Structural and cellular insights into DCTPP1 antagonists and their synergistic action with DNMT inhibitors Authors: Glenn Hauk, Jianyong Liu, William G. Nelson, Srinivasan Yegnasubramanian, James M. Berger Published: 2026 ABSTRACT DCTPP1 is a nucleotide pyrophosphatase that helps preserve genomic stability and epigenetic programming by hydrolyzing and preventing the misincorporation of methylated base-modified deoxycytosine triphosphates into DNA. Through this role, DCTPP1 can degrade the efficacy of nucleotide analog-based DNA methyltransferase inhibitors and thus represents a compelling therapeutic target in cancer treatment. To identify prospective antagonists of DCTPP1, we conducted a high-throughput chemical screen against the enzyme, identifying both existing and previously unreported inhibitor classes with potent submicromolar activity. Structural characterization using X-ray crystallography revealed that the inhibitors all occupy DCTPP1’s nucleotide-binding pocket, associating primarily with a pair of tryptophans and two critical histidine residues that mimic interactions observed with natural substrates. Biochemical assays using modified chemical scaffolds confirmed the relevancy of the observed DCTPP1–antagonist interactions, while cell-based experiments demonstrated significant synergy between the lead inhibitors and the nucleoside analog decitabine in blocking the growth of prostate cancer cells. The specificity and efficacy of the compounds were further validated through loss- and gain-of-function studies, confirming the dependence of their therapeutic synergy on DCTPP1 activity. These findings advance our understanding of DCTPP1 as a therapeutic target while uncovering chemical scaffolds that can potentiate the action of existing nucleotide-based cancer therapies. The genetic material of cells is routinely subject to chemical modification through a variety of endogenous and exogenous factors (1–4). For example, nucleobases can be readily altered through events such as oxidation, alkylation, or deamidation. These changes can occur after DNA synthesis or by modifying precursor nucleotides that are then incorporated into DNA. Free nucleotides are methylated at much higher rates than polymerized nucleic acid chains (5). Accordingly, a variety of enzymes have been identified that are responsible for cleaning nucleotide pools by removing phosphate groups from modified nucleotides and preventing their incorporation into DNA (1, 6, 7). The presence of modified nucleotides in the genome can lead to toxicity, mutagenesis, and epigenetic reprogramming (7–9). Inappropriate incorporation of methylated cytosine poses a particular challenge for cells, as it can lead to mutagenesis through deamination and because cytosine is a natural substrate for postincorporation methylation by specific methyl-transferases such as DNMT1 (10, 11). This chemical mark can in turn change the local structural properties of DNA, as well as serve as a recognition determinant for proteins that regulate transcription (11–13). The appropriate control of DNA methylation status is critical for epigenetic programming and the avoidance of disease. For example, in higher eukaryotes, approximately 60 to 80% of all CpG sites are methylated (14). In vertebrates, an enzyme known as DCTPP1 is responsible for helping to ensure that methylated cytosine triphosphates are broken down into a monophosphate form with the liberation of pyrophosphate (15). DCTPP1 activity depletes modified cytosine nucleotides from the general pool to minimize their incorporation into DNA by polymerases during replication (16). Consequently, DCTPP1 helps maintain genetic and epigenetic integrity. Numerous studies have tied aberrant DCTPP1 action to cellular dysfunction and cancer. For example, inappropriate DCTPP1 levels have been linked to poor cancer prognosis (17–19). DCTPP1 has also emerged as an agent that degrades clinically valuable nucleotide-based drugs such as decitabine (20, 21). As a result, DCTPP1 can directly modulate the efficacy of some therapeutics, contributing to diminished potency and/or the potential release of undesirable side products (21). The diverse roles of DCTTP1 make it an attractive target for small molecule intervention to selectively control its activity. Previous studies have identified several chemical scaffolds that can inhibit the enzyme (22–26); however, until recently, no structures of DCTPP1 bound to small molecule antagonists were available. Our lab recently solved the structure of DCTPP1 bound to triptolide, a natural product inhibitor of DCTPP1 (21). Triptolide contains reactive epoxides that can form covalent adducts with proteins such as XPB (27), likely making it unsuitable as a clinically relevant small molecule inhibitor of DCTPP1. To expand the physical basis for inhibitor interactions with human DCTPP1 and further map out possible chemical scaffolds that could guide medicinal chemistry efforts, we conducted a screen of the protein against a 10,000-compound diversity library of small molecules. In addition to re-identifying one previously observed class of inhibitor, several new classes of chemically distinct antagonists were found, including some with submicromolar potency for the enzyme. We determined structures of three of these compounds in conjunction with DCTPP1, showing that they all work by associating with its nucleotide binding pocket. Cell-based studies showed that these agents, through their DCTPP1 inhibitory activity, potentiated decitabine mediated inhibition of cancer cell growth/survival. Our findings demonstrate that novel classes of DCTTP1 inhibitors can serve as modulators of chemotherapeutic efficacy and provide a useful set of scaffolds for advancing targeted drug discovery efforts against the enzyme. Results A Diversity Screen against DCTPP1 Yields Previously Unreported Inhibitory Agents. A variety of DCTPP1 inhibitors have been previously discovered using target-agnostic chemical crosslinking approaches (22, 23) and through DCTPP1-centered screening efforts using chemical libraries (24–26). Because DCTPP1 has proven tractable to chemical screening, we conducted our own independent screen using a commercially available 10,000 compound library from ChemDiv (Materials and Methods). For our screening approach, we used a previously established enzyme-coupled colorimetric assay in which the pyrophosphate generated by DCTPP1 in cleaving dCTP is converted to free phosphate by inorganic phosphatase, which is then detected by malachite green (24). The assay produced Z’-scores >0.5 for the no-enzyme (Z’ = 0.77) and plus-EDTA (Z’ = 0.55) negative controls, giving a suitable signal-to-noise profile for the identification of inhibitors. From the initial screen, ~90 compounds were identified that showed at least 85% inhibition of activity at 20 μM concentration compared to positive (inhibitor-free) controls (Fig. 1A). Of these, 55 were selected and rescreened at lower concentrations (5 μM, 1 μM, 200 nM, 40 nM, and 8 nM) to rank-order the hits by potency (Fig. 1B). The screen revealed previously identified DCTPP1-inhibiting chemical scaffolds, including 1,2,3 triazoles, [1,2,4]trialzolo[1,5-a] pyrimidines, and aryl piperazine containing compounds (Fig. 1A). However, new chemical scaffolds were also identified, including isobenzofurans and substituted pyrimidines. By far, the most highly represented chemophore among the top 55 hits was a nitrobenzofurazan-type of inhibitor (31 out of 55). Remarkably, multiple distinct groups of molecules (SI Appendix, Fig. S1) were obtained from a relatively limited screening of chemical space, confirming that DCTTP1 is highly amenable to ligand discovery efforts. Structural Characterization of DCTPP1 Inhibitors. We next sought to understand the physical details by which different classes of inhibitors act against human DCTPP1 using structural studies. From our top scoring hits, we manually selected a diverse set of 16 inhibitors that included examples from each major chemotype and set up cocrystallization screens with human DCTPP1 using conditions previously employed to determine a triptolide-bound structure of the enzyme (21) (Materials and Methods). Other conditions were also tested but typically resulted in empty, inhibitor-free structures (a “solvent-induced” state) in which precipitant (e.g., MPD) was present in the ligand binding pocket. Of the 16 compounds screened, crystals were obtained for 10. Overall, complete datasets were successfully collected for three chemically distinct inhibitor classes. From these data, we determined three different models of DCTPP1 bound to an inhibitory ligand from two different space groups, all at 2.3 Å resolution or better (Table 1 and SI Appendix, Fig. S2). A single compound-free structure of human DCTTP1 was also determined for comparison. In all instances, DCTPP1 formed a canonical homotetramer in the crystal (Fig. 1C), whose structure comports with prior models determined for mouse DCTPP1 (28, 29) and for human DCTPP1 in the presence of triptolide (21). In each of the three determined structures, the inhibitors occupy the substrate binding pocket of DCTPP1 (Fig. 1C). Previously solved structures of mouse DCTPP1 have defined the residues responsible for coordinating dCTP (28, 29). The active site consists of a cytosine base-binding pocket and a series of charged residues responsible for coordinating the phosphates of dCTP along with two magnesium ions required for catalysis (SI Appendix, Fig. S3). The inhibitors in all three structures make contacts with the cytosine base-binding pocket, which consists in part of a hydrophobic sandwich formed by two tryptophans (Trp47 and Trp73 from an adjacent monomer, Fig. 2, Left). The pocket defined by the tryptophan sandwich is also lined by Phe41, Phe50, Ile101, Tyr102, and Tyr129, providing ample hydrophobicity to the region. Within the pocket there is a pair of histidine residues that coordinate ligands through hydrogen bonding interactions (His38 and His51) (Fig. 2, Middle); these side chains normally associate with the carbonyl oxygen and N-3 nitrogen of cytosine when that base is bound (28, 29) (SI Appendix, Fig. S3). All three inhibitors share the binding pocket formed by the hydrophobic tryptophan sandwich, but each makes distinct interactions with the active site histidines. Each type of inhibitor also forms unique interactions with the enzyme that likely account for their different binding affinities. Class I Interactions (Nitrobenzofurazans). The class I compounds within our top 55 hits consist of a nitro-substituted benzene ring fused to a furazan ring (nitrobenzofurazan, SI Appendix, Fig. S2A). In addition to the shared central scaffold, which is structurally similar to nitro-substituted benzimidazole compounds identified previously (24), many of the compounds also share similar substitutions to the benzene ring, namely hydrophobic substituents para to the nitro group (which in many cases includes a piperazine ring) and an amino-alcohol group at the ortho position of the benzene moiety. In our structure of a representative class I inhibitor (Cmpd # K781–8268, SI Appendix, Fig. S1), the nitrobenzofurazan and piperazine groups fit within the Trp47/73 sandwich, with the piperazine and phenyl rings extending the contact within the hydrophobic pocket to include Phe50 (Fig. 2A). His38 and His51 act as hydrogen bond donors, similar to what is seen in structures of mouse DCTPP1 bound to dCTP analogs, but here hydrogen bond contacts are made with the carbonyl oxygen of the nitro-substituent of the nitrobenzofurazan benzene ring and the nitrogen of the furazan ring, respectively, rather than the base of dCTP. The aminoethanol substituent of the nitrobenzofurazan is positioned to take advantage of a water-mediated hydrogen bond to Asp98 [which is seen to hydrogen bond to the hydroxyl group of deoxyribose in dCTP analog structures (28, 29)]. In the class I compound structure, there are eight monomer chains within the asymmetric unit of the crystal. For two of these eight, the compound extends across a nonphysiological dimer interface; within these protomers, the active site magnesium ions are displaced, leading the compound to lose its water-mediated contact with the aminoethanol chain and gain new, direct hydrogen bonding interactions with Asp98 (SI Appendix, Fig. S4). Due to the nonphysiological interface and magnesium ion displacement, these direct hydrogen bonds are likely not representative of the binding mode of the class I compounds that is observed in the other subunits. Class II Interactions (1,2,3 Triazoles). Class II-type inhibitors have been previously identified in DCTPP1 compound screens (26). Class II inhibitors are characterized by a 1,2,3 triazole ring, often substituted with an oxadiazole group at 4-position of the triazole and further hydrophobic substitutions at the 1 or 5-positions. In our representative inhibitor of this class (Cmpd # S457-0718, SI Appendix, Fig. S1), there is a central 1,2,3 triazole moiety substituted with a propyl-oxadiazole group at the 4-position of the ring and a piperidine ring linked at the 1-position (Fig. 2B and SI Appendix, Fig. S2B). In our structure, we could not resolve density for the pyridine ring that is attached through a ketone linkage to the piperidine ring nitrogen; consequently, this group was not included in the final model. The triazole and attached piperidine ring sit between the tryptophan sandwich, with the central triazole making π–π stacking interactions with Trp47 and Trp73. The propyl group attached to the oxadiazole ring allows the class II inhibitor to extend further into the binding pocket, overlapping with the deoxyribose binding pocket observed in dCTP analog-bound structures. As with the piperazine ring in the class I inhibitors, the piperidine moiety extends the hydrophobic contact area to include Phe50. His38 and His51 make hydrogen bond contacts with the N-2 nitrogen of the oxadiazole ring and the N-3 nitrogen of the triazole ring, respectively. Class III Interactions (Pyrimidines). The class III inhibitor is unique among our inhibitor classes and bears little resemblance to previously identified DCTPP1 inhibitors (Cmpd # D490–0336, SI Appendix, Fig. S1). This antagonist consists of a central pyrimidine ring decorated with a methylsulfide at the C-2 position, carbonitrile at the C-5 position, chlorobenzene at the C-4 position, and a propanol-substituted piperazine ring at the C-6 position of the pyrimidine (SI Appendix, Fig. S2C). Although dCTP and the class III inhibitor share a pyrimidine ring, the contacts made in our structure are completely different than those seen between mouse DCTPP1 and various dCTP analogs (Fig. 2C). Here, the pyrimidine ring π–π stacks with the tryptophan sandwich and, along with the piperazine ring, extends out into the hydrophobic pocket formed by Trp47, Trp73, and Phe50. Although class I and class III inhibitors both contain an amino alcohol group, this appendage sits on opposite sides of the central rings in the respective structures. The carbonitrile group located at the C-5 position of the pyrimidine ring is positioned to act as a hydrogen bond acceptor for Tyr129, which in dCTP analog structures forms a hydrogen bond with an α-phosphate. The chlorophenyl substituent of our class III inhibitor resides within the deoxyribose-binding region of DCTPP1, extending into a pocket formed by Phe41, Ile101, Tyr102, and Tyr129. The chloro group forms a halogen bond (X-bond) with Asp98, which is a hydrogen bond acceptor for a deoxyribose hydroxyl group when dCTP is bound. Halogen bonds occur between electronegative centers and a “σ hole” present in halogenated hydrocarbons (with the halogen moiety acting analogously to a hydrogen-bond donor) and have been observed in a variety of biological contexts (30–32). As with dCTP analogs and class I and class II inhibitors, His38 again serves as a hydrogen bond donor. In the class III compounds, the N-1 nitrogen of the pyrimidine ring acts as a hydrogen bond acceptor, taking the place of a carbonyl oxygen that is present when dCTP is bound. The most significant difference within the binding pocket when compared to dCTP analogs, class I, and class III inhibitors is the position of His51. For the class III agent, the methylsulfide group located at position C-2 of the pyrimidine ring forms a distinctive space filling arrangement with His51 rather than engaging in hydrogen bonding interactions. This configuration results in a subtle repositioning of the Cα position for His51 and the adoption of an alternative rotamer that expands the pocket to accommodate the methylsulfide. “Solvent-Induced” (2-Methyl-2,4-Pentanediol) Interactions. In the process of conducting sparse matrix crystallization screens for human DCTPP1, we also obtained crystals in conditions containing 30 to 40% 2-methyl-2,4-pentanediol (MPD) (21). Although at the time DCTPP1 was crystallized in the presence of 2 mM triptolide [an agent used in a different study probing synergy with the chemotherapeutic decitabine (21)], no compound was observed in the active site. Instead, two MPD molecules were found to occupy this region. Interestingly, MPD-1 and MPD-2 both contact many of the same residues that the inhibitors exploit for binding to DCTPP1 (Fig. 2D and SI Appendix, Fig. S2D). The hydroxyls of MPD-1 contact His38 and His51, in which the oxygen from each acts as a hydrogen bond acceptor. Because MPD has a relatively short aliphatic chain between the hydroxyl groups, there is space for Trp73 to rotate and form a “T”-shaped π–π stacking interaction with Trp47, rather than the sandwich-type interaction observed in the class I, II, and III inhibitors. The second MPD molecule within the active site sits close to the ribose binding pocket and hydrogen bonds to Asp98, an interaction akin to that formed with the hydroxyl of the aminoethanol group in the class I inhibitors or the halogen bond formed with the chlorine of the class III inhibitor. IC50 Determinations for Crystallized Compounds. Having established the physical mode of binding to DCTPP1 for each of the three inhibitor classes, we next sought to better characterize the biochemical activity of the agents as a prelude to cell-based studies. Titrations of one representative from each class were conducted against DCTPP1 in the presence of dCTP and enzyme activity assessed using the malachite green assay to generate IC50s (Fig. 3 A–C. The class I inhibitor proved most active (IC50 of 200 ± 50 nM), followed by the class 2 agent (1.5 ± 0.25 μM) and the class 3 compound (3.0 ± 0.5 μM). We next conducted a substructure search for each of these compounds and identified related molecules that failed to inhibit DCTPP1 activity at the 20 μM screening concentration and that we predicted could form unfavorable interactions with DCTPP1 based on docking studies (SI Appendix, Fig. S5). For class I agents, we selected a molecule that replaces the aminoethanol substituent off the nitrobenzofurazan with a piperazine group that appeared too bulky to fit in the binding site (Fig. 3D). For the class II molecules, we chose a compound that replaced an iso-butyl group on the oxadiazole with a methyl substituent (ablating interactions with a hydrophobic pocket) and that altered the geometry of the tryptophan sandwich by swapping a 6-membered piperidine group for a 5-membered pyrrolidine ring (Fig. 3E). Finally, for the class III compounds, we found a derivative that replaced the methylsulfide found in the active molecule with a bulkier N-methyl-phenyl group and the ethanol substituent from the piperazine ring with a methanesulfonyl group (Fig. 3F). When tested for activity in vitro, the resultant IC50 values for the “inert” compounds exceeded 100 μM in each case (Fig. 3 A–C). Although the class I agent started to show some inhibitory activity at higher concentrations (>10 μM), its potency differs from that of its active analog by more than 1,000-fold. By comparison, the inert class II and class III reagents showed no inhibitory activity, even at near millimolar concentrations. Collectively, this study not only provides useful SAR validating the binding poses observed in our structural studies but also established a set of negative controls for subsequent cell-based assays. DCTPP1 Inhibitors Sensitize Cancer Cells to the Nucleoside DNMT Inhibitor Decitabine. DCTPP1 has been shown to be a critical mediator of primary resistance to nucleoside analog DNA methyltransferase inhibitors (DNMTIs) including decitabine, azacitidine, and related drugs (20, 21). Moreover, disruption of DCTPP1 function leads to significant sensitization to nucleoside analog-type DNMTIs in multiple cancer cell types (21). Based on these findings, we reasoned that the DCTPP1 inhibitors identified here would show strong synergy with the nucleoside analog decitabine in inhibiting growth of cancer cells. To test this hypothesis, we tested the class II and class III compounds that were used in our structural studies and that each showed strong biochemical inhibition of DCTPP1 enzymatic activity (Cmpd # S457-0718, Class II, and Cmpd # D490–0336, class III; SI Appendix, Fig. S1), along with two close analogs of each of these compounds that failed to show any appreciable DCTPP1 inhibition (Fig. 3 B and C). We then assessed the ability of the class II and III compounds to improve the potency of decitabine for inhibiting the growth of DU-145 prostate cancer cells, which have been shown to have some degree of primary resistance to low dose decitabine exposure (21). Class I compounds were not tested in this assay due to solubility problems at the high end of titration ranges for the representative compounds that were assessed. By first measuring DU-145 cancer cell growth against a fixed amount of decitabine for 200 h, we observed that the ability of the drug to inhibit proliferation was improved by the presence of the prototypical class II and III compounds but not their inactive derivatives (Fig. 4A). We then examined DU-145 growth across a titrated dose matrix of decitabine vs. the class II or class III compounds and their inactive variants. A dose–response analysis of the resultant cell growth data again demonstrated that both the class II and class III agents significantly enhanced the potency (reduced Log IC50) of decitabine, whereas their inactive versions did not (Fig. 4B). A heatmap of the percent change in viable cells compared to vehicle treatment is shown in Fig. 4 C and D for each of the four compounds, respectively; the calculated Bliss independence score for each dose combination is shown alongside (21). Positive Bliss scores indicate greater than additive (synergistic) effect, while negative scores indicate less than additive (antagonistic) effect. The Bliss sum across all dose combinations provides a succinct summary of the degree of synergy or antagonism for the drug combination. These analyses confirmed that the class II and class III molecules exhibit true synergy with decitabine across the dose matrix whereas their biochemically inactive variants do not (Bliss sum = 2.6 vs. −0.56 for class II and its inactive counterpart, respectively; Bliss sum = 1.75 vs. −0.82, for class III and its inactive version, respectively). To confirm that the observed synergy was due to DCTPP1 inhibition, we examined the inhibitory effect of decitabine and the class III compound on cell growth alone and in combination across a full dose matrix in parental DU-145 cells, and compared these responses to cells in which DCTPP1 was knocked out by CRISPR-CAS (DCTPP1-KO) or those in which DCTPP1 was overexpressed (DCTPP1-OE) [these isogenic models were developed and validated in a prior study (21)]. Analysis of the data revealed that ablation of DCTPP1 abolishes the inhibitory synergy between the class II molecule and decitabine (Bliss sum = −1.50) while DCTPP1-OE enhances it (Bliss sum = 2.36) (Fig. 4E). Collectively, these experiments demonstrate that the class II and class III agents are DCTPP1 inhibitors in cells and that they can synergize with a known DNMT inhibitor, decitabine, in disrupting cancer cell growth. Discussion Dysregulation of DNA methylation status and nucleotide pools can lead to inappropriate nucleotide incorporation, which in turn can promote mutation and epigenetic reprogramming. DCTPP1 is a nucleotide pyrophosphatase that is involved in regulating phosphorylated deoxycytidine nucleotide pools (33) and in breaking down modified variants of dCTP to prevent their use by DNA polymerases and help maintain genetic and epigenetic integrity (15, 16). In addition, the enzyme has been suggested to diminish the utility of certain nucleotide-analog drugs that are used as cancer chemotherapeutics (20, 21). The activities of DCTTP1 have suggested it could be useful as a target for antiproliferative compounds. Using a 10,000-compound small molecule screen, we identified several classes of inhibitors of human DCTPP1, including nitrobenzofurazans (class I), 1,2,3 triazoles (class II), and pyrimidines (class III) (Fig. 1 and SI Appendix, Fig. S1). We then determined the chemical basis of inhibition for these three inhibitor classes using X-ray crystallography (Fig. 2), assessed their potency in blocking DCTPP1 activity in vitro (Fig. 3), and demonstrated their ability to act synergistically with decitabine, a clinically used anticancer agent, to kill cancer cells (Fig. 4). Although the three classes of compounds identified here are chemically distinct, they all act at a common chemical binding pocket on DCTPP1 (Fig. 2 and SI Appendix, Fig. S3). The pocket, which comprises the enzyme active site, spans three regions, including the deoxyribose binding site, the central nucleobase binding site, and a cavity that extends from the base-binding region. The deoxyribose binding site is partially filled by an aminoethanol group in the most potent class I inhibitor (Fig. 2A), a propyl group in the top class II inhibitor (Fig. 2B), and a chlorobenzene in the sole class III compound (Fig. 2C). The central base-binding region consists of a hydrophobic cleft composed primarily of a pair of tryptophans (Trp47 and Trp73) that engage in π–π stacking interactions with ligands containing conjugated ring systems (Fig. 2, Left column). Specificity for natural, dCTP-based ligands has been hypothesized to be dependent on His38 and His51 as selectivity filters, which are hydrogen bond donors for the carbonyl oxygen and N-3 nitrogen of cytosine, respectively (29). Thymine and uracil, which are not preferred substrates for DCTPP1, both contain a protonated nitrogen at the N-3 position and would therefore act as a hydrogen bond acceptor for His51, rather than a hydrogen bond donor. These two histidines, which appear to function as selectivity filters, also serve as hydrogen bond donors for class I and class II inhibitors (and for class III agents, although in this instance only His38 forms hydrogen bonding interactions) (Fig. 2, Middle column). The ability of His51 to act as a space filling contact with the class III inhibitor, rather than as a hydrogen bond donor, highlights the chemical adaptability of the dCTP-binding pocket of DCTPP1 (Fig. 5). Further evidence for the flexibility of the DCTPP1 active site is observed in the solvent-induced MPD-bound structure. In previous structures of mouse DCTPP1, solved in the absence of ligand (28), Trp47 and Trp73 adopt conformations similar to substrate-bound structures of DCTPP1 (28, 29). However, in the MPD structure, Trp73 rotates to close off the DCTPP1-binding pocket, forming a collapsed, T-shaped π–π stacking conformation with Trp47 (Fig. 2D). Interestingly, the conformational flexibility of Trp47 in this context is close to the C5 position of modified dCTP substrates (SI Appendix, Fig. S3A). This movement positions Trp47 to act as an additional selectivity filter for dCTP substrates modified at the C5 position, and perhaps contributes to the substrate promiscuity of DCTPP1 for a variety of C5-modified dCTP analogs observed by others (16). The hydrogen bonding pattern of His38 and His51 accounts for the repeated appearance of an oxadiazole group coupled to a 1,2,3 triazole in both our and a previous screen (26). The conjugated ring of the central triazole allows for π–π stacking with the tryptophan sandwich, while the position of the N-2 nitrogen of the oxadiazole ring and the N-3 nitrogen of the triazole ring provide the appropriate spacing to form hydrogen bond contacts with the histidine dyad (Fig. 2B). The ability of His51 to adopt alternate conformations in the presence of an inhibitor, as exemplified with class III compounds (Fig. 2 C, Middle), suggests that the rotamer position of these residues may not be strictly defined; however, it is interesting to note the class III inhibitor did display one of the highest IC50s of the other inhibitor classes tested here (Fig. 3C), suggesting that there may be an energetic cost to local conformational rearrangements in the binding pocket. Interestingly, none of the identified inhibitors for DCTPP1 for which a structure could be determined (including triptolide (21) exploit the phosphate-binding pocket of the enzyme. This absence suggests that it may be possible to identify or construct inhibitors to engage the region to increase compound potency and/or specificity. Opportunities also exist to maximize contacts within the ribose binding pocket and the hydrophobic pocket that extends from the base-binding region. Filling this latter region, which extends to Phe50, may be the most tractable route to improving inhibitor characteristics, although there may be other residues in this region (such as from the flexible C-terminus, which was removed in our crystallization construct) that could contribute to affinity. Further studies will be required to more precisely map out the full suite of determinants for inhibitor binding to DCTPP1; however, this current effort provides foundational scaffolds and structure/activity relationships for further antagonist development. One major prospective application of selective DCTPP1 inhibitors is in the cancer therapeutic space, where they could be used to enhance the effect of nucleoside analog-based DNA methyltransferase inhibitors. While DNMTIs have had significant success in the treatment of myelodysplasia and acute myeloid leukemia (34–36), there has been limited utility for such agents in the majority of nonhematological malignancies (37). Recent work has shown that this shortcoming may be in part due to DCTPP1-mediated inactivation of the active triphosphorylated metabolite generated from inhibitory nucleoside analogs (21). Consistent with this concept, genetic disruption of DCTPP1, or inhibition of the enzyme with nonselective agents, has been shown to significantly sensitize multiple cancer cell models to DNMTIs (21). Building on this concept, we show that the newly identified selective DCTPP1 inhibitors developed here exhibit strong synergy with and can greatly sensitize cancer cells to DNMTIs. This synergy was not evident when using inhibitor analogs that did not antagonize DCTPP1 activity or when DCTPP1 was knocked out in a prostate-cancer cell model, confirming that the mechanism of synergy is truly mediated by DCTPP1 inhibition. It is important to note that the selected compounds tested in cancer cell models for synergy with decitabine were direct hits from initial screening efforts to identify novel classes of DCTPP1 inhibitors from a diversity compound library. While the DCTPP1-active compounds (but not the DCTPP1-inactive analogs), demonstrated utility in synergizing with decitabine in our cancer cell line studies, it is likely that these raw hit compounds may still have other off-target effects. For example, both the DCTPP1-active and -inactive compounds exhibited some single agent growth inhibitory effect, and this may be attributable to off-target effects of these classes of compounds. As such, further efforts to refine these compounds to minimize off-target effects and optimize their use in vivo from a pharmacological and safety perspective will be necessary before advancing them to additional in vivo safety and efficacy studies toward translation in human clinical settings. These findings nominate the newly identified DCTPP1 inhibitors, or pharmacologically optimized analogs, for use as effective combinations with DNMTIs in the treatment of human cancers. Future studies will be needed to capitalize on the prospective utility of this approach. Materials and Methods Protein Expression and Purification. Full-length human DCTPP1 (DCTPP1FL) and a truncated form (residues 21–130, DCTPP121−130) were independently subcloned into a modified pET vector containing an N-terminal hexahistidine SUMO tag as described previously (21). DCTPP1FL and DCTPP121−130 were expressed individually in BL21[DE3]-RIL Escherichia coli, grown in 2XTY at 37 °C to OD600 of 0.5 before lowering the temperature to 18 °C for 30 min, followed by induction with 300 μM IPTG. Cells were harvested after 16 h at 18 °C and resuspended in Buffer A [50 mM Tris-HCl (pH 7.8); 800 mM NaCl; 10% glycerol; 10 mM imidazole; 0.5 mM TCEP-HCl (pH 7.0); 1 mM PMSF; 1 μg/mL Leupeptin; 1 μg/mL Pepstatin]. DCTPP1FL and DCTPP121−130 were purified as described previously (21). Briefly, cells were lysed by sonication and clarified by centrifugation before passing the lysate over a 5 mL HisTrap-HP column (www.cytivalifesciences.com) equilibrated in Buffer A. Eluted protein was cleaved with hexahistidine-tagged human SENP protease (expressed and purified in house) while dialyzing overnight against Buffer A to remove the imidazole. Cleaved protein was passaged again over a 5 mL HisTrap-HP column equilibrated in Buffer A to remove any uncleaved protein and the tagged protease, while the flowthrough containing cleaved protein was collected. Cleaved protein was applied to a Sephacryl S-300 column (www.cytivalifesciences.com) equilibrated in Buffer C [50 mM Tris (pH 7.8); 500 mM NaCl; 10% glycerol; 0.5 mM TCEP (pH 7.0)]. Peak fractions were collected, assessed for purity by SDS-PAGE, and concentrated to 20 mg/mL. Concentrated proteins were supplemented with glycerol to a final concentration of 30% glycerol before flash freezing in liquid nitrogen and storing at −80 °C. 10K Diversity Library Screen. Inhibitor screening against human DCTPP1FL was conducted at the Johns Hopkins School of Medicine ChemCORE. A 10,000-compound subset of a 100,000 compound ChemDiv Representative Diversity Library (www.chemdiv.com) was screened at a concentration of 20 μM per agent in 384-well polystyrene plates using a malachite green based assay for detecting free phosphate (38). Final assay conditions for 20 μL reactions were as follows: 50 mM Tris-HCl (pH 7.5); 50 mM KCl; 5 mM MgCl2; 0.05% Brij-35; 0.2% DMSO (from compound screen or as vehicle control); 0.3 μM E. coli inorganic pyrophosphatase [ecPPI, expressed and purified in house as previously described (39)]; 80 nM DCTPP1FL; 0.2 mM dCTP. Wells in columns 3–23 of clear bottomed, 384-well polystyrene plates were seeded with 40nL of 10 mM compound in 100% DMSO using acoustic dispensing. In each plate, column 1 was reserved for a negative control containing a final assay concentration of 10 mM EDTA, column 24 was reserved for a negative control containing no enzyme, and columns 2 and 23 were positive control wells containing a final assay concentration of 0.2% DMSO as a vehicle control. Buffer and enzyme mix were added to wells and the reaction was initiated by the addition of dCTP. Plates were mixed by shaking and centrifuged, followed by incubation at room temperature for 20 min. Reactions were quenched by the addition of 20 μL 30 mM EDTA, bringing the final concentration of EDTA in the quenched reactions to 15 mM. 20 μL of freshly prepared malachite green phosphate detection reagent (0.16% H2SO4; 1.06 mg/mL malachite green; 1.5% ammonium molybdate; 0.176% TWEEN-20) was then added to each well, plates were mixed by shaking and centrifuged, followed by incubation at room temperature for 60 min.Absorbance at 655 nm was measured using a Tecan Safire II plate reader (www.tecan.com). To calculate Z’ score, the mean (μ) and SD (σ) for A655 absorbance values were determined across all vehicle-containing (0.2% DMSO) positive controls wells, along with the mean and SD for negative control wells containing either EDTA or lacking enzyme. Z’ scores were calculated independently for EDTA containing or no enzyme controls according to the following equation: Z’ = 1 − (3σpositive controls + 3σnegative controls)/|μpositive controls − μnegative controls|. A total of 32 384-well plates were used, with the assay performed in two batches (plates 1–16 and plates 17–32) on the same day. The mean A655 absorbance values from all vehicle-containing (0.2% DMSO) positive control wells within each batch were used to calculate fractional activity across all treatment conditions and compounds with a fractional activity <0.15 were considered as hits for follow-up assays. X-Ray Crystallography. Previous sparse matrix screening efforts to cocrystallize human DCTPP121−130 with the natural product inhibitor triptolide (21) identified several conditions that readily produced well diffracting crystals. However, upon structure determination and inspection of the DCTPP1 nucleotide binding pocket, crystals that formed in conditions containing MPD as a precipitant [100 mM Tris-HCl (pH 7.8); 80 mM NaCl; 80 mM MgCl2; 35 to 45% MPD] contained only MPD in the active site of the enzyme, despite the presence of 2 mM triptolide. These crystallization conditions also produced crystals in the absence of compound or DMSO, which were cryoprotected in 50% MPD, 80 mM NaCl; 100 mM MgCl2; 100 mM Tris-HCl (pH 7.8); 1 mM TCEP (pH 7.0) and diffracted to better than 2.0 Å resolution in the absence of compound. Crystallization conditions containing high concentrations of MPD appeared to be incompatible with compound binding in the active site of DCTPP1. Consequently, conditions containing longer chain polymer precipitants such as PEG3350 were prioritized for further efforts to crystallize DCTPP1 with inhibitors. Following a previous protocol in which a complex with triptolide was obtained (21), purified DCTPP121−130 was dialyzed into buffer containing 20 mM Tris-HCl (pH 7.8), 500 mM NaCl, and 0.5 mM TCEP (pH 7.0). Dialyzed protein was mixed with selected hits from the diversity screen (2 μL 50 mM compound in 100% DMSO; 98 μL 11 mg/mL DCTPP121−130), resulting in a final compound concentration of 1 mM in 2% DMSO. The protein/compound mixtures were incubated at room temperature for 30 min before setting hanging drop crystal trays by mixing 0.4 μL of each reaction with 0.4 μL precipitant using a SPTLabtech mosquito liquid handling robot (www.sptlabtech.com). Compounds were crystallized with DCTPP121−130 in the following conditions: Class I compound (ChemDiv K781-8268)-100 mM Tris-HCl (pH 7.5); 10% PEG3350; 400 mM proline, Class II compound (ChemDiv S457-0718)-100 mM Tris-HCl (pH 7.5); 10% PEG3350; 200 mM proline, Class III compound (ChemDiv D490-0336)-10 mM Tris-HCl (pH 7.5); 7% PEG3350; 400 mM NaOAc. Crystals were cryoprotected in mother liquor conditions supplemented with 1 mM of each respective compound, PEG3350 to a final concentration of 16%, and glycerol to a final concentration of 25% before plunge freezing in liquid nitrogen. Diffraction data were collected using the 17-ID-2 (FMX) and 17-ID1 (AMX) beamlines at Brookhaven NSLS2 (40, 41). Data were autoprocessed using FASTDP (42). Molecular replacement for all structures was conducted using PHASER (43) and PDB-ID 2OIG [mouse DCTPP1 (28)] as a search model. Refinement of all structures was performed with PHENIX (44) and model building was conducted using COOT (45). Table 1 reports the data collection and refinement statistics for each model. IC50 Determinations for Crystallized Compounds. A similarity search using JChem (www.chemaxon.com) was conducted against the 10,000-compound library used in the high-throughput screen to identify compounds that shared chemical features with the inhibitors we identified. We then filtered the results for compounds that demonstrated little to no inhibition in the screen. Compounds were chosen that closely matched the core chemical scaffold of our inhibitors and were manually docked into their respective DCTPP1•inhibitor structures. A candidate compound was considered inert if it did not inhibit DCTPP1 in the 10,000-compound screen and if it appeared to be sterically incompatible with binding to DCTPP1 based on superposition. From this analysis, three inert compounds were selected: Class I (ChemDiv 4061–0044), Class II (ChemDiv S460-0072), and Class III (ChemDiv D575-0056). To compare inhibition efficiencies for the active and inert compounds, a malachite green assay for phosphate release was employed as described above for the high-throughput screen. Inhibitors and putative inert compounds were formulated as 10× stocks (5 mM) in 50% DMSO, except for the Class I inhibitor, which was not soluble at 5 mM, therefore 2.5 mM was the highest stock concentration used. Dilutions were made in 50% DMSO and titrated from 0 to 500 μM (2 μL compound in 20 μL final reaction volumes for a final concentration of 5% DMSO in all reactions). After quenching, fractional activity was calculated by comparing A655 absorbance values in treatment conditions to vehicle control values containing 5% DMSO and no compound. Three independent assays were conducted for each compound and the SD between fractional activity values were calculated. Nonlinear regression curve fitting to calculate IC50 values was conducted using the following equation: Y = Min + X × (Max − Min)/(IC50 + X) in Prism GraphPad. Cell Models and Growth Curve Assay. DU-145 cells, and isogenic sublines with DCTPP1-KO or DCTPP1-OE were developed and maintained in culture as described in detail previously (21). 500 DU-145 cancer cells or isogenic sublines were seeded per well in 48-well cell culture plates (Corning) 1 d before compound treatment. Decitabine and other drugs were freshly prepared in DMSO and added to culture plates in the dose ranges indicated in Fig. 4B. For growth curve experiments, growth data were determined by imaging cell confluence at 6 h intervals in 16 fields per well using the IncuCyte™ ZOOM System (Essen Bioscience, MA) in the Johns Hopkins SKCCC Imaging Core Facility. The percent confluence was calculated with Incucyte built-in software. Statistics. All statistical analyses were performed using Prism GraphPad (San Diego, CA), unless described separately above. The a priori level of significance was set at alpha = 0.05. Growth curves were analyzed and constructed with Graphpad Prism. For determination of IC50 values, the time at which cells reached 90% confluence in the vehicle control treated cells was used to assess degree of growth for each of the other treatments/conditions. The dose of decitabine at which growth was reduced by 50% at this timepoint was calculated through curve fitting using Graphpad Prism and used to calculate the IC50 and 95% CI. The degree of synergy of combinations of decitabine and GH1 or GH1-i was calculated as the Bliss synergy score across the full dose ranges (Bliss values > 0 indicate a greater-than-additive effect). Bliss independence index was calculated for each dose combination using GraphPad Prism software, and the Bliss Synergy score represents the sum of the Bliss independence index across the full combination dose–response matrix tested. In Fig. 4B, the Wilcoxon nonparametric test was used to assess if the median difference in the Bliss index across all doses tested was significantly different between GH1 and GH1-i. Statement of Work. All authors contributed to the design of the experiments, data analysis, and writing of the manuscript. GH performed protein purification, chemical screening, X-ray crystallography, and in vitro biochemical essay studies; JL carried out the cell-based inhibition and synergy studies. Supplementary Material This article contains supporting information online at https://www.pnas.org/lookup/suppl/doi:10.1073/pnas.2534029123/-/DCSupplemental.