NAMPT is a prognostic factor and a potential therapeutic target against 3D-spheroids with cancer stemness features
To morphologically mimic floating cell clusters in the peritoneal cavity of patients with OV [19], we obtained various non-adherent cell lines that proliferated while suspended in the medium using ULA plates (Supplementary Fig. S1A). All the cell lines investigated in this study (A2780, IGROV1, OVCAR3, OVCAR8, ES2, SKOV3) formed 3D-spheroids [14]. We opted to focus on studying the metabolic pathways in A2780 and IGROV1, which are representative models for poorly differentiated OV with a high likelihood of relapse and consistently exhibit high genomic instability [20, 21]. To confirm that the spheroids have the characteristics of CSCs functionally and morphologically, we assessed the expression of CD44 and CD133 proteins, which are stem cell markers that contribute to sphere formation and tumor cell biogenesis [22, 23], using western blotting. CD44 was significantly more abundant in 3D-spheroids relative than in conventional 2D-cultured cells in IGROV1 (Fig. 1A); little or no CD44 expression was observed in A2780 cells grown under either condition. 3D-spheroids showed higher CD133 expression in both cell lines than planar growth (Fig. 1A). These results suggest that 3D-spheroids obtained using ULA plates are enriched in stemness markers and mimic floating cell clusters in the peritoneal cavity, both morphologically and functionally.
One of the reasons for the therapy-resistance of CSCs is elevated NAD+ production [24]. We hypothesized that 3D-spheroids maintain higher NAD+ levels than conventional 2D-cultured cells. In line with this hypothesis, 3D-spheroids exhibited significantly increased NAD+ levels compared to 2D-cultured cells across all six cell lines used in this study (Fig. 1B). NAD+ is produced via various pathways and acts as co-factors in various reactions [12, 25]. Key enzymes for the NAD+ production include NAPRT (Preiss-Handler pathway), QPRT (de novo pathway), NADSYN1 (common to Preiss-Handler and de novo pathways), and NAMPT (salvage pathway) (Supplementary Fig. S1B). To understand the relevance of enzymes associated with NAD+ production in OV patients, we analyzed patient data from The Cancer Genome Atlas (TCGA) to assess the impact of these enzymes on patient survival. Notably, high NAMPT expression in tumors was correlated with a significantly worse prognosis (Fig. 1C). Conversely, there was no statistical difference in the expression levels of the other NAD+-producing enzymes (Supplementary Fig. S1C). Furthermore, we evaluated the correlation between NAD+ levels and the gene expression of the above-mentioned enzymes in OV cell lines using the public database of the Cancer Dependency Map (DepMap) portal data explorer. Consistent with the overall survival data, NAMPT showed a strong positive correlation with NAD+ (Fig. 1D), whereas other enzymes did not show a clear positive correlation (Supplementary Fig. S1D). To better understand the high-risk role of NAMPT, we assessed its prognostic impact in other cancer types. Importantly, OV had the lowest p-value rank, suggesting that elevated NAMPT expression may promote tumor growth by increasing NAD+ production in the OV (Fig. 1E). We tested this hypothesis by evaluating NAMPT expression by western blot analysis. NAMPT expression was consistently more abundant in 3D-spheroids than in 2D-cultured cells across all cell lines (Fig. 1F and 1G), whereas the expression of other enzymes varied depending on cell lines and was not significantly differences between 3D-speroids and 2D-cultured cells (Fig. 1F). To investigate the impact of NAMPT silencing on NAD+ production, we quantified NAD+ levels after individually suppressing enzyme expression using siRNAs. Protein levels of each enzyme were effectively suppressed in both 3D- and 2D-cultured cells (Fig. 1H and Supplementary Fig. S1E), and silencing NAMPT effectively inhibited NAD+ production compared to control siRNA, whereas silencing of the other enzymes did not (Fig. 1I and Supplementary Fig. S1F). Consistent with the known NAD+ production pathways, silencing of NAMPT also suppressed NADPH and ATP production (Fig. 1J and 1K). To determine whether silencing of NAMPT also impaired mitochondrial function, we quantified the oxygen consumption rate (OCR), which is an indicator of mitochondrial respiration. OCR was most efficiently suppressed by silencing NAMPT (Fig. 1L). Silencing NAMPT also reduced the extracellular acidification rate (ECAR), indicating a reduction in glycolytic flux (Fig. 1L). Conversely, suppression of OCR and ECAR was not observed in 2D-cultured cells (Supplementary Fig. S1G). Overall, we observed that 3D-spheroids exhibit CSC characteristics with elevated NAD+, and NAMPT is the key enzyme that supports NAD+ production. These findings suggest that NAMPT is a potential therapeutic target for OV.
3D-spheroids oxidize glucose via the TCA cycle for elevated ATP production
Given that ATP is produced via glycolysis and the TCA cycle using NAD+ as a co-factor [26], we evaluated the total ATP production in 2D- and 3D-cultured cells. Consistent with total NAD (Fig. 1B), we observed more ATP in 3D-spheroids (Fig. 2A). To determine the mechanism of elevated ATP production, we evaluated the function and mass of mitochondria. MitoTracker Orange CMTMRos, which stains mitochondria in living cells in a membrane potential-dependent fashion [26], and MitoTracker Green FM, which is membrane potential-independent and is useful for measuring mitochondrial mass [27], were quantified in A2780 and IGROV1 cells. We observed a significantly higher intensity of MitoTracker Orange CMTMRos in both 3D-cultured cell lines (Fig. 2B). In contrast, MitoTracker Green FM was significantly lower in 3D-spheroids (Fig. 2B). This suggests that 3D-spheroids possess mitochondria with higher ATP production capacity.
We investigated whether the higher ATP production in 3D-spheroids was a result of modified central carbon metabolism. To this end, we used IGROV1 cells, which rely on glucose for growth (Supplementary Fig. S2A). The cells were cultured in a tracer medium containing 13C6-glucose to investigate the fate of glucose in central metabolic pathways. We evaluated 13C6-glucose utilization in 2D- and 3D-cultured IGROV1 cells by NMR and ion chromatography-mass spectrometry (IC-MS) (Supplementary Fig. S2B). 1H-13C HSQC NMR analysis revealed that 3D-cultured cells exhibited increased biosynthesis of glutamine, glutamate, and glutathione from 13C6-glucose, indicating the enhanced entry and cycling of glucose in the TCA cycle (Supplementary Fig. S2C). Furthermore, resonances arising from de novo-synthesized purine nucleotides (AXP) and components of the hexosamine pathway (e.g., UDP-GlcNAc, UDP-GalNAc, GlcNAc-1-P), which contribute to resistance of cancer stem cells to chemotherapy [28, 29], indicate enhanced distribution of glycolytic intermediates in the pentose phosphate pathway (via glucose-6-phosphate (G6P)) and hexosamine biosynthetic pathway (via fructose-6-phosphate (F6P)). 2D-cultured cells demonstrated 13C6-glucose labeling of lactate, which exceeded that of alanine, whereas in 3D-cultured cells, 13C6-glucose labeling of alanine was in excess of that of lactate (Supplementary Fig. S2C), suggesting a shift away from lactate fermentation in the 3D-cultured cells. Although there was no significant difference in the glucose consumption in the culture medium (Supplementary Fig. S2D), 3D-cultured cells exhibited decreased intracellular 13C-glucose/G6P, which is consistent with the enhanced entry and distribution of 13C-glucose into the pathways outlined above (Supplementary Fig. S2C). Furthermore, the protein levels of hexokinase 2, known for its upregulated expression in cancer cells [30], were significantly higher in 3D-cultured cells (Supplementary Fig. S2F).
IC-MS of the polar extracts revealed no significant differences between 2D- and 3D-cultured cells in the abundance or isotopological distribution of the glycolytic intermediates G6P, F6P, and pyruvate (Fig. 2C). Conversely, 3- and 2-phosphoglycerate (3PG_2PG), phosphoenolpyruvate (PEP), and lactate levels were higher in 2D-cultured cells (Fig. 2C). Furthermore, more lactate was secreted into the culture medium of the 2D-cultures, consistent with elevated lactate fermentation in 2D-cultured cells compared to 3D-cultured cells (Fig. 2C). m + 3 isotopologues of citrate were significantly elevated in 3D-spheroids (Fig. 2C), consistent with pyruvate carboxylase (PC)-mediated entry of glucose-derived carbon into TCA cycle [31, 32] (Supplementary Fig. S3A and S3B) and elevated m + 4 and m + 5 citrate isotopologues indicated multiple turns of the TCA cycle. Consistent with this result, we found that PC expression was enhanced in the 3D-spheroids (Supplementary Fig. S3C). Additionally, the total levels as well as 13C-enrichment of α-KG, glutamate, fumarate, malate, and aspartate were significantly higher in the 3D-spheroids (Fig. 2C). These results revealed robust glucose oxidation in the 3D-spheroids relative to 2D-cultured cells and demonstrated increased TCA cycle activity and capacity for ATP production via mitochondrial oxidative phosphorylation.
Enhanced nucleotide synthesis and phosphorylation in 3D-spheroids
To comprehensively understand the metabolic differences between 2D- and 3D-cultured cells, we evaluated the levels of representative metabolites using a heat map and a volcano plot (Fig. 3A and 3B). Interestingly, the metabolite that exhibited the greatest increase in 3D-spheroids was phosphocreatine, whose main function was to serve as an energy buffer to maintain cellular ATP levels [33, 34] (Fig. 3C and 3D). Furthermore, 3D-spheroids contained higher levels of phosphoribosyl pyrophosphate (PRPP; Fig. 3E), a crucial component of nucleotide synthesis, which is derived from the pentose phosphate pathway (PPP) [35]. 13C-enrichment of PRPP was significantly greater in 3D-spheroids relative than in 2D-cultured cells, providing evidence of an enhanced capacity for nucleotide synthesis (Fig. 3E). Notably, the 13C-enrichment of nucleotides containing three phosphate groups, ATP, GTP, CTP, UTP, and TTP, was significantly higher in 3D-spheroids, while mono-phosphorylated nucleotides (IMP, AMP, GMP, CMP, UMP, and TMP) were less 13C-enriched in the 3D-spheroids (Fig. 3E). Both purines and pyrimidines exhibited robust m + 5 13C-enrichment, indicating that their ribose moieties were derived from oxidative PPP. Pyrimidines, including UTP, also exhibited increased fractional enrichments of m + 6, m + 7, and m + 8 isotopologues (Fig. 3E), consistent with de novo pyrimidine synthesis from aspartate and orotate isotopologues partially labeled with 13C6-glucose (Fig. 2B and Supplementary Fig. S3D). We then examined the ratio of triphosphates/monophosphates to assess energy levels and observed that the CTP/CMP, UTP/UMP, ATP/AMP, and GTP/GMP ratios were notably higher, while the TTP/TMP ratio showed a tendency to increase (Fig. 3F). Overall, these results suggest that 3D-spheroids augmented their high-energy substrate levels by upregulating not only the synthesis, but also the phosphorylation of both purine and pyrimidine nucleotides, including ATP (Fig. 3G).
NAMPT inhibitors are a potential strategy against NAMPT-dependent OV
To further examine whether NAMPT may be an effective therapeutic target in OV, we first evaluated the effect of the NAMPT inhibitor GNE-617 on NAD+ production. We observed that GNE-617 effectively reduced NAD+ levels in 3D-cultured A2780 and IGROV1 cells in a concentration-dependent manner (Fig. 4A), similar to silencing NAMPT (Fig. 1H – 1K). GNE-617 significantly decreased intracellular NADPH (Fig. 4B) and ATP levels (Fig. 4C). In agreement with these results, GNE-617 and other NAMPT inhibitors, such as GNE-618, FK-866, and KPT-9274, significantly inhibited the growth of 3D-spheroids (Fig. 4D).
Notably, the RPMI-1640 medium used in this study contained nicotinamide (NAM) and tryptophan (Trp) but no nicotinic acid (NA). Since human plasma contains NA, we next evaluated the effect of NA on the anti-tumor effect of GNE-617 to understand NAD+ production kinetics in a more biologically relevant environment (Supplementary Fig. S4A). The addition of NA did not rescue the anti-proliferative effect of NAMPT inhibitors in A2780 and IGROV1 cells, but completely restored the effect in OVCAR3, OVCAR8, and ES2 cells (Fig. 4E and Supplementary Fig. S4B). These results indicate that these latter cell lines can use NAPRT in the Preiss-Handler pathway for NAD+ production, whereas A2780 and IGROV1 depend only on NAMPT in the salvage pathway (Supplementary Fig. S1B). As expected, the addition of nicotinamide mononucleotide (NMN), a derivative of NAM, rescued the anti-tumor effect of GNE-617 across all cell lines (Fig. 4E), since this molecule is downstream of NAMPT in the salvage pathway (Supplementary Fig. S1B). Taken together, our findings indicate that NAMPT inhibitors may be potent therapeutics against OV dependent on NAMPT for NAD+ production.
Given that NADPH is an essential co-enzyme for the synthesis of cholesterol, polypeptides, and fatty acids [36], we hypothesized that NAMPT inhibitors would also inhibit these processes (Fig. 4F). Protein levels of ATP citrate lyase (ACLY) and 3-hydroxy-3-methylglutaryl-CoA synthase (HMGCS1), which are essential enzymes for mevalonate and lipid synthesis, increased in 3D-cultured cells (Fig. 4G). Moreover, the cells had significantly higher cholesterol and polypeptide levels as measured by o-propargyl-puromycin (OPP), which reflects protein synthesis capacity, and increased Lipi-Blue staining, which reflects lipid-producing capacity (Fig. 4H, 4I, and 4J). As anticipated, GNE-617 inhibited cholesterol synthesis in the 3D-spheroids (Fig. 4K). Similarly, the number and intensity of OPP-containing polypeptides and Lipi-Blue were significantly suppressed by GNE-617 treatment (Fig. 4L and 4M). Together, these results suggest that NAMPT inhibitors induce anti-tumor effects in NAMPT-dependent cancers by downregulating the production of NAD+, NADPH, and ATP, thereby affecting cholesterol, protein, and lipid synthesis.
The combination of NAMPT inhibitors and disulfiram synergistically inhibits cell viability in vitro
To enhance anti-tumor effects, delay drug resistance, and minimize side effects, we sought a drug that synergizes with NAMPT inhibitors. Previously, we showed that disulfiram (Fig. 5A), an ALDH inhibitor used to promote alcohol abstinence [37], inhibits ALDH activity, which is associated with stemness [38, 39]. ALDH consumes NAD+ to catalyze the oxidation of aldehyde substrates. Thus, we hypothesized that combining disulfiram with NAMPT inhibitors would yield synergistic outcomes. Disulfiram inhibited growth in 3D-spheroids alone (Fig. 5B), and in combination with GNE-617 demonstrated highly potent and synergistic anti-tumor effects in 3D-cultured IGROV1 and A2780 (Fig. 5C, 5D, and Supplementary Fig. S5A). Disulfiram also showed synergistic effects with other NAMPT inhibitors, including FK-866, GNE-618, and KPT-9274 (Supplementary Fig. S5A and S5B).
To determine the effect of combinatorial therapy targeting ALDH activity, we assessed ALDH activity using an ALDEFLUOR assay. The combinatorial treatment inhibited ALDH activity most strongly in 3D-cultured IGROV1 cells (Fig. 5E), which are known to have high ALDH activity. A2780 cells exhibited low baseline ALDH activity (Supplementary Fig. S5C), suggesting that additional mechanisms contribute to the synergistic effects of A2780. To evaluate the synergistic effects of ALDH enzyme activity and understand the mechanism in detail, we also tested NCT-505, a specific inhibitor of ALDH1A1, which primarily contributes to ALDH activity [39], in combination with GNE-617 (Fig. 5F). Previous reports showed the ALDH inhibitory activity of NCT-505 at an IC50 of 2 nM in 3D-spheroids of the OV cell lines OVCAR3 and OV90 [39], and ALDH activity was significantly suppressed at 10 and 100 nM of NCT-505 in 3D-cultured IGROV1 cells (Fig. 5G). Despite strong suppression of ALDH activity, NCT-505 showed no obvious anti-proliferative effect on itself inIGROV1 (Fig. 5H). When NCT-505 was combined with GNE-617, an antagonistic effect, rather than a synergistic effect, was observed in both IGROV1 (high ALDH activity-high) and A2780 (low ALDH activity) (Fig. 5I). Our findings not only suggest that the combination of disulfiram and NAMPT inhibitors could be an effective strategy against OV but also indicate an unknown ALDH activity-independent mechanism that drives the synergistic effects of disulfiram and NAMPT inhibitors.
GNE-617 inhibits the GAPDH-mediated response and PAR production, while disulfiram suppresses the TCA cycle
To elucidate the mechanisms that induce the synergistic effect of combinatorial treatment, we tested metabolite changes induced by GNE-617, disulfiram, and combinatorial treatment in 3D-spheroids grown in the presence of 13C6-glucose for 24h during treatments (Supplementary Fig. S6A). Principal component analysis revealed significant differences between the four treatment groups (Supplementary Fig. S6B). We evaluated the levels of representative metabolites of glycolysis and the TCA cycle (Fig. 6A). To identify differentially expressed metabolites, we compared the effects of each treatment (Fig. 6B and 6C). GNE-617 markedly downregulated ADP-ribose produced from NAD+ in the mitochondria (Fig. 6B, 6C, and 6D). Given that poly(ADP-ribos)ylation (PARylation), a repair process for damaged DNA, is mediated by NAD+ as a co-factor [40] (Fig. 6D), we hypothesized that GNE-617 would also inhibit PARylation. As anticipated, GNE-617 treatment markedly suppressed PAR protein expression (Fig. 6E). Fructose 1-6-bisphosphate levels increased 10-fold following GNE-617 treatment (Fig. 6A). To test our hypothesis that GNE-617 inhibits the flux from glyceraldehyde-3-phosphate to 1-3-bisphosphoglycerate via GAPDH, which results in the accumulation of fructose 1-6-bisphosphate, we quantified the GAPDH-mediated response. GNE-617 downregulated GAPDH-mediated flux in a concentration-dependent manner, and this effect was rescued by NMN (Fig. 6F). This suggested that GNE-617 blocked the conversion of fructose 1-6-bisphosphate to glyceraldehyde-3-phosphate (Fig. 6G). Compared with the control, disulfiram significantly reduced the metabolite levels of α-KG, fumarate, malate, and aspartate (Fig. 6A), as well as a notable decrease in glucose contribution to these metabolites (Supplementary Fig. S6C), indicating decreased capacity for glucose oxidation via the TCA cycle. In further support of this, we found that the levels of citrate and cis-aconitate in the disulfiram-treated group were more than 20 times higher than in the control group (Fig. 6A). To test the hypothesis that citrate and cis-aconitate accumulation is caused by the inhibition of aconitase, an enzyme that catalyzes the isomerization of citrate to isocitrate via cis-aconitate [31], we directly measured aconitase in disulfiram-treated cells. Disulfiram treatment significantly reduced the aconitase activity in both A2780 and IGROV1 cells (Fig. 6H and 6I). Moreover, disulfiram, alone or in combination, significantly diminished the glucose contribution to m + 3 and m + 4 isotopologues of citrate, whereas the m + 2 citrate isotopologue increased, consistent with a block of the TCA cycle at the level of aconitase [31] (Supplementary Fig. S6C). Taken together, these findings suggest that GNE-617 inhibits ADP-ribose production derived from the mitochondria, while disulfiram acts as an aconitase inhibitor within the TCA cycle. This indicates an anti-tumor mechanism involving the suppression of mitochondrial function.
Combinatorial treatment induces a synergistic anti-tumor effect by suppressing mitochondrial function in vitro
To evaluate whether the combinatorial treatment downregulated mitochondrial function, we first quantified the anti-tumor effect of GNE-617 and disulfiram on mitochondrial respiration in 3D-spheroids (Fig. 7A). We observed that the combinatorial treatment significantly suppressed not only OCR, but also maximal ECAR, reflecting glycolysis (Fig. 7B). Additionally, we found reduced expression of complex-V, an ATP synthase in the mitochondria [41] (Supplementary Fig. S7A). Consistent with the reduction in OCR and complex-V, we observed that the combinatorial treatment inhibited ATP production in 3D-spheroids (Fig. 7C). We next used MitoSOX Red to quantify mitochondrial-derived reactive oxygen species (ROS), which reflects the presence of damaged mitochondria [42]. The expression of MitoSOX Red was enhanced by the combinatorial treatment compared to monotherapy, suggesting that combinatorial therapy induces mitochondrial damage (Fig. 7D and Supplementary Fig. S7B and S7C). Moreover, cleaved caspase-3/7, a key executioner caspase from mitochondria [43], was significantly higher in 3D-spheroids treated with the combinatorial therapy than in the controls (Fig. 7E and Supplementary Fig. S7D). To further confirm that the combinatorial treatment induced cell death, we quantified viability, cytotoxicity, and apoptosis induction with or without the pan-caspase inhibitor benzyloxycarbonyl-Val-Ala-Asp-fluoromethyl ketone (Z-VAD-FMK) [44]. The combinatorial treatment suppressed cell viability and induced cytotoxicity as well as cleaved caspase-3/7 luminescence (Fig. 7F and Supplementary Fig. S7E). Pre-treatment with Z-VAD-FMK inhibited only cleaved caspase-3/7 secretion but had no considerable effect on cytotoxicity (Fig. 7G, 7H, and Supplementary Fig. S7F). In summary, our findings suggest that combination therapies synergistically induce mitochondrial dysfunction, leading to death in OV cells. We also suggested that this mechanism of cell death is independent of caspase-dependent apoptosis.
Combinatorial treatment significantly inhibits tumor growth and extends survival in vivo
We tested the effects of the combinatorial therapy in two immunocompromised mouse models. In the first experiment, IGROV1 spheroids were subcutaneously implanted, and in the second experiment, A2780 spheroids were intraperitoneally injected into immunocompromised mice (Fig. 8A and 8B). In both experiments, the mice were randomly divided into four groups: vehicle control, GNE-617, disulfiram, or the combinatorial treatment for five continuous days (Fig. 8A and 8B). Within the group receiving combinatorial treatment in the intraperitoneal model, two mice exhibited a transient reduction in body weight post-therapy, which subsequently normalized (Supplementary Fig. S8A and S8B). We observed a trend for decreased tumor growth in subcutaneous tumors with combinatorial treatment (Fig. 8C). In the intraperitoneal model, all untreated mice exhibited intraperitoneal dissemination and ovarian enlargement (Fig. 8D), leading to euthanasia 3–5 weeks after injection. Notably, the combinatorial treatment significantly increased mouse survival (Fig. 8E and 8F). Taken together, our findings indicate that the synergistic effect of GNE-617 and disulfiram reduces tumor growth and enhances survival in OV-bearing mice. This suggests that such a combinatorial approach may offer a novel and potent therapeutic strategy for the management of recurrent OV.