Pharmacologic Inhibition of NT5C2 Reverses Genetic and Nongenetic Drivers of 6-MP Resistance in Acute Lymphoblastic Leukemia

High-intensity multiagent chemotherapy protocols combining glucocorticoids, DNA-damaging agents, mitotic poisons, and L-asparaginase, followed by prolonged lower intensity maintenance therapy with oral 6-mercaptopurine (6-MP) has become the gold-standard treatment for acute lymphoblastic leukemia (ALL; refs. 1, 2). These regimens achieve over 80% cure rates in children and almost a 50% long-term remission in adults with this disease (3, 4). However, despite these overall favorable results, patients with incomplete responses and the even more sizable number who experience relapse after a transient remission face a very poor prognosis. The unfavorable outcomes of relapsed ALL are linked to the development of secondary chemotherapy resistance, which hampers the efficacy of salvage therapies (5, 6). Genomic analyses have identified specific genetic drivers of chemotherapy resistance and relapse in support of a Darwinian selection of resistance mutation–bearing clones at relapse (7–13). Prominent among these relapse-associated genetic alterations, activating mutations in the 5′-nucleotidase cytosolic II (NT5C2) gene are present in ∼10% of relapsed B-precursor ALL cases and ∼20% of T-ALLs (7, 8) and are selectively more frequent in early relapses occurring during or early after maintenance therapy (7, 8). Detailed genetic, enzymatic, structural, and functional analyses of relapsed leukemia–associated NT5C2 mutant alleles identified three distinct NT5C2 gain-of-function mechanisms implicating direct stabilization of the catalytically active enzymatic configuration, the disruption of an intramolecular switch-off mechanism responsible for returning the enzyme to its basal inactive state following allosteric activation, and finally enhanced allosteric activation normally limited by the stabilization of the basal inactive form of the enzyme by the insertion of the C-terminal acidic tail in the basic interface separating the two NT5C2 dimer subunits (14–16). Despite this molecular heterogeneity, structural analyses of wild-type and relapse-associated NT5C2 mutant proteins reveal a markedly convergent configuration for NT5C2 mutant alleles overlapping with that of the allosterically activated wild-type enzyme (14–16). Mechanistically, relapse-associated gain-of-function NT5C2 mutant proteins induce resistance to treatment with 6-MP by facilitating the dephosphorylation and subsequent degradation of thiopurine nucleotides generated via incorporation of 6-MP in the salvage pathway of purine biosynthesis (17, 18). Based on these results, we hypothesized that small-molecule NT5C2 inhibitors could overcome thiopurine resistance at relapse and enhance the efficacy of 6-MP maintenance therapy in the clinic.

Besides its role in the inactivation of cytotoxic metabolites of 6-MP and 6-thioguanine, NT5C2 regulates the purine nucleotide intracellular pool by dephosphorylating 6-hydroxypurine nucleotide monophosphates such as inosine monophosphate (IMP), guanosine monophosphate (GMP), and xanthosine monophosphate (XMP; refs. 19, 20). To identify small-molecule compounds with NT5C2 inhibitory activity, we screened a chemical library against recombinant NT5C2 R367Q mutant protein, the most prevalent relapsed ALL-associated NT5C2 allele present in over 90% of NT5C2 mutant relapsed leukemia samples (ref. 14; Fig. 1A–C). Analysis of 60,640 small molecules enriched in compounds with drug-like features in a high-throughput malachite green–based nucleotidase assay using IMP as substrate identified 225 potential active small molecules with 33.21% inhibitory activity or greater. Following the secondary screening of compounds with the highest Z’ scores, we validated the activity of top hits in dose–response curves (Fig. 1D). Independent analysis of top candidates with newly sourced compounds further confirmed the NT5C2 inhibitory activity of these compounds (Fig. 1E) and identified N-(3-carbamoyl-4,5,6,7-tetrahydrobenzo[b]thiophen-2-yl)-1H-benzo[d]imidazole-5-carboxamide (hereafter termed CRCD2) as the most active NT5C2 inhibitor in this screen. To further evaluate the specific inhibitory activity of CRCD2 and to verify its chemical structure, we tested an additional independently commercially sourced lot of this molecule and two batches of in-house synthesized compound generated using two different synthesis routes. All four sources of CRCD2 showed uniform dose-dependent inhibitory activity against NT5C2 R367Q (Fig. 1F).

Expression of the Nt5c2 R367Q relapse-associated mutation in ALL cells results in an NT5C2-mediated increase in degradation of purine monophosphate nucleotides with consequent depletion of IMP and accumulation of purine degradation products compared with isogenic wild-type controls (21). In this model, CRCD2 treatment of Nt5c2 R367Q–expressing lymphoblasts resulted in increased levels of IMP and decreased accumulation of deoxyxanthosine congruent with the NT5C2 inhibitory activity of this compound (Supplementary Fig. S1A–S1C). Furthermore, and in concert with a direct inhibitory effect, surface plasmon resonance analyses demonstrated direct binding of CRCD2 to NT5C2 R367Q recombinant protein, with a K_d of 70.9 μmol/L (Fig. 1G).

Increased nucleotidase activity in the NT5C2 R367Q mutant results from the loss of an intramolecular switch-off mechanism responsible for returning the enzyme to its basal inactive configuration following allosteric activation (14). This is in contrast with rare but highly active class I NT5C2 mutants, which reconfigure the environment of the catalytic center to lock this region in its active open configuration, resulting in high levels of constitutive NT5C2 activity in the absence of allosteric effectors (14). To test the span of NT5C2 inhibition by CRCD2 (Fig. 2A), we evaluated the activity of this compound against NT5C2 R367Q in comparison with wild-type NT5C2 and the class I NT5C2 K359Q mutant protein, the most active relapse-associated form of NT5C2. In these assays, CRCD2 displayed dose-dependent low micromolar inhibitory capacity against both NT5C2 wild-type and the R367Q mutant and significantly reduced NT5C2 K359Q nucleotidase activity, albeit with lower potency (Fig. 2B).

To evaluate the mechanism of NT5C2 inhibition by CRCD2, we tested the effect of this compound against NT5C2 R367Q in the presence of increasing concentrations of substrate (IMP). Michaelis–Menten curve analyses of these results showed a reduction in the maximum rate of reaction (V_max) and in the Michaelis constant (Km) in support of an uncompetitive mode of action (Fig. 2C) in which inhibitory effects are enhanced in the presence of substrate. Consistent with this model, surface plasmon resonance analyses of CRCD2 binding to NT5C2 R367Q in the presence of IMP revealed a 14.5% decrease in K_d compared with substrate-free conditions (Fig. 2D). In agreement, in vitro enzymatic assays testing CRCD2 against NT5C2 wild-type, R367Q, and K359Q recombinant proteins in the presence of increasing concentrations of IMP showed a significantly higher dose-dependent inhibitory capacity in the presence of substrate (Fig. 2E and F). To get a deeper insight into this mechanism of action, we performed a hydrogen–deuterium exchange mass spectrometry (HDX-MS) analysis of NT5C2 R367Q in the presence of IMP as substrate and in the presence or absence of CRCD2 (Supplementary Fig. S2A and S2B). Consistent with their high dynamic activity, we observed high deuterium exchange rates in regulatory regions of the protein implicated in allosteric activation and return to the basal inactive state including the arm domain and the N-terminal segment (Supplementary Fig. S2C). Moreover, we detected several peptides showing a different deuterium exchange kinetics in the presence of CRCD2 (Supplementary Fig. S2D). Most of these were located in the arm domain of NT5C2, a result consistent with decreased dynamics of this region consequent to decreased enzymatic activity in the presence of inhibitor. In addition, and most interestingly, we also observed changes in deuterium exchange kinetics in four peptides in the HAD III catalytic domain of NT5C2 (Supplementary Fig. S2E), a region critical for NT5C2 activity located near the substrate binding area (Supplementary Fig. S2F). These results suggest that CRCD2 binds to or induces a reconfiguration of the catalytic domain environment in the presence of substrate, a mechanism congruent with the uncompetitive nature of this inhibitor in enzymatic assays. In all, these results identify CRCD2 as a first-in-class small-molecule uncompetitive inhibitor of wild-type and relapsed leukemia–associated gain-of-function mutant forms of NT5C2.

Exploration of structure–activity relationships of CRCD2 led to the identification of several active analogues and revealed relevant moieties needed for NT5C2 inhibitory activity. Methylation of the benzimidazole ring on C-2 (CRCD47) or nitrogen (CRCD48) gave active analogues. However, the replacement of one of the benzimidazole ring nitrogens with sulfur (CRCD18) or CH (CRCD74) was not tolerated (Supplementary Fig. S3A). Replacement of the benzimidazole ring with a diamino phenyl group (CRCD54) or a phenylpyrrolidine-2,5-dione (CRCD9) led to active analogues; however, replacement with benzotriazole (CRCD57) or 2-aminobenzimidazole (CRCD58) led to loss of potency (Supplementary Fig. S3B). Methylation of the amide nitrogen in CRCD2 led to an analogue (CRCD40) with a loss of potency, as did replacement of the amide with a methyl ester (CRCD43) or carboxylic acid moiety (CRCD76). In contrast, replacement of the amide group in CRCD2 with a nitrile (CRCD22) gave an active analogue. Modification of the saturated six-membered ring fused to the thiofuran with a five-membered ring gave an active analogue (CRCD41); however, the seven-membered ring analogue (CRCD49) was not potent. Other modifications such as conversion of the six-membered ring to a pyran ring (CRCD42) or methylation of the ring (CRCD39) were also not tolerated (Supplementary Fig. S3C). Removal of the saturated six-membered ring (CRCD50) or replacement with one (CRCD51 and CRCD53) or two methyl groups (CRCD52) also led to loss of potency (Supplementary Fig. S3D). Exploration of replacement of the 4,5,6,7-tetrahydrobenzo[b]thiophene moiety with other moieties (e.g., alkyl, aromatic, and heteroaromatic systems) did not lead to potent analogues (CRCD59–CRCD73; Supplementary Fig. S3E).

These structure–activity relationship analyses reveal several regions of the CRCD2 scaffold that can be further modified to enhance potency and physicochemical properties (Supplementary Fig. S3F). For example, the benzimidazole ring can be substituted at carbon or nitrogen (CRD47 and CRCD48, respectively), and functionalized substituents could make additional binding interactions such as electrostatic or hydrogen bonding with the enzyme to increase potency and/or contain water solubilizing groups to enhance solubility. The amide on the thiophene ring of CRCD2 can be replaced with a nitrile group (CRCD22), suggesting that exploration of other substituents at this position could lead to increased activity. Similarly, the amines in active analogue CRCD54 could be alkylated or acylated to explore additional chemical space with the potential to increase potency. In addition, other modifications like fused aromatics, use of other heterocycles (furan and pyrrole), or linking longer groups or with other functionalities (ether and alkyl) are worth being explored.

Following structure–activity relationship analyses and given that the original hit compound CRCD2 showed the highest level of NT5C2 inhibitory activity, we evaluated the capacity of this small molecule to reverse 6-MP resistance driven by relapsed leukemia–associated NT5C2 mutations in cellular assays. Toward this goal, we analyzed the response to 6-MP in NT5C2 wild-type Jurkat and CUTLL1 T-ALL cells and in PEER and BE13 lines, which harbor the relapse-associated activating NT5C2 R29Q mutation. These analyses revealed increased sensitivity to 6-MP following CRCD2 treatment of wild-type cells and effective reversal of 6-MP resistance driven by the NT5C2 R29Q allele (Fig. 3A and B). Similar, results were obtained in REH (NT5C2 wild-type) and 697 (NT5C2 R368W mutant) B-precursor ALL cell lines (Fig. 3C). In addition, CRCD2 treatment effectively reversed 6-MP resistance induced by lentiviral expression of a K359Q, L375F, R367Q, and D407A mutant forms of NT5C2 in Jurkat cells (Fig. 3D–G). Moreover, CRCD2 treatment of NT5C2 R367Q mutant patient-derived xenograft cells from a pediatric T-ALL patient at relapse (ref. 11; Supplementary Table S1) induced increased sensitivity to 6-MP treatment (Fig. 3H).

Increased sensitivity to 6-MP in wild-type and NT5C2 mutant–expressing cells is consistent with a broad inhibitory activity of CRCD2 against wild-type and different relapse-associated resistance-driving forms of NT5C2 and supports that basal levels of NT5C2 in wild-type cells negatively affect the cytotoxic capacity of 6-MP. To further test this possibility and to evaluate the specificity of the effects of CRCD2 in 6-MP resistance, we tested the thiopurine sensitivity of isogenic NT5C2 wild-type and knockout CUTLL1 cells (Supplementary Fig. S4A and S4B) in the presence and absence of inhibitor. In these experiments, treatment with CRCD2 and genetic inactivation of NT5C2 induced similar increased sensitivity to 6-MP, further supporting a role for the wild-type enzyme in modulating the response to thiopurine therapy (Supplementary Fig. S4C–S4E). Moreover, NT5C2 knockout cells showed similar response to 6-MP in the presence and absence of CRCD2 (Supplementary Fig. S4C–S4E). The epistatic interaction of CRCD2 with NT5C2 knockout in the response to 6-MP corroborates the NT5C2-specific mechanism of action of this inhibitor.

Next, we aimed to directly test the differential interaction of CRCD2 with 6-MP in the context of isogenic Nt5c2 wild-type and Nt5c2 R367Q mutant primary leukemia lymphoblasts. Toward this goal, we isolated mouse leukemia cells generated via bone marrow transplantation of hematopoietic progenitors from conditional inducible Nt5c2 knockin mice (Rosa26^+/CreERT2 Nt5c2^+/co-R367Q) infected with retrovirus particles expressing a constitutively active form of NOTCH1 (ΔE-NOTCH1). In this validated model of resistance (21), tamoxifen treatment activates the expression of the Nt5c2 R67Q mutant knockin allele in leukemia lymphoblasts, resulting in overt resistance to 6-MP (Fig. 4A). In this experiment, treatment of Nt5c2 wild-type ALL cells with CRCD2 induced increased sensitivity to 6-MP, and treatment of their isogenic Nt5c2 R367Q counterparts resulted in effective reversal of 6-MP resistance (Fig. 4B–D). Further analyses of drug response curves across multiple dose levels revealed mild synergism (combination index = 0.8) between CRCD2 and 6-MP in drug-sensitive Nt5c2 wild-type cells and a strong synergistic interaction in 6-MP-resistant tamoxifen-treated Nt5c2 R367Q lymphoblasts (combination index = 0.32; Fig. 4E–G).

Following on these in vitro results, we evaluated the therapeutic activity of CRCD2 for the reversal of 6-MP resistance in vivo. Analysis of plasma clearance kinetics following CRCD2 administration revealed favorable pharmacokinetics with a half-life of 3.2 hours in female C57/BL6 mice (Supplementary Fig. S5A). Consistently, we documented a significant reduction in plasma 5′-nucleotidase activity in vivo 1 hour after CRCD2 injection (Supplementary Fig. S5B). In addition, intraperitoneal injection of CRCD2 in a 5 day on–2 day off dose-escalation scheme revealed no weight loss or apparent toxicities at the limit of solubility of this drug (Supplementary Fig. S5C and S5D). Moreover, histopathologic analysis of CRCD2-treated animals revealed no specific alterations in the intestine, kidney, liver, and brain compared with controls in support of a favorable safety profile (Supplementary Fig. S5E). Evaluation of the toxicity profile of CRCD2 with 6-MP in combination over 5 days of treatment (Supplementary Fig. S5F) compared with 6-MP alone revealed that adding CRCD2 to thiopurine treatment results in a modest effect in body weight and no differences (Supplementary Fig. S5G) in bone marrow cellularity (Supplementary Fig. S5H) and peripheral blood cell counts (Supplementary Fig. S5I and S5J). We noted no specific histologic changes in the kidney and intestine. Finally, similar alterations including a decrease in bone marrow cellularity and focal, mild sinusoidal dilatation, and congestion in the liver (Supplementary Fig. S5K), consistent with the toxicity profile of 6-MP, were noted in both groups.

To test the therapeutic activity of CRCD2, we allografted luciferized NOTCH1-induced conditional inducible Nt5c2 R367Q (Rosa26^+/CreERT2 Nt5c2^+/co-R367Q) T-ALL tumor cells into isogenic mice and treated these with vehicle only or tamoxifen to generate isogenic wild-type and Nt5c2 R367Q leukemias, respectively (21). Consistent with the role of Nt5c2 R367Q in driving 6-MP resistance, treatment of Nt5c2 wild-type leukemia–bearing mice with 6-MP at 50 mg/kg and 100 mg/kg induced overt and complete therapeutic responses, respectively, whereas treatment of Nt5c2 R367Q–expressing tumors resulted in significantly decreased therapeutic effects (Supplementary Fig. S6A and S6B; ref. 21). In this setting, treatment of Nt5c2 wild-type leukemias with vehicle only, 6-MP, CRCD2, and 6-MP plus CRCD2 in combination (Fig. 5A) revealed significantly improved antitumor activity in the CRCD2 plus 6-MP cotreatment arm after 5 days of treatment, as evidenced by quantification of tumor burden by luciferase bioimaging, spleen size, and spleen and bone marrow tumor cell content (Fig. 5B–E). In addition, treatment of mice harboring Nt5c2 R367Q tumors documented effective reversal of 6-MP resistance in animals cotreated with 6-MP and CRCD2 in combination (Fig. 5F–I). Moreover, in vivo treatment of our relapsed leukemia–derived T-ALL xenograft harboring the R367Q NT5C2 mutation (Supplementary Fig. S7A) verified enhanced antitumor response for the 6-MP plus CRCD2 combination (Supplementary Fig. S7B and S7C) when compared with 6-MP treatment alone.

Altogether, these results demonstrate effective CRCD2-mediated inhibition of Nt5c2 in wild-type and R367Q mutant–expressing leukemia cells in vivo, which results in increased therapeutic response to 6-MP and effective reversal of 6-MP resistance, respectively.

Early relapsed ALLs, those with disease progression under 6-MP chemotherapy, have a particularly high prevalence of NT5C2 mutations pointing to increased NT5C2 activity as a prominent mechanism of resistance. However, about 65% of early relapsed ALL cases are wild-type for NT5C2 and are devoid of other known 6-MP resistance–driver mutations. Across our experimental therapeutic assays, we observed that even though NT5C2 wild-type protein shows limited enzymatic activity in cell-free systems, treatment with CRCD2 resulted in increased sensitivity to 6-MP in NT5C2 wild-type leukemia cells in vitro and in vivo. Given the dynamic role of NT5C2 configuration changes in the regulation of NT5C2 nucleotidase activity, we hypothesized that posttranslational modifications involving NT5C2 regulatory regions could induce increased nucleotidase activity, phenocopy the effects of NT5C2 mutations, and drive 6-MP resistance by convergent, therapeutically relevant nongenetic mechanisms in ALL that could also be pharmacologically targeted for enhanced efficacy during maintenance.

To explore this possibility, we performed mass spectrometry analysis of NT5C2 wild-type protein immunoprecipitated from Jurkat human ALL cells. These analyses identified Ser418 and Ser502 as NT5C2 phosphorylation sites and residues Lys217 and Lys344 as modified by lysine acetylation (Supplementary Fig. S8A and S8B). Each of these NT5C2 phosphorylation and acetylation sites is highly evolutionarily conserved across vertebrate species (Supplementary Fig. S8C). To evaluate the potential regulatory role and functional relevance of these modifications as drivers of 6-MP resistance, we expressed acetylation mimic (K217Q, K344Q), acetylation disruptive (K217R, K344R), phosphorylation mimic (S418D, S502D), and phosphorylation disruptive (S418A, S502A) mutant forms of NT5C2 in Jurkat ALL cells and assessed their impact in response to 6-MP (Supplementary Fig. S8D). These analyses revealed no significant changes in the response to 6-MP in cells expressing K217, K344, and S418 modification mutant forms of NT5C2 (Supplementary Fig. S8E). In contrast, the expression of Ser502 disruptive (S502A) and phosphomimic (S502D) NT5C2 mutants in this system resulted in 6-MP resistance in support of a regulatory role for Ser502 in NT5C2 activity (Fig. 6A–C). Consistently, enzymatic analysis of recombinant NT5C2 S502D protein revealed increased nucleotidase activity in response to allosteric activation compared with wild-type NT5C2 control, further supporting a role for Ser502 phosphorylation in the regulation of NT5C2 function (Fig. 6D). NT5C2 is a tetrameric protein organized as a dimer of dimers with well-defined regulatory elements implicated in allosteric activation and the resolution of enzymatic activity and the return of the protein to its basal inactive closed configuration (14). Close examination of Ser502 in the crystal structure of wild-type NT5C2 revealed a hydrogen bond between this residue and Asp229 in the neighbor NT5C2 subunit selectively present in the absence of allosteric activators (Fig. 6E). Moreover, analysis of the crystal structure of the NT5C2 Ser502 phosphomimic mutant revealed a conformational change in which residues 494 to 561 in the C-terminal region of NT5C2 adopt a disordered conformation (Fig. 6F). These results suggest that the Ser502–Asp229 hydrogen bond contributes to stabilize this protein in its inactive closed configuration. To test the potential regulatory role of this Ser502–Asp229 hydrogen bond in NT5C2 regulation, we generated NT5C2 mutants disrupting this interaction and tested their capacity to induce resistance to 6-MP when expressed in ALL cells. In these experiments, the expression of hydrogen bond–disruptive mutants (S502D, D229A) induced resistance to 6-MP compared with the wild-type protein (Fig. 6G–I). In contrast, a dual mutant (S502D/D229S) in which the Ser and Asp residues are swapped to preserve the formation of the intermonomeric hydrogen bond failed to induce 6-MP resistance over the wild type (Fig. 6J). In all, these results identify the Ser502–Asp229 intermolecular hydrogen bond as a novel regulatory element in the control of NT5C2 nucleotidase activity and support a role of NT5C2 Ser502 phosphorylation in the regulation of 6-MP metabolism and clearance.

To further explore the potential role of NT5C2 Ser502 phosphorylation as a nongenetic mechanism of NT5C2 activation and the relevance of this posttranslational modification as a driver of 6-MP resistance in ALL, we generated and characterized an NT5C2 Ser502 phospho-specific antibody. Western blot analysis of Jurkat cell lysates with anti–pSer502-NT5C2 revealed a single band corresponding to the predicted size of NT5C2 (65 kDa), which was ablated by preincubation with alkaline phosphatase in support of a specific phosphor­ylation-dependent signal (Fig. 7A). In addition, we observed no reactivity against NT5C2 S502A or S502D mutant proteins immunoprecipitated from Jurkat cells expressing Flag-NT5C2 constructs, further supporting the specificity of this antibody against NT5C2 Ser502 phosphorylation (Fig. 7B). Quantitative western blot analysis of matched diagnosis and relapse patient-derived xenografts revealed increased NT5C2 Ser502 phosphorylation levels in 50% (6/12) of relapsed ALL xenografts analyzed compared with the respective diagnosis xenograft counterpart samples (Fig. 7C and D). We observed no effects of CRCD2 or 6-MP treatment on NT5C2 Ser502 phosphorylation with shorter treatment times (Fig. 7E). In this context, the increased sensitivity to 6-MP of ALL cells expressing NT5C2 wild-type following treatment with CRCD2 could result from inhibition of NT5C2 activity induced at least in part by Ser502 phosphorylation (Fig. 7F and G). To formally test this possibility, we evaluated the ability of CRCD2 to enhance the therapeutic response to 6-MP in resistant cells expressing the gain-of-function NT5C2 S502D phosphomimic mutant. In these experiments, NT5C2 inhibition with CRCD2 effectively reversed 6-MP resistance induced by NT5C2 S502D to the same extent as that induced by expression of the relapse-associated NT5C2 R367Q mutant allele (Fig. 7F and G). These results support a role for the nongenetic activation of NT5C2 nucleotidase activity as a contributing factor to impaired response to 6-MP in ALL and argue for a therapeutic role of NT5C2 inhibition with CRCD2 in the reversal of genetic and nongenetic mechanisms of 6-MP resistance in this disease.

Low-dose maintenance therapy with 6-MP effectively curtails the risk of relapse in ALL following high-dose combination chemotherapy and represents a core component of the treatment of this disease (22–25). The critical role of 6-MP in the clinic is highlighted by the importance of duration, dose intensity, and compliance during maintenance for the successful control of relapse-driving leukemia clones (26, 27). Indeed, therapy compliance monitoring has established decreased adherence to 6-MP therapy during maintenance as a prominent risk factor of relapse (2, 28). In concert with this central role of 6-MP in the control of disease progression, relapsed ALL frequently show a positive selection of resistance-driving mutations antagonizing the cellular effects of this agent (7, 8, 29). These include, most prominently, gain-of-function point mutations in the NT5C2 gene (7, 8, 29–31) and in rare cases activating mutations in PRPS1 (29) and haploinsufficiency of the mismatch DNA repair gene MSH6 (30). In this context, NT5C2 stands out as a clear therapeutic target to improve the efficacy of ALL maintenance therapy by curtailing the emergence of 6-MP resistance clones responsible for relapse. However, and despite much effort (32–36), no specific NT5C2 inhibitor with validated activity in the reversal of 6-MP resistance in vivo has been identified to date. Here, we leveraged a high-throughput inhibitor screen against the relapse-associated NT5C2 R367Q mutant protein and an array of preclinical cellular and animal models of NT5C2-driven 6-MP resistance toward the identification of CRCD2 as a first-in-class NT5C2-specific inhibitor with validated activity for reversal of 6-MP resistance in vitro and in vivo.

Interestingly, CRCD2 behaved as an uncompetitive inhibitor with increased NT5C2 binding and inhibitory activity in the presence of substrate, suggesting that this small molecule engages the NT5C2 protein in its active (substrate accessible) configuration and may facilitate the transition from the active state to inactive basal conformation. Across cell-free system analyses, we observed a broad inhibitory activity of CRCD2 against wild-type NT5C2 protein and against relapse-associated mutations, which induce increased NT5C2 activity as a result of the reconfiguration of the catalytic center environment (K359Q) or disruption of intramolecular self-inactivation mechanisms (R367Q; ref. 14). This observation is consistent with the largely overlapping structures of these relapse-associated NT5C2 mutant proteins, with that of the wild-type enzyme following allosteric activation (14). Even though a mutant-specific inhibitor could offer, in principle, an improved therapeutic window, an important consideration here is that the NT5C2 protein adopts a tetrameric configuration and NT5C2 mutations are characteristically heterozygous, which results in the expression of an array of wild-type and mutant NT5C2 heteromers with variable composition (7, 8, 16). Moreover, analysis of hetero-oligomeric complexes combining wild-type and R367Q mutant NT5C2 subunits supports that the enzymatic activation induced by this amino acid substitution can be transmitted from the mutated to the wild-type subunit (16). Thus, the capacity of CRCD2 to inhibit wild-type NT5C2 activity may be of relevance to overcome the overactivation of the wild-type subunit in tumors harboring the R367Q NT5C2 mutation.

Treatment with CRCD2 resulted in increased sensitivity to 6-MP not only in ALL cells expressing mutant forms of NT5C2 but also in NT5C2 wild-type leukemia, which is in agreement with the proposed role for the basal activity of the wild-type enzyme in metabolizing the thiopurine mononucleotide metabolites of 6-MP (17, 18). Indeed, genome-wide CRISPR screen–based drug–gene interaction mapping of modulators of 6-MP response in ALL demonstrates that genetic inactivation of NT5C2 can increase the sensitivity of NT5C2 wild-type cells to 6-MP in support of therapeutically relevant nongenetic mechanisms of NT5C2 activation in ALL (11), which could also be pharmacologically targeted for enhanced efficacy during maintenance. In this setting, it is worth noting that NT5C2 is sensitive to allosteric activation by phosphate-containing cellular metabolites such as diadenosine polyphosphates, 2,3-bisphosphoglycerate, and ATP, which connect its function with the metabolic state of the cell (37). Moreover, increased levels of NT5C2 expression and 6-MP metabolism have been reported in association with single-nucleotide polymorphism germline variants resulting in increased enhancer activity at the NT5C2 locus (38, 39). In addition, posttranslational modifications could play an important role in the control of NT5C2 function, regulating intramolecular interactions between effector and regulatory domains, multiprotein complex assembly, subcellular localization, and protein turnover. In this regard, our identification of NT5C2 Ser502 phosphorylation as a prevalent modification resulting in increased nucleotidase activity and resistance to 6-MP argues for a relevant role of nongenetic mechanisms of NT5C2 activation as drivers of reduced therapeutic response to this drug. The ability of CRCD2 to inhibit both wild-type and mutant forms of NT5C2 and to counteract the effects of NT5C2 Ser502 phosphorylation offers the opportunity to target both genetic and nongenetic mechanisms of NT5C2 activation in the clinic.

Finally, the identification of NT5C2 Ser502 phosphorylation as a relevant regulatory mechanism that influences the response to 6-MP argues that leukemia cell persistence during maintenance therapy may be initially mediated by nongenetic mechanisms resulting in increased NT5C2 activity, creating an opportunity for the occurrence of secondary activating mutations in NT5C2, which would then constitutively enhance enzymatic activity enabling the emergence of resistant disease, progression, and relapse (10, 21). In this setting, we propose a therapeutic role for NT5C2 inhibitors in combination with 6-MP for the treatment of NT5C2 mutant leukemia, but also to prevent the occurrence of relapsed disease by enhancing the antileukemic effects of 6-MP against NT5C2 wild-type tumors before NT5C2 mutant clones emerge. A relevant open question for further research is the identification of the kinases responsible for NT5C2 Ser502 phosphorylation, as these could potentially serve as therapeutic targets complementing NT5C2 inhibitor therapies. The proposed physiologic role of NT5C2 is to balance intracellular nucleotide pools via degradation of excess purine nucleotide monophosphate nucleotides. It is plausible that NT5C2 Ser502 phosphorylation functions downstream of signaling networks implicated in the regulation of purine metabolism, cell cycle, and DNA synthesis and repair.

We propose that in the context of heterogeneous leukemia cell populations with variable levels of NT5C2 Ser502 phosphorylation, 6-MP treatment will favor the positive selection of cells with increased levels of this posttranslational modification. However, it did not escape our attention that while a significant fraction of relapsed ALL xenografts show increased NT5C2 Ser502 phosphorylation, other samples displayed reduced levels of this modification. NT5C2 activity is tightly regulated, as excess purine degradation in the context of deregulated NT5C2 enzymatic activity impairs leukemia cell growth and leukemia-initiating cell activity (21). As a result, expression of the Nt5c2 R367Q resistance-driving allele is negatively selected in the absence of 6-MP (21). Thus, it is possible that downregulation of NT5C2 Ser502 phosphorylation may happen during xenograft expansion when cells are not exposed to 6-MP. However, it is possible that decreased NT5C2 Ser502 in these cases may reflect an adaptive mechanism to maintain cell homeostasis in relapsed samples in which other factors such as increased levels of allosteric regulators drive NT5C2 activation. Further biochemical characterization of nucleotide pools, NT5C2 posttranslational modifications, and NT5C2 activity in serial primary samples from patients treated with 6-MP could help clarify the mechanisms at play.

Altogether, our results provide a framework for the development of new combination therapies aimed at curtailing the emergence of thiopurine-resistant, relapse-driving clones in ALL and support the development of CRCD2 as the first resistance-directed targeted therapy for the treatment of ALL.

We treated human cell lines with inducible expression of NT5C2 for 48 hours with doxycycline (1 mg/mL). We purchased 4-hydroxy­tamoxifen from Sigma/Santa Cruz Biotechnology (#SC-3542) and dissolved it in 100% ethanol for in vitro assays. We purchased 6-MP monohydrate (#AC226520050) from Thermo Fisher and inosine 5′-monophosphate disodium salt hydrate (IMP; #57510), adenosine 5′-triphosphate disodium salt hydrate (ATP; #A3377), and tamoxifen (#T5648-1G) from Sigma-Aldrich. For in vitro assays, we dissolved 6-MP in DMSO, 4-hydroxytamoxifen in 100% ethanol, and ATP and IMP in Tris gel filtration buffer (50 mmol/L Tris-HCl, 100 mmol/L NaCl, 10% glycerol, and 5 mmol/L β-mercaptoethanol). For intraperitoneal injections of tamoxifen, we resuspended 100 mg tamoxifen in 100 μL of ethanol and added corn oil to reach a final concentration of 30 mg/mL. We then rotated the tamoxifen suspension for 1 hour at 55°C and froze it in aliquots at −20°C. We administered tamoxifen as a single 100 μL intraperitoneal injection per mouse. For in vivo studies with 6-MP, we prepared frozen aliquots of 5 mg/mL 6-MP in 0.1 M NaOH, and immediately before each round of treatment, we prepared fresh final solutions of 6-MP by buffering the stock solution down to pH 8 with 0.2 M NaH₂PO₄. This resulted in a 6-MP concentration of 3.53 mg/mL, which we diluted to various final concentrations using a solution made from 0.05 M NaOH and 0.2 M NaH₂PO₄ adjusted to pH 8. We administered 6-MP as 50 or 100 mg/kg twice a day. We prepared the vehicle by dissolving 0.254 g NaCl in 50 mL 0.05 M NaOH and adjusting the pH to 8 with 0.2 M NaH₂PO₄. We dissolved CRCD2 in DMSO for in vitro studies. For in vivo experiments, we dissolved CRCD2 to 50 mmol/L in DMSO, and then we further dissolved it to 5 mmol/L in a 30% PEG PBS solution. We adjusted injection volume to correct for any differences in weight between individual mice.

We purchased CRCD2 from Enamine (Z27358589) and synthesized it using two different routes. CRCD2 was synthesized by treatment of 1H-benzo[d]imidazole-5-carbonyl chloride with 2-amino-4,5,6,7-tetrahydrobenzo[b]thiophene-3-carboxamide and diisopropylethylamine in DMF followed by purification by reverse-phase high-performance liquid chromatography (HPLC; Supplementary Fig. S9A). Alternatively, we obtained CRCD2 by treating the diamine CRCD54 with formic acid (Supplementary Fig. S9B). The diamine CRCD54 was synthesized by reaction of 2-amino-4,5,6,7-tetrahydrobenzo[b]thiophene-3-carboxamide with 3,4-dinitrobenzoyl chloride and triethylamine in dichloromethane followed by reduction of the dinitro compound with hydrogen and palladium on carbon in methanol (Supplementary Fig. S9C). Treatment of the diamine CRCD54 with NaNO₂ in HOAc gave the benzotriazole CRCD57, and treatment with cyanogen bromide in aqueous methanol gave the 2-aminobenzoimidazole CRCD58 (Supplementary Fig. S9D). The remaining CRCD2 analogues were purchased from commercial suppliers (Enamine, ChemBridge) and synthesized by treatment of acid chlorides with 2-amino-4,5,6,7-tetrahydrobenzo[b]thiophene-3-carboxamide or analogous amines in DMF in the presence of Hunig's base followed by purification by reverse-phase HPLC. The acid chlorides were commercially available, or the commercially available carboxylic acid was converted to the corresponding acid chloride by treatment with oxalyl chloride and catalytic DMF in dichloromethane.

We generated and expanded ALL xenograft cells by intravenous injection of relapsed ALL lymphoblasts in NOD.Cg-Rag1tm1Mom Il2rgtm1Wjl/SzJ (NRG) immunodeficient mice (The Jackson Laboratory) from a relapsed ALL sample provided by the Children's Oncology Group and Princess Maxima Center (Netherlands) leukemia tissue banks. Written informed consent was obtained at study entry, and samples were collected under the supervision of local Institutional Review Boards for participating institutions and analyzed under the supervision of the Columbia University Medical Center Institutional Review Board (Protocol Number: IRB-AAAB3250) and in compliance with ethical regulations.

We performed cell culture in a humidified atmosphere at 37°C under 5% CO₂, and we regularly tested for Mycoplasma contamination. We purchased HEK293T cells for viral production from American Type Culture Collection (ATCC) and grew them in DMEM supplemented with 10% FBS, 100 U/mL penicillin G, and 100 μg/mL streptomycin for up to 2 weeks. The CUTLL1 cell line, which was generated by continuous culture of T-cell lymphoblastic pleural effusion cells from a patient at relapse, has been characterized and reported before (40). We obtained Jurkat and REH cells from ATCC and PEER, BE13, and 697 cells from Deutsche Sammlung von Mikroorganismen und Zellkulturen (DSMZ). We cultured CUTLL1, Jurkat, and REH cells in RPMI 1640 media supplemented with 10% FBS, 100 U/mL penicillin G, and 100 μg/mL streptomycin and PEER, BE13, and 697 cells in RPMI 1640 media supplemented with 20% FBS, 100 U/mL penicillin G, and 100 μg/mL streptomycin. Mouse Rosa26^+/CreERT2Nt5c2^+/co-R367Q T-ALL tumor cells were previously described (14) and were cultured in OptiMEM media supplemented with 10% FBS, 100 U/mL penicillin G, 100 μg/mL streptomycin, 55 μmol/L β-mercaptoethanol, and 10 ng/mL mouse IL7. Primary human xenograft ALL cells were passaged and collected from the spleens of NRG mice (The Jackson Laboratory) and cultured in RPMI medium supplemented with 20% FBS, 100 U/mL penicillin G, 100 μg/mL streptomycin, and 10 ng/mL human IL7.

We obtained the pET28aLIC (plasmid #26094) and pL-CRISPR.efs.gfp (plasmid #57818) plasmids from Addgene and pLVXTRE3GZsGreen1 vector from Clontech. We amplified the coding sequence of the NT5C2 cDNA from pLOC-NT5C2 (7) and cloned it into the pET28aLIC and pLVXTRE3GzsGreen1 vectors using In-fusion cloning with the In-Fusion HD Cloning Kit (Clontech) following the manufacturer's guidelines. We generated lentiviral vectors expressing CAS9 and gRNAs targeting exon 3 or 8 of Nt5c2 by cloning the corresponding gRNA oligonucleotides (Sigma-Aldrich) into pL-CRISPR.efs.gfp as reported (41). We cloned NT5C2 R238W, K359Q, R367Q, L375F, D407A, K217R, K217Q, K344R, K344Q, S418A, S418D, S502A, S502D, D229A, and D229S mutations into pLOC-NT5C2 (7) or pET28aLIC-NT5C2 by site-directed mutagenesis using the QuikChange II XL Site-Directed Mutagenesis kit (Agilent Technologies) according to the manufacturer's guidelines.

We transfected lentiviral plasmids together with gag-pol (pCMV ΔR8.91) and V-SVG (pMD.G VSVG) expressing vectors into HEK293T cells using JetPEI transfection reagent (Polyplus). We collected viral supernatants after 48 hours and used them to infect the CUTLL1 human cell line by spinoculation with 4 μg/mL polybrene infection/transfection reagent (Fisher Scientific). We selected infected human cell lines either with 1 mg/mL blasticidin (InvivoGen, #ant-bl-1) for 14 days or with 1 mg/mL puromycin (Sigma-Aldrich, #P8833) for 7 days.

For 5′-nucleotidase assays in the absence and presence of allo­steric activators, we cloned, expressed, and purified recombinant wild-type and mutant NT5C2 proteins as previously described (14). Briefly, we cloned full-length NT5C2 cDNA constructs with an N-terminal hexahistidine (His6) tag in the pET28a-LIC expression vector. We expressed recombinant proteins from Rosetta 2 (DE3) Escherichia coli cells by induction with 0.5 mmol/L isopropyl-β-D-thiogalactopyranoside overnight at 16°C. We resuspended harvested cells in lysis buffer (50 mmol/L Tris-HCl pH 7.4, 500 mmol/L sodium chloride, 10% glycerol, 0.5 mmol/L TCEP, 20 mmol/L imidazole) supplemented with cOmplete EDTA-free protease inhibitor (Roche) and lysed cells by sonication. We purified recombinant proteins using an ÄKTA fast protein liquid chromatography system (GE Healthcare) using a two-step protocol adapted from one previously described (42). We first performed affinity chromatography using a 1 mL Ni²⁺-charged His-Trap HP column (GE Healthcare) equilibrated in lysis buffer. We eluted NT5C2 proteins from the His-Trap column in a step-wise method with elution buffer (lysis buffer with 500 mmol/L imidazole) by first setting the buffer ratio to 25% elution buffer for 8 column volumes and then switching to a linear gradient to 100% elution buffer over 10 column volumes. We pooled NT5C2-containing fractions and purified them further by size exclusion chromatography using a HiLoad 16/60 Superdex 200 gel filtration column (GE Healthcare) equilibrated in 50 mmol/L Tris-HCl, pH 7.4, 100 mmol/L NaCl, 10% glycerol, and 0.5 mmol/L TCEP (for in vitro nucleotidase assays using the malachite green kit) or in 50 mmol/L sodium phosphate, pH 7.4, 100 mmol/L NaCl, 10% glycerol, and 0.5 mmol/L TCEP (for protein crystallization or nucleotidase assays using the Diazyme kit). We assessed protein expression and purity by SDS-PAGE and Coomassie staining. For crystallography studies, we concentrated protein samples to 4 to 9 mg/mL.

For high-throughput screenings, the malachite green enzymatic assay was adapted from a previously published protocol (32) and optimized for NT5C2 R367Q enzyme kinetics. To optimize the assay, initial velocity and linear reaction conditions were performed following NIH enzyme assay guidelines (43). Briefly, we incubated 0.02 μmol/L or 2.6 ng/mL purified recombinant NT5C2 R367Q protein with 100 μmol/L IMP as substrate for 15 minutes at 37°C. We terminated the enzymatic reaction (IMP→ Pi +inosine) by adding the malachite green reagent following the manufacturer's guidelines (Sigma-Aldrich, #MAK307). Colorimetric analysis of malachite green reaction with free phosphate (Pi) was measured after 15 minutes at 600 nmol/L. We performed these assays in Tris gel filtration buffer (50 mmol/L Tris-HCl, 100 mmol/L NaCl, 10% glycerol, 5 mmol/L β-mercaptoethanol) with a final concentration of 10 mmol/L MgCl₂. For in vitro nucleotidase assays with inhibitors or small-molecule compounds, we added compounds dissolved in DMSO to the protein and incubated the protein with the compounds for 10 minutes at room temperature prior to adding IMP substrate. We next incubated the plate at 37°C for 10 minutes and measured the malachite green reaction after 30 minutes. For Michaelis–Menten curve analyses, we used a phosphate standard curve following the manufacturer's guidelines. We assessed 5′-nucleotidase activity of purified recombinant wild-type and mutant NT5C2 proteins using the 5′-NT Enzymatic Test Kit (Diazyme) according to the manufacturer's instructions as described previously (7). We calculated 5′-nucleotidase activity levels using a calibrator of known 5′-nucleotidase activity as standard. We performed assays in triplicate in a Glomax Multidetection System plate reader (Promega). For assays with allosteric activators, ATP was dissolved directly in Reagent 2 of the test kit (containing the substrate IMP) and made serial dilutions to achieve a range of concentrations. We evaluated 5′-nucleotidase activity in plasma from mice treated with 17 mg/kg CRCD2 using the 5′-NT Enzymatic Test Kit (Diazyme) according to the manufacturer's instructions.

The Enamine Library consists of 60,640 compounds filtered for traditional physicochemical descriptors such as the rule of five, rotatable bond count, topological polar surface, and suitable aqueous solubility. The Enamine Library was curated by assigning a score for each compound relative to the calculated value in relationship to a standard value, a method modified from previous works (44). For example, the number of rotatable bonds was calculated for each compound, and then using a standard value of 6, deviations from that value are penalized based on distance from 6. After all criteria are calculated, the distribution of scores is then used to stratify the population with the highest scoring compounds being eliminated. We clustered the resulting compounds and selected two compounds from each cluster for purchase.

We performed the inhibitor screen on a Cell:Explorer robotic station (PerkinElmer) using the following modules: Janus liquid handling platform with NanoHead, FlexDrop liquid dispenser, Liconic microplate hotel, Liconic 500 microplate incubator (37°C), and Envision plate reader. Briefly, we plated 20 μL of 0.4 μmol/L NT5C2 R367Q recombinant protein in Tris gel filtration buffer or buffer alone controls in clear bottom, low binding, black 384-well microplates (Greiner, #781906). We delivered compounds from the Enamine Library to the plates by NanoHead from 10 mmol/L stock library microplates to a final concentration of 100 nmol/L, and we incubated the plates at room temperature in the Liconic incubator for 10 minutes. Then, we added 20 μL of 200 mmol/L IMP substrate to the wells with the FlexDrop, and plates were incubated at 37°C in the Liconic incubator for 15 minutes. Following incubation, we added 5 μL of the malachite green reagent (Sigma, #MAK-307) to all of the wells. We incubated the plates for 15 minutes and then we read on an Envision plate reader at 650 nmol/L. We added MgCl₂ to all buffers for a final concentration of 10 mmol/L in the assay. In screening the Enamine Library, each plate had two columns for a nontreatment control containing all elements of the assay without the addition of library compounds and two columns serving as negative controls that had neither substrate nor drugs added. We normalized the raw data for all compound-containing wells to the average of all the wells screened. From the primary screen, 3 standard deviations away from the mean was used as a cutoff to select 225 compounds that had 33.21% inhibitory activity or higher. Next, those 225 compounds were subject to confirmatory assay by cherry-picking them from the original library and testing them again in the same assay. We confirmed 28 compounds out of 225. Next, we chose six compounds that showed best inhibition and tested them in dose–response curves.

Surface plasmon resonance analyses were performed by Creative Biolabs using a Biacore T200 instrument (Cytiva). NT5C2 D52N R367Q recombinant protein was directly immobilized on the CM5 chip using an amine coupling kit (Cytiva). Before immobilization, the CM5 sensor surface was activated using a mixture of 400 mmol/L 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide and 100 mmol/L N-hydroxysuccinimide. Then, 50 μg/mL of NT5C2 protein in immobilization buffer (10 mmol/L NaAc, pH 4.0) was injected into the Fc2 sample channel at a flow rate of 10 μL/minute. The amount of ligand immobilized was about 15,000 resonance units (RU). The chip was deactivated by 1 M ethanolamine hydrochloride-NaOH (GE Healthcare Life Sciences) at a flow rate of 10 μL/minute for 420 seconds. The reference Fc1 channel underwent similar procedures but without injecting the ligand. The analyte (CRCD2) was serially diluted with the running buffer in the absence or presence of 500 μmol/L IMP to give a concentration of 62.5, 31.25, 15.625, 7.813, 3.906, 1.953, and 0 μmol/L, respectively. Different concentrations of analytes were then injected into the Fc2–Fc1 of channels at a flow rate of 30 μL/minute, with a contact time of 60 seconds, followed by a dissociation time of 90 seconds. Data analysis was performed on the Biacore T200 computer and with the Biacore T200 evaluation software using the steady-state affinity model.

HDX-MS studies were performed at the City University of New York Mass Spectrometry Facility. All subsequent sample handling was performed in an ice bath. The quenched sample was digested online using the Enzymate BEH Pepsin column (Waters). The digestion was performed at a flow rate of 0.15 mL/minute using 0.15% formic acid/3% acetonitrile as the mobile phase. The resulting peptides were collected and desalted with an in-line 4 μL C8-Opti-lynx II trap cartridge (Optimize Technologies) and then eluted through a C-18 column (Thermo Fisher Scientific, 50 × 1 mm Hypersil Gold C-18) using a rapid gradient from 2% to 90% acetonitrile containing 0.15% formic acid and a flow rate of 0.04 mL/minute, leading directly into a maXis-II ETD ESI-QqTOF mass spectrometer. The total time for the digest and desalting was 3 minutes, and all peptides had eluted from the C-18 column by 15 minutes. To avoid cross-contamination from carry-over peptides, comprehensive pepsin and C-18 column wash steps were included after each run. The peptide fragments were identified using Bruker Compass and Biotools software packages. The level of deuterium incorporation was assessed using the commercial software HDExaminer (Trajan Scientific). ​

The protein solution of the full-length mutant (S502D) NT5C2 at a concentration of 2.5 mg/mL in a protein buffer comprising 50 mmol/L ammonium phosphate (pH 7.5), 100 mmol/L sodium chloride, 10% (v/v) glycerol, and 1 mmol/L TCP was initially subjected to extensive robotic screening at the High-Throughput Crystallization Screening Center (45) of the Hauptman-Woodward Medical Research Institute (https://hwi.buffalo.edu/high-throughput-crystallization-center/). The only crystal hit was reproduced using the under-oil microbatch method at 4°C.

Small block-shaped crystals of NT5C2 appeared after 3 weeks in a crystallization condition comprising 100 mmol/L sodium acetate trihydrate (pH 4.6), 30% (v/v) MPD, and 20 mmol/L calcium chloride dihydrate with protein to crystallization reagent at a ratio 2:1 μL. The crystals were subsequently transferred into a similar crystallization reagent that was supplemented by 20% (v/v) glycerol and flash-frozen in liquid nitrogen. A similar methodology was used for growing crystals of truncated (537X) mutant (S502D) NT5C2. A native data set was collected on each crystal of the full-length NT5C2 at the NE-CAT24-ID-C beam line of Advanced Photon Source in Lemont, and NYX beam line of NSLSII was used for data collection on crystals of the truncated (537X) mutant (S502D) NT5C2. The best crystal for the full-length and truncated NT5C2 diffracted the X-ray beam to resolutions 2.77 Å and 2.12 Å, respectively. The images were processed and scaled in space group C2 using XDS (46). Both structures were determined by the molecular replacement method using the program MOLREP (47), and the crystal structure of wild-type full-length [Protein Data Bank (PDB) ID: 6DDO] and truncated NT5C2 (PDB ID: 6DDC) was used as a search model for each. The geometry of each crystal structure was subsequently fixed and modeled by programs XtalView (48) and Coot (49) and refined using Phenix (50). There are two protomers of the full-length and truncated NT5C2 in the asymmetric unit of each crystal. The crystallographic statistics are shown in Supplementary Table S2.

For the conditional tamoxifen-inducible expression of R367Q NT5C2 in vitro, we treated cells with 1 μmol/L 4-hydroxytamoxifen for 48 hours and tested induction of the R367Q allele by PCR as previously described (21). We next added 100 μL MeOH to pellets containing at least 1 × 10⁶ live cells and flash-froze the samples. Analysis of purine and pyrimidine metabolites was performed at the UVic-Genome BC Proteomics Centre (Victoria, Canada). Serially diluted standard solutions containing standard substances of the targeted compounds were prepared. Each sample was lysed on an MM 400 mill mixer with the aid of two beads at a shaking frequency of 30 Hz for 1 minute, twice. The mixtures were homogenized and then placed at −20°C for 1 hour, followed by centrifugation at 21,000 × g and 0°C for 10 minutes. Protein pellets were used for protein quantification and normalization. Then, the supernatant was dried under a nitrogen gas flow and reconstituted in 100 μL of the internal standard solution. Ten microliter aliquots were injected into a C-18 LC column (2.1 × 150 mm, 1.8 μm) to run ultra-performance liquid chromatography electrospray ionization-tandem mass spectrometry in multiple reactions monitoring mode (UPLC-MRM/MS) on a Waters Acquity UPLC system coupled to a Sciex QTRAP 6500 Plus mass spectrometer operated in the negative-ion mode for detection of nucleotides. The mobile phase was a tributylamine buffer (A) and acetonitrile/methanol (B) for binary gradient elution (5% to 40% B in 25 minutes) at 0.25 mL/minutes and 45°C. For the quantitation of nucleosides and nucleobases, 10 μL aliquots of the sample solutions and standard solutions were injected onto a polar reversed-phase C-18 column (2.1 × 100 mm, 2.0 μm) to run UPLC-MRM/MS on a Waters Acquity UPLC system coupled to a Sciex QTRAP 6500 Plus mass spectrometer operated in the positive-ion mode. The mobile phase was 0.1% formic acid (A) and methanol (B) for binary gradient elution (0% to 60% B in 15 minutes) at 0.30 mL/minute and 40°C. Concentrations of detected analytes were calculated with internal standard calibration by interpolating the constructed linear-regression curves of individual compounds, with the analyte-to-internal standard peak area ratios measured from the sample solutions.

We analyzed chemotherapy responses of human leukemia cell lines or murine mouse lymphoblasts following a 72-hour incubation with increasing concentrations of 6-MP or CRCD2 by measurement of the metabolic reduction of the tetrazolium salt MTT using the Cell Proliferation Kit I (Roche) following the manufacturer's instructions. We performed isobologram synergy assays as above with 6-MP, CRCD2, or the combination of 6-MP and CRCD2 at a consistent ratio. We performed isobologram analysis using Calcusyn software.

CUTLL1 cells were infected with pL-CRISPR.efs.gfp lentiviral particles containing an Nt5c2 exon 3 targeted sgRNA (GCAAAGCTG AGCAACTCCTG), Nt5c2 exon 8 targeted sgRNA (GTCCTACCGGA GTATGTTCC), or empty vector controls. We sorted infected cells based on GFP expression using a SONY SH800S cell sorter (SONY) and subsequently grew single-cell clones. We confirmed NT5C2 knockout via western blot analysis.

We generated rabbit polyclonal antisera directed against Keyhole Limpet Hemocyanin–conjugated NT5C2 peptides corresponding to the Ser502 region (CDINEMESPLATR). Phospho-specific immunoglobulins were purified from rabbit sera by positive affinity purification using the corresponding immobilized peptide columns (Covance).

We lysed cells in RIPA buffer and cleared them of cell debris. We performed BCA protein quantification according to the manufacturer's guidelines (BCA Protein Assay Kit, Fisher Scientific). For immunoprecipitation of FLAG-tag, we rotated 1 mg lysate with Flag affinity beads (EZview Red ANTI-FLAG M2 Affinity Gel clone M2, Sigma-Aldrich) overnight and then washed the unbound material with RIPA buffer. For western blot analysis, we loaded immunoprecipitated material or equal amounts of lysate onto a 4% to 12% Bis-Tris gel (Life Technologies), separated by SDS-PAGE, and transferred it to a nitrocellulose membrane for western blot analysis. We detected NT5C2 with mouse anti-NT5C2 (Sigma-Aldrich, #WH0022978M2) and rabbit anti–pSer502-NT5C2 (dilution 1:1,000; Covance) antibodies and β-actin with a mouse monoclonal anti–β-actin antibody (Sigma-Aldrich, #A5441).

We collected 500 million Jurkat cells expressing HA-Flag-NT5C2 in lysis buffer (50 mmol/L Tris-HCl, 100 mmol/L NaCl, 1 mmol/L EDTA, 1% Triton, 5% glycerol), lysed the cells for 30 minutes at 4°C in a rotating shaker, and spun them down in ultracentrifuge at 2,500 × g for 1 hour at 4°C. Then, we incubated the lysate with anti-FLAG beads (EZview Red ANTI-FLAG M2 Affinity Gel clone M2; Sigma-Aldrich, #F2426-1ML; 1 mg/mL) for more than 8 hours at 4°C and eluted twice with FLAG peptide (1 mg/mL). We next incubated FLAG eluate anti-HA beads (EZview Red Anti-HA Affinity Gel; Sigma-Aldrich, #E6779-1ML; 1 mg/mL) for more than 8 hours at 4°C and eluted twice with HA peptide (1 mg/mL).

For the identification of posttranslational modifications in NT5C2, HA eluates were analyzed by mass spectrometry at the Taplin Mass Spectrometry Facility at Harvard Medical School. Briefly, we electrophoresed the eluates on a 4% to 12% Bis-Tris gel, stained with Simply Blue Stain (Invitrogen), excised, reduced with dithiothreitol, alkylated with iodoacetamide, and digested with trypsin. Peptides were later extracted by removing the ammonium bicarbonate solution and then dried in a speed-vac (∼1 hour). On the day of analysis, the samples were reconstituted in 5 to 10 μL of HPLC solvent A (2.5% acetonitrile, 0.1% formic acid) and eluted using a nanoscale reverse-phase HPLC capillary column and increasing concentrations of solvent B (97.5% acetonitrile, 0.1% formic acid). As each peptide was eluted, it was subjected to electrospray ionization and then entered into an LTQ Orbitrap Velos Pro ion-trap mass spectrometer (Thermo Fisher Scientific). Eluting peptides were detected, isolated, and fragmented to produce a tandem mass spectrum of specific fragment ions for each peptide. Peptide sequences (and hence protein identity) were determined by matching protein or translated nucleotide databases with the acquired fragmentation pattern by the software program Sequest (ThermoFinnigan). Modification assignments were determined by the Ascore algorithm (51). All databases include a reversed version of all the sequences, and the data were filtered to between a 1% and 2% peptide false discovery rate.

We fixed mouse organs in 10% buffered formalin. The Molecular Pathology shared resource facility at the Herbert Irving Cancer Comprehensive Center proceeded to the embedding of fixed mouse organs in paraffin blocks, sectioning, and hematoxylin and eosin staining by following standard procedures. Slides were scanned using a Leica SCN 400 scanner, and photomicrographs were examined with Aperio ImageScope software (Leica Biosystems).

All animals were maintained in specific pathogen–free facilities at the Irving Cancer Research Center at Columbia University Medical Center. The Columbia University Institutional Animal Care and Use Committee approved all animal procedures. Animal experiments were conducted in compliance with all relevant ethical regulations. Animals were euthanized upon showing symptoms of clinically overt disease (not feeding, lack of activity, abnormal grooming behavior, hunched back posture) or excessive weight loss (15% body-weight loss over a week).

For systemic toxicity evaluation in vivo, we treated C57BL/6 mice with increasing concentrations of CRCD2 (8.5, 17, 34 mg/kg/day) in a 5 day on–2 day off schedule until we reached the maximum deliverable dose. We monitored the mice for low activity and weight loss. To evaluate the toxicity of the combination of 6-MP and CRCD2 in vivo, we treated C57BL/6 mice with therapeutic doses of 6-MP (50 mg/kg) and CRCD2 (34 mg/kg) for 5 consecutive days and evaluated blood cell counts, bone marrow cellularity, and tissue histopathology.

We performed a pharmacokinetic analysis of CRCD2 in plasma as previously described (52). We injected 17 mg/kg into C57BL/6 mice and collected samples at 0 (before the injection), 5 minutes, 10 minutes, 15 minutes, 30 minutes, 1 hour, 2 hours, 4 hours, 8 hours, 24 hours, and 48 hours from three mice per time point. We collected blood from the mouse via cardiac puncture and placed it into tubes containing EDTA anticoagulant on ice. We next centrifuged EDTA tubes at 2,100 × g for 10 minutes at 4°C, flash-froze samples, and stored them at −80°C. Then, we added 900 μL of acetonitrile to 100 μL of plasma, mixed the sample, and sonicated it. After spinning for 10 minutes at 4,000 × g at 4°C, the supernatant was collected and analyzed by liquid chromatography/mass spectrometry on a platform comprising a Thermo Scientific Dionex Ultimate 3000RS controlled by Chromeleon (Dionex) and a Bruker Amazon SL ESI ion-trap mass spectrometer. Chromatographic separation was performed at 20°C on an Agilent Eclipse Plus C-18 column (2.1 × 50 mm, 3.5 mm) at 20°C over a 12-minute gradient elution. Mobile phase A consisted of water with 0.1% acetic acid v/v, and mobile phase B was methanol with 0.1% acetic acid v/v. Mass spectrometry analysis was performed on a Bruker Amazon SL in positive ESI mode. Trap Control was used to control the ESI settings with the inlet capillary held at −4,500 V and the end plate offset at −500 V. Nitrogen was used as the desolvation gas. Hystar v3.2 was used to integrate the ultra-high performance liquid chromatography and MS applications, and data analysis was performed with Compass DataAnalysis software. The base peak chromatogram at m/z 655.2 with a width of ±0.1 was integrated, and the peak area was quantified by a standard curve. Pharmacokinetics of CRCD2 was assessed using Prism fitted with lognormal of one phase exponential decay.

For experimental therapeutics treatment studies, we allografted luciferized NOTCH1-induced conditional inducible Nt5c2 R367Q (Nt5c2^+/co-R367Q Rosa26^+/CreERT2) T-ALL tumor cells into isogenic mice. We harvested fresh luciferized tumor cells and transplanted them into sublethally irradiated (500 Rads) C57BL/6 recipients by retro-orbital injection. We monitored tumor development by in vivo luminescence bioimaging with the In Vivo Imaging System (IVIS; Xenogen). Once mice had a detectable baseline tumor burden by bioluminescence, we randomly assigned animals into different treatment groups and treated them with tamoxifen or corn oil vehicle by intraperitoneal injection as described above. Two days later, we initiated treatment with vehicle only, 34 mg/kg of CRCD2, 50 or 100 mg/kg of 6-MP, or the combination treatment via intraperitoneal injection for 5 consecutive days (n = 5 mice per group). We monitored disease progression and response to chemotherapy by bioluminescence imaging on days 0 and 5 after the start of treatment. We euthanized mice on day 5 and analyzed bioluminescence, GFP⁺ tumor infiltration in the spleen and bone marrow by flow cytometry, and spleen weight.

For therapeutic studies in primary human leukemia xenografts, a luciferized NT5C2 R367Q–bearing T-ALL patient-derived xenograft was transplanted into secondary NRG recipients by retro-orbital injection. We monitored tumor development by in vivo luminescence bioimaging with IVIS (Xenogen). Once mice had a detectable baseline tumor burden by bioluminescence, we randomly assigned animals into different treatment groups and introduced treatment with vehicle only, 34 mg/kg of CRCD2, 100 mg/kg of 6-MP, or the combination via intraperitoneal injection for 5 days. After this, first-cycle mice were allowed to recover for 9 days, and then we reintroduced therapy at the same level for 2 consecutive days, evaluating disease progression and therapeutic response by in vivo bioimaging on day 16.

We conducted statistical analyses using Prism software v8.0 (GraphPad software) and considered P < 0.05 statistically significant. We reported results as mean ± SD, with significance annotated by P value calculated as indicated in the figure legends using Student t tests assuming equal variance and normal distribution or using ANOVA and Dunnett multiple comparison tests. The investigators were not blinded to allocation during the experiments and outcome assessment. The experiments were not randomized. No data were excluded from the analyses.

No data sets were generated or analyzed during the current study.

C.L. Dieck reports a patent for NT5C2 inhibitors for the treatment of chemotherapy-resistant acute lymphoblastic leukemia pending. A. Zask reports grants from the NCI during the conduct of the study, as well as a pending patent for US20220105105A1. J.M. Gastier-Foster reports grants from the NCI during the conduct of the study. L. Tong reports grants and personal fees from Kintor Pharmaceuticals, Nimbus Therapeutics, and RADD Pharmaceuticals outside the submitted work. B.R. Stockwell reports grants from SMPO and personal fees from Inzen Therapeutics, Exarta Therapeutics, ProJenX Inc., Weatherwax Biotechnologies Corporation, and Akin Gump Strauss Hauer & Feld LLP outside the submitted work. T. Palomero reports grants from the NCI during the conduct of the study, as well as grants from Kura Oncology outside the submitted work. A.A. Ferrando reports grants from the NCI, NIH, the Leukemia & Lymphoma Society, and Alex's Lemonade Stand Foundation during the conduct of the study; personal fees from Regeneron Genomics Center, Bristol Myers Squibb, and VantAI outside the submitted work; and a patent for NT5C2 inhibitors for the treatment of chemotherapy-resistant acute lymphoblastic leukemia pending. No disclosures were reported by the other authors.

C. Reglero: Validation, investigation, visualization, writing–original draft, writing–review and editing. C.L. Dieck: Validation, investigation, visualization, writing–original draft, writing–review and editing. A. Zask: Investigation, visualization, writing–original draft, writing–review and editing. F. Forouhar: Investigation, visualization, writing–original draft, writing–review and editing. A.P. Laurent: Investigation. W.-H.W. Lin: Investigation. R. Albero: Investigation. H.I. Miller: Investigation. C. Ma: Investigation. J.M. Gastier-Foster: Resources. M.L. Loh: Resources. L. Tong: Conceptualization, investigation, writing–review and editing. B.R. Stockwell: Conceptualization, supervision, funding acquisition, writing–review and editing. T. Palomero: Conceptualization, funding acquisition, project administration, writing–review and editing. A.A. Ferrando: Conceptualization, supervision, funding acquisition, writing–original draft, writing–review and editing.

This work was supported by the Chemotherapy Foundation (A.A. Ferrando); NIH grants P30 CA013696 (Genomics and High-Throughput Screen Shared Resource, Oncology Precision Therapeutics Shared Resource), R35 CA210065 (A.A. Ferrando), R01 CA206501 (A.A. Ferrando and B.R. Stockwell), U10 CA98543 (J.M. Gastier-Foster and M.L. Loh), and Human Specimen Banking Grant U24 CA114766 (J.M. Gastier-Foster); a Translational Research Grant (A.A. Ferrando, LLS 6455-15) and a Screen to Lead grant (8011-18, A.A. Ferrando) by the Leukemia & Lymphoma Society; an Innovative Research Award from Alex's Lemonade Stand Foundation (A.A. Ferrando); and an Accelerating Cancer Therapeutics Pilot Award by the Irving Institute for Clinical and Translational Research of Columbia University. C. Reglero is supported by a Leukemia & Lymphoma Society Special Fellow award. R. Albero is supported by the Leukemia & Lymphoma Society Postdoctoral Fellowship award.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.