measures. There was a significant difference in genotype distribution between the RA patients and controls. Of the RA patients, 24.7% had the 1298CC genotype, as compared with 12.8% of the controls (P < 0.001). There was no association between disease activity and any of the 1298 genotypes (AA, AC, and CC). However, the 1298CC genotype appeared to be protective from MTX-related adverse effects in RA patients; 33% of patients without side effects had the 1298CC genotype, whereas only 9.1% of those with side effects carried this genotype (P = 0.035). Patients with the wild-type 1298AA genotype were 5 times more likely to develop MTX-related adverse effects than were patients with the 1298CC genotype (OR 5.24 [95% CI 1.38–20]). Patients with the 1298CC genotype had higher plasma homocysteine levels than did patients with the AA or AC genotype; homocysteine levels were not influenced by folate levels. There was no association between the C677T polymorphism and MTX-associated side effects in this study. The authors concluded that RA may be more common among 1298CC homozygotes in their population and that 1298CC homozygosity may confer protection from MTX-related side effects through a homocysteine-dependent mechanism (62).

A retrospective study evaluated 106 patients with RA who were taking MTX or had discontinued MTX because of adverse events (63). All patients were genotyped for the MTHFR C677T and A1298C polymorphisms, which were then correlated with the efficacy and toxicity of MTX. Plasma homocysteine levels were not measured in this study. Patients with the A1298C polymorphism (homozygous or heterozygous) were taking lower doses of MTX as compared with patients without the polymorphism (P < 0.05, RR 2.18 [95% CI 1.17–4.06]) and showed improvements in C-reactive protein (CRP) levels and the erythrocyte sedimentation rate (P < 0.05), but not in the tender or swollen joint counts. No such associations were observed with the C677T polymorphism, but this SNP (homozygous or heterozygous) was associated with overall MTX toxicity, such as an increase in transaminase levels, stomatitis, nausea, vomiting, hair loss, fatigue, and rash (P < 0.05, RR 1.25 [95% CI 1.05–1.49]). The A1298C polymorphism had no effect on MTX toxicity. Thus, both SNPs in the MTHFR gene had very different effects; the C677T polymorphism influenced MTX toxicity, and the A1298C polymorphism influenced MTX efficacy (63).

A more recent study genotyped 63 patients with autoimmune diseases who were taking MTX and folate supplements for the MTHFR C677T polymorphism (64). Toxicity information was collected retrospectively. Of these patients, 36.5% had the CC, 49.2% the CT, and 14.2% had the TT genotypes. Fifteen patients (23.8%) had adverse effects, and 6 of these patients discontinued MTX because of toxicity. Surprisingly, toxicity was more frequent in patients with the wild-type (CC) genotype as compared with those who carried the polymorphism (P = 0.005). The authors postulated a protective effect of the MTHFR C677T SNP on MTX toxicity, although it should be noted that the number of patients in this study was small; it is unclear how many of the patients had RA, and no details on MTX dosage and toxicity were provided (64).


TYMS, a key enzyme in de novo thymidylate synthesis, converts dUMP to dTMP. TYMS is inhibited directly by polyglutamated MTX and indirectly by folate cofactor depletion induced by MTX (Figure 1). Inhibition of TYMS leads to dTMP depletion and increased uracil misincorporation into nucleic acid, which in turn, leads to chromosome damage and cell death. A tandem repeat sequence has been identified in the 5′-untranslated region (5′-UTR) of the TYMS gene (65). This sequence is polymorphic, with a variable number of 28-bp repeat elements (5′-UTR 28-bp repeat). The repeat element appears to function as an enhancer, because TYMS messenger RNA (mRNA) expression and TYMS enzyme activity are increased with an increasing number of repeat sequences in vitro (66–68). Patients homozygous for the triple repeat allele (TSER*3/*3) have higher TYMS mRNA expression than those homozygous for a double repeat allele (TSER*2/*2) (67, 68). Another polymorphism in the TYMS gene, consisting of a 6-bp deletion of the sequence TTAAAG at nucleotide 1494 in the 3′-UTR (3′-UTR 6-bp deletion), has been described (69). Although the function of this polymorphism is not fully known, there is evidence to suggest that the 3′-UTR deletion is associated with decreased TYMS mRNA stability and expression (69, 70).

In a study analyzing the effects of the TYMS and MTHFR genotypes on MTX efficacy and toxicity, 167 patients with RA, 115 of whom had been treated with MTX, were recruited (71) (Table 3). The mean ± SD weekly dose of MTX was 5.7 ± 2.3 mg. Toxicity information was extracted from the medical records. Genotyping for the TYMS 5′-UTR 28-bp repeat, the 3′-UTR 6-bp deletion, and the MTHFR C677T and A1298C polymorphisms was performed. The allele frequencies of all 4 polymorphisms were similar between the MTX and non-MTX groups. Forty-five percent of patients treated with MTX experienced toxicity; none of the TYMS or MTHFR polymorphisms influenced this. The weekly MTX dosage and the CRP levels were used to assess efficacy. Patients homozygous for the TSER*3/*3 repeat allele required higher dosages of MTX (>6 mg/week) compared with those homozygous for the TSER*2/*2 repeat allele (P = 0.033). In contrast, patients homozygous for the deletion allele (0 bp/0 bp) had ≥50% improvement in their CRP levels after treatment with MTX as compared with pretreatment levels (P = 0.024). Based on these findings, the authors speculated that by increasing TYMS mRNA expression, the TSER*3/*3 polymorphism lowered MTX efficacy, whereas by decreasing TYMS expression, the deletion SNP (0 bp) made patients more sensitive to MTX. The MTHFR polymorphisms had no effects on MTX toxicity or efficacy in this study (71).

Table 3. MTX pathway pharmacogenetics
GeneRole in MTX pathwayPolymorphismEffects on gene product/enzymeClinical effectsReference
  1. * MTX = methotrexate; ATIC = aminoimidazole carboxamide ribonucleotide transformylase; AICAR = aminoimidazole carboxamide ribonucleotide; TYMS = thymidylate synthase; 5′-UTR = 5′-untranslated region.

ATICConversion of AICAR to 10-formyl-AICAR; target of MTXC347GMay decrease ATIC activity and affect AICAR accumulation and adenosine releaseMay affect MTX efficacy and toxicity72, 74
TYMSConversion of dUMP to dTMP; target of MTX5′-UTR 28-bp repeatMay increase TYMS enzyme activityMay affect MTX efficacy and toxicity71, 74
  3′-UTR 6-bp deletionMay decrease TYMS mRNA stability and expressionMay affect MTX efficacy71

Thus, there is fairly good evidence to suggest an association between the C677T SNP in MTHFR and MTX toxicity, presumably through its effects on homocysteine metabolism (60, 61, 63). One study showed an opposite effect, that is, the C677T SNP protected patients from MTX toxicity, but that study had several limitations as described above (64). The role of the A1298C polymorphism in influencing response to MTX is more nebulous, with available data indicating that it may make patients more responsive to MTX (63) and may possibly protect them from side effects (62). However, it should be mentioned here that the associations between these polymorphisms and toxicity or response in these studies were not robust. Some of these studies were retrospective, which may have led to errors in the estimation of MTX effects, particularly adverse effects. Thus, it is unclear whether data from these studies will be of clinical utility in predicting outcomes in an individual patient taking MTX. Variants in the TYMS gene also appear to influence MTX efficacy (71). Although that study concluded that MTHFR SNPs did not affect MTX efficacy or toxicity, it is noteworthy that the dosages of MTX used in the study were small (6 mg/week), which may have leveled off the differences in MTX response, and MTX efficacy was not assessed using standardized measures of disease activity. Also important to consider is whether ethnicity influenced the response to MTX in patients in this Japanese study (71).

MTX pathway pharmacogenetics

Since MTX is influenced by many gene products, it is not surprising that polygenic analyses have begun to be reported.


AICAR transformylase (ATIC), which converts AICAR to 10-formyl-AICAR, is directly inhibited by MTX, leading to the accumulation of AICAR and adenosine, an antiinflammatory purine (Figure 1). A study by Dervieux et al (72) examined the combined effects of the C347G SNP in ATIC, the TYMS 5′-UTR 28-bp repeat polymorphism (TSER*2/*2), and the G80A polymorphism in RFC-1 (a MTX transporter) in 108 patients with RA treated with MTX for >3 months (Table 3). All patients were genotyped for these SNPs, and a pharmacogenetic index corresponding to the sum of the homozygous variant genotypes (RFC-1 80AA, ATIC 347GG, and TSER*2/*2) carried by the patients was calculated. Red blood cell long-chain MTXPG concentrations and response to MTX were assessed. Patients were classified as MTX responders (VAS score ≤2 cm) and MTX nonresponders (VAS score >2 cm), based on the physician's assessment of the patient's response to MTX. The results showed that the presence of at least 1 homozygous variant genotype (RFC-1 80AA, ATIC 347GG, or TSER*2/*2) and/or a higher pharmacogenetic index was associated with a greater response to MTX (VAS ≤2 cm; P < 0.001), as were increased MTXPG levels. Patients with at least 1 homozygous variant were 3.7 times more likely than patients without a homozygous variant to respond to MTX (P = 0.01 [95% CI 1.7–9.1]). Thus, homozygosity for ≥1 of these alleles appeared to confer a favorable “response genotype” to MTX (72).


Serine hydroxymethyltransferase (SHMT) encodes a vitamin B6–dependent enzyme that is important in the synthesis of 5,10-methylene-THF. A C1420T polymorphism in this gene that influences red blood cell folate levels has been described (73). In 1 study, 214 RA patients taking MTX were genotyped for MTHFR C677T, TYMS 5′-UTR 28-bp repeat (TSER*2/*2), ATIC C347G, and SHMT C1420T polymorphisms. MTX side effects at the time of a single study visit were recorded. A sum of the homozygous variants for each gene was calculated to generate the toxicogenetic index for each patient. Sixty-eight patients (32%) experienced a side effect. Specific genotypes were associated with specific toxicities. MTHFR 677TT (P < 0.01, OR 3.3 [95% CI 1.7–6.4]) and SHMT 1420CC (P = 0.019, OR 2.4 [95% CI 1.2–4.8]) were associated with headache and lethargy. ATIC 347GG (P < 0.01, OR 3.0 [95% CI 1.6–5.4]) was associated with GI side effects. TSER*2/*2 and SHMT 1420CC were associated with alopecia (P < 0.01, OR 5.6 [95% CI 3.0–9.5] and P < 0.01, OR 3.2 [95% CI 2.2–4.5], respectively). The toxicogenetic index (MTHFR 677TT plus SHMT 1420CC plus TSER*2/*2 plus ATIC 347GG) ranged from 0 to 3. An increase in each unit of the index resulted in a 1.9-fold increase in the likelihood of side effects (P = 0.004 [95% CI 1.1–3.1]). Thus, homozygosity for ≥1 of these alleles appeared to confer a “toxicity genotype” to MTX (74).

Other genes with pharmacogenetic implications

Polymorphisms have been described in other genes in the MTX pathway, although their functional and clinical significance remain largely unknown (Table 4). GGH catalyzes the removal of γ-linked polyglutamates, converting long-chain MTXPGs into shorter-chain MTXPGs and, eventually, to MTX, which can diffuse out of the cell. Hence, GGH may be important in mediating resistance to MTX. A C401T promoter polymorphism in GGH has been described (75), and it influences MTXPG levels in RA patients. Individuals carrying the GGH 401TT genotype had lower MTXPG levels (P = 0.002, OR 4.8 [95% CI 1.8–13.0]) as compared with the other genotype groups (47). A C452T polymorphism in the GGH gene, altering the binding affinity of GGH to long-chain MTXPGs and resulting in low GGH activity, has also been described (76). DHFR is a major target of MTX, and polymorphisms in this enzyme affecting either enzyme function or binding to MTX may impact the response to MTX. Two SNPs in the 3′-UTR of the DHFR gene, T721A and C829T, have been described in a Japanese population (77). The 3′-UTR of DHFR may be an important regulatory component of the gene, since subjects homozygous for the C829T SNP (TT genotype) had greater expression of DHFR (median expression 2.53) as compared with CT heterozygotes (median expression 0.41) or CC wild-type subjects (median expression 0.18) (P < 0.001) (77). The allele frequencies of these SNPs in Caucasians are currently unknown.

Table 4. Other genes with potential pharmacogenetic implications in the MTX pathway*
GeneRole in MTX pathwayPolymorphismEffects on gene product/enzymePostulated clinical effectsReference
  • *

    MTX = methotrexate; GGH = γ-glutamyl hydrolase; MTXPGs = methotrexate polyglutamates; DHFR = dihydrofolate reductase; DHF = dihydrofolate; THF = tetrahydrofolate; 3′-UTR = 3′-untranslated region; MS = methionine synthase; MTRR = methionine synthase reductase.

GGHConversion of long-chain MTXPGs to short-chain MTXPGs by removal of glutamatesC452TDecreased binding affinity of GGH for MTXPGsMay affect MTX efficacy76
  C401TAffects MTXPG levels47
DHFRReduction of DHF to THF; target of MTX3′-UTR T721A and C829TMay increase DHFR expressionMay affect MTX efficacy77
MSMethylation of homocysteine to methionineA2756GMay decrease MS activity; increase homocysteine levelsMay affect MTX toxicity79, 80
MTRRMethylation of cobalamin cofactor required for the action of MSA66GMay decrease MTRR activity; increase homocysteine levelsMay affect MTX toxicity81, 82

Methionine synthase (MS) is required for the methylation of homocysteine to methionine in the presence of a cobalamin cofactor. A polymorphism in the MS gene, A2756G, resulting in an aspartic acid-to-glycine change at codon 919 (D919G), has been described (78). This SNP may be an activating mutation, since individuals homozygous for the SNP (DD genotype) have high homocysteine levels as compared with individuals with the wild-type (GG) genotype (79, 80). Methionine synthase reductase (MTRR) is an enzyme important for the methylation of the cobalamin cofactor of MS. A A66G SNP in MTRR, leading to substitution of methionine for isoleucine at codon 22, with the MTRR 66GG genotype conferring the risk of elevated homocysteine levels, has been identified (81, 82).

Gene–gene interactions

It is unlikely that a single genetic locus will be sufficient to adequately predict response to a drug in a polygenic disease such as RA. Rather, given the effects of MTX on several metabolic pathways, a composite of multiple “risk” loci will identify the patients likely to have a poor outcome with MTX therapy and may be difficult to detect. Haplotype analysis may be another useful tool for determining associations with measurements of outcomes (83). Such composite allele and haplotype analyses may identify distinct subsets of RA patients who may have a differential response to MTX.

The inhibition of de novo purine and pyrimidine biosynthesis by MTX is mediated by both the depletion of folate cofactors (such as 5,10-methylene-THF, 10-formyl-THF) and direct enzyme inhibition (such as TYMS and ATIC). TYMS requires 5,10-methylene-THF for its activity, whereas ATIC requires 10-formyl-THF for its activity (Figure 1). Hence, variants of genes that control the adequacy of folate cofactor pools (such as MTHFR) may be synergistic with variants of genes that encode these enzymes (TYMS and ATIC) and may have more profound effects on MTX response. This concept was partly validated in the study by Dervieux et al (72), in which homozygosity for variants in the ATIC and TYMS genes was linked to a better response to MTX. Lower levels of TYMS, ATIC, 5,10-methylene-THF, and 10-formyl-THF would facilitate the action of MTX, while higher levels would confer MTX resistance.

Similarly, published data indicate that the C677T polymorphism in the MTHFR gene may be associated with an increased risk of MTX toxicity by causing elevations in plasma homocysteine levels (60, 61, 63). Since the A2756G SNP in the MS gene and the A66G SNP in the MTRR gene have also been shown to cause increased plasma homocysteine levels, individuals with a combination of such “risk” alleles (MTHFR 677T, MS 2756A, and MTRR 66G) who are taking MTX may experience more toxicity from the drug.

Future directions

Studies published so far on the pharmacogenetics of MTX in RA offer exciting insights into the varied mechanisms of action of this drug in RA and the potential for exploiting genetic interindividual differences in such mechanisms for the benefit of the individual patient. However, the data from these studies are not convincing enough to draw far-reaching conclusions about the applicability of MTX pharmacogenetics in clinical practice. It should be emphasized that prospective multicenter trials with large numbers of patients conducted with the collaborative efforts of research funding agencies and investigators interested in this area of research are needed to establish associations between specific genotypes and drug toxicity/efficacy and to define predictors of treatment response. This is particularly true for drugs used in RA, such as MTX, leflunomide, and the biologic agents which cause toxicity only in a small proportion of patients treated. Criteria for response need to be strictly defined in such trials, since traditional criteria, such as ACR 20% response rates, may be insufficient to distinguish differences in response on a pharmacogenetic basis. Establishment of multi-SNP haplotype maps for genes of interest for these drugs will be another mechanism to facilitate genotype-guided therapy in the near future. The commitment of major funding agencies to pharmacogenetics research is evident through the establishment of the International HapMap Consortium ( and the Pharmacogenetics Research Network ( by the National Institutes of Health.

Another major consideration that will affect the integration of pharmacogenetics into clinical practice will be the cost-effectiveness of these approaches, which may be influenced by several factors (84). Drugs with a narrow therapeutic index with severe, expensive side effects are ideal candidates for pharmacogenetic testing. Drugs for which there are no established methods for monitoring adverse events (such as the biologic agents used in RA) are also most suited for pharmacogenomic analyses. However, for such approaches to be cost-effective, a well-established association should exist between genotype and clinical phenotype, and the frequency of the variant gene should be high. For example, if the frequency of a variant allele is only 0.5%, then ∼200 patients will have to be tested in order to identify 1 patient with the variant allele. Similarly, the strength of association between genotype and clinical phenotype will be important. If half the number of patients with a genetic variant have a serious adverse event from a drug, withholding the drug from all patients with the variant would unnecessarily deprive the rest of the patients of the drug.

Thus, several issues need to be considered before pharmacogenetics can be fully integrated into clinical practice. The ideal pharmacogenetic assay would quickly, accurately, and inexpensively provide composite genotypes for an individual patient in order to allow selection of the best drug for that patient. Although such assays are not yet readily available in clinical practice, ongoing research in this field in rheumatology is sure to bring one of the promises of the human genome project to fruition soon, that being individualized drug therapy.