Developed over 70 years ago as an anti-folate chemotherapy agent, methotrexate (MTX) is a WHO ‘essential medicine’ that is now widely employed as a first-line treatment in auto-immune, inflammatory diseases such as rheumatoid arthritis (RA), psoriasis and Crone's disease. When used for these diseases patients typically take a once weekly low-dose of MTX — a therapy which provides effective inflammatory control to tens of millions of people worldwide. While undoubtedly effective, our understanding of the anti-inflammatory mechanism-of-action of low-dose MTX is incomplete. In particular, the long-held dogma that this disease-modifying anti-rheumatic drug (DMARD) acts via the folate pathway does not appear to hold up to scrutiny. Recently, MTX has been identified as an inhibitor of JAK/STAT pathway activity, a suggestion supported by many independent threads of evidence. Intriguingly, the JAK/STAT pathway is central to both the inflammatory and immune systems and is a pathway already targeted by other RA treatments. We suggest that the DMARD activity of MTX is likely to be largely mediated by its inhibition of JAK/STAT pathway signalling while many of its side effects are likely associated with the folate pathway. This insight into the mechanism-of-action of MTX opens the possibility for repurposing this low cost, safe and effective drug for the treatment of other JAK/STAT pathway-associated diseases.
Originally published over 70 years ago, the use of anti-folates as a treatment for juvenile acute lymphocytic leukaemia represented a novel therapeutic approach that heralded the dawn of chemotherapy as an anti-neoplastic strategy . By replacing the pteridine ring hydroxyl group of folic acid with an amino group (Figure 1), Faber and colleagues had unwittingly generated a tightly binding, but reversible inhibitor of dihydrofolate reductase (DHFR), an enzyme at the heart of the folate pathway and an essential component in the production of the purines and pyrimidine nucleotides required for DNA replication and repair (Figure 2) . The search for analogues with lower toxicity and more straightforward synthesis soon led to the development of methotrexate (MTX) (Figure 1).
Structure of MTX-related molecules.
Interactions of MTX with the folate pathway.
In vivo both MTX and its extracellular metabolite, 7-hydroxy-MTX (Figure 1), are actively transported into cells by Folate Binding Protein and the Reduced Folate Carrier. Once intracellular, much of the MTX is polyglutamated to give MTXglu1–7 (Figure 2). Both MTX and MTXglun bind to and inhibit DHFR in a stoichiometric manner, with a Ki for MTXglu5 of 0.004 nM . In addition, MTXglu5 also inhibits the folate pathway components thymidylate synthase (Ki = 47 nM), glycinamide adenine ribonucleotide formyl transferase (Ki = 2500 nM) and aminoimidazole carboxamide ribonucleotide transferase (AICAR-T) with a Ki of 56 nM (Figure 2) . Although the much higher Ki's of these secondary interactions suggests that most folate pathway inhibition occurs via DHFR.
Although now increasingly superseded as a chemotherapy option, MTX has been an indispensable agent for the treatment of trophoblastic disease, gastric, bladder, and head and neck cancers. As part of combination therapies, MTX was central for the treatment of lymphoma and a mainstay of adjuvant therapy for breast cancer, as well as a more generally employed treatment of metastatic disease . Given as a bolus dose of up to 12 g/m2  high-dose MTX therapy for the treatment of cancer is an aggressive treatment with significant side effects often requiring intensive supportive care.
In addition to its rapid take-up in the field of oncology, a 1951 study reported a striking response in six of seven rheumatoid arthritis (RA) patients who demonstrated an improvement in inflammatory symptoms when treated with aminopterin . These observations were subsequently extended to patients with psoriasis and to the newly developed MTX analogue . Although adopted in the dermatology field for the treatment of psoriasis, the discovery of glucocorticoids and their anti-inflammatory effects eclipsed the nascent repurposing of MTX for RA for some years and even the effectiveness of intramuscular low-dose MTX at 10–15 mg per week in 25 of 29 patients  was not widely followed up. However, by the mid-1980s, many randomised, placebo-controlled studies in patients with refractory RA were underway and ultimately, significant and sustained clinical responses, improvements in functional status and favourable long-term retention rates were widely reported . Subsequently, large-scale studies comparing MTX to other therapies and therapy combinations, including biologics such as TNFα inhibitors, highlighted the effectiveness of the drug in both mono and combinatorial therapies .
Now widely used as a first-line treatment for RA, MTX is a mainstay in the rheumatology field with well-established safety and efficacy profiles and internationally developed treatment guidelines . In this context, MTX is generally dosed at between 7.5 and 25 mg once a week and experience over the last 20 years has demonstrated that supplementation of 1–5 mg/day of folic or folinic acid helps to control side effects — although folic acid supplementation is not advised on the day of the MTX dose so as not to saturate cellular folate transporters.
The ability of MTX to inhibit DHFR is central to its mechanism-of-action as an anti-neoplastic agent. With this in mind, and in the absence of alternative molecular mechanisms, folate pathway-related mechanisms-of-action have also been the focus of studies investigating its activity as an anti-inflammatory and DMARD (disease-modifying anti-rheumatic drug). Mechanisms including inhibition of purine and pyrimidine synthesis, suppression of transmethylation reactions, reduction in antigen-dependent T-cell proliferation, interference with glutathione metabolism leading to alterations in the recruitment of monocytes and the promotion of adenosine release have all been suggested .
Of these hypotheses, a reduction in inflammation via the increase in levels of extracellular adenosine has attracted the most study. In this model, polyglutamated MTX-mediated inhibition of aminoimidazole-carboxamide ribonucleotide transformylase (AICAR-T) leads to an increase in the intracellular levels of AICAR (Figure 2) — a molecule that is detectably increased in RA patients taking MTX. AICAR then inhibits AMP Deaminase leading to the release of intracellular adenine nucleotides (ATP) which are subsequently dephosphorylated to adenosine extracellularly. Extracellular adenosine binds to a family of G-protein-coupled adenosine receptors which then modulate the activity of neutrophils via as yet poorly understood mechanisms. While this relatively complex pathway has been studied quite extensively, little direct evidence is available to support the contention that adenosine represents the primary mechanism-of-action of low-dose MTX in vivo.
Can folate pathway inhibition really explain MTX DMARD activity?
Although a long-standing dogma within the field and the subject of significant research, the in vivo DMARD mechanism-of-action of MTX is not entirely clear. In particular, the suggestion that DMARD activity is a consequence of MTX-mediated inhibition of folate pathway enzymes raises many key questions. These include:
- Doses of MTX used in RA are between 1 and 0.1% of those used in chemotherapy, where DHFR inhibition is targeted. Typically, RA is dosed once a week at between 7.5 and 20 mg while chemotherapy protocols define high-dose as >1 g/m2 and range up to 12 g/m2 . Similarly, serum levels of MTX in chemotherapy peak at over 1500 µM , while peak serum levels of MTX measured in RA patients are between 0.4 and 0.8 μM [12,13]. Given these significant differences in dose, it seems unlikely that both regimes are functioning via the same inhibitory pathways in vivo.
- It is suggested that inhibition of AICAR-T by MTXglun reduces inflammation via an indirect mechanism that ultimately increases levels of extracellular adenosine. However, inhibition kinetics suggests that MTX preferentially inhibits DHFR over AICAR-T. For example, MTXglu5 has a Ki of 56 nM to AICAR-T and 0.004 nM to DHFR (Figure 2; ref. ) — a difference of almost 14 000 times.
- The half-life of MTX is ∼3–10 h and is undetectable in RA patient serum after 24 h . Similarly, the half-life of extracellular adenosine is <10 s . However, MTX in RA is dosed once weekly and its effects often take several weeks to develop following the start of treatment. The dichotomy of rapidly acting adenosine and long-term effects of MTX therapy does not appear to be mutually compatible.
- The consequence of adenosine receptor inhibition by the receptor antagonist caffeine on outcomes for MTX-treated patients is not always consistent with an adenosine mediated mechanism-of-action. While at least one study suggests that high caffeine intakes are associated with reduced MTX efficacy , other studies have found no significant effects . Were adenosines signalling central to the MTX DMARD mechanism of action one would imagine that caffeine would be strongly contra-indicated.
- Most clinical RA protocols include a folic or folinic acid supplement taken on non-MTX treatment days. Biochemically, this supplement bypasses the block in folate metabolism produced by DHFR inhibition by directly providing tetrahydrofolate, the product of DHFR activity in vivo . This suggests that antagonism cannot be central to the drugs anti-inflammatory activity.
- MTX side effects, such as mouth ulcers and nausea, are relieved by folic/folinic acid supplements. Furthermore, this symptomatic relief does not affect DMARD activity . This suggests that the mechanism-of-action for MTX as a DMARD is independent of the mechanism producing the side effects.
- Patients with SNPs in components of the folate pathway report increased MTX-related toxicities, but no loss of MTX DMARD activity. Using retrospective approaches in RA patients, studies have investigated polymorphisms in folate-related genes, including changes that modify MTX transporters and enzymes central to the metabolic pathway [17–19]. Strikingly, these genetic polymorphisms modify drug toxicity, but do not seem to influence the efficacy of MTX as an anti-inflammatory/DMARD. Similarly, a study in 374 patients with psoriasis, identified polymorphisms in four folate pathway-linked enzymes and showed that variations in these loci are not predictive of response to MTX therapy . As such, in both RA and psoriasis, it appears that side effects/toxicity to MTX and its effectiveness as an anti-inflammatory/DMARD are mediated by distinct and possibly separable molecular mechanisms.
Evidence for MTX as a JAK/STAT inhibitor
Conserved across evolution, JAK/STAT pathway signalling is central to a range of biological processes, including haematopoiesis, immunity and inflammation [21,22]. In the pathway, extracellular ligands bind to transmembrane receptors causing the activation of receptor-associated JAK tyrosine kinases. Subsequent phosphorylation of both receptor and JAKs create docking sites for cytosolic STAT molecules that are themselves activated by phosphorylation of an invariant C-terminal tyrosine reside that triggers STAT dimerisation, nuclear translocation, DNA binding and the expression of pathway target genes (Figure 3A and reviewed in [21,22]).
The JAK/STAT signalling pathway and predicted MTX interactions.
Using a cell-based Drosophila JAK/STAT pathway reporter assay previously validated in genome-scale RNAi screens , we screened a small molecule library of 2000 pure natural products, agrochemicals and FDA-approved drugs . This screen independently identified both aminopterin and MTX as inhibitors of Drosophila JAK/STAT pathway signalling. Furthermore, dosage-dependent inhibition was confirmed for the two compounds following pathway stimulation by both ligand and gain-of-function JAK alleles , suggesting that inhibition is being mediated at or below the level of the JAK kinase. Furthermore, RNAi mediated knock-down of folate pathway enzymes, including DHFR, Thymidylate Synthase and Glycinamide Adenine Ribonucleotide Formyl Transferase did not affect JAK/STAT pathway signalling — suggesting that MTX-mediated inhibition was independent of the folate pathway.
Consistent with this finding, independent groups have also used Drosophila screening approaches which identified MTX as a heterochromatin promoting drug and inhibitor of JAK/STAT signalling in vivo .
Following on from this initial identification in the Drosophila system we showed that the ability of MTX to inhibit JAK/STAT pathway signalling is maintained in human lymphoma-derived HDLM2 cells, reducing STAT1, STAT5, JAK1 and JAK2 phosphorylation levels. In addition, erythroid leukaemia-derived HEL cells carrying the gain-of-function JAK2 V617F mutation demonstrate reduced STAT3 and STAT5 phosphorylation following MTX treatment. In both cases, effects are dose-dependent and occur at concentrations consistent with the 0.4–0.8 μM peak levels of MTX detectable in the serum of RA patients [13,25]. In contrast, no change in the phosphorylation state of pAKT, pJun or pERK1/2 was detectable, suggesting that reduction in JAK and STAT phosphorylation is specific to this pathway and not the result of non-specific interactions .
Subsequent to this initial identification, in vivo analysis of MTX activity used a mouse model of polycythemia vera (PV) expressing the human JAK2 V617F gain-of-function mutation. Featuring increased haemoglobin and haematocrit levels, this disease model is driven by gain-of-function mutations in haematopoietic stem cells which constitutively activate JAK/STAT pathway signalling [27,28]. We found that treatment with MTX doses previously shown to alleviate murine induced RA do not reduce blood counts in wild type littermate controls — indicating that MTX does not cause non-specific myelosuppression. However, both the haematological and splenic overgrowth disease phenotypes in mutant mice are relieved by low-dose MTX, with results comparable to the JAK1/2 inhibitor Ruxolitinib used as a positive control. In addition, the level of phosphorylated STAT3 and STAT5, as well as the mRNA levels of the JAK/STAT target gene PIM1 are also reduced in drug-treated spleens .
Finally, molecular modelling was undertaken to examine the hypothesis that MTX could at as a competitive inhibitor of the JAK tyrosine kinase domain. Consistent with the in vivo data, in silico docking to a high-resolution model of the JAK2 tyrosine kinase domain suggested that MTX can occupy the ATP-binding pocket with a greater affinity than ATP itself, implying that MTX may act as a competitive kinase inhibitor (Figure 3B,C and ). It would be intriguing to extend these models to structurally related molecules and other JAK-kinases to explore this mechanism-of-action more thoroughly.
In addition to the cell-based, mouse model and in silico findings, data from patients also supports the hypothesis that MTX acts as a JAK/STAT pathway inhibitor. In particular, a recent transcriptomic analysis of CD4+ cells from DMARD-naïve RA patients randomised to tocilizumab plus MTX, tocilizumab or MTX therapy has proven insightful. In particular, analysis of differentially expressed mRNAs analysed using the KEGG database in the MTX arm identified p53 signalling (P = 8.44E–06) and JAK–STAT signalling (P = 2.22E–04) as the only significantly affected molecular pathways . This suggests that low-dose MTX is capable of inhibiting JAK/STAT pathway signalling in RA patients.
Finally, it should also be noted that a recent large-scale study of 4786 cardiac patients treated with low-dose MTX to reduce inflammation associated with atherosclerotic events showed a statistically significant increase in non-basal-cell-skin carcinoma from 0.2 events per 100 patient-years in the control group to 0.61 events per 100 patient-years in the MTX-treated group . Although a definitive cause for this increase will require further research, it is possible that this is a consequence of a reduction in JAK-mediated cancer immunosurveillance — a hypothesis that has also been proposed to explain similar effects following treatment with the JAK1/2 inhibitor ruxolitinib .
Why might JAK inhibition explain the DMARD activity of MTX?
Although further research will be required to dissect the details underlying MTX-mediated JAK/STAT inhibition, an increasing body of evidence generated by independent laboratories in a range of model systems consistently suggests that MTX is indeed a JAK/STAT pathway antagonist. What is perhaps most striking in the context of low-dose MTX, is the role played by the JAK/STAT pathway in the regulation of the immune system and signalling both upstream and downstream of inflammatory cytokines . For example, Interferon-gamma (IFNγ) and Interleukins (IL)-6 and 10 are all potent pro-inflammatory cytokines whose levels and activity are significantly increased in inflammatory disease and whose mechanism-of-action signals directly through receptors associated with JAKs and downstream STAT activation (Figure 3A and ).
Given this background, it is not surprising that the link between JAK/STAT pathway signalling and RA has been the subject of considerable study  with many drug development efforts having targeted the pathway. This has led to the development and licencing of the JAK1/3 inhibitor Tofacitinib  and Tocilizumab a blocking antibody targeting the receptor of the pro-inflammatory cytokine IL-6 [35,36]. Both therapies have proven to be effective in large-scale clinical trials and have now gained regulatory approval.
Implications for the repurposing of MTX
On the World Health Organisation's list of ‘Essential Drugs’, and with patent and intellectual property protections long since lapsed, MTX is available at very low cost via health care systems around the world and is taken weekly by millions of RA patients. It is generally effective and well-tolerated, with a well-characterised safety profile and high retention rates. Indeed, many patients remaining on MTX mono-therapy for years.
Given this background, the possibility that low-dose MTX may be acting via inhibition of JAK/STAT signalling provides insight into the previously ambiguous molecular mechanism-of-action and opens up the possibility of drug repurposing strategies.
Proposed as a rapid, cost-effective and efficient way of identifying treatments for novel indications, drug repurposing, repositioning and rescue efforts generally seek to use existing drugs, with known safety, pharmacodynamics and pharmacokinetic properties for the treatment of new diseases. These efforts are often based on novel insights into drug mechanism-of-action, improved understanding of the molecular basis of disease, serendipitous observations or targeted analysis of ‘big data’, including in silico modelling and machine learning-based approaches .
In the light of the discovery that MTX acts as a JAK/STAT pathway inhibitor, repurposing to treat other diseases associated with inappropriate JAK/STAT activation is a logical next step. While inappropriate JAK/STAT signalling has been associated with inflammatory disease, immune-dysregulation and multiple malignancies , perhaps the best-characterised examples of pathway-associated diseases are the myeloproliferative neoplasms (MPNs). This group of rare blood cancers including polycythemia vera (PV), featuring an increased erythrocyte levels, essential thrombocythemia (ET), characterised by increased platelet counts and myelofibrosis (MF) in which bone marrow exhaustion and fibrosis leads to pan-cytopaenia. Strikingly, almost all MPN patients carry clonally related haematopoietic stem cell mutations that constitutively activate the JAK/STAT pathway at the level of the receptor or the JAK kinase itself . Given the relatively recent molecular identification of the pathway activated in MPNs, the use of pathway inhibitors as a potentially valuable therapeutic approach has been rapidly tested and the JAK1/2 inhibitor Ruxolitinib is now licenced for MF and a subset of PV patients . However, side effects and the very high cost of this on-patent treatment have limited availability in many healthcare systems.
Given the role of JAK/STAT signalling in MPNs and the identification of MTX as a pathway inhibitor, we have therefore proposed that MTX could be repurposed as a therapy to treat this group of diseases [23,39]. Supportive of this proposal, mouse-based models of PV respond well to low-dose MTX  and preliminary studies suggest effectiveness in PV and ET patients [40,41]. Although clinical trials to demonstrate safety and efficacy in the MPN patient population are still needed, it will be fascinating in the longer term to see whether discovering the mechanism-of-action of MTX will ultimately lead to the development of new treatments for this significant patient population.
Knowledge of the mechanism-of-action of existing drugs is an essential prerequisite that allows us to identify repurposing opportunities for the treatment of other diseases.
The fact that high-dose MTX acts as an inhibitor of DHFR has dominated thinking about its use as an anti-inflammatory drug in RA — despite RA dosing being 1/100th to 1/1000th of chemotherapy levels and unlikely to act via this mechanism.
Recent insights suggest that low-dose MTX is more likely to exert its anti-inflammatory effects via inhibition of the JAK/STAT pathway. A pathway central to inflammation and a pharmacological target of significant potential interest.
The authors declare that there are no competing interests associated with the manuscript.
Funding for the Zeidler laboratory was provided by an MRC Confidence in Concept award and a clinical PhD studentship from Cancer Research UK/Yorkshire Cancer Research. Adel Alqarni is supported by a fellowship from the Prince Sattam Bin Abdulaziz University and the Kingdom of Saudi Arabia.
The authors wish to thank Dr Sally Thomas and Dr Sebastian Francis for stimulating discussions and valuable patient insights.
Aminoimidazole Carboxamide Ribonucleotide
Aminoimidazole Carboxamide Ribonucleotide Transferase
disease-modifying anti-rheumatic drug
formamidoimidazole carboxamide ribonucleotide
small nucleotide polymorphisms