TPMT & NUDT15 genotype-guided prediction of severe azathioprine / 6-MP myelosuppression — from blood draw to CPIC dosing recommendation
Azathioprine and 6-mercaptopurine (6-MP) remain first-line immunosuppressants for inflammatory bowel disease, autoimmune hepatitis, and a backbone of maintenance chemotherapy for pediatric acute lymphoblastic leukemia (ALL). Their therapeutic window is narrow: the same thioguanine nucleotide metabolites that kill leukemic blasts or dampen autoreactive lymphocytes will just as readily kill bone marrow precursors if they accumulate unchecked. Since 2004 (TPMT) and 2019 (NUDT15), FDA labeling and CPIC guidelines have recommended genotyping before the first dose rather than discovering deficiency through a neutropenic fever admission.
Azathioprine is a prodrug, non-enzymatically cleaved by glutathione-S-transferase to release 6-mercaptopurine (6-MP) in vivo. 6-MP itself sits at a metabolic branch point: one path bioactivates it into cytotoxic thioguanine nucleotides (TGN) that get incorporated into DNA and RNA; two independent detoxification enzymes — thiopurine S-methyltransferase (TPMT) and NUDT15 (nucleotide diphosphatase, also called MTH2) — divert or sanitize this flux before it becomes lethal to rapidly dividing bone marrow precursors.
When either enzyme is genetically deficient, standard weight-based dosing (azathioprine 2–3 mg/kg/day, or 6-MP 75 mg/m²/day) delivers a functional overdose: TGN accumulates 2–10× higher than in normal metabolizers, and severe, sometimes fatal, myelosuppression can occur within the first 4–8 weeks of therapy — well before a routine CBC would normally flag it.
• TPMT-deficient patients (homozygous poor metabolizers, ~0.3% of most populations) who receive standard doses have a >90% risk of severe, life-threatening myelosuppression • NUDT15-deficient patients (homozygous *3/*3, up to ~2% of East Asian and Hispanic populations) show an equally severe phenotype but were undetectable by TPMT testing alone until the gene was characterized in 2014–2016 • Because both enzymes act on independent, non-redundant steps of the same pathway, a patient can be TPMT-normal but NUDT15-deficient (or vice-versa) and still be at high risk — neither test alone is sufficient
Modern pharmacogenomic-enabled health systems embed TPMT/NUDT15 genotyping directly into the thiopurine prescribing order set:
1. Clinical trigger: gastroenterologist or oncologist places an order for azathioprine, 6-MP, or 6-thioguanine 2. Automatic reflex: the EHR clinical decision support (CDS) fires a pre-emptive pharmacogenomic panel order if no prior result exists on file 3. Specimen: 3–4 mL peripheral blood in an EDTA (purple-top) tube — germline DNA is stable, so testing does not need to be repeated once on record 4. Turnaround: 3–5 business days for a targeted genotyping panel; result is banked in the EHR for life, since germline genotype does not change 5. Dosing hold: for elective indications (IBD induction), the first dose is typically deferred until results return; for urgent oncologic indications, empiric dosing may begin with early CBC surveillance while genotyping is pending
This pre-emptive model — testing once and reusing the result for every future thiopurine prescription — is now standard practice at large pharmacogenomics-implementing centers such as St. Jude Children's Research Hospital (PG4KDS program) and the Vanderbilt PREDICT program.
Genomic DNA extracted from the EDTA sample is screened for the handful of common reduced-function variants that account for the vast majority of clinically relevant TPMT and NUDT15 deficiency. Most clinical laboratories use real-time PCR allelic-discrimination (TaqMan) assays for speed and cost, reflexing to targeted next-generation sequencing or Sanger confirmation when an unusual or ambiguous signal is seen.
TPMT is inherited as an autosomal co-dominant trait; roughly 90% of most populations are homozygous wild-type (normal metabolizers, *1/*1), ~10% are heterozygous (intermediate metabolizers), and ~0.3% are homozygous or compound heterozygous for two reduced-function alleles (poor metabolizers).
• TPMT*2 (rs1800462, c.238G>A, p.Ala80Pro) — rare, reduced activity • TPMT*3A (c.460G>A + c.719A>G, p.Ala154Thr + p.Tyr240Cys) — the most common reduced-function haplotype in European-ancestry populations, allele frequency ~3–5% • TPMT*3B (c.460G>A only) and TPMT*3C (c.719A>G only, rs1142345) — TPMT*3C predominates in East Asian and African populations • TPMT*4, *8 and other rare alleles — captured only by broader sequencing panels
NUDT15 deficiency, discovered more recently (Yang et al., Nature Genetics 2014; Moriyama et al., Nature Genetics 2016), is far more clinically important in Asian and Hispanic populations than TPMT:
• NUDT15*3 (rs116855232, c.415C>T, p.Arg139Cys) — the dominant reduced-function allele, ~10% allele frequency in East Asians, ~4% in Hispanics/Latinos, <1% in Europeans and Africans • NUDT15*2 and *5/*6 — additional reduced-function variants captured on expanded panels • Homozygous NUDT15*3/*3 individuals (~2% of East Asians) are poor metabolizers who tolerate only ~10% of standard thiopurine dose
TaqMan SNP genotyping uses two fluorescently labeled allele-specific probes (FAM for the reference allele, VIC/HEX for the variant allele) in a single competitive real-time PCR reaction:
• Both probes anneal to the target SNP locus; only the perfectly matched probe is cleaved by Taq polymerase's 5′ exonuclease activity during extension, releasing its fluorophore from a quencher • End-point fluorescence intensity is plotted as an FAM-vs-VIC scatter: homozygous reference clusters high-FAM/low-VIC, homozygous variant clusters low-FAM/high-VIC, and heterozygotes cluster at an intermediate diagonal position • No-template controls and known genotype controls are run on every plate to validate allele calling
When a specimen shows an ambiguous or borderline fluorescence signal, or a phenotype/genotype mismatch is suspected clinically, the laboratory reflexes to targeted amplicon-based next-generation sequencing (covering the full TPMT and NUDT15 coding sequence at ~500× depth) or Sanger sequencing to resolve rare or novel variants that panel-based SNP genotyping would otherwise miss. Some reference laboratories now run NGS as the primary platform, since it simultaneously captures common star alleles and rare loss-of-function variants in one assay.
Understanding thiopurine toxicity risk requires tracing the biochemical fate of 6-MP through three competing enzymatic routes. TPMT and NUDT15 do not act on the same molecule at the same step — they intercept the pathway at two different, non-redundant points — which is exactly why a patient must be genotyped for both genes rather than either one alone.
Once inside a cell, 6-MP can be shunted down three enzymatic pathways that compete for the same substrate pool:
1. Activation (HGPRT pathway): hypoxanthine-guanine phosphoribosyltransferase converts 6-MP to thioinosine monophosphate (TIMP), which is further metabolized by IMPDH and GMPS into thioguanine nucleotides (TGN: thioGDP, thioGTP, thio-dGTP). Thio-dGTP is incorporated into DNA in place of guanine during replication, triggering futile mismatch-repair cycling, DNA strand breaks, and apoptosis — this is both the intended cytotoxic mechanism in leukemic blasts and the unintended mechanism of bone-marrow toxicity.
2. Inactivation (TPMT pathway): thiopurine S-methyltransferase, using S-adenosylmethionine (SAM) as methyl donor, methylates 6-MP and TIMP into 6-methylmercaptopurine (6-MeMP) and methyl-TIMP — inactive metabolites that cannot be incorporated into DNA. This is the dominant detoxification route for 6-MP itself, and TPMT activity is what is measured by the classic RBC TPMT enzyme assay (radiochemical, normal >25 U/mL packed RBC).
3. Inactivation (xanthine oxidase pathway): hepatic and intestinal xanthine oxidase (XO) oxidizes 6-MP directly to 6-thiouric acid, an inactive metabolite excreted renally — this is why co-administration of allopurinol (an XO inhibitor) forces more substrate down the TGN-generating HGPRT pathway and requires an empiric 65–75% thiopurine dose reduction.
NUDT15 acts at a completely different point in the pathway than TPMT — after TGN has already been synthesized, not before. NUDT15 hydrolyzes thio-GTP and thio-dGTP to their monophosphate forms (thio-GMP, thio-dGMP), which cannot be used as substrates for DNA or RNA polymerases. In effect, NUDT15 is a nucleotide pool "sanitizing" enzyme, structurally related to the MutT/Nudix hydrolase family that removes oxidized or mismatched nucleotides from the triphosphate pool before they are incorporated into nucleic acids.
Because TPMT acts upstream (on 6-MP/TIMP, before TGN is made) and NUDT15 acts downstream (on TGN itself, after it is made), a patient can have completely normal TPMT activity but still accumulate cytotoxic thio-dGTP in DNA if NUDT15 is deficient — and vice versa. This is precisely why the two genes provide non-overlapping, additive risk information, and why the 2019 CPIC guideline update explicitly combines both genotypes into a single composite phenotype rather than treating them as interchangeable tests.
The Clinical Pharmacogenetics Implementation Consortium (CPIC) publishes an evidence-graded, machine-readable guideline that converts TPMT and NUDT15 diplotypes into discrete metabolizer phenotypes, then combines the two into a single starting-dose recommendation — the last computational step before a prescriber sees an actionable number.
Each gene's diplotype (the pair of star alleles inherited from each parent) is first translated independently into an activity phenotype:
TPMT phenotype: • Normal metabolizer (NM): *1/*1 — two fully functional alleles • Intermediate metabolizer (IM): one normal + one reduced-function allele (e.g., *1/*3A) — roughly 10% of most populations • Poor metabolizer (PM): two reduced-function alleles (e.g., *3A/*3A) — roughly 1 in 300 individuals
NUDT15 phenotype follows the identical logic using its own allele set: • Normal metabolizer: *1/*1 • Intermediate metabolizer: one normal + one reduced-function allele (e.g., *1/*3) • Poor metabolizer: two reduced-function alleles (e.g., *3/*3) — up to ~2% of East Asian and Hispanic patients, versus <0.1% of European-ancestry patients
Because RBC TPMT enzyme activity can also be measured directly (radiochemical or HPLC assay) as a phenotypic confirmation independent of genotype, some laboratories report a combined genotype+phenotype result, particularly useful in patients who have had a recent blood transfusion (which can transiently mask genotype-predicted enzyme activity in a phenotypic assay).
CPIC combines the TPMT and NUDT15 phenotypes into a single composite recommendation, since either gene alone being deficient is sufficient to markedly increase toxicity risk:
• Both NM (normal/normal): start at 100% of the standard target dose (e.g., azathioprine 2–3 mg/kg/day; 6-MP 75 mg/m²/day), titrate to efficacy and tolerance • One IM, other NM: reduce starting dose to roughly 30–80% of standard, titrate over 2–4 weeks based on CBC and, where available, TGN levels • Both IM, or one PM with other NM: reduce starting dose to approximately 10% of standard, given 3× weekly rather than daily, with dose titrated very slowly based on tolerance • Either gene PM (homozygous deficient), especially if the other is also reduced: consider an alternative non-thiopurine immunosuppressant; if a thiopurine must be used, start at drastically reduced dose (~10% or less) with intensive hematologic monitoring and therapeutic drug monitoring (TGN levels)
This composite table is what CPIC assigns its highest ("Strong") evidence grade to, reflecting consistent replication across multiple independent cohorts spanning European, East Asian, and admixed Hispanic populations.
A patient who is TPMT-intermediate AND NUDT15-intermediate is treated by CPIC as equivalent in risk to a patient who is homozygous poor metabolizer at a single gene — the two independent, non-redundant detoxification deficits are additive. This composite logic is the single most clinically important reason both genes must be tested together rather than substituting one test for the other.
A genotype-guided starting dose is not a "set and forget" prescription — it is the first data point in an iterative titration protocol built around complete blood count surveillance, with the genotype result setting the starting point rather than the final answer. Multiple prospective cohorts have now quantified exactly how much this pre-emptive approach reduces severe hematologic toxicity compared to empiric weight-based dosing alone.
Even with a genotype-informed starting dose, individual variability in absorption, comedication, disease activity, and unmeasured genetic factors means dosing is refined iteratively:
• Weeks 1–4: CBC with differential drawn weekly; absolute neutrophil count (ANC) and platelet count are the primary safety signals • Target ANC generally kept above 1,000–1,500/µL for IBD maintenance therapy, with protocol-specific lower thresholds tolerated in ALL maintenance chemotherapy where controlled myelosuppression is therapeutically intended • If ANC falls below threshold: dose is held and reduced by 25–50% upon recovery; if severe (ANC <500/µL) or accompanied by fever, thiopurine is held entirely and alternative therapy considered • Once stable, monitoring interval extends to every 1–3 months for the duration of therapy • Where available, TGN and 6-MeMP metabolite levels (measured by HPLC or LC-MS/MS on red blood cell lysate) provide a direct pharmacodynamic readout: therapeutic TGN range is roughly 235–450 pmol/8×10⁸ RBC for IBD; levels above this range correlate strongly with myelosuppression risk independent of genotype
Retrospective and prospective cohorts published over the past decade consistently show that pre-emptive TPMT/NUDT15 genotyping, paired with CPIC-guided starting doses, substantially reduces severe hematologic toxicity without compromising efficacy:
• Moriyama et al. (Nature Genetics, 2016) — the pivotal NUDT15 discovery and validation study in pediatric ALL cohorts, showing NUDT15*3/*3 patients tolerate only ~8% of the standard mercaptopurine dose, and that genotype-guided dose reduction eliminates the previously unexplained excess of severe leukopenia in East Asian and Hispanic patients • Relling et al. and the St. Jude PG4KDS program — pre-emptive genotyping embedded in the EHR at the point of prescribing, with documented reductions in dose-limiting toxicity episodes and fewer treatment interruptions • Multiple IBD cohorts (European and North American) — TPMT pre-emptive testing associated with significantly lower rates of severe leukopenia in the first 8 weeks of azathioprine therapy compared to historical un-genotyped controls
Collectively, more than 20 published cohorts across diverse ancestries now support the same conclusion: composite TPMT + NUDT15 genotyping, applied before the first dose, converts an unpredictable and occasionally fatal idiosyncratic toxicity into a proactively manageable, mostly preventable adverse drug reaction — one of the clearest success stories in translating pharmacogenomics into routine clinical practice.