NUDT15 Hydrolyzes 6-Thio-DeoxyGTP to Mediate the Anticancer Efficacy of 6-Thioguanine

Interfering with nucleotide metabolism is one of the most successful treatment strategies against cancer. Nucleoside analogues target rapidly proliferating cancer cells by disrupting processes related to RNA and DNA synthesis. Even after half a century of clinical research (1), thiopurines remain one of the most effective maintenance therapies against childhood leukemia [acute lymphoblastic leukemia (ALL); ref. 2] and are also used as anti-inflammatory and immunosuppressant drugs (3).

Three thiopurines are used in the clinic: 6-thioguanine (6-TG), azathioprine (AZA-T), and 6-mercaptopurine (6-MP). The metabolism of thiopurines is complex and involves a number of enzymes and intermediates that may improve or impair effective treatment (4). The nucleoside triphosphates, 6-thio-GTP and 6-thio-dGTP, are the primary active metabolites. 6-Thio-dGTP is a substrate for DNA polymerases leading to substitution of 0.01% to 0.1% of canonical guanine bases for thiobases (5–8). Incorporation of 6-thio-dGTP is neither particularly toxic nor mutagenic (6, 9–11). However, methylation, most likely by S-adenosylmethionine (SAM; ref. 12), and a second round of replication generate the Me-6-thio-dG:T mispair, which is detected by the mismatch repair (MMR) machinery (12, 13). Processing of this lesion presumably leads to futile attempts of repair, irreparable DNA damage, and cell death (14–16). Because this cascade requires 2 rounds of DNA replication, the antiproliferative effects of thiopurines are noticeably delayed (14). MMR-deficient cells are significantly less sensitive to thiopurines (14), yet residual sensitivity in these cells indicates that thiopurines may act via additional mechanisms. The complexity of thiopurine metabolism complicates treatment regimens and imposes a restrictive therapeutic window that requires close monitoring (2, 3, 10).

In searching for novel thiopurine sensitivity factors, several research groups have identified a missense variant in NUDT15 (rs116855232) that results in an arginine to cysteine substitution at position 139 (R139C) and thiopurine intolerance. Patients with inflammatory bowel disease (IBD) having this mutation showed increased side effects after thiopurine treatment (17, 18). Accordingly, this NUDT15 variant was found to influence the sensitivity to 6-MP in children with ALL (19–21). NUDT15 [nudix hydrolase 15, also known as MutT homolog 2 (MTH2)] belongs to the NUDIX hydrolase family and was initially described as an oxidized nucleotide sanitation enzyme, similar to MTH1 (NUDT1 or nudix hydrolase 1; ref. 22). A comparison of MTH1 and NUDT15 revealed that NUDT15 had minimal activity toward oxidized nucleotides in vitro and in cells. However, a substrate screen identified that NUDT15 could hydrolyze 6-thio-dGTP and 6-thio-GTP, which, like the pharmacogenetic studies, suggested a possible role for NUDT15 in thiopurine metabolism (23).

Herein, we present that NUDT15 is a key enzyme in thiopurine metabolism and tempers the activity of thiopurine drugs by catalyzing the hydrolysis of 6-thio-dGTP and 6-thio-GTP. We resolve the first crystal structure of NUDT15 in complex with a reaction product, 6-thio-GMP, thereby elucidating key determinants for the observed substrate selectivity. With in vitro and cellular experiments, we explore the activity of NUDT15 R139C and find that the protein maintains catalytic activity but is unstable under physiologic conditions. In cells, NUDT15 R139C is rapidly degraded, likely causing thiopurine sensitivity in patients with this mutation. In line with this, depletion of the protein with NUDT15-specific RNAi in cancer cells increases the sensitivity to 6-TG. Altogether, our data suggest that NUDT15 is integral to thiopurine metabolism and acts as a barrier to therapeutic efficacy.

The construct pNIC28hNUDT15 for bacterial expression of NUDT15 was a gift from the Structural Genome Consortium (Stockholm, Sweden). NUDT15 wild-type (WT) was expressed and purified as earlier described (23). Site-directed mutagenesis for the R139C, R139S, R139A, and R139K mutants was performed as described by Li and colleagues (24). The mutants were expressed from pNIC28 in BL21DE3 at 37°C and grown for 4 hours after induction by 1 mmol/L isopropyl β-d-1-thiogalactopyranoside (IPTG) before harvesting by centrifugation. Bacteria were lysed using Bugbuster (Merck-Millipore), fortified by benzonase (2.5 U/mL, Merck-Millipore), and cOmplete Mini, EDTA-free protease inhibitor (Roche) was added to the cleared lysate. NUDT15 R139C was purified using HisTrap HP (GE Healthcare) and 100 mmol/L HEPES (pH 7.5), 500 mmol/L NaCl, 10 mmol/L imidazole, and 10% glycerol as starting buffer. Proteins were eluted using an imidazole gradient (0–500 mmol/L), and the His-tag was removed by using TEV protease and passing the protein over a HisTrap HP column. Following dialysis with ion exchange chromatography starting buffer [20 mmol/L HEPES (pH 7.5), 20 mmol/L NaCl, 10% glycerol], NUDT15 R139C was further purified by MonoQ HP (GE Healthcare) and eluted with a NaCl gradient (10–500 mmol/L). All purification was performed in the absence of reducing agent. SDS-PAGE and Coomassie staining confirmed purity of the proteins, and protein concentration was determined by NanoDrop (Thermo Fisher Scientific) A₂₈₀ measurement.

High-performance liquid chromatography (HPLC) analysis was performed similar to previously reported conditions (25). Briefly, NUDT15 WT (10 nmol/L) was incubated with 6-thio-GTP or 6-thio-dGTP (50 μmol/L) in buffer containing 100 mmol/L Tris-HCl (pH 7.5), 40 mmol/L NaCl, 10 mmol/L MgCl₂, 1 mmol/L dithiothreitol (DTT), and 0.005% Tween-20. Mixtures were incubated at 37°C and stirred using magnetic stir bars. At each time point, 40-μL reaction mixture was added to 60 μL of −20°C methanol. Samples were stored at −20°C for at least 20 minutes before centrifugation. The supernatant (90 μL) was moved to fresh 1.5-mL tubes and evaporated by vacuum centrifugation at 60°C. The samples were resuspended in 90 μL ddH₂O, moved into a sealed 96-well plate, and loaded for HPLC analysis (samples were kept at 4°C prior to injection).

NUDT15-mediated hydrolysis of 5 μmol/L 6-thio-dGTP, 6-thio-GTP, dGTP, and GTP was monitored after 10, 20, and 30 minutes of incubation at 22°C in assay buffer (100 mmol/L Tris-acetate, pH 7.5, 40 mmol/L NaCl, 10 mmol/L magnesium acetate, 1 mmol/L DTT) using 4 nmol/L NUDT15 WT. For determination of kinetic parameters, initial rates were determined using reaction buffer, 4 nmol/L NUDT15 WT or R139C and concentrations of the substrates ranging from 0 to 40 μmol/L for 6-thio-dGTP and 6-thio-GTP and 0 to 50 μmol/L for dGTP and GTP (or 0–400 μmol/L for R139C). NUDT15 R139C reactions were performed in the absence or presence of 1 mmol/L DTT. Formed pyrophosphate (PPi) was detected using PPiLight Inorganic Pyrophosphate Assay (Lonza) as described previously (23).

Full-length NUDT15 (15 mg/mL) was crystallized with 2 mmol/L 6-thio-GTP in 15 mmol/L HEPES, 225 mmol/L NaCl, 7.5% glycerol, and 1.5 mmol/L TCEP (pH 7.5). Hanging drop vapor diffusion experiments at 4°C were performed, and NUDT15 was mixed with reservoir solution (30% PEG3350, 0.1 mol/L Tris, pH 8.5, and 0.24 mol/L MgCl₂) in a 1:1 ratio. Diffraction quality crystals appeared overnight and grew to full size within 3 days. The crystals were extracted quickly without additional cryoprotectant and flash-frozen in liquid nitrogen. Data collection was performed at beam line ID29 at ESRF, Grenoble, at 100 K and wavelength 1.072 Å equipped with a Pilatus 3 6M detector (Dectris). Data reduction and processing were carried out using Mosflm (25) and Aimless (26) program from the CCP4 suite (27). The structure was solved by molecular replacement of the template structure file with PDB ID 5BON using Phaser (28). The resultant models were refined using Refmac5 (29) and Arp/wARP (30) was used for initial addition of the water molecules. Manual adjustments of the model were carried out using Coot (31). Validation was conducted with Molprobity (32). Relevant statistics can be found in Supplementary Table S1. All figures were drawn with PyMOL (Schrödinger, LLC). The coordinates and structure factors were deposited in the PDB with the accession code 5LPG.

Cells were cultured in a humidified incubator at 37°C with 5% CO₂. HCT116 and HCT116 3-6 human colon carcinoma cells were obtained from Dr. Bert Vogelstein (2001, Johns Hopkins, Baltimore, MD) and HEK293T transformed human embryonic kidney cells from ATCC (2010). For HCT116 and HCT116 3-6, McCoy 5A GlutaMAX (Life Technologies) was used and for HEK293T DMEM, GlutaMAX (Life Technologies). All media were supplemented with 10% FBS, penicillin (50 U/mL), and streptomycin (50 μg/mL). Mycoplasma contamination was screened using the MycoAlert Mycoplasma Detection Kit (Lonza).

Doxycycline hydrochloride (Sigma-Aldrich) was dissolved in MilliQ H₂O and used at 1 μg/mL. 6-TG (Sigma-Aldrich) was dissolved in DMSO to a stock concentration of 10 mmol/L immediately before use and was protected from light. MG-132 (Z-Leu-Leu-Leu-al, Sigma-Aldrich) was dissolved in DMSO.

HEK293T cells were transfected with lentiviral plasmids as described before (33). Selection for Tet-pLKO-puro–containing cells was achieved by using 1 μg/mL puromycin (Sigma-Aldrich) and pInducer20-containing cells with 400 μg/mL neomycin (G418, Sigma-Aldrich).

Cells were lysed and prepared for Western blotting as described previously (23). Primary and secondary antibodies are listed in Supplementary Methods. IRDye secondary antibodies (LI-COR) were used, and blots were visualized and analyzed using an Odyssey Fc Imager and Image Studio Software (LI-COR).

Protein unfolding was detected by differential scanning fluorimetry (DSF; ref. 34). The conditions used were 4 μmol/L enzyme in assay buffer and SYPRO Orange (Thermo Fisher Scientific) with or without 1 mmol/L TCEP. A CFX96 Touch Real-Time PCR Detection System (Bio-Rad) was used to increase the temperature from 25°C to 95°C in 1°C/min increments, and fluorescence intensity was measured at each step. The melting temperature (T_m) was calculated by CFX Manager Software (Bio-Rad).

Clonogenic survival experiments were based on those from Meyers and colleagues (35). Cells were treated with doxycycline for 48 hours and then replated as single cells. The following day, 6-TG or DMSO was added (containing doxycycline). Medium was replaced every 3 days containing fresh 6-TG/DMSO and doxycycline. After 9 days, the colonies (>50 cells) were fixed/stained with methylene blue (4 g/L) in methanol and scored by eye. The plating efficiencies (PE) were determined from control cells (DMSO-treated) and used to calculate the surviving fraction (SF) after 6-TG treatments.

Site-directed mutagenesis primers, HPLC instrument settings, circular dichroism, size exclusion chromatography, cloning of lentiviral constructs, RT-qPCR, primary antibodies, siRNA transfection, and resazurin assay methods can be found in Supplementary Methods.

With substantial clinical evidence and preliminary biochemical data supporting a role for NUDT15 in thiopurine metabolism (23), we conducted thorough biochemical analyses with NUDT15 WT and the protein derivative R139C expressed and purified from bacterial lysates under nonreducing conditions (Supplementary Fig. S1). To analyze the formed products, NUDT15 activity toward 6-thio-dGTP and 6-thio-GTP was assessed by HPLC. NUDT15 was able to hydrolyze 6-thio-dGTP and 6-thio-GTP to their corresponding monophosphate species (Fig. 1A). Interestingly, 6-thioguanosine was also detected as a product, albeit in minor amounts. Thus, it appears that NUDT15 may have some ambiguity with respect to which phosphate bond is hydrolyzed in thioguanosine triphosphates.

We measured time-dependent hydrolysis of dGTP at 5 μmol/L, close to reported cellular concentrations (36, 37), as well as with GTP, 6-thio-dGTP, and 6-thio-GTP by analyzing PPi formation. Pronounced hydrolysis of 6-thio-dGTP and 6-thio-GTP, but undetectable activity with dGTP and GTP, was observed (Supplementary Fig. S2), indicating a preference for the thionylated substrates.

We performed in-depth kinetic analyses with increasing concentrations of 6-thio-dGTP, dGTP, 6-thio-GTP, and GTP to better understand the substrate preferences of NUDT15 WT and R139C, using the PPiLight assay. NUDT15 showed higher affinity for thioguanosine triphosphates over canonical guanine substrates with K_M = 1.9 and 1.8 μmol/L for 6-thio-dGTP and 6-thio-GTP, respectively, compared with 43 and 254 μmol/L for dGTP and GTP, respectively (Fig. 1B and Table 1). Catalytic turnover (k_cat) was similar for all substrates (Table 1). A comparison of the catalytic efficiencies (k_cat/K_M) shows that 6-thio-dGTP and 6-thio-GTP are preferred substrates for NUDT15 over canonical guanine substrates, with 243- to 290-fold higher catalytic efficiency than for GTP and 13- to 15-fold higher efficiency than for dGTP, respectively (Table 1).

Because patients with the R139C mutation are particularly sensitive to thioguanine treatments, we hypothesized that the enzyme is unable to hydrolyze thioguanosine triphosphates efficiently. Surprisingly, NUDT15 R139C could also hydrolyze all species with similar activity as the WT protein when activity was measured under reducing conditions at 22°C (Fig. 1C). Kinetic analyses indicated similar K_M values and catalytic turnover rates for WT and mutant NUDT15 (Table 1). The resultant catalytic efficiency was only marginally affected by the R139C mutation with respect to the thioguanine nucleotides. In fact, NUDT15 R139C had slightly higher activity than WT with 6-thio-dGTP as the substrate and only 3-fold lower catalytic efficiency with 6-thio-GTP.

To understand the observed substrate specificity of NUDT15 for thioguanine substrates, we crystallized NUDT15 in the presence of Mg²⁺ and 6-thio-GTP. We determined the 1.7 Å resolution structure of NUDT15 in complex with the product, 6-thio-GMP, and Mg²⁺ (Fig. 2A and B). The overall structure is highly similar to the NUDT15 apo structure with RMSDs between 0.34 and 0.6 Å. There are no major structural differences, except in the loop formed by residues Arg34-Gly41 (Supplementary Fig. S3). Molecule A of the NUDT15 dimer has bound the hydrolysis product 6-thio-GMP and one Mg²⁺ ion coordinated by NUDIX box residues (Fig. 2C). Molecule B of the dimer did not bind 6-thio-GMP and instead displayed coordination of 4 Mg²⁺ ions as previously observed in the apo structure (23). Hydrogen bonding between the thioguanine base and the backbones of Leu45 and Gly137 coordinate the substrate in the active site (Fig. 2D). The α-phosphate of 6-thio-GMP forms one direct hydrogen bond to His49. In addition, multiple water-mediated hydrogen bonds between 6-thio-GMP and surrounding residues contribute to the binding (Fig. 2C and D). The 6-thio moiety is well accommodated in a hydrophobic pocket formed by Phe135, Leu138, Gln44, and Leu45 of NUDT15, explaining the lower K_M values for 6-thio-GTP and 6-thio-dGTP compared with the canonical guanosine triphosphates (Table 1). The steric fit and the hydrophobic interactions of the sulfur further stabilize the binding (Fig. 2D).

We next sought to understand how R139C protein can be active in vitro but induce thiopurine sensitivity in its carriers. Therefore, we overexpressed HA-tagged WT or R139C NUDT15 using doxycycline-inducible expression constructs in HCT116 cells, which are carrying endogenous WT NUDT15. Overexpression of the HA-tagged proteins was assessed with an anti-HA or anti-NUDT15 antibody. When analyzing protein levels, overexpressed NUDT15 WT was robustly induced upon doxycycline addition, but the overexpressed R139C mutant was barely detectable (Fig. 3A). No alterations of endogenous NUDT15 were observed. This was not a time-dependent event, as overexpressed WT and R139C protein did not fluctuate over the course of 72 hours (Fig. 3B). Interestingly, measurements of total NUDT15 mRNA showed roughly equal expression after induction by doxycycline over 72 hours (Fig. 3C), indicating a stark deviation from the protein levels. This suggested that something inherent to the R139C protein was limiting its accumulation.

In the cell, chaperone proteins first mediate the fate of structurally unstable or misfolded proteins, but, ultimately, chronically misfolded proteins are either degraded by autophagy or the proteasome system to prevent detrimental aggregation (38). To determine whether NUDT15 R139C is degraded by the proteasome, we performed a time course experiment with the proteasome inhibitor, MG-132. After 72 hours of doxycycline-induced overexpression of NUDT15 WT or R139C, doxycycline was removed. MG-132 (5 μmol/L) was added for the indicated time points starting with the longest treatment and all samples were harvested at the same time point (Fig. 3D). Accrual of p53, which rapidly accumulates following proteasome inhibition, was monitored as a control (39). While inhibition of the proteasome did not lead to increased accumulation of overexpressed NUDT15 WT protein, treatment with MG-132 led to a clear and distinct accumulation of R139C in a time-dependent manner (Fig. 3D). Thus, while the NUDT15 R139C mutant is expressed normally in cells, the protein is rapidly degraded.

Because NUDT15 R139C showed activity under reducing conditions at 22°C in vitro but little protein was detected in a cellular setting, we investigated protein stability in vitro using a DSF assay and compared nonreducing and reducing conditions (Fig. 4A and B). NUDT15 WT had a melting point (T_m) of 59°C and 58°C with or without TCEP, respectively. This was much higher than the NUDT15 R139C mutant, which had a melting point of 48°C without TCEP and 51°C with TCEP. Interestingly, NUDT15 R139C showed higher background fluorescence levels, which indicates that the SYPRO Orange has more hydrophobic surface to interact with (Fig. 4A). This may suggest that the protein exists in a more open form, where more of the NUDT15 R139C interior is accessible, and could be due to alterations in secondary structure or disruption of protein dimerization. Circular dichroism experiments, however, indicated that the R139C secondary structures did not deviate significantly from WT protein (Supplementary Fig. S4A). In addition, R139C was still able to form a homodimer (23), as seen by size exclusion chromatography (Supplementary Fig. S4B). Thus, NUDT15 R139C is less stable than the WT protein.

Because NUDT15 R139C melting temperature is slightly influenced by nonreducing conditions, we explored potential sources of instability. Interestingly, a second cysteine residue is located adjacent to residue 139 at position 140, which could poise these residues to form a disulfide bridge.

If a disulfide bond occurs, we might expect a decrease in NUDT15 R139C enzymatic activity under nonreducing conditions. To test this, NUDT15 R139C was subjected to kinetic analysis in the absence or presence of reducing agent (Supplementary Fig. S5A and S5B). Surprisingly, the R139C mutant maintained activity under nonreducing conditions, and there was only a 2.3- to 2.5-fold increase in catalytic efficiency when DTT was added (Supplementary Fig. S5C). Thus, with little difference between the activity of WT and R139C NUDT15 under reducing conditions, it was not apparently clear, from a biochemical standpoint, what contribution a potential disulfide bond has to R139C stability.

Residue 139 is contained within helix α2, whose N terminus makes up the base of the active site (Fig. 2B and C). We hypothesized that perturbations in this helix could influence the stability of NUDT15. Arg139 is able to make key hydrogen bonds within helix α2 (Glu143) and a potentially crucial ionic interaction with Asp132, which is adjacent to helix α2 (Fig. 4C). Deviations from arginine at position 139 could, therefore, have significant structural implications. In addition, Arg139 is proximal to the negatively charged absolute C-terminus (Leu164), further requiring a positive charge to stabilize this network (Fig. 2B). To assess whether the loss of Arg139 or the introduction of a cysteine disturbs NUDT15 protein stability, the mutants R139A, R139S, R139C, and R139K were produced under nonreducing conditions and their stability was assayed using DSF (Supplementary Fig. S1 and Fig. 4D and E). Compared with WT, only the R139K mutant had a similar melting point (T_m = 54.5°C). This can be explained by the fact that lysine is able to make the same intramolecular ionic bond as arginine (Fig. 4C and Supplementary Fig. S6). The R139C, R139S, and R139A mutants had significantly lower melting points (T_m = 46°C, 44.9°C, and 44.5°C, respectively). While their contributions to hydrogen bonding within helix α2 vary, each of these residues would be unable to make the ionic connection to Asp132, implying that this interaction is important for NUDT15 stability. Taken together, the data suggest that R139 has a stabilizing effect on helix α2 in NUDT15, and the alteration of this residue is likely to contribute to inherent instability.

Because our in vitro experiments showed activity of NUDT15 with 6-thio-dGTP and 6-thio-GTP, we wanted to investigate whether depletion of the protein increases efficacy of 6-TG in a cellular setting, similar to the conditions expected in patients carrying the R139C mutant. Therefore, we depleted NUDT15 in HCT116 and HCT116 3-6 cells using doxycycline-inducible NUDT15-specific shRNA (shNUDT15) and compared it with a nontargeting shRNA control. HCT116 cells and their isogenic, MMR-proficient counterpart, HCT116 3-6 (40), have been used extensively to study MMR-dependent mechanisms, such as those involving thiopurines (41). shNUDT15 efficiently silenced endogenous NUDT15 mRNA and protein in both cell lines after a 72-hour incubation with doxycycline (Fig. 5A and B). Subsequently, clonogenic survival was assessed with increasing doses of 6-TG. As expected, MMR-deficient HCT116 cells were more resistant to 6-TG, whereas MMR-proficient HCT116 3-6 cells were more sensitive. Depletion of NUDT15 dramatically increased the sensitivity of HCT116 3-6 cells to 6-TG and, to a lesser extent, parental HCT116 cells (Fig. 5C). This sensitization was also seen with 3 different siRNA oligos against NUDT15 by resazurin assay, which confirmed target specificity (Supplementary Fig. S7).

Futile repair cycling and prolonged G₂ arrest are signatures of 6-TG DNA incorporation in MMR-proficient cells (11, 12), which activates the ATR-Chk1 checkpoint and accounts for most of the toxic effects of thiopurines (14, 15). To determine whether NUDT15 knockdown cells are more sensitive to 6-TG–induced checkpoint activation, HCT116 3-6 cells were depleted of NUDT15 and submitted to a time course treatment with low-dose 6-TG (Fig. 5D). At 150 nmol/L 6-TG, cells without doxycycline-induced depletion of NUDT15 (−DOX) had minor increases in Chk1 and Chk2 phosphorylation, γH2A.X, and CDK inhibitory phosphorylation, an indicator of cells in G₂ phase (42). Depletion of NUDT15 had a striking increase on these markers, indicating a much stronger response to 6-TG treatment. Importantly, Chk1 phosphorylation was followed by Chk2 phosphorylation and occurred approximately 2 cell cycles after addition of 6-TG, as described previously (42). Altogether, these experiments show that NUDT15 reduces the efficacy of 6-TG treatment.

After a half century in clinical applications, thiopurines are still a cornerstone therapy for childhood ALL and autoimmune disorders, such as IBD (2, 3, 10). Several pharmacogenetic studies have linked NUDT15 Arg139 mutations to thiopurine sensitivity in patients with leukemia and IBD (17–21). Further supporting a role for NUDT15 in thiopurine metabolism was the identification of 6-thio-dGTP and 6-thio-GTP as substrates (23).

Our previously reported substrate screen performed at 50 μmol/L concentration identified that NUDT15 hydrolyses 6-thio-dGTP, 6-thio-GTP, and dGTP (23), which we now know to be saturating. The in-depth kinetic analysis presented here highlights that NUDT15 hydrolyzes 6-thio-dGTP and 6-thio-GTP more efficiently than the canonical guanosine triphosphates, GTP and dGTP, due to higher substrate affinity (lower K_M). To explain the observed substrate preferences, we solved the first crystal structure of NUDT15 with 6-thio-GMP bound to the active site. Surprisingly, the NUDT15 R139C mutant hydrolyzed these substrates at similar efficiencies as NUDT15 WT, indicating that sensitivity in patients may be independent of enzymatic capability. When expressed in cells, R139C was rapidly degraded, suggesting that structural abnormalities could be a factor. Thermal instability of the R139C mutant was confirmed in vitro, and analysis of a variety of Arg139 mutants indicated that the loss of arginine and disruption of the ionic bonding network destabilizes the protein. Our attempts to crystalize NUDT15 R139C have been unsuccessful so far, which is likely an additional consequence of decreased stability. Because NUDT15 R139C was rapidly degraded in cells, as presumably is the case in patients carrying this variant, we wanted to mimic this situation with NUDT15-knockdown experiments. Depletion of NUDT15 clearly sensitized cells to 6-TG, similar to the response of patients carrying the R139C mutation. This sensitization was enhanced by functional MMR. Taken together, the data confirm the importance of NUDT15 in thiopurine metabolism and efficacy (Fig. 6).

These data support a recent study that demonstrated a role for NUDT15 in thiopurine metabolism and explored the biologic ramifications of missense polymorphisms (43). Our results extend the implications of this study through further mechanistic and biochemical characterizations, including a high-resolution structure of NUDT15 with the 6-thio-GTP hydrolysis product, 6-thio-GMP, and analysis of the structural contribution of Arg139 to protein stability. Furthermore, we show that NUDT15 depletion sensitizes both MMR-proficient and -deficient cells to thioguanine, which is indicative of increased DNA incorporation, as well as contribution of other MMR-independent toxicity mechanisms. For example, defective RNA production (11), influence on GTPases, like RAC1 (44), and interference in purine biosynthesis have been suggested as additional sources of activity and potential toxicity of thiopurines (4, 45).

We contend that sensitivity in patients with the R139C mutation is due to protein instability that results in NUDT15 R139C degradation (Fig. 3, 4). This implies that sensitivity results from limited NUDT15 availability and is not a direct consequence of enzymatic capacity (Table 1). Notably, we found that, when overexpressed, R139C protein levels were significantly lower than WT protein levels despite having essentially the same mRNA expression. Treatment with proteasome inhibitor was able to restore this difference, suggesting the ubiquitin/proteasome system flags the protein for destruction (38), and subsequent evidence supported that loss of structural integrity is the likely cause.

Pharmacogenetics studies have emphasized the importance of screening patients for NUDT15 missense variants before administering thiopurines, as the tolerated dose can be less than 10% of that for unaffected patients (17, 19, 43). Screening for mutations to thiopurine S-methyltransferase (TPMT), an enzyme that methylates thiopurines and prevents their DNA incorporation, is a common routine before initiating thiopurine treatments. However, TPMT variations are less common in the Asian population than in Europeans (17, 46), and only a quarter of European patients with IBD suffering from thiopurine-induced leukopenia carry mutated TPMT (17, 47, 48). Other enzymes, such as ITPase (ITPA or inosine triphosphatase; refs. 49, 50) and NT5C2 (5′-nucleotidase, cytosolic II; refs. 51, 52), have also been suggested to influence sensitivity to thiopurines. Thus, thiopurine sensitivity and resistance factors are multifaceted.

Currently, the physiologic role of NUDT15 remains unclear—as thiopurines are not endogenous metabolites. Because cellular concentrations of dGTP have been reported to be around 5 μmol/L (36, 37), we measured time-dependent catalysis at this concentration (Supplementary Fig. S2) and performed a thorough kinetic analysis (Fig. 1B and C). No detectable NUDT15-catalyzed hydrolysis of dGTP at 5 μmol/L would suggest that it is not a relevant substrate in the cell; however, the prospect that NUDT15 may have a role in dGTP metabolism under certain situational contexts cannot be excluded at this time. The metabolism of dGTP is regulated by multiple enzymes with substantial overlap, which further complicates evaluating the role for NUDT15 in these processes. Nonetheless, depletion of NUDT15 alone does not have a discernable effect on cell-cycle progression, DNA damage, or cell viability (23), nor has the R139C mutation been identified as a risk factor for disease or sensitivity to other therapeutics in patients.

In the presence of thiopurines, however, NUDT15 appears to be a key mediator of efficacy as our and other's results clearly suggest (43). Intriguingly, we found NUDT15 mRNA expression to be upregulated in clinical ALL samples (Supplementary Fig. S8; ref. 53), which raises the enticing possibility of modulating thiopurine dosing by specifically targeting NUDT15.

No potential conflicts of interest were disclosed.

Conception and design: N.C.K. Valerie, A. Hagenkort, M. Carter, P. Stenmark, A.-S. Jemth, T. Helleday

Development of methodology: M. Carter, P. Herr

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): N.C.K. Valerie, A. Hagenkort, B.D.G. Page, G. Masuyer, D. Rehling, M. Carter, L. Bevc, A.-S. Jemth

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): N.C.K. Valerie, A. Hagenkort, B.D.G. Page, G. Masuyer, M. Carter, L. Bevc, E. Homan, N.G. Sheppard, P. Stenmark, A.-S. Jemth, T. Helleday

Writing, review, and/or revision of the manuscript: N.C.K. Valerie, A. Hagenkort, B.D.G. Page, G. Masuyer, D. Rehling, M. Carter, N.G. Sheppard, P. Stenmark, A.-S. Jemth, T. Helleday

Study supervision: P. Stenmark, T. Helleday

We thank Prof. Astrid Gräslund and Axel Abelein for help and advice regarding the circular dichroism analysis. We thank the beamline scientists at BESSY, Berlin, ESRF, Grenoble for their support in data collection, PSF for protein purification and Biostruct-X for support. We also thank Kristina Edfeldt and Sabina Eriksson for their support in the Helleday Lab, Elisée Wiita for help with DSF experiments and Saeed Eshtad for providing us with the non-targeting shRNA construct.

This project is primarily supported by The Knut and Alice Wallenberg Foundation (T. Helleday, P. Stenmark). Further support was received from the Swedish Research Council (T. Helleday, P. Stenmark), the European Research Council (T. Helleday), Göran Gustafsson Foundation (T. Helleday), Swedish Cancer Society (T. Helleday, P. Stenmark), the Swedish Children's Cancer Foundation (N.G. Sheppard, T. Helleday), the Swedish Pain Relief Foundation (T. Helleday), the Wenner-Gren Foundation (P. Herr, P. Stenmark), the Åke Wiberg Foundation (P. Stenmark), the Torsten and Ragnar Söderberg Foundation (T. Helleday), the Canadian Institutes of Health Research and the Breast Cancer Society of Canada (B.D.G. Page), and the David and Astrid Hagelén Foundation (B.D.G. Page).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.