Blockade of glutamine-dependent cell survival augments antitumor efficacy of CPI-613 in head and neck cancer

Human HNSCC cell lines (HN6, HN12, HN13, HN30, and HN31) were maintained in our lab and used for experiments before passage 10. All cell lines were cultured in DMEM medium containing 10% fetal bovine serum at 37 °C in a humidified incubator with 5% CO₂. Western blotting, plasmid transfection, lentiviral infection, cell proliferation, and colony formation assays were carried out as we previously described [7, 21, 22].

Texas-red phalloidin was purchased from Invitrogen (Carlsbad, CA). CPI-613 and CB-839 were obtained from Selleckchem (Houston, TX). Antibodies against GLS1, GLUD1, HKI, HKII, PKM1, PKM2, PFKP, PDH, α-KGDH and p-PDHA1 (Ser293) were purchased from Cell Signaling Technology (Beverly, MA). 3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) and an antibody that recognizes β-Actin were obtained from Sigma-Aldrich (St Louis, MO). For gene knockdown, the pLKO.1-puro TRC control shRNA targeting the gene encoding green fluorescent protein (GFP) and shRNAs targeting the GLS1 gene were purchased from Horizon Discovery (Waterbeach, UK).

For cultured cells, apoptosis was assessed by flow cytometry using fluorescein isothiocyanate (FITC) Annexin V Apoptosis Detection Kit (BD Biosciences, San Jose, CA) with propidium iodide (PI). In xenograft tumor sections, apoptosis was determined by the DeadEnd™ Fluorometric TUNEL System (Promega, Madison, WI). All TUNEL-positive cells in each section were counted in twenty randomly selected fields using a fluorescence microscope (Zeiss, Oberkochen, Germany).

The activity of PDH in HNSCC cells was measured by a colorimetric assay according to the previously described method [23]. Briefly, 1 × 10⁴ cells were seeded into 96-well plate overnight before being permeabilized with 0.5% Triton X-100. Cells were incubated in 50 mM Tris-HCl (pH 7.8) buffer supplemented with 3-bromopyruvate (5 mM), MgCl₂ (1 mM), EDTA (0.05 mM), Triton X-100 (0.0025%), thiamine diphosphate (0.3 mM), rotenone (10 μM), pyruvate (10 mM), NAD⁺ (3 mM), Co-A (1 mM), nitroblue tetrazolium (0.75 mM) and phenazine methosulfate (0.05 mM) for 45 min at 37 °C. Insoluble blue formazan precipitates were dissolved in 10% SDS in 0.01 M HCl overnight and absorbance was monitored at OD 570 nm.

CellTiter-Glo® 2.0 Assay Kit was used to determine the ATP amount in cell lysates. Glucose-Glo™ Assay Kit and Lactate-Glo™ Assay Kit (Promega, Madison, WI) were used to measure glucose uptake and lactate levels in cell supernatants, respectively. Glutamine/Glutamate-Glo Assay (Promega) was used to assess the changes in glutamine and glutamate content in cells with indicated treatments. Changes in the level of total intracellular GSH were determined using Glutathione Detection Assay Kit (Fluorometric, Abcam, MA). For GSH determination, HN6 and HN31 cells were treated with or without 100 μM CPI-613 for 24 h, and the lysates were collected and incubated on ice for 30 min. The supernatants were collected after centrifugation at 12,000×g for 10 min and processed according to the kit protocol, and fluorescence was measured at 380/461 nm (excitation/emission wavelengths).

Oxygen consumption rate (OCR) was measured using Seahorse XF Cell Energy Phenotype Test Kit on a Seahorse XF96 Extracellular Flux analyzer (Agilent Technology). Briefly, 4 × 10³ cancer cells were seeded on each well of poly-D-lysine coated XF96 miniplates and cultured in DMEM containing 10% FBS overnight, then kept in 100 μl XF medium for 45 mins (XF base medium containing 10 mM glucose, 2 mM L-glutamine, and 1 mM sodium pyruvate). Three OCR measurements were obtained under basal conditions and upon sequential injection of 1 μM oligomycin (Oligo) and 1 μM fluoro-carbonyl-cyanide phenylhydrazone (FCCP). OCR values were calculated from 3-min measurement cycles and adjusted to cell numbers.

Dissociated single HN6 and HN31 cells were plated on 6-well ultra-low attachment plates (Corning, New York, NY) at a density of 1 × 10⁵ cells per ml and grown in serum-free DMEM/F-12 medium supplemented with B27 (Invitrogen, Waltham, MA), 20 ng/ml epidermal growth factor (EGF), 20 ng/ml basic fibroblast growth factor (b-FGF), and 5 μg/ml insulin (Peprotech, Rocky Hill, NJ). Two weeks after incubation, the numbers of spheres were counted using a Zeiss microscope.

Approximately 1.0 × 10⁵ HNSCC cells were seeded into SeedEZ™ scaffold (Lena Bioscience, Atlanta, GA) for 7 days before treatment with CPI-613 and CB-839 alone or in combination. After one-week of drug treatment, cell viability in SeedEZ™ scaffold was measured by alamarBlue Cell Viability Reagent (Thermo Fisher Scientific, Waltham, MA) at 545/590 nm ex/em, followed by staining with Texas-red phalloidin and imaging.

Six-week-old NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice were purchased from the Jackson Laboratory (Bar Harbor, ME). An orthotopic tongue tumor model was generated as we previously described [22, 24]. To determine the function of GLS1, 1 × 10⁵ GLS1 knockdown and control HN6 cells were suspended in 50 μl of PBS/Matrigel (3:1) and injected into the anterior ~ 1/3 tongue of NSG mice under anesthesia. To determine the therapeutic efficacy of single or combined drug treatment 7 days after HN6 cell implantation, tumor-bearing mice were randomized into four groups to receive vehicle (PBS), CPI-613, CB-839, or the combination of CPI-613 and CB-839 for a total of 2 weeks. CPI-613 was administered by intraperitoneal injection (i.p.) every 3 days at the dose of 25 mg/kg body weight, and CB-839 was administered by oral gavage twice a day at the dose of 200 mg/kg body weight. Mice were imaged for bioluminescent luciferase signal every week using a Xenogen IVIS-200 In Vivo Imaging System (PerkinElmer, Waltham, MA). When experiments were terminated, primary tongue xenografts and major organs (including the heart, intestine, kidney, liver, and lung) from the mice were excised and processed for H&E staining and IHC. Tumor volume measurement was performed by digital caliper according to the formula V = length × width² × 1/2. If the mice were implanted with luciferase-expressing cancer cells, tumor burden was assessed using bioluminescence imaging. All animal experiments were approved by the Institutional Animal Care and Use Committee (IACUC) of Augusta University. Sections of paraffin-embedded mouse tongue tumors that received different treatments were deparaffinized and rehydrated by standard pathology laboratory methods, followed by heat-induced antigen retrieval and IHC with anti-Ki67 antibody (Abcam, Cambridge, UK). The sections were developed with DAB substrate kit (Vector Laboratories, Burlingame, CA) and counterstained with hematoxylin. IHC staining was quantified using Image pro-Plus6.0 software (Media Cybernetics, Silver Springs, MD, USA) and presented as integrated optical density (IOD).

Data on GLS1 gene expression for 33 cancer types and adjacent non-carcinoma tissues were extracted from TCGA (https://​tcga.​xenahubs.​net) and used to generate an expression matrix. Hazard ratios (HRs) from TCGA pan-cancer data were used to measure the prognostic significance of GLS1. Student’s t-test was used for comparison of two groups and analysis of variance (ANOVA) was used for comparison of multiple groups. The data obtained from three independent repetitions were presented in the form of an average and a standard deviation. All statistical calculation graphs were generated by GraphPad Prism 9.0 and a p-value of < 0.05 was set as the criterion for statistical significance. To assess the drug combination effect, combination index (CI) values were calculated using the software ComboSyn [25]. CI values of less than 1, equal to 1, and greater than 1 indicated synergistic, additive, and antagonistic effects, respectively.

To determine the anti-HNSCC effect of CPI-613, five HNSCC cell lines (HN6, HN12, HN13, HN30, and HN31) were treated with varying doses of CPI-613 for 3 days. Although all cell lines examined showed dose-dependent decreases in cell viability, HN6 and HN31 were more resistant to CPI-613 treatment than the other three cell lines (Fig. 1A). The TCA cycle metabolizes acetate derived from carbohydrates, proteins, and fats to generate ATP [26, 27]. As CPI-613 treatment resulted in significant reduction of metabolic flux through the TCA cycle [7], we asked whether CPI-613 could block ATP production in HNSCC cells. As shown in Supplementary Fig. S1, CPI-613 reduced the amount of ATP in both cell lines cultured in either glucose or pyruvate, but the reduction in ATP amount in HN6 cells was less than that in HN12 cells under the same conditions. This result was consistent with the effect of CPI-613 on proliferation of these two cell lines (Fig. 1A). Because 3D cell cultures more faithfully recapitulate the complex aspects of cancer growth and drug response in vitro [28], we also cultured HNSCC cells in SeedEZ™ 3D scaffolds, which are composed of completely inert and transparent glass microfibers, for 7 days before CPI-613 treatment. In line with the results from traditional 2D culture, HN6 and HN31 were more resistant to CPI-613 than the other cell lines examined in this study (Fig. 1B). One possible reason for the low cytotoxicity of CPI-613 in certain HNSCC cells was that the targeting of PDH/α-KGDH redirected cell growth and survival to other metabolic pathways. To assess this possibility, we determined the changes in metabolic enzymes in HN6 and HN31 cells in the presence or absence of CPI-613. This analysis identified that GLS1, a mitochondrial enzyme that hydrolyzes glutamine into glutamate to fuel rapid cancer cell proliferation, was the only molecule upregulated in both CPI-613-treated HN6 and HN31 cells, regardless of their growth in 2D culture dishes or SeedEZ™ scaffolds (Fig. 1C-E), suggesting that glutaminolysis plays a compensatory role in cell survival upon CPI-613 treatment. GLUD1 was also upregulated in CPI-613-treated 2D cells; however, it remained at the same levels in cells cultured in 3D cultures with or without CPI-613 treatment (Fig. 1D). These findings motivated our further study of GLS1. Next, we determined GLS1 levels in two CPI-613 sensitive cell lines, HN12 and HN30, in the presence and absence of CPI-613. GLS1 levels were unchanged in these two cell lines treated with and without CPI-613 (Fig. 1F), indicating that upregulation of GLS1 is one major mechanism associated with low treatment efficacy of CPI-613.

Glutamine is an effective precursor of glutamate for GSH synthesis in many cell types, including cancer cells [29]. Consistent with this notion, increased GSH levels were observed in CPI-613 treated HN6 and HN31 cells (Fig. 1G), indicating that HNSCC cells experience enhanced glutaminolysis upon CPI-613 treatment. To determine whether HNSCC cells are glutamate-dependent, we cultured HN6, HN30, and HN31 cells in medium with or without 4 mM glutamate. Colony formation assay showed that glutamate deprivation impaired the proliferation of all three HNSCC cell lines (Fig. 1H), but HN6 and HN31 cells were more addicted to glutamate than HN30 cells (Fig. 1H), which may explain their greater resistance to CPI-613 treatment.

TCGA comprises over 20,000 samples from 33 types of cancer and corresponding non-carcinoma samples. Using this data, we found that GLS1 expression is elevated in HNSCC tissues relative to matched non-carcinoma tissues (Fig. 2A). We then integrated HRs for overall survival in the form of a meta-analysis to assess the prognostic value of GLS1 score in patients cross cancer types. Increased GLS1 risk score correlated with overall survival in HNSCC patients [HR: 1.215, 95% confidence interval (CI): 1.011–1.460, p = 0.038] (Fig. 2B), suggesting that GLS1 is a prognosis-associated gene for this cancer type. High levels of GLS1 also indicate poor patient survival in liver hepatocellular carcinoma (LIHC, HR: 1.313, 95% CI: 1.098–1.569, p = 0.003) and uterine corpus endometrial carcinoma (UCEC, HR: 1.349; 95% CI: 1.116–1.630, p = 0.002). To interrogate the role of GLS1 in HNSCC cells, we depleted its expression in HN6 and HN31 cells using shRNAs (Fig. 2C). Compared with the knockdown controls, GLS1 depletion dramatically decreased the levels of secreted glutamate and reduced glutamine uptake (Fig. 2D), leading to repression of HNSCC cell proliferation (Fig. 2E). Moreover, GLS1 knockdown in both HN6 and HN31 cells reduced the efficiency of tumorsphere formation as evidenced by lower number and smaller size of tumorspheres relative to the knockdown control cells (Fig. 2F and G). This result is consistent with a previous report regarding the function of GLS1 in hepatocellular carcinoma [30], supporting the role of GLS1 in promoting HNSCC cell stemness.

To validate these observations in animals, GLS1 knockdown and its control HN6 cells were individually injected into the anterior tongue of NSG mice to generate an orthotopic tongue tumor mouse model (n = 5/group). Four weeks after inoculation, a marked decrease in tumor volume, along with a significant reduction in Ki67 positive tumor cells, were observed in mice implanted with GLS1 knockdown cells (Fig. 2H and I), We also performed TUNEL assays to determine apoptosis; however, no significant change was seen in apoptotic rate with or without GLS1 loss in xenograft tumors (data not shown). These data support the pivotal role of GLS1 in sustaining HNSCC cell growth.

To determine the functional association of GLS1 with cellular energy metabolism, we determined changes in the levels of PDH and α-KGDH in GLS1 knockdown and its control HN6 and HN12 cells. GLS1 depletion decreased p-PDHA1 levels but did not affect its total protein levels or α-KGDH (Fig. 3A). As the enzymatic activity of PDH is negatively regulated by phosphorylation at several serine sites [31], we then determined PDH activity in HN6 and HN12 cells with or without GLS1 knockdown. As shown in Fig. 3B, loss of GLS1 increased PDH activity in both cell lines, and this increase was significantly abrogated by additional CPI-613 treatment (Fig. 3C). These findings prompted us to study the effect of CPI-613 in combination with GLS1 knockdown in HNSCC cells. Compared with GLS1 depletion or CPI-613 treatment alone, a greater inhibitory effect on cell viability was seen when GLS1 knockdown cells were treated with CPI-613 (Fig. 3D). The same results were obtained in HN6 and HN31 cells cultured in SeedEZ™ scaffold (Fig. 3E), suggesting that synergistic blockade of glutaminolysis and energy metabolism results in a superior cytotoxic effect in HNSCC cells.

CB-839 is the first-in-clinic and orally bioavailable noncompetitive inhibitor of GLS1 splice variants, kidney-type (KGA) and glutaminase C (GAC), which convert glutamine into glutamate. There is highly credible evidence indicating the significant anti-proliferative activity of CB-839 in many types of cancers [32‐35], including lung adenocarcinoma, chondrosarcoma, liver cancer, and lymphoma. However, it is still unclear whether CB-839 possesses therapeutic potential in HNSCC. To address this, we first evaluated the effect of CB-839 on cell proliferation. As shown in Fig. 4A, CB-839 dose-dependently inhibited HN6 and HN31 cell growth with an IC50 of around 1 μM. Decreased glucose uptake and lactate levels were also seen in these cells when treated with CB-839 (Fig. 4B), suggesting the effect of CB-613 in coupling glutaminolysis inhibition with decreased glucose consumption. Consistent with the data from GLS1 knockdown, CB-839 led to a reduction in secreted glutamate and glutamine consumption in HNSCC cells (Fig. 4C). Moreover, CB-839 also induced apoptosis, as evidenced by the appearance of cleaved PARP and caspase 3 in cells following its administration (Fig. 4D). In HN6-derived tumor-bearing mice, two-week treatment with CB-839 did not reduce the body weight of mice but resulted in significantly smaller tumor xenografts relative to those found in vehicle-treated animals (Fig. 4E and F). Consistently, fewer Ki67-positive tumor cells along with increased apoptosis were seen in mice receiving CB-839 compared with vehicle (Fig. 4G and H). Collectively, these observations suggest that CB-839 could be a safe and potent anti-HNSCC agent.

The above data prompted us to assess the addition of CB-839 to CPI-613 treatment. We first investigated the role of this combination in metabolic alterations. Consistent with the enrichment in glucose utilization in cancer cells with PDH depletion [36], CPI-613 treatment alone increased glucose uptake and lactate production in both HN6 and HN31 cells (Fig. 5A and B), suggesting that inhibition of mitochondrial metabolism by CPI-613 may elicit increasing compensatory glycolytic activity. The addition of CB-839 to CPI-613 treatment attenuated the CPI-613-induced increase in glucose consumption and lactate production (Fig. 5A and B). To measure the impact of these treatments on mitochondrial metabolic function, we assessed OCR, an indicator of mitochondrial oxidative phosphorylation (OXPHOS). Remarkably, the combination of CB-839 and CPI-613 demonstrated the lowest OCR normalized to cell number among the four treatment groups (Fig. 5C). Consistent with this phenotype, co-treatment suppressed cell viability and impaired colony formation ability more efficiently than the other three treatments (Fig. 5D and E). Moreover, a strong synergistic induction of apoptosis was seen in the dual drug-treated HN6 and HN31 cells, as indicated by an increase in PARP cleavage and rates of apoptosis (Fig. 5F and G). The inhibitory effect of the CB-839 and CPI-613 combination on 3D growth was also evident compared with single-drug treatments (Fig. 5H). Taken together, these observations indicate that CB-839 augments the cytotoxic effect of CPI-613 in HNSCC cells.

To evaluate the effectiveness of the CB-839 and CPI-613 combination in vivo, luciferase-containing HN6 cells were implanted into the tongues of NSG mice to establish tumors. Mice were treated with vehicle, CB-839, CPI-613, or the combination when tongue tumors were visualized. After a two-week treatment, tumor burden was reduced in each single-arm treatment, as evidenced by lower bioluminescent signals (Fig. 6A). Most importantly, the combination of CB-839 and CPI-613 achieved an anticancer effect that was superior to either of the drug treatments alone (Fig. 6A). Histopathological analysis by H&E staining showed no morphologic changes in the major organs (Fig. 6B), suggesting this combination does not produce detectable systemic toxicities. To assess the effect of the treatments on tumor xenografts, tissue sections were immunostained with an anti-Ki67 antibody. Compared with single-arm treatment, a significant reduction in Ki67-positive cells was seen in the mice receiving the dual-drug treatment (Fig. 6C). The drug combination also led to a more robust induction of tumor cell apoptosis than single-arm treatment, as evidenced by greater numbers of TUNEL-positive cells in the tumor tissues obtained from mice in this group (Fig. 6D). These novel findings indicate that the combination of CB-839 and CPI-613 is a safe and more effective therapeutic strategy for HNSCC treatment.