Glutamine metabolism genes are upregulated in PDA
Enzymes utilized for glutamine metabolism in PDA, GLS1, GOT1/2, and malic enzyme 1 (ME1) (see pathway, Fig. 1a), in addition to NQO1, were significantly upregulated in PDA compared to 17 other cancers when assessed using the Oncomine webtool (Fig. 1b) [8]. This was apparent in both cell lines and tumor samples. In contrast, GLUD1, which diverts glutamine carbon away from the GOT2–GOT1–ME1 pathway into an alternate metabolic pathway, was not upregulated in PDA relative to other cancer types (Fig. 1a). Additionally, glutamine metabolic enzymes, NQO1, and GLUD1 were found to be significantly upregulated in PDA relative to normal pancreatic tissue (Fig. 1a). To determine the clinical relevance associated with the PDA glutamine metabolic pathway relative to other enzymes involved in glutamine metabolism, we evaluated the association of individual gene expression levels with overall survival in the data set that contained clinical follow-up information [23]. Kaplan–Meier analysis did not show a significant difference in outcome when patients were separated into high and low expression levels of the genes of interest (data not shown). However, when GOT1 and GOT2 were normalized to GLUD1 expression, we found that a high GOT1 to GLUD1 or GOT2 to GLUD1 ratio was significantly associated with poor outcome (Fig. 1c, Additional file 1: Figure S1B). These data suggest that gene expression of GOT isoforms are generally elevated in PDA and have prognostic and functional significance relative to GLUD1 expression.
Inhibiting glutamine metabolism sensitizes PDA to ß-lap
Given the reliance of PDA on glutamine metabolism for redox balance, we hypothesized that glutamine deprivation of NQO1-overexpressing PDA cells would sensitize them to ß-lap exposure by lowering anti-oxidant defenses and increasing NQO1-induced ROS damage. MiaPaCa2 cells were grown in Gln-free or Gln-containing (2 mM) media for 16 h and then exposed to ß-lap for 2 h. Short-term Gln deprivation did not significantly alter clonogenic survival on its own (Fig. 2a) but did sensitize MiaPaCa2 cells to ß-lap at sub-lethal and higher doses of the drug (Fig. 2a). To confirm these results, we repeated this experiment in five other PDA cell lines: ASPC1, MPanc96, HPAFII, SW1990, and DAN-G (Fig. 2b–f). Additionally, we demonstrated that the observed cytotoxicity was NQO1-dependent, and as addition of the potent NQO1 inhibitor, dicoumarol (DIC), spared cells from lethality (Fig. 2b–f) [24–26].
Next, RNAi-mediated knockdown of glutamine metabolism enzymes revealed that GLS1, GOT1, and ME1 dramatically sensitized MiaPaCa2 and ASPC1 PDA cell lines to ß-lap, relative to non-targeting control (scramble small interfering RNA (siScr)) (Fig. 2g–k). Consistent with the mechanism by which glutamine is metabolized in PDA to maintain redox balance [8], knockdown of GLUD1 had no effect on ß-lap sensitivity (Fig. 2i). Sensitization of ß-lap-treated MiaPaCa2 cells by GLS1 knockdown was rescued by replenishing metabolic substrates of the glutamine metabolism pathway that are downstream of the GLS1 reaction, namely, oxaloacetate (OAA) or cell-permeable dimethyl malate (Fig. 2j, k). These data indicate that PDA cells have an increased reliance on glutamine to generate NADPH (Fig. 1a) in the presence of ß-lap-induced ROS stress.
GLS1 inhibition by BPTES sensitizes PDA to ß-lap in an NQO1-dependent manner
To pharmacologically replicate the ß-lap sensitization to inhibition of Gln metabolism, MiaPaCa2 cells were treated with a sub-lethal dose of the mitochondrial GLS1 inhibitor, BPTES (500 nM, 48 h, Additional file 2: Figure S2) and then exposed to various doses of ß-lap for 2 h, with or without DIC (Fig. 3a). BPTES pre-treatment in combination with ß-lap significantly reduced clonogenic survival versus ß-lap alone, while addition of DIC spared the lethality (Fig. 3a). To confirm that our results were due to inhibition of the glutamine-dependent transamination pathway and not the alternative glutamine metabolism pathway through GLUD1 , we pre-treated MiaPaCa2 cells with epigallocatechin gallate (EGCG), an inhibitor of GLUD1 [27] for 48 h and then exposed them to ß-lap. Consistent with our RNAi results (Fig. 2i), we found that GLUD1 inhibition by EGCG had no effect on ß-lap sensitivity (Fig. 3b). Furthermore, normal human IMR-90 embryonic lung fibroblasts, which have low NQO1 levels [26] were not affected by ß-lap, with or without BPTES treatments (Fig. 3c). Replenishing the NADPH-producing transamination pathway with the addition of OAA or dimethyl malate, metabolites downstream of GLS1, rescued BPTES-dependent hypersensitivity to ß-lap in MiaPaCa2, ASPC1, and HPAFII PDA cells (Fig. 3d).
Mutant KRAS signaling drives NQO1 expression and glutamine dependence [7, 8, 28, 29], thus we sought to assess the generality of mutant KRAS expression on the effects of BPTES and ß-lap. BPTES pre-treatment sensitized a variety of NQO1-expressing KRAS mutant cancer cell lines to ß-lap, including lung, triple-negative breast, and additional PDA cell lines. In contrast, NQO1-deficient KRAS mutant lines remained resistant to ß-lap, whether or not BPTES was added (Fig. 3e). In addition, pre-treatment with 500 nM BPTES for 48 h did not increase the sensitivities of ß-lap-responsive, NQO1-expressing KRAS wild-type lung, breast, or pancreatic cancer cell lines, consistent with reported literature [8, 30]. Collectively, these results illustrate that in order for this targeted combination of agents to be effective, sensitive cells must exhibit both mutant KRAS-driven Gln dependence and NQO1 expression. The latter is a feature of most, but not all, mutant KRAS-transformed cancer types.
GLS1 inhibition attenuates anti-oxidant defenses and increases susceptibility to ß-lap-induced DNA damage
We observed a dose-dependent increase in NADP+/NADPH ratios, a proxy for the cell’s oxidative state, in MiaPaCa2 cells exposed to ß-lap alone, reaching fourfold higher levels versus baseline found in untreated cells (Fig. 4a). With BPTES pre-treatment, we noted a sevenfold increase in NADP+/NADPH ratios in ß-lap-exposed cells versus baseline levels in untreated MiaPaCa2 cells (Fig. 4a) and an ~twofold higher level than in cells exposed to ß-lap alone. BPTES pre-treatment (500 nM, 48 h) also significantly lowered reduced glutathione (GSH) levels in MiaPaCa2 cells compared to DMSO vehicle alone (>twofold; Additional file 3: Figure S3A), and extracellular H2O2 production was dramatically increased after BPTES plus ß-lap treatment in a time- and dose-dependent manner (Fig. 4b, Additional file 3: Figure S3B). Additionally, total intracellular ROS levels were dramatically increased after BPTES plus ß-lap treatment (Additional file 3: Figure S3C). Consistent with the kinetics of H2O2 production, we observed a decrease in the minimum time to death [31, 32] for ß-lap-induced lethality in MiaPaCa2 with BPTES pre-treatment (Fig. 4c). This effect on clonogenic survival could be rescued with the anti-oxidant reduced diethyl-ester GSH in ß-lap-exposed MiaPaCa2, ASPC1, HPAFII, and MPanc96 cells after 48 h pre-treatment with BPTES plus 2 h with ß-lap (Fig. 4d).
Pre-treatment with BPTES followed by exposure to ß-lap synergistically increased total DNA lesions in ASPC1 cells, as assessed by alkaline comet assay immediately after 2-h treatment, and DNA double-strand break (DSB) formation in MiaPaCa2 cells, monitored by 53BP1 foci formation 24 h after treatment (Figs. 4e–g). Furthermore, pre-treatment with BPTES followed by treatment with ß-lap dramatically increased PARP hyperactivation noted by concomitant NAD+ depletion and PAR formation, which was blocked with the addition of the PARP inhibitor, Rucaparib (AG014699), in MiaPaCa2 cells compared to either treatment alone (Fig. 4h, i). These data indicate that PDA cells are reliant on glutamine for redox balance and that the disrupted redox state enhances ß-lap-induced DNA damage and PARP-driven metabolic catastrophe.
GLS1 inhibition sensitizes PDA to ß-lap in vivo
To determine whether pharmacologic inhibition of GLS1 in combination with ß-lap would lead to synergistic inhibition of PDA tumor growth in vivo, we utilized the clinical formulation of ß-lap (ARQ761), hydroxypropyl beta cyclodextrin travel (HPßCD)-ß-lap, and the orally available GLS1 inhibitor, CB-839, provided by Calithera Biosciences. Both compounds are in separate phase I/II clinical trials for a variety of cancer types (NCT01502800, NCT02071862, NCT02071888, and NCT02071927) [11]. CB-839 was employed for these studies because BPTES has poor metabolic stability and low solubility in vivo [11]. First, we confirmed that CB-839 pre-treatment, like BPTES, also sensitized PDA cell lines in vitro to ß-lap in the MiaPaCa2 and ASPC1 lines (Fig. 5a). Next, we generated subcutaneous tumors from human MiaPaCa2 cells injected into the right hind limb in Nu/Nu female athymic mice and allowed the tumors to grow to a volume of 100 mm3 before beginning treatment. The mice were sacrificed when tumor volumes reached 1000 mm3, as per the Institutional Animal Care and Use Committee (IACUC) regulations.
The animals received either vehicle (i.e., HPßCD) or CB-839 (200 mg/kg) by oral gavage twice a day for 10 days, with or without a sub-efficacious dose of ß-lap (25 mg/kg) administrated intravenously (IV) every other day (Fig. 5b, arrows) [33]. After only one regimen of treatment (10 days), we found that the mice treated with CB-839 plus ß-lap displayed significantly delayed tumor growth compared to either agent alone through day 60. Importantly, we noted that neither CB-839 (200 mg/kg) nor ß-lap (25 mg/kg) administered alone significantly altered tumor growth (Fig. 5b). Individual tumors are presented in waterfall plots in Additional file 4: Figure S4A. Notably, combination treatment did not decrease mouse weights when compared to vehicle-treated mice (Additional file 4: Figure S4B). Long- or short-term toxicities, including hemolysis and methemoglobinemia, were not observed.
The mice were sacrificed when their original body weights dropped by one third, tumor volumes exceeded 1000 mm3, or when tumors began to ulcerate or impede normal motion. Kaplan–Meier curves showed a significant anti-tumor effect of the drug combination (Fig. 5c). The treatment with a sub-efficacious dose of ß-lap (25 mg/kg) resulted in a median survival of 49 days, while the vehicle-treated group displayed a median survival time of 39.5 days (Fig. 5c). The treatment with a sub-efficacious dose of CB-839 resulted in a median survival time of 42 days. The treatment with the ß-lap plus CB-839 combination resulted in a median survival of 64 days, significantly extending median survival by 24.5 days compared to the vehicle (HPßCD)-treated group (Fig. 5c).
Importantly, to ensure that the anti-tumor efficacy we observed was due to on-target effects of both drugs and by the same mechanism of action observed in vitro, we analyzed the pharmacodynamics profile of each agent alone and in combination. Briefly, the MiaPaCa2 tumor-bearing mice received either vehicle (HPßCD) alone or 200 mg/kg CB-839 by oral gavage twice a day for 4 days, with or without a single sub-efficacious dose of ß-lap at 25 mg/kg IV on day 4. After the last dose of CB-839 and 30 min after ß-lap injection, the mice were sacrificed and the tumor tissue was harvested. Tumor glutamate levels were measured from multiple animals for each treatment condition (Additional file 4: Figure S4C). CB-839 and CB-839 plus ß-lap-treated mice displayed significantly lower overall tumor glutamate levels when compared to vehicle or ß-lap alone treatments, consistent with GLS1 inhibition in vivo (Additional file 4: Figure S4C).
We then immunoblotted tumor tissue lysates for PAR polymer formation as a proxy for PARP hyperactivity and γH2AX to assess DNA DSBs (Fig. 5d, see quantification in Additional file 4: Figure S4D) [33]. Consistent with our results in vitro, we found that the mice treated with the combination of ß-lap plus CB-839 displayed dramatically increased PAR and γH2AX formation relative to all other groups (Fig. 5d). Additionally, we harvested liver tissue (“normal tissue” in Fig. 5d) from the mice after exposure to the combination-treated mice and found no evidence of PAR or γH2AX formation (Fig. 5d), consistent with a lack of response to ß-lap in normal tissues [24, 33]. Notably, the liver contains the highest levels of NQO1 in the mouse (not in human livers) and is used as a surrogate for normal tissue responses to NQO1-bioactivatable drugs.
Next, to determine the redox status of tumors after treatment, we measured the GSH to glutathione disulfide (GSSG) ratio of tumor lysates after vehicle alone or CB-839, with or without 30-min ß-lap treatment. Interestingly, we found that the GSH to GSSG ratios were significantly decreased in the single-arm CB-839 or ß-lap-treated mice (Fig. 5e). Moreover, the combination treatment resulted in an even greater decrease in GSH, monitored by the GSH to GSSG ratio (Fig. 5e). Taking our pharmacodynamics observations and anti-tumor studies together, these data demonstrate that modulating glutamine metabolism in PDA in vivo results in a significantly decreased anti-oxidant defense state. As a result, the drug combination significantly sensitized NQO1-expressing tumors, but not associated normal tissue, to ROS induction from ß-lap leading to DNA damage, PARP hyperactivation, and tumor-selective death.