First-in-Class NADH/Ubiquinone Oxidoreductase Core Subunit S7 (NDUFS7) Antagonist for the Treatment of Pancreatic Cancer

Abstract

Pancreatic cancer cells adapt to nutrient-scarce metabolic conditions by increasing their oxidative phosphorylation reserve to survive. Here, we present a first-in-class small-molecule NDUFS7 antagonist that inhibits oxidative phosphorylation (OXPHOS) for the treatment of pancreatic cancer. The lead compound, DX2-201, suppresses the proliferation of a panel of cell lines, and a metabolically stable analogue, DX3-213B, shows significant efficacy in a syngeneic model of pancreatic cancer. Exome sequencing of six out of six clones resistant to DX2-201 revealed a pV91M mutation in NDUFS7, providing direct evidence of its drug-binding site. In combination studies, DX2-201 showed synergy with multiple metabolic modulators, select OXPHOS inhibitors, and PARP inhibitors. Importantly, a combination with 2-deoxyglucose overcomes drug resistance in vivo. This study demonstrates that an efficacious treatment for pancreatic cancer can be achieved through inhibition of OXPHOS and direct binding to NDUFS7, providing a novel therapeutic strategy for this hard-to-treat cancer.

Oxidative phosphorylation (OXPHOS) plays an important role in mitochondria function by mediating bioenergetics,¹⁻³ biomass production, and redox balance, all essential processes for tumor progression, spreading, and survival. Glycolysis-deficient cells,⁴ PTEN-null cells,⁵ and SWI/SNF complex mutated cancer cells⁶ are especially dependent on intact OXPHOS machinery for growth. Enhanced OXPHOS dependency is also a hallmark of cancer stem cells (CSCs)^7,8 and KRAS ablation-resistant cells in pancreatic ductal adenocarcinoma (PDAC).⁹ OXPHOS inhibition can alleviate cellular hypoxia,¹⁰ overcome resistance to chemotherapy,¹¹⁻¹³ and tyrosine kinase inhibitors.^14,15 Thus, targeting OXPHOS holds great promise to treat select cancers.

Pancreatic cancer is a notoriously lethal disease and responds poorly to current therapies. Metformin, a clinically used OXPHOS inhibitor, shows a positive correlation with survival in pancreatic cancer in multiple clinical cohorts,¹⁶ suggesting OXPHOS inhibitors can be used to treat pancreatic cancer. A clearer picture is now emerging on the essential role of OXPHOS in pancreatic cancer progression: overexpression of OXPHOS genes is correlated with poor prognosis in pancreatic cancer patients,¹⁷ and suppression of mitochondrial oxygen consumption significantly retards pancreatic cancer progression.¹⁸ Drug-resistant pancreatic cancer cells ablated for KRAS are heavily dependent on OXPHOS and are sensitive to OXPHOS inhibition.⁹ Pancreatic-tumor-initiating cells also show a strong reliance on OXPHOS.^19,20 Importantly, OXPHOS inhibitors are synergistic with gemcitabine specifically in cells with high expression of OXPHOS-related genes,²¹ suggesting that targeting OXPHOS can overcome drug resistance in PDAC patient subpopulations. Pancreatic cancer is also characterized by prominent genetic alterations that have relevant therapeutic implications associated with OXPHOS signaling. For example, GNAS^R201H/C mutations, occurring in 8% of PDAC patients²² and over 50% of intraductal papillary mucinous neoplasm (IPMN) patients,²³ drive pancreatic tumorigenesis and are critical for pancreatic tumor maintenance. Mechanistically, GNAS^R201H/C reprograms lipid metabolism and makes PDAC cells more sensitive toward fatty acid metabolism inhibition.²⁴ As β-oxidation in fatty acid metabolism is strongly dependent on OXPHOS function,²⁵ GNAS-mutated PDAC tumors may expose vulnerability to OXPHOS inhibition. The high frequency of somatic mtDNA mutations in pancreatic cancer²⁶ provides opportunities for tumor heterogeneity and at the same time vulnerability to mitochondrion-targeted therapy. A reduction of mitochondrial oxygen consumption caused by dichloroacetate (DCA) retards tumor progression.¹⁸ Taken together, OXPHOS inhibitors can be a promising strategy to treat subtypes of PDAC.

The OXPHOS machinery consists of five multiprotein complexes (complexes I–V), transferring electrons, maintaining the redox balance, and creating an electrochemical proton gradient to drive the synthesis of adenosine triphosphate (ATP). Complex I initiates the transfer of electrons from NADH to ubiquinone that further transfers the electrons to the respiratory chain unit. Specifically, NADH is oxidized to NAD⁺ on the hydrophilic arm by flavin mononucleotide (FMN) and the electrons are transferred by eight Fe–S clusters to ubiquinone at its binding pocket located at the interface of NDUFS2 and NDUFS7 subunits.²⁷⁻²⁹ Mutagenesis and photoaffinity labeling experiments suggest that the ubiquinone-binding pocket is essential for the function of complex I.^30,31 As a key component of the ubiquinone-binding pocket in complex I, the NDUFS7 gene encodes the NADH/ubiquinone oxidoreductase core subunit S7 and is preferentially required for growth in low concentrations of glucose,⁴ implicating its essential role in OXPHOS function. The interface of NDUFS7 and NDUFS2 contains a novel binding pocket for pharmacological inhibition of complex I and provides unique opportunities for the development of novel complex I inhibitors.

In this study, we present a potent and specific OXPHOS inhibitor, DX2-201, which targets NDUFS7, an essential component of complex I. By inhibiting complex I activity, DX2-201 suppresses mitochondrial function and impedes proliferation of a panel of pancreatic cancer cell lines. Importantly, its metabolically stable analogue DX3-213B significantly delays tumor growth in vivo. DX2-201 is synergistic with radiation, PARP inhibitors, and several metabolic modulators. A combination of DX2-201 with 2-deoxyglucose (2-DG) sensitizes pancreatic cancer cells in vitro and in vivo. Our in-depth integrated analysis of cells resistant to complex I treatment reveals GNAS as a potential biomarker for responsive patient populations. In conclusion, we validated NDUFS7 as a novel therapeutic target and discovered DX2-201 as the first-in-class NDUFS7 inhibitor, which shows significant single-agent efficacy in a syngeneic pancreatic cancer model and remarkable synergy with select drugs to overcome drug resistance.

Results

Discovery of DX2-201 from a Phenotypic Screen Specific for Galactose-Dependent Cell Growth

We performed a phenotypic screen of a highly diverse collection of our in-house compounds for selective cytotoxicity in media containing galactose versus glucose. Cells grown in the galactose-containing medium rely almost exclusively on mitochondria for their ATP production and are therefore especially sensitive to OXPHOS inhibition (Figure 1A).³² The positive control, metformin, is > 10-fold more potent in killing pancreatic cancer cells grown in galactose than in a glucose-containing medium (Figure 1B). Our high-throughput screen led to the discovery of a lead compound, N18230, with a benzene-1,4-disulfonamide core structure (IC₅₀ = 2.3 μM) (Figure 1B). N18230 is a mixture of R- and S- enantiomers due to the presence of a chiral center at the ethyl piperidine-3-carboxylate moiety. Since the absolute configuration greatly influences the overall shape and orientation of the molecule, we synthesized R- (DX2-201) and S-enantiomers (DX2-202) and evaluated their cytotoxicity in UM16 cells, an early-passaged patient-derived pancreatic cancer cell line (Figure 1C). In the galactose-containing medium, metformin showed an IC₅₀ value of 1000 μM and was practically inactive in the glucose-containing medium (IC₅₀ > 1000 μM). Under similar conditions, DX2-201 (IC₅₀ = 0.31 μM) was >3000-fold more potent than metformin. Interestingly, DX2-201 is > 10 times more potent than DX2-202 (Figure 1D), suggesting binding to a specific pocket on its cellular target. DX2-201 and its derivatives were more cytotoxic to UM16 cells than normal HFF-1 cells (Figure S1), providing a reasonable therapeutic window. Thus, DX2-201 was selected for in-depth mechanistic studies.^33,34

Figure 1.

Identification of the lead OXPHOS inhibitor N18230 in a phenotypic screen. (A) Cells grown in media enriched with galactose rely on mitochondria for growth and are remarkably sensitive to OXPHOS inhibitors. (B) Differential activities of metformin and N18230 in cells grown in a glucose- vs galactose-containing medium. UM16 cells (5000 cells/well in 96-well plates) were treated with indicated concentrations of metformin or N18230 for 72 h, and the cell viability was determined using the MTT assay. (C) Structure of N18230 (a mixture of R- and S-enantiomers), DX2-201 (R-enantiomer), and DX2-202 (S-enantiomer). (D) Dose–response curves for N18230 (mixture of R- and S-enantiomers), DX2-201 (R-enantiomer), and DX2-202 (S-enantiomer) in UM16 cells (200 cells/well in 96-well plates) after 7 days of treatment in the glucose-containing medium.

Multiomics Analysis Reveals Changes in Cellular Metabolism in Response to DX2-201 Treatment

To provide insights into the mechanism of action of DX2-201, we performed nascent RNA Bru-seq (4 h treatment) and proteomics (72 h treatment) of UM16 cells treated with 5 μM DX2-201. The top 25 genes and proteins dysregulated upon DX2-201 treatment are listed in Tables S1–S4. Gene set enrichment analysis of Bru-seq data showed a strong transcriptional perturbation of metabolic gene sets, including upregulation of glycolysis, glucose catabolic probes, and the pyruvate metabolic process. Partial lists of dysregulated pathways are shown in Tables S5–S10 (Hallmark, KEGG, and GO). Accordingly, the “oxidative phosphorylation” in both KEGG and Hallmark gene sets was enriched in downregulated gene sets. The “pyrimidine and purine metabolism” gene set was also downregulated (Figures 2A,B and S2A). Both transcriptomics and proteomics analysis revealed that similar gene sets change in drug-treated cells (Figures 2C and S2B,S2C), including upregulation of “glycolysis” and downregulation of “purine and pyrimidine metabolism”. Partial lists of dysregulated pathways with FDR <0.1 are shown in Tables S11 and S12. Additionally, gene sets such as “cholesterol homeostasis” were upregulated and “DNA repair” and “E2F targets” were downregulated in response to DX2-201 treatment in both platforms (Figure 2C,D). Collectively, DX2-201 treatment in pancreatic cancer cells causes upregulation of glycolysis and cholesterol catabolic pathways and suppression of transcription of genes involved in OXPHOS, leading to cell death.

Figure 2.

DX2-201 suppresses mitochondrial function. (A) DX2-201 treatment significantly downregulates transcription of gene sets related to oxidative phosphorylation, pyrimidine, and purine metabolism and upregulates transcription of gene sets related to glycolysis and pyruvate metabolism. UM16 cells were treated with 5 μM DX2-201 for 4 h and processed for Bru-seq, followed by GSEA. (B) GSEA plots for oxidative phosphorylation and glycolysis gene sets are shown. NES = normalized enrichment score; FDR = false discovery rate. (C) Proteomics (MS) and transcriptomics (Bru-seq) comparisons of UM16 cells in response to DX2-201 treatment revealed common enrichment of gene sets involved in cellular metabolism and DNA repair. UM16 cells were treated with 5 μM DX2-201 for 72 h and processed with MS-based proteomics. (D) Glycolysis and cholesterol metabolism pathways were upregulated, while DNA repair, pyrimidine, and purine metabolism were downregulated in response to DX2-201 treatment in multiomics analysis. (E) DX2-201 significantly inhibits ATP production in the galactose-containing medium as compared to the glucose-containing medium. MIA PaCa-2 cells (5000 cells/well in 96-well plates) were treated with indicated concentrations of DX2-201 in glucose- and galactose-containing media for 16 h. ATP production was determined by CellTiter-Glo Luminescent Cell Viability Assay. (F) DX2-201 significantly increased the expression of p-AMPK upon 24 h of treatment in BxPC-3 and MIA PaCa-2 cells. (G) DX2-201 decreased the levels of DHFR in a time-dependent manner in MIA PaCa-2 cells. (H) DX3-182B, a BODIPY-linked DX2-201 analogue, showed differential activity in glucose- and galactose-containing media at 5 μM. (I) DX3-182B accumulates in mitochondria. Confocal microscopy images of MIA PaCa-2 cells treated with DX3-182B (5 μM) and Mito-Tracker for 4 h.

DX2-201 Accumulates in Mitochondria and Suppresses Its Function

DX2-201 clearly causes upregulation of genes involved in glucose utilization and suppression of multiple metabolic pathways including purine and pyrimidine metabolism. Accordingly, DX2-201 is less toxic in a high-glucose-containing medium (Figures 2E and S3). Importantly, AMP-activated protein kinase (AMPK), an important energy biogenesis sensor, was significantly activated upon DX2-201 treatment in pancreatic cancer cell lines (Figure 2F). In addition, DX2-201 time-dependently suppresses the expression of dihydrofolate reductase (DHFR) (Figure 2G), which catalyzes the conversion of folic acid and affects subsequent metabolic reactions including purine nucleotide biosynthesis. We examined the status of the mitochondrial membrane potential using JC-1 staining, and after 4 h of DX2-201 treatment, no significant changes in membrane potential were observed. Moreover, DX2-201 shows no effect on the cell cycle (Figure S4A) and is not a spontaneous reactive oxygen species (ROS) inducer (Figure S4B). Thus, the downregulation of DHFR followed by suppression of purine and pyrimidine synthesis further implicates DX2-201’s role in blocking cell metabolism. This property can be uniquely exploited to produce synergistic cell kill when combined with standard-of-care chemotherapy for pancreatic cancer.

We further investigated the subcellular distribution of DX2-201 using a specific BODIPY-labeled DX2-201 probe (Figure 2H). Similar to DX2-201, this BODIPY probe showed significant differential cytotoxicity in glucose- versus galactose-containing media (Figure 2H). DX2-201-BODIPY colocalizes with MitoTracker (Figure 2I), suggesting uptake into mitochondria.

In summary, our results show that DX2-201 accumulates in mitochondria, suppresses OXPHOS, and inhibits select metabolic pathways including nucleotide synthesis, leading to cell death.

NDUFS7 Is the Direct Target of DX2-201

Using HCT116 cells, naturally defective in mismatch repair, we generated multiple single clones resistant to DX2-201 (DXGR) in cells growing in glucose- (Figure 3A) as well as galactose-containing media (Figure 3B). All clones were insensitive to DX2-201 and its analogues at concentrations as high as 30 μM in the glucose-containing medium and were at least 1/30th as sensitive as the parental HCT116 cells in the galactose-containing medium (Figure S5). None of the clones were resistant to paclitaxel, ruling out the possibility of resistance due to a nonspecific upregulation of efflux pumps (Figure S6). Importantly, none of the clones were resistant to other OXPHOS inhibitors, including complex I inhibitors IACS, phenformin, and rotenone and complex III inhibitor antimycin A (Figure 3C,D). This suggests that the target of DX2-201 is distinctly different from all other known OXPHOS inhibitors.

Figure 3.

DX2-201 is a complex I inhibitor. (A) Dose–response curves for the HCT116-parental cell line and six HCT116 clones resistant to DX2-201 (DXGR clones 1–6) (200 cells/well in 96-well plates) in the glucose-containing medium after 7 days of treatment. The cell viability was determined using an MTT assay. (B) Dose–response curves for the HCT116-parental cell line and six HCT116 clones resistant to DX2-201 (DXGR clones 1–6) in the galactose-containing medium (3000 cells/well in 96-well plates). The cell viability was determined using an MTT assay. (C) Colony formation assay of the HCT116-parental cell line and DXGR clones treated with DX2-201, DX3-134B, or IACS (200 cells/well in 96-well plates). Cells were harvested and stained with crystal violet. See Figure S5 for the results of all six resistant clones. (D) IC₅₀ values of indicated compounds in the HCT116-parental cell line and DXGR cell lines in glucose- and galactose-containing media as determined in the MTT assay as described above. (E) Gene mutation frequency in DXGR clones as obtained through exome sequencing. Eight genes were mutated in all six clones. GATK Mutect2 and MuSE v1.0rc were used to analyze the nonsynonymous variants in exonic, splicing, or upstream regions. (F) pV91M mutation in NDUFS7 was present in all 6 DXGR clones and occurred in 50% of the genomic DNA (heterozygous mutation). (G) Illustration of the structure of complex I with the NDUFS7 subunit highlighted in cyan and the NDUFS2 subunit highlighted in yellow. The pV91M mutation in NDUFS7 is located at the interface with NDUFS2 at the end of the ubiquinone-binding channel. Residues and structures were generated using the highly homologous structure of bovine complex I (5LC5.pdb). (H) DX2-201 inhibits complex I activity in a cell-free ubiquinone-dependent assay. (I) Effect of indicated concentrations of DX2-201 (24 h, MIA PaCa-2 cells) on the NAD/NADH concentration ratio. (J) Effect of indicated concentrations of DX2-201 (24 h) on NAD/NADH concentration ratio in HCT116-parental and DXGR cell lines. (K) Silencing of NDUFS7 significantly reduced the cell viability of HCT116 cells in the galactose-containing medium in 72 h. (L) ATP production of MIA PaCa-2 cells in galactose-containing medium compared to glucose-containing medium upon knockdown of NDUFS7. (M) Knockdown of NDUFS7 sensitizes HCT116 cells to DX2-201 treatment. HCT116 cells cultured in the galactose-containing medium were transfected with siNDUFS7, and compounds were added the next day for 3 days. Cell viability was determined using an MTT assay.

Exome sequencing of all clones identified 487 missense mutations that were present in at least one DX2-201-resistant clone (Figure 3E). Eight genes were mutated in six out of the six clones sequenced (Table S17). Mutated genes in five out of six clones are listed in Table S18. Of the eight genes mutated in all six clones, NDUFS7 was unique because all six clones contained the same NDUSF7 mutation (p. V91M) with a 50% read support, implying that one allele was mutated (from G-A mutation) in all resistant clones (Figure 3F). Thus, the identification of NDUFS7 mutant alleles in multiple independently derived resistant clones directly implicates NDUFS7 as the biologically relevant target of DX2-201.

The common single pV91M mutation in all DX2-201 resistant clones suggests that the valine at position 91 is the binding site of DX2-201. This specific amino acid resides on the interface of NDUFS2 and NDUFS7 subunits (Figure 3G), is important for ubiquinone binding,³⁵ and point mutations of this valine dramatically decrease the activity of complex I.³⁶ Therefore, we propose a model wherein DX2-201 binds to V91 at the NDUFS7 and NDUFS2 interface to block the binding of ubiquinone, thereby inhibiting the electron transfer of complex I.

DX2-201 Is an OXPHOS Inhibitor

To further validate the mechanism of action of DX2-201 in blocking ubiquinone binding and inhibiting complex I, we assessed the effect of DX2-201 on complex I activity in a cell-free ubiquinone-dependent assay (Figure 3H). Consistent with the cytotoxicity results, DX2-201 inhibits the function of complex I with an IC₅₀ of 312 nM. Because complex I is the main source of producing NAD⁺ through NADH and is essential for keeping the redox balance, we further determined if DX2-201 affects the cellular NADH/NAD⁺ ratio. As expected, DX2-201 significantly increased the NADH/NAD⁺ ratio at 24 h in multiple cell lines (Figure 3I,J) but not in the resistant DXGR cells (Figure 3J). V91 is located at the interface of NDUFS7 and NDFS2, and V91M mutation is sufficient for overcoming the complex I inhibition caused by DX2-201. Therefore, this site provides a unique binding pocket for DX2-201 and all its analogues. Similar to IACS, another specific OXPHOS complex I inhibitor, the cytotoxicity of DX2-201 can be partially rescued by pyruvate and aspartic acid in MIA PaCa-2 cells (Figure S7), further suggesting that these two compounds have similar mechanisms of action. In conclusion, DX2-201 is a potent OXPHOS inhibitor that significantly inhibits complex I activity and decreases cellular redox balance.

Partial Knockdown of NDUFS7 Sensitizes Cells to DX2-201 Treatment

To further understand the function of NDUFS7 on the effect of DX2-201, we knocked down NDUFS7 in HCT116 and in MIA PaCa-2 cells and evaluated the ATP production and response to DX2-201 treatment (Figure 3K,L). Consistent with the DX2-201 mechanism of action in inhibiting NDUFS7, knocking down NDUFS7 also showed a stronger growth suppression in the galactose-containing medium as compared to the glucose-containing medium (Figure 3M). As expected, cells with reduced NDUFS7 expression are more sensitive to DX2-201 treatment, supporting NDUFS7 as the cellular target for DX2-201. Taken together, these data are consistent with DX2-201 binding to NDUFS7 and thus blocking ubiquinone binding to inhibit complex I activity.

DX2-201 Induces ROS, Reduces Mitochondrial Mass, and Causes Apoptosis

We profiled 105 cancer cell lines for their sensitivity to DX2-201 treatment to identify cells that are naturally resistant to DX2-201 (Figure 4A,B). DX2-201 was also tested in the NCI60 panel of cell lines and additional in-house cells, demonstrating selectivity for select cell types (not shown). Of note, both IACS and DX2-201 produced a similar cytotoxicity profile, providing further support for complex I inhibition (Figure S7). DX2-201 showed significant cytotoxicity in leukemia cell lines with 10 out of 12 cell lines responding. Five out of seven pancreatic cancer cell lines were sensitive to DX2-201. Importantly, DX2-201 was not toxic to HPDE, a normal immortalized pancreatic cell line (IC₅₀ > 10 μM, Figure 4C). DX2-201 also showed significant growth inhibition in a panel of low-passaged PDAC patient primary cell lines (Figure 4D), further supporting the utility of OXPHOS inhibition for the treatment of PDAC.

Figure 4.

DX2-201 selectively inhibits pancreatic cancer cell lines, reduces mitochondrial mass, induces mitochondrial fusion, and suppresses tumor growth in a Pan02 syngeneic mouse model. (A) Cytotoxicity of DX2-201 was assessed in a panel of cell lines across 12 cancer types. Number of sensitive cell lines per total tested cell lines is shown for each disease type. Cell lines with an IC₅₀ lower than 2 μM were defined as sensitive. (B) Waterfall plot showing the adjusted IC₅₀ values of DX2-201 in diverse cancer cell lines. IC₅₀ was determined by MTT in a glucose-containing medium after 7 days of continuous treatment. (C) Dose–response curves of DX2-201 in a panel of pancreatic cancer cell lines in the glucose-containing medium (7 days of treatment). The effect of DX2-201 in HPDE cells is also shown. (D) Dose–response curves of DX2-201 in a panel of patient-derived-pancreatic cancer cell lines in the glucose-containing medium after 7 days of treatment. (E) DX2-201 (1 μM) and DX3-213B (0.1 μM) enhanced ROS production upon 24 h of treatment. CM-DCFDA was used to stain the cells in Hank’s balanced salt solution for 30 min, followed by flow cytometry. (F) Mitochondrial membrane potential of KPC cells upon DX2-201 (5 μM) treatment for 2 days quantified by ImageJ. Microscopy images of mitochondria are shown in Figure S9. After 2 days of compound treatment, cells were incubated with 10 μg/ml JC-1 for 10 min in the dark followed by microscopy imaging. (G) Apoptosis induction upon 2 days of DX2-201 treatment. Cells were stained with PI and Annexin V in the dark for 10 min and followed by flow cytometry. (H) DX2-201 treatment reduces hypoxia in a 3D pancreatic tumor spheroid model after 4 h. Pan02 cells (10,000 cells/well) were cultured in RPMI with 5% collagen and 10% FBS in ultralow attached U-shaped-bottom 96-well plates. After compound treatment, the cells were stained with 5 μM Image-iT Green Hypoxia Reagent for 1 h and imaged by fluorescence microscopy. (I) Cytotoxicity of DX2-201 in a panel of murine pancreatic cancer cell lines. Cells were treated with indicated concentrations of DX2-201 for 7 continuous days. The cell viability was determined using MTT. (J) A metabolic stable analogue of DX2-201 (DX3-213B) and IACS-010759 show in vivo efficacy in a Pan02 syngeneic model. The tumor volume at the end point of the study was analyzed using Student’s t-test, and ** represents p < 0.01. Changes in body weight in each group are shown in the lower panel.

Complex I is one of the main sources of ROS production due to electrons leaking from the mitochondrial outer matrix.³⁷ Multiple complex I inhibitors have been reported to induce ROS,^38,39 leading to apoptosis.⁴⁰ Accordingly, we also observed significantly enhanced ROS production upon DX2-201 treatment (Figure 4E). This phenotype was accompanied by a reduction in mitochondrial mass (Figures 4F and S9), indicating a suppression of mitochondrial function caused by complex I inhibition. Importantly, DX2-201 treatment leads to significant apoptosis (Figure 4G). Thus, by inhibiting complex I function, DX2-201 enhances ROS production, suppresses mitochondrial function, and induces apoptosis.

DX2-201 Alleviates Hypoxia and Significantly Inhibits Tumor Growth in a Syngeneic Mouse Model

Pharmacological inhibition of complex I alleviates hypoxia in different disease models.^41,42 We utilized pancreatic cancer 3D tumor spheroid models to simulate hypoxia and observed that DX2-201 significantly reduced hypoxia upon 4 h of treatment (Figure 4H), suggesting that DX2-201 effectively modulates the hypoxic microenvironment in tumors.

We determined cytotoxicity of DX2-201 in a panel of murine pancreatic cancer cell lines (Figure 4I). Since NDUFS7 is highly conserved between Homo sapiens and murine, DX2-201 also inhibits the murine complex I. Among the five murine cell lines, PAN02 is one of the most sensitive cell lines to DX2-201 treatment. Thus, we further validated the efficacy of our complex I inhibitors in a PAN02 allograft pancreatic cancer model. A metabolic stable analogue of DX2-201 (DX3-213B) was used for the in vivo studies. DX3-213B is one of the most potent and metabolic stable analogues of DX2-201 as we discovered in our lead optimization campaign.³⁴ DX3-213B was given orally at 1, 2.5, and 7.5 mg/kg (n = 5 per group) for 28 consecutive days to PAN02 tumor-bearing mice. The tumor volume was monitored twice a week. All doses of DX3-213B were well tolerated with no obvious loss in body weight (Figure 4J). Importantly, DX3-213B showed dose-dependent inhibition of tumor growth, supporting DX3-213B as a promising drug candidate to treat pancreatic cancer.

DX2-201 Resistant Clones Selected from the Glucose-Containing Medium Upregulate Genes Implicated in OXPHOS and Purine Metabolism

To further elucidate the mechanisms of resistance to DX2-201, we also generated resistant clones (DXR) to DX2-201 in cells grown in a glucose-enriched medium. Six single clones were isolated after short-term selection with lethal doses of DX2-201 (Figures 5A and S10). Compared to parental HCT116 cells, DXR cells are only resistant to DX2-201 in the glucose-containing medium but not in the galactose-containing medium (Figure 5B), suggesting the presence of unique mutations responsible for resistance. Moreover, these resistant clones are not only resistant to DX2-201 and its analogues but also resistant to IACS (Figure 5C), suggesting a similar and broad mechanism of drug resistance. These results suggest that although IACS and DX2-201 bind to different targets, the resistance is perhaps due to the upregulation of alternate signaling pathways, a transporter, or a unique escape mechanism.

Figure 5.

DX2-201-treated cells and DX-resistant clones (DXR) show opposite transcriptomic profiles. (A) Dose–response curves for HCT116-parental cells and six HCT116 clones resistant to DX2-201 (DXR clone1-6) in the glucose-containing medium (7 days of treatment). The cell viability was determined using an MTT assay. (B) Dose–response curves for HCT116-parental cells and six HCT116 clones resistant to DX2-201 (DXR clone1-6) in the galactose-containing medium. 3000 cells/well in 96-well plates were treated with indicated concentrations of the compound in a galactose-containing medium, and the cell viability was determined using an MTT assay. (C) Colony formation assay of HCT116-parental cells and DXGR clones treated with DX2-201, DX3-134B, and IACS. 200 cells/well in 96-well plates were treated with indicated concentrations of compounds for 7 days in a glucose-containing medium, then stained with crystal violet, and imaged. (D) Common significantly changed gene sets shared by downregulated gene sets in response to DX2-201 treatment (Bru-seq) and upregulated gene sets in DXR cells (RNA-seq). (E) Gene sets upregulated in DXR cells were downregulated upon DX2-201 treatment. H: Hallmark; K: KEGG. (F) GSEA plots for select gene sets showing opposite trends in DXR cells and in response to DX2-201 treatment. NES = normalized enrichment score; FDR = false discovery rate. (G) Gene mutation frequency in DXR clones. GATK Mutect2 and MuSE v1.0rc were used to analyze the nonsynonymous variants in exonic, splicing, or upstream regions. (H) GNAS expression after siGNAS transfection for 3 days. (I) The ATP production ratio between the galactose- and glucose-containing media was compared with siScramble after GNAS knockdown. The ATP production was determined using the CellTiter-Glo Luminescent Cell Viability Assay 3 days after siRNA transfection. (J) ATP production after 48 h of DX2-201 treatment 3 days after GNAS siRNA transfection. ATP production was determined using the CellTiter-Glo Luminescent Cell Viability Assay.

To further understand this resistance mechanism, we performed RNA-seq of resistant clones (DXR) and the parental HCT116 cells. Surprisingly, we did not observe upregulation of the glycolysis pathway, which can be compensatorily enhanced during OXPHOS inhibition. The top 30 up- and downregulated genes are listed in Tables S13 and S14, and the volcano plot highlights differentially expressed genes (Figure S11). GSEA analysis revealed that >20 gene sets were upregulated in DXR2 (clone 2) in comparison to three gene sets that were significantly downregulated (Tables S15 and S16). A comparison of up- or downregulated gene sets with those modulated by DX2-201 (Tables S1–S4) revealed a significant number of gene sets moving in the opposite direction (p < 1.5 × 10^–14) (Figure 5D). Eleven gene sets suppressed by DX2-201 including “OXPHOS” (Hallmark and KEGG datasets), “Myc targets”, “E2F targets”, “DNA repair”, and “purine metabolism” were upregulated in DXR2 (Figure 5E,F), suggesting that enhancement of these pathways is important to overcome DX2-201 toxicity. In summary, the opposite transcriptional profile of DX2-201 resistant cells to DX2-201 treated cells reinforces the significance of select pathways in DX2-201’s mechanism of action, shedding light on the uniqueness of complex I inhibitors.

GNAS Mutations Observed in DX2-201 Resistant Clones Can Serve as Potential Predictive Biomarkers

To further elucidate the genetic mutations that drive the resistance to DX2-201 in the glucose-containing medium, we sequenced the whole exome of all six resistant clones. We identified a total of 1500 missense mutations in the six DXR clones. Three mutations, including guanine nucleotide-binding protein alpha stimulating activity polypeptide 1 (GNAS), notch receptor 3 (NOTCH3), and synaptotagmin-like 3 (SYTL3) mutations, were shared by three out of six clones (Figure 5G). Among these, GNAS is one of the driver genes in pancreatic cancer²⁴ with an incidence of 8% in PDAC,²² ∼60% in IPMN, and ∼30–50% in PDAC, concomitant with or derived from IPMN.²³ Interestingly, upon knockdown of GNAS, we observed a significant drop in ATP production in the galactose-containing medium but not in the glucose-containing medium, suggesting that GNAS downregulation leads to suppression of OXPHOS (Figure 5H,I). Accordingly, DX2-201 was less sensitive to ATP inhibition in the glucose-containing medium when GNAS was downregulated, indicating that GNAS is important in the sensitivity of DX2-201 (Figure 5J). Although further experiments are needed to validate GNAS mutation in sensitizing cells to complex I inhibitors, our study sheds light on its potential role as a predictive biomarker for future patient selection.

DX2-201 Is Synergistic with Select Metabolic Modulators and PARP Inhibitors and Sensitizes Cancer Cells to Radiation

The plasticity of cancer cells facilitates metabolic adaptation when OXPHOS is inhibited. Thus, combining DX2-201 with select metabolic modulators should have a better therapeutic outcome than single-agent therapy. Our bioinformatics-guided drug combination studies focused on various metabolic modulators targeting compensatory pathways (Figure 6A). For example, combining DX2-201 with a glycolysis inhibitor (2-DG) led to significant synthetic lethality due to shutting down bioenergetic processes. In addition, DX2-201 showed significant synergy with the glutaminase inhibitor (CB-839) and pyruvate dehydrogenase complex inhibitor (CPI-613), suggesting that suppression of the TCA cycle increases the sensitivity to OXPHOS inhibitors (Figure 6B,C). Interestingly, complex I inhibitors including IACS, metformin, and phenformin also showed significant synergy with DX2-201 (Figure 6B,C), further supporting the nonoverlapping targets of these inhibitors.

Figure 6.

DX2-201 is synergistic with select drugs and shows synergy with 2-DG in vivo. (A) Model illustrating the mechanism of action of the synergistic effect between DX2-201 and metabolic modulators targeting different processes of cancer cell metabolism. (B) Pair of compounds showing a combination index (CI) < 1 are considered synergistic. (C) MIA PaCa-2 cells were treated with indicated concentrations of 2-DG, IACS, antimycin, CPI-613, CB-839, and phenformin in combination with DX2-201 for 7 days in the glucose-containing medium and stained with crystal violet. (D) GSEA plots for DNA repair gene sets from Bru-seq and proteomics results. NES = normalized enrichment score; FDR = false discovery rate. (E) Representative images of colonies and surviving fraction after treatment with DX2-201 combined with radiation. MIA PaCa-2 cells were treated with indicated doses of radiation and plated for clonogenic survival analysis. DX2-201 was added the day after seeding for 8 days of continuous treatment. Cells were stained with crystal violet. (F) Combinations with 2-DG overcome the resistance of KPC-2 cells to DX2-201 treatment. Both compounds with indicated concentrations were used to treat KPC-2 cells (200 cells/well in 96-well plates) for 7 days in the glucose-containing medium. Cells were stained with crystal violet. (G) In vivo efficacy of DX3-213B and metformin combined with 2-DG in a KPC-2 allograft mouse model. 2-DG (500 mg/kg), DX3-213B (5 mg/kg from day 1 to day 7; 7.5 mg/kg from day 8 to day 25), and their combination were given to mice through i.p. injection. (H) Tumor weight of allografts at the end of the experiments. (I) In vivo efficacy of sequential treatment in the KPC-2 allograft mouse model. Mice were treated with 2-DG from day 1 to day 10 until the tumor volume reached 300 mm³ and then dosed with 7.5 mg/kg DX3-213B or 250 mg/kg metformin. Compounds were given daily through i.p. injection. The tumor volume at the end of the study was analyzed using Student’s t-test. (J) Body weight of all the treatment groups.

Resistance to radiation features a strong reliance on mitochondrial respiration and diminished dependence on glycolysis. This is perhaps the major reason for the failure of radiotherapy in the clinic.⁴³ There is an urgent need for the discovery of novel radiosensitizers with clinical utility. Our multiomics analysis revealed significant downregulation of DNA repair genes upon treatment with DX2-201 (Figure 6D), supporting the potential benefit of combination treatments with radiation. In MIA PaCa-2 cells, we observed an enhancement ratio of 1.57 (p = 0.0002) when we combined DX2-201 with radiation (Figure 6E). In addition, DX2-201 is synergistic with PARP inhibitors in multiple pancreatic cancer cell lines (Figure S12). These results further reinforce the notion that DX2-201 significantly suppresses the DNA repair machinery. In summary, our mechanism-based drug combination studies revealed unique opportunities to use DX2-201 for combination treatments of pancreatic cancer.

Combination of DX2-201 with 2-DG Overcomes Resistance of Pancreatic Cancers to OXPHOS Inhibitors In Vitro and In Vivo

Metabolic reprograming in PDAC is heterogeneous, and at least three distinct metabolic subtypes were identified, signifying metabolic levels associated with glycolysis, lipogenesis, and redox pathways.⁴⁴ We also observed that not all PDAC cell lines are responsive to OXPHOS inhibitors (Figure 4C,D,H). As synthetic lethality is a promising strategy to overcome drug resistance, we tested the combination of glycolysis inhibitor 2-DG with our OXPHOS inhibitors in vitro and in vivo. Although KPC-2 cells are not sensitive to DX2-201 treatment alone, combination with 2-DG significantly sensitized the cells to DX2-201 treatment (Figure 6F). Due to the dual inhibition of ATP production, the combination regimen showed strong cytotoxicity after 3 days of treatment. Importantly, the combination of 2-DG with DX3-213B was well tolerated in mice as no obvious toxicity or drop in body weight was observed when 10 mg/kg DX3-213B was combined with 500 mg/kg of 2-DG. Metformin (250 mg/kg) was used in parallel experiments for comparison. Consistent with our in vitro studies, neither DX3-213B nor metformin showed tumor growth delay in vivo as a single agent, suggesting that KPC-2 cells are not sensitive to OXPHOS inhibition (Figure 6G). Strikingly, a combination of DX3-213B with 2-DG significantly delayed tumor growth in both treatment groups (Figure 6G,H), suggesting an effective approach to overcome resistance to OXPHOS inhibitors. In addition, as a potent OXPHOS inhibitor in vitro and in vivo, DX3-213B warrants further development.

We also investigated sequential single-agent treatment by treating the mice with 2-DG first and sequentially adding DX3-213B or metformin (Figure 6I). We hypothesized that the tumor cells will adapt to a high dependency on OXPHOS when treated with 2-DG and become sensitive to OXPHOS inhibition. Consistent with our hypothesis, we observed slowed tumor growth when DX3-213B was administered first followed by 2-DG treatment. 2-DG treatment followed by metformin also resulted in a smaller tumor size, but this difference was not statistically significant. All the treatments were well tolerated, and no obvious drop in body weight was observed (Figure 6J).

Taken together, these experiments demonstrate that a combination of 2-DG and sequential treatment is effective to overcome resistance to OXPHOS inhibitors, providing a potential strategy to sensitize the OXPHOS inhibitors.

Discussion

In this study, we report the identification of a first-in-class NDUFS7 inhibitor, DX2-201, for the treatment of pancreatic cancer. DX2-201 inhibits OXPHOS by blocking ubiquinone binding at the NDUFS2 and NDUFS7 interface, suppresses cell proliferation in the nanomolar range, and acts in synergy with ionizing radiation and PARP inhibitors. Importantly, a metabolically stable derivative of DX2-201, DX3-213B, demonstrated significant single-agent efficacy in a mouse syngeneic model of pancreatic cancer. Combination with 2-DG overcame resistance to OXPHOS inhibitors both in vitro and in vivo. Treatment of pancreatic cancer cells with DX2-201 reversed the highly hypoxic and immune-suppressive microenvironment, two hallmarks of pancreatic cancer. In addition, we identified GNAS as a potential predictive biomarker to identify a patient population that would especially benefit from OXPHOS inhibition. Cumulatively, our studies provide a potential strategy to treat pancreatic cancer through OXPHOS inhibition.

Although OXPHOS is a promising target for the treatment of select types of cancers, only a few OXPHOS inhibitors have been tested in patients so far. Unfortunately, the mode of action of many OXPHOS inhibitors remains elusive. For example, biguanides are prototypical complex I inhibitors that have been safely used to treat diabetes;⁴⁵⁻⁴⁸ however, they require millimolar concentrations to show efficacy as anticancer agents. At such high concentrations, their off-target effect becomes pronounced and limits their clinical utility. In addition, the exact targets of bisguanides still remain unclear. Therefore, more potent, and selective OXPHOS inhibitors with well-established mitochondrial targets are in urgent need to effectively treat select hard-to-treat cancers that are uniquely vulnerable to OXPHOS inhibition.

Pancreatic cancer is characterized by a tumor microenvironment that is highly hypoxic and nutrient-deprived, which suppresses antitumor immunity and enhances immune evasion. Severe hypoxic regions within pancreatic cancer closely correlate with tumor progression and poor prognosis. Inhibition of OXPHOS can limit biomass and energy supply to alleviate hypoxia,^10,15 and OXPHOS inhibitors are more effective in tumor cells at the core region of multicellular tumor spheroids with low oxygen and nutrient gradients.^49,50 Accordingly, our study revealed that DX2-201 significantly alleviates hypoxia in 3D tumor spheroids, indicating that complex 1 inhibition can effectively reverse the highly hypoxic environment in pancreatic cancer.

Although most pancreatic cancers are resistant or immune-quiescent tumors and not responsive to single-agent checkpoint treatment, combining OXPHOS inhibitors with immune checkpoint inhibitors targeting different pathways in immune cells can improve efficacy. High PD-L1 levels in pancreatic cancers are strongly correlated with unfavorable prognoses.⁵¹ Our study demonstrates that DX2-201 can reduce PD-L1 protein levels, suggesting that it may maintain high cytotoxic T lymphocyte immune surveillance (not shown). Importantly, immune cells penetrating pancreatic cancers are metabolically heterogeneous, which provides an opportunity to modulate immune cell function by selectively targeting their metabolic dependency.⁵² Glycolysis is essential for the differentiation and function of effector CD8⁺ T cells, Th1, Th2, and Th17 CD4⁺ T cells,⁵³ while OXPHOS is important for immunosuppressive M2 macrophages, regulatory T cells, and myeloid-derived suppressor cells.⁵³⁻⁵⁵ Inhibition of OXPHOS leads to the upregulation of glycolysis, which facilitates the activation of cytotoxic T cells and M1 macrophages while suppressing the function of immunosuppressive cells, leading to enhanced antitumor effects. Our future work will focus on understanding how OXPHOS inhibitors modulate antitumor immune functions and their potential to combine with immune checkpoint inhibitors. We predict that by alleviating hypoxia and targeting immune-suppressive cells, our OXPHOS inhibitors hold great promise to exert enhanced antitumor efficacy by modulating the harsh tumor microenvironment in pancreatic cancer.

Discovery of predictive biomarkers is important for the selection of patients who would best respond to treatment due to the limited therapeutic window of OXPHOS inhibitors. We show for the first time that downregulation of GNAS sensitizes PDAC cells to DX2-201. Therefore, pancreatic cancer patients with GNAS^R201H/C mutations would be predicted to be sensitive to complex I inhibition. However, this hypothesis requires validation on a larger panel of patient-derived cell lines and tissues.

Plasticity of cancer cells as well as CSCs facilitates metabolic reprograming to enable tumor growth. Under such conditions, OXPHOS inhibition is uniquely suited for combination treatments with metabolic modulators to overcome drug resistance. In fact, DX2-201 shows significant synergy with the glutaminase inhibitor CB-839. Previously, it was shown that IACS decreases glucose flux through the TCA cycle, leading to a compensatory increase in glutamine consumption to fuel the TCA cycle.⁵⁶ Our studies show for the first time that a combination of CB-839 with DX2-201 can substantially shut down this bioenergetic process, leading to cell death. Such synergistic drug combination can be quite effective in overcoming drug resistance but requires validation in multiple in vivo models.

DNA damage is energetically costly. PARP-1 activation upon significant DNA damage directly leads to NAD⁺ depletion and subsequently to ATP depletion.⁵⁷ DX2-201 suppresses NAD⁺ and ATP production, leading to impaired DNA repair mechanisms. As a result, inhibition of complex I results in sensitization of cells to radiation and PARP inhibition and this agrees with previous studies.^58,59 However, such a combination modality has not been previously exploited as an effective approach to treat pancreatic cancer.

Achilles’ heel of OXPHOS inhibition is potential toxicity to normal cells requiring intracellular oxygenation. As a result, only a limited number of OXPHOS inhibitors are in clinical trials and none has been approved by the FDA. The recent failed clinical trial of IACS010759, a highly potent and selective complex I inhibitor, indicates the challenges of OXPHOS inhibitors to achieve an acceptable therapeutic index in patients.⁶⁰ However, the establishment of recommended phase 2 dose (RP2D) of IM156, a novel potent OXPHOS inhibitor, for the treatment of solid tumors brings the first evidence that the OXPHOS inhibitors could be administered safely.⁶¹ In our studies, the DX compounds were well tolerated when dosed orally (p.o.) or by intraperitoneal injection (i.p.) at an efficacious dose, and long-term dosing did not lead to obvious organ toxicity in mice. However, the side effects of DX analogues should be carefully monitored when given to patients considering the inconsistency of the toxicity interpretations in preclinical models and humans.

In conclusion, we report the identification of a novel OXPHOS inhibitor that shows efficacy in syngeneic models of PDAC as a single agent and in combination with 2-DG. In-depth mechanistic studies revealed a novel binding pocket of complex I at the interface of NDUFS2 and NDUFS7, providing new mechanistic insights as to how the function of complex I is inhibited by our compound. Our study also provides evidence that OXPHOS inhibitors suppress pancreatic cancer through not only the direct killing of tumor cells but also indirectly by improving the tumor microenvironment. We further address the potential clinical implication of our complex I inhibitor by exploring the synergism with standard-of-care chemotherapy and several FDA-approved drugs. Our in-depth study of resistant cells reveals novel resistant mechanisms and provides potential biomarkers to select sensitive tumor types for future clinical studies. Importantly, our study demonstrates that a combination with 2-DG could overcome resistance, which provides an effective strategy to enhance the efficacy of OXPHOS inhibitors. Together, our studies validate inhibition of OXPHOS through NDUFS7 for the treatment of pancreatic cancer. Using synthetic lethality and predictive biomarkers as strategies to increase the efficacy of OXPHOS inhibitors is feasible and can be further evaluated in clinical trials.

Materials and Methods

Cell Lines

Unless otherwise specified, all cell lines (MIA PaCa-2, BxPC-3, PANC-1, KPC-2, UM16, HFF, HL-60, K-562, Raji, CT26, HCT116, ID8, OVCAR-3, OVCAR-8, HEY, and SKOV-3) were cultured in RPMI 1640 supplemented with 10% fetal bovine serum (FBS, Gibco). Pancreatic cancer cell lines MIA PaCa-2, BxPC-3, and PANC-1 were obtained from ATCC, and the ovarian cancer OVCAR-3, OVCAR-8, and SKOV-3 cells were obtained from the National Cancer Institute, Developmental Therapeutics Program. U-87 MG cells were provided by Dr. Alan L. Epstein (University of Southern California) and were cultured in Dulbecco’s minimal essential medium (DMEM) supplemented with 10% FBS. iKRAS cells provided by Dr. Alnawaz Rehemtulla (University of Michigan) were cultured in DMEM with 10% FBS and 1 μg/ml doxycycline. All UM16 and HPDE cells were provided by Dr. Diane Simeone (New York University). HPDE cells were maintained in a keratinocyte serum-free medium, supplemented (Invitrogen) with EGF and bovine pituitary extract. The high-glucose medium was replaced with 10 mM galactose where indicated. Cells were kept at 37 °C in a humidified atmosphere of 5% CO₂. All cells were maintained in culture under 30 passages and tested regularly for Mycoplasma contamination using PlasmoTest (InvivoGen).

Cell Viability Assays

Cells were seeded in 96-well plates at 200–500 cells/well overnight and treated with the indicated concentration of compounds or DMSO for 7 days. For glucose- and galactose-containing media, cells were seeded in 25 mM glucose or 10 mM galactose and treated with the indicated concentration of compounds or DMSO the next day. Cells were then grown for 72 h, and cell viability was assessed using the MTT assay.

Generation of DX2-201-Resistant Clones

To generate DX2-201-resistant clones in a glucose-containing medium (DXR), HCT116 cells were seeded at a density of 100 cells/well in 96-well plates in a glucose-containing medium and treated with 5 μM DX2-201 for 3 weeks. The medium with 5 μM DX2-201 was changed twice a week. Twenty-seven resistant clones were isolated and were seeded at 200 cells/well in 96-well plates in a 100 μL glucose-containing medium. The indicated concentration of DX2-201 was added the next day, and the cells were grown for 7 days. Colony formation and MTT assays were performed to determine cell viability, and 10 clones showed resistance to DX2-201. Four of the 10 resistant clones were further excluded due to significant differences in the growth rate.

To generate DX2-201 resistant clones in the galactose-containing medium (DXGR), HCT116 cells (1 × 10⁶ cells/plate) were seeded in 10 cm dishes in the galactose-containing growth medium and treated with 5 μM DX2-201 for 3 weeks. The medium with 5 μM DX2-201 was changed twice a week. Twelve resistant clones were isolated and were seeded at 5 × 10³ cells/well in 96-well plates in a 100 μL galactose-containing medium. After cells became fully attached, indicated concentrations of DX2-201 or IACS-010759 were added for 3 days. Cell viability was then assessed by the MTT assay. Six of the resistant clones were picked and further seeded at 200 cells/well in 96-well plates in a 100 μL glucose-containing medium. Indicated concentrations of DX2-201, IACS-010759, and other OXPHOS inhibitors were added the next day, and the cells were grown for 7 days. MTT was performed to determine cell viability.

Whole Exome Sequencing

Genomic DNA purification was performed using the DNeasy Blood & Tissue Kit (Qiagen, 69504). For whole exome sequencing, 1.0 μg of genomic DNA per sample was used as the input material for DNA sample preparation. Sequencing libraries were generated using the Agilent SureSelect Human All ExonV6 kit (Agilent Technologies, CA, USA) following the manufacturer’s recommendations, and index codes were added to attribute sequences to each sample. In brief, fragmentation was carried out by a hydrodynamic shearing system (Covaris) to generate 180–280 bp fragments. The remaining overhangs were converted into blunt ends via exonuclease/polymerase activities, and then enzymes were removed. After adenylation of 3′ ends of DNA fragments, adapter oligonucleotides were ligated. DNA fragments with ligated adapter molecules on both ends were selectively enriched in a PCR. Captured libraries were enriched in a PCR to add index tags to prepare for hybridization. Products were purified using the AMPure XP system (Beckman Coulter) and quantified using the Agilent high-sensitivity DNA assay on an Agilent Bioanalyzer 2100 system. Exome libraries were subjected to 100X exome sequencing (Novogene).

Exome-Seq Analysis

Six DX2-201 resistant clones and two parental cell lines with the same treatment were profiled so that variant calling could be run in resistant vs wild-type cells. Picard 2.4.1 was first used to convert fastq files to unaligned .bam files. BWA mem 0.7.15 was used to align reads to GRCh38-v29.⁶² Picard was used to mark and remove duplicate reads, and GATK 3.6 was used to recalibrate base quality scores.⁶³ GATK Mutect2 and MuSE v1.0rc were used as part of GATK 3.6 to call variants present in resistant clones versus parental cell lines for genomic intervals covering 1/10 of each chromosome.^63,64 Standard Mutect2 and MuSE variant caller-specific prefiltering was used, and then additional filtering was performed requiring calls to be supported by 10 total reads in both resistant and wild-type samples, with at least 5 reads supporting the variant and no more than 1 read supporting the reference sequence. Annovar Feb 2016 was used to annotate variants using Apr 2019 refgene annotations.⁶⁵ Only nonsynonymous variants in exonic, splicing, or upstream regions were considered. Our analysis focused on variants called with both Mutect2 and MuSE.

Confocal Microscopy Imaging

Cells were seeded in chamber slides at low density, and live cells were loaded with 100 nM MitoTracker Red CMXRos (M7512, Thermo Fisher Scientific) for 45 min after 4 h of treatment of the BODIPY-labeled DX2-201 probe. The probe was synthesized using similar methods as described before.⁶⁶ After several washes, cells were stained with Hoechst 33342 (H3570, Thermo Fisher Scientific) for 15 min and then fixed in 4% paraformaldehyde in PBS. Coverslips were mounted in the proLong gold antifade mountant (P36934, Thermo Fisher Scientific). Cells were visualized under an A1Si laser scanning confocal microscope (Nikon) and processed using ImageJ software (NIH).

Clonogenic Assay

MIA PaCa-2 cells were irradiated with indicated doses and plated the following day at a clonogenic density in a fresh medium with indicated concentrations of DX2-201 for 7 to 14 days before the colonies were fixed with 4% paraformaldehyde (Electron Microscopy Science) and stained using 0.1% crystal violet (Sigma-Aldrich).

Bru-Seq Analysis of Nascent RNA

Bru-seq analysis was performed as previously described.⁶⁷ In brief, after UM16 cells (80–90% confluence) were treated with DX2-201 for 3.5 h, bromouridine (final concentration of 2 mM) was added to the media to label newly synthesized nascent RNA for 30 min. Cells were collected in TRIzol (Invitrogen), and total RNA was isolated. Bru-labeled, nascent RNA was isolated using anti-BrdU antibody capture, converted into cDNA libraries, and sequenced using an Illumina HiSeq 2000 sequencer at the Advanced Genomics Core, University of Michigan. Sequencing reads were mapped to the hg38 reference sequence. Ensemble gene identifiers were mapped to HGNC symbols and entrez identifiers using Gencode v27 annotations.⁶⁸ Only measurements mapping to protein-coding genes with entrez identifiers were considered, and gene changes with count >100 and mean RPKM > 0.1 were analyzed.

TMT Labeling and Liquid Chromatography–Mass Spectrometry Analysis

UM16 cells treated with 5 μM DX2-201 and DMSO were lysed with RIPA buffer (Thermo Fisher Scientific, 89901), followed by sonication and centrifugation at 10,000g for 15 min at 4 °C. The supernatant was saved, and protein concentration was determined by the BCA assay (Thermo Fisher Scientific, 23225). 75 μg of protein was used to perform tandem Mass Tag (TMT) labeling with the TMT 6-plex isobaric labeling kit (Thermo Fisher Scientific) according to the manufacturer’s protocol. In brief, upon reduction and alkylation of cysteines, the proteins were precipitated by adding 6 volumes of ice-cold acetone followed by overnight incubation at −20° C. The precipitate was spun down, and the pellet was allowed to air-dry. The pellet was resuspended in 0.1 M TEAB and digested overnight with trypsin (1:50; enzyme/protein) at 37 °C with constant mixing using a thermomixer. The TMT 6-plex reagents were dissolved in 41 μL of anhydrous acetonitrile, and labeling was performed by transferring the entire digest to the TMT reagent vial and incubating at room temperature for 1 h. The reaction was quenched by adding 8 μL of 5% hydroxyl amine and a further 15 min of incubation. Labeled samples were mixed together and dried using a vacufuge. An offline fractionation of the combined sample (∼200 μg) into eight fractions was performed using a high pH reversed-phase peptide fractionation kit according to the manufacturer’s protocol (Pierce, 84868). Fractions were dried and reconstituted in 9 μL of 0.1% formic acid/2% acetonitrile in preparation for LC–MS/MS analysis.

To obtain high quantitation accuracy, we employed multinotch-MS3 (McAlister GC) which minimizes the reporter ion ratio distortion resulting from the fragmentation of coisolated peptides during MS analysis. Orbitrap Fusion (Thermo Fisher Scientific) and RSLC Ultimate 3000 nano-UPLC (Dionex) were used to acquire the data. 2 μL of the sample was resolved on a PepMap RSLC C18 column (75 μm i.d. × 50 cm; Thermo Scientific) at the flow rate of 300 nL/min using a 0.1% formic acid/acetonitrile gradient system (2–22% acetonitrile in 150 min; 22–32% acetonitrile in 40 min; 20 min wash at 90%, followed by 50 min of re-equilibration) and directly sprayed onto the mass spectrometer using an EasySpray source (Thermo Fisher Scientific). The mass spectrometer was set to collect one MS1 scan (Orbitrap; 120K resolution; AGC target 2 × 105; max IT 100 ms) followed by data-dependent, “Top Speed” (3 s) MS2 scans (collision-induced dissociation; ion trap; NCE 35; AGC 5 × 10³; max IT 100 ms). For multinotch-MS3, the top 10 precursors from each MS2 were fragmented by HCD, followed by Orbitrap analysis (NCE 55; 60K resolution; AGC 5 × 104; max IT 120 ms, 100–500 m/z scan range).

Proteome Discoverer (v2.4, Thermo Fisher) was used for data analysis. MS2 spectra were searched against the SwissProt human protein database using the following search parameters: the MS1 and MS2 tolerances were set to 10 ppm and 0.6 Da, respectively; carbamidomethylation of cysteines (57.02146 Da) and TMT labeling of lysine and N-termini of peptides (229.16293 Da) were considered static modifications; oxidation of methionine (15.9949 Da) and deamidation of asparagine and glutamine (0.98401 Da) were considered variable. Identified proteins and peptides were filtered to retain only those with FDR ≤ 1%. Quantitation was performed using high-quality MS3 spectra (average signal-to-noise ratio of 10 and <30% isolation interference). Protein changes with absolute fold change >1.5 and qval <0.1 were considered significant.

RNA-Seq Profiling

HCT116 parental and DXR cells were lysed with TRIzol Reagent (Thermo Fisher Scientific) at room temperature. RNA was further purified with the DirectZol kit (Zymo Research). RNA quality was assessed using the TapeStation (Agilent Technologies). Samples with RNA integrity numbers of 8 or greater were prepared with TruSeq Stranded mRNA Library Prep (Illumina) as per the supplier’s protocol with 1 μg of RNA and 12 cycles of PCR amplification. Libraries were checked for size on the TapeStation and quantified using the Kapa Biosystems library quantification kit (Illumina). The libraries were barcoded, pooled, and sequenced using paired-end 150 bp sequencing (Novogene). Reads were mapped to GRCh38 using STAR v2.5.2,⁶⁹ and gene quantifications were calculated using Cufflinks v2.2.1⁷⁰ to quantify refGene annotations. Gene read counts were calculated using featureCounts⁷¹ v1.6.1 and were used to evaluate the differential expression using DESeq2 v1.18.1.⁷² Protein coding genes were considered significantly differentially expressed with a mean FPKM > 0.5, an absolute fold change > 1.5, and an FDR adjusted p-value < 0.05. A volcano plot was generated using R with fold changes and q-values sourced from DESeq2 output. All gene readouts were required to be mappable to both an HGNC and entrez identifier to be considered for gene set enrichment analyses. This data has been deposited in NCBI’s Gene Expression Omnibus and is accessible through GEO Series accession number GSExxx (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSExxx).⁷³

Bru-Seq and RNA-Seq Gene Set Enrichment

Gene set enrichment was performed using a rlogFC rank-ordered list with genes >10 normalized counts. GSEAv2.2.3 was used with v6.1 gene sets sourced from MSigDB to identify gene sets significantly up- or downregulated. 10,000 gene set permutations were performed using weighted-mode scoring of genes ranked using the log fold-change APEGLM method.⁷⁴ Gene sets with FDR-adjusted p-values < 0.1 were considered significant.

Gene Knockdown

Two siRNA oligonucleotides specific for human NDUFS7 (hs.Ri.NDUFS7.13.1 and hs.Ri.NDUFS7.13.2) were purchased from Integrated DNA Technologies. Lipofectamine RNAiMAX Transfection Reagent (Thermo Fisher Scientific) was used for siRNA transfection according to the transfection workflow and protocol described by the manufacturer.

Western Blot

Western blotting was performed to assess protein expression changes in response to DX2-201 treatment. Cells were lysed in RIPA buffer (Sigma-Aldrich, R0278) containing protease and phosphatase inhibitors.

Determination of the NAD⁺/NADH Ratio

MIA PaCa-2 cells were seeded at a density of 5 × 10³ cells/well in flat-bottom 96-well plates. The next day, cells were treated with indicated concentrations of compounds diluted in RPMI 1640 with 10% FBS for 48 h. The medium was aspirated, and cells were washed with PBS. After suspension in 100 μL of PBS, cells were lysed by adding 100 μL of bicarbonate base buffer with 1% DTAB. Then, samples were divided into separate wells for acid and base treatments as described in the NAD/NADH-Glo Assay protocol (G9071). In brief, to measure NAD⁺, 25 μL of 0.4 N HCl was added to 50 μL of the cell lysate and heated at 60 °C for 15 min. To measure NADH, 50 μL of the cell lysate was heated at 60 °C for 15 min directly. After heating, both samples were incubated at room temperature for 10 min. Trizma base (25 μL) was added to each well of 0.4 N HCl-treated samples (NAD⁺), and the HCl/Trizma solution (50 μL) was added to each well of untreated samples (NADH). The intensity of NAD⁺ or NADH was measured using the NAD/NADH-Glo assay. Both NAD⁺ and NADH samples (10 μL) were incubated with 10 μL of the NAD/NADH-Glo detection reagent in white 384-well luminometer plates. After a 30 min incubation, luminescence was measured on a Synergy H1 Hybrid Multi-Mode Reader (BioTek).

Animal Studies

Pan02 syngeneic model: 5 × 10⁶ Pan02 cells in 100 μL of RPMI containing 50% Matrigel were implanted subcutaneously into the right flanks of 6–10 week-old female C57BL/6 mice. Once the tumor size reached 100 mm³, the animals were randomized and treated with 7.5, 2.5, or 1 mg/kg of DX3-213B and 7.5 mg/kg of IACS-010759. DX3-213B was dissolved in 10% DMSO, 50% PG, and 40% saline, while IACS-010759 was suspended in 0.5% methylcellulose. Compounds were dosed daily for 28 consecutive days. Tumor growth was monitored twice a week by a digital caliper, and tumor volumes were calculated by the (length × width²)/2 equation. The body weights of mice were monitored before dosing and twice a week. Mice were sacrificed after 28 days of compound administration. The University of Michigan Institutional Animal Care and Use Committee approved all animal experiments.

KPC-2 syngeneic model: 1 × 10⁶ KPC-2 cells in RPMI without FBS were implanted subcutaneously into the right flanks of 6 week-old female FVB mice. Once the tumor size reached 100 mm³, the animals were randomized and treated with DX3-213B (5 mg/kg from day 1 to day 7; 7.5 mg/kg from day 8 to day 25) and metformin (250 mg/kg), as a single agent or in combination with 2-DG (500 mg/kg). DX3-213B was dissolved in 10% DMSO, 50% PG, and 40% saline, while metformin and 2-DG were dissolved in saline. For sequential combinations, mice were dosed with 2-DG from day 1 to day 10 until the tumor volume reached 300 mm³ and then dosed with 7.5 mg/kg DX3-213B or 250 mg/kg metformin. Compounds were dosed i.p. daily. Tumor growth was monitored twice a week by a digital caliper, and tumor volumes were calculated by the (length × width²)/2 equation. The body weights of mice were monitored before dosing and twice a week. Mice were sacrificed after 25 days of compound administration.

Statistics

Statistical analyses were performed using the two-tailed Student’s t-test. A p-value < 0.05 was considered statistically significant. The 50% inhibitory concentration values (IC₅₀) were determined by analyzing the log of the concentration–response curves by nonlinear regression analysis using GraphPad Prism (version 5).

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsptsci.3c00069.

• Biology data further validating the mechanism of action of DX2-201, bioinformatics analysis, and list of top up- and downregulated genes and proteins in response to DX2-201 treatment (PDF)

Author Contributions

Y.X. and D.X. contributed equally. N.N., Y.X., and D.X. conceived the project. Y.X., D.X., A.K., and J.R. performed experiments. Y.X., D.X., A.K., A.B., and M.L. analyzed the data and generated figures. Y.X., D.X., and A.B. wrote the manuscript. N.N., Y.X., A.B., and M.L. reviewed and edited the paper. N.N. designed the study, acquired funding, and supervised the project.

This work was supported by the NIH grants R01 CA188252 and R01 CA272641.

The authors declare no competing financial interest.