6-Phosphofructo-2-Kinase/Fructose-2,6-Biphosphatase-2 Regulates TP53-Dependent Paclitaxel Sensitivity in Ovarian and Breast Cancers Free

Paclitaxel has emerged as an important agent for adjuvant treatment of both ovarian and breast cancers. When used as a single agent, however, less than half of these cancers respond (1–3). Enhancing the response to primary therapy with paclitaxel could improve outcomes for women with both diseases. The tumor suppressor TP53 is frequently mutated in human cancers including breast and ovarian cancers. Somatic TP53 mutations occur in over 50% of all tumors (4, 5), across nearly all types of cancer. TP53 mediates multiple processes, of which the best understood include cell-cycle arrest, DNA repair, and apoptosis (6, 7). Recent work indicates that TP53 has an important role in modulating metabolic processes including glycolysis, oxidative phosphorylation, and the pentose–phosphate pathway (PPP; refs. 8, 9).

Cancer cells can exhibit enhanced glycolysis and lactate production even under normoxic conditions, that is, the Warburg effect (10, 11). Cancer cells often exhibit increased expression of glycolytic enzymes and glucose uptake, which result in markedly enhanced glycolytic rates. In addition to producing additional ATP, increased glycolysis also generates three-carbon–building blocks for the biosynthesis of proteins, lipids, and nucleic acids in rapidly growing cells (12). Many previous studies have shown that oncogenic signaling pathways regulate the activity of metabolic enzymes to support macromolecular synthesis in cancer cells, which is required for their rapid proliferation. Enzymes involved in altered glucose metabolism might provide important targets for cancer therapy (13, 14). The PPP consists of oxidative and nonoxidative branches. The oxidative branch is a major source of reduced NADPH, a key reducing equivalent for lipid, nucleotide, and aromatic amino acid biosynthesis (15). NADPH also plays a key role in maintaining intracellular redox homeostasis by maintaining a pool of reduced glutathione, which neutralizes reactive oxygen species (ROS). Cancer cells frequently have an increased burden of oxidative stress (16, 17) and are likely to be more sensitive to the additional oxidative damage promoted by ROS. Multiple studies indicate that tumors are particularly dependent on redox-regulating pathways and that this vulnerability can be exploited by targeted therapeutics (18, 19). In recent years, several kinases have been identified that regulate the sensitivity of cancer cells to paclitaxel by inhibiting centrosome splitting (20) or enhancing microtubule stability (21, 22). Much less attention has been given to kinases that affect paclitaxel sensitivity by modulating cancer cell metabolism.

In earlier studies, we found that 20% of potential kinase targets whose knockdown modulates paclitaxel sensitivity participate in glucose and energy metabolism (23). In this study, we targeted genes engaged in glucose and energy metabolism in ovarian and breast cancer cell lines and were able to validate four genes (CKMT1B, CKMT2, PFKFB2, and PFKFB4), whose knockdown reduced cancer cell survival and enhanced paclitaxel sensitivity in at least three of six cancer cell lines tested. CKMT1B and CKMT2 are mitochondrial creatine kinase isoenzymes, which catalyze the reversible transfer of phosphate groups from phosphocreatine to ADP to yield ATP and creatine. CKMT1B and CKMT2 are the key molecules for oxidative phosphorylation and apoptosis in mitochondria. PFKFB2 and PFKFB4 belong to a family of bifunctional enzymes that control the levels of fructose 2,6-bisphosphate. Several studies have demonstrated that some cancer cell lines produce markedly elevated level of fructose-2,6 bisphosphate (24, 25). Among these four genes, a leading candidate was 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase 2 (PFKFB2), an isoform of the glycolytic enzyme phosphofructokinase 2 (PFK2). Analysis of this gene's function in ovarian and breast cancers revealed that PFKFB2 maintains intracellular redox balance, which attenuates p53-dependent apoptotic cell death. Knockdown of PFKFB2 increased the glycolysis rate, but decreased the flow of intermediates through the PPP in cancer cells with wild-type TP53 (wtTP53), decreasing NADPH. ROS accumulated after PFKFB2 knockdown, which stimulated Jun N-terminal kinase (JNK) and p53 phosphorylation, induced apoptosis that depended upon upregulation of p21 and Puma.

Cell lines used in this article are listed in Supplementary Table S1. The identity of all cell lines was confirmed with short tandem repeat DNA fingerprinting (December 18, 2018) in the MDACC Characterized Cell Line core (supported by NCI P30CA016672). Cell line Mycoplasma test was performed using an ATCC PCR kit on November 20, 2018.

A confirmatory screen with 12 positive hits (CKM, CKMT1B, CKMT2, GCK, PDK3, PDK4, PFKFB2, PFKFB3, PFKFB4, PFKL, PFKM, and PGK10), which were selected from our initial screen, was conducted in four ovarian cancer (A2780, HeyA8, OC316, and OVCAR8) and two breast cancer (MCF-7 and ZR-75-1) cell lines. The same screening procedures were applied in the primary screen (23) and reagents used in the screen are listed in Supplementary Table S2.

ON-TARGETplus PFKFB2 siRNA #6, #7, #8, and #9, TP53 siRNA #14, #15, and #16, SMARTpool G6PD siRNA, and nontargeting siRNA #2 were purchased from Dharmacon. siRNA (30 nmol/L) and 1.5% DharmaFECT 4 (Dharmacon) were diluted in OPTI-MEM medium individually and then mixed together for 20 minutes at room temperature. Cells were laid on top of siRNA-DharmaFECT mixture in 6-well plates and incubated for 24, 48, or 72 hours before analyses. We validated all siRNA oligos in multiple cell lines and discovered the knockdown efficiency of siPFKFB2 #6 and #7 is better than #8 and #9, and siTP53 #14 is better than #15 and #16.

A total of 3–8 × 10³ cells were reversely transfected with a mixture of 20 nmol/L of siRNA and 1.5% DharmaFECT 4 in triplicate wells of 96-well plates. Twenty-four hours post transfection, cells were treated with a serial dilution of paclitaxel and incubated for another 72 hours. Cells were fixed with 10% trichloroacetic acid at 4°C for 30 minutes. Plates were rinsed with distilled H₂O and 100 μL of 0.4% sulforhodamine B (SRB) in 1% acetic acid was added to each well. Plates were incubated at room temperature for 30 minutes, and then rinsed with 1% acetic acid to remove excessive dye. SRB was solubilized with 100 μL of 10 mmol/L Tris buffer at pH = 8 for 5 minutes with shaking. Absorbance was determined at 570 nm on a Synergy 2 Microplate Reader from BioTek. Paclitaxel dose-response curves were plotted and IC₅₀ was determined by GraphPad Prism 7.

A total of 400–800 cells were seeded in triplicate wells of 6-well plates 24 hours post siRNA transfection. Cells were incubated for up to 14 days until colonies became visible. Cells were washed twice with PBS, fixed in 0.1% Brilliant Blue R with 10% v/v acetic acid and 30% v/v methanol for 1 minute and washed with tap water until background was clear. Colonies with more than 50 cells were counted.

A total of 1.5–3 × 10⁵ cells were transfected with control or PFKFB2 siRNA for 24–72 hours and then incubated either with 10 |{\rm{\rmu mol}}/{\rm{L}}$| CM-H2DCFDA (5- and 6-chloromethyl-2′,7′-dichlorodihydrofluorescein diacetate acetyl ester, Thermo Fisher Scientific) in PBS for 30 minutes or with 10 |{\rm{\rmu mol}}/{\rm{L}}$| dihydroethidium (Thermo Fisher Scientific) in PBS for 1 hour in a 37°C incubator. Cells were then washed and suspended in 0.5 mL of PBS. Florescence was measured in a flow cytometer using the FITC and PE-Texas Red channels.

Apoptosis was detected with the Alexa Fluor 488 Annexin V/Dead Cell Apoptosis Kit (Thermo Fisher Scientific). Cells were reversely transfected with siRNA for 24 hours and treated with paclitaxel or diluent for 48 hours. Cells were stained with 0.1 mL of Annexin binding buffer containing 5 μL of Annexin V–Alexa 488 conjugate and 1 μg/mL of propidium iodide for 15 minutes at room temperature (Thermo Fisher Scientific). Stained cells were read on Gallios Analyzer (Beckman Coulter) and 2 × 10⁴ events were counted.

Forty-eight hours post siRNA transfection, cells were lysed in QIAzol and total RNA was extracted using an RNeasy Kit (Qiagen) according to the manufacturer's instructions. cDNA was synthesized from 1 μg of RNA using the Superscript II First Strand Synthesis Kit (Thermo Fisher Scientific). qPCR was performed using CFX Connect Real-time System (Bio-Rad) in a total volume of 20 μL, which included 10 μL of 2× iTaq Universal PCR master mix, 300 nmol/L of each forward and reverse primers, and 40 ng of cDNA. Puma primers are listed as follows: forward 5′-CACCTAATTGGGCTCCATCTC-3′ and reverse 5′-GACGACCTCAACGCACAGTA-3′. GAPDH primers are listed as follows: forward 5′-TCGACAGTCAGCCGCATCTTCTTT-3′ and reverse 5′-ACCAAATCCGTTGACTCCGACCTT-3′.

Extracellular acidification rates (ECAR) were measured using a Seahorse XF24 Analyzer (Agilent Technologies). A total of 1.5–4 × 10⁴ cells were seeded into XF24 cell culture microplates and then transfected with siRNA at 24, 48, and 72 hours prior to analysis. Cells were gently washed and incubated in XF assay media with pH 7.4 at 37°C without CO₂ for 30 minutes. ECAR was detected under basal conditions followed by the sequential addition of 10 mmol/L glucose, 2 μmol/L oligomycin, and 50 mmol/L 2-deoxy-glucose (2-DG; XF Glycolysis Stress Test Kit, Agilent). pH levels were measured three times using XF24 extracellular flux analyzer after the injection of stressing reagents. ECAR data were normalized to cell number of each sample and glycolysis rate was calculated with a formula: glycolysis = ECAR after addition of glucose – ECAR after 2-DG treatment (26).

Sample preparation and LC/MS-MS analysis followed a published protocol with modifications (27). Cells were transfected with control or PFKFB2 siRNA for 28 hours and then washed twice with cold PBS. Cells were lysed on dry ice with 3 mL of cold extraction solvent (80% methanol containing ¹³C6-G6P and ¹³C3-pyruvate as internal standard) and were incubated at −80°C for 30 minutes. Cells were then scraped and lysates were transferred to 15 mL prechilled conical tubes. Cells were pelleted at 2,000 rpm for 5 minutes and the supernatant was transferred into prechilled 15 mL conical tubes. The pellet was washed with 1 mL of the cold extraction solvent and then centrifuged again at 2,000 rpm for 5 minutes. Supernatant (1 mL) was transferred into Eppendorf tubes and dried down in a speed vacuum concentrator. The residues were reconstituted with 50% ACN/H₂O (100 μL), vortexed, and centrifuged for 10 minutes at 15,000 rpm at 4°C. The supernatant (80 μL) was transferred into high-performance liquid chromatography vials for LC/MS-MS analysis. Samples were analyzed on an Agilent 1290 Infinity LC System coupled with an Agilent 6460 triple quadrupole mass spectrometer operated in a negative mode.

The enzyme activity of glucose-6-phosphate dehydrogenase was measured using a G6PD Activity Colorimetric Assay Kit (BioVision). Cells were homogenized in 500 μL of cold PBS. Cell lysates were assayed following the manufacturer's instructions. The concentrations of NADP⁺ and NADPH were determined with a NADP/NADPH-Glo Assay (Promega). Cells were transfected with siRNA in 96-well plates. Forty-eight hours post transfection, cells were lysed and treated under heat and acid conditions to separate NADPH and NADP⁺ components and lysates were incubated with luminescent substrate. NADPH and NADP⁺ concentrations were determined by NADPH and NADP⁺ standard curves, respectively. The NADPH/NADP⁺ ratio was calculated to represent cellular redox potential.

Tissue slides were subjected to IHC staining according to the manufacturer's protocol (Biocare Medical). Six-micron sections were cut. Slides were baked at 60°C overnight, and deparaffinized in two changes of xylene. After rehydration, antigen was retrieved in 10 mmol/L sodium citrate buffer (pH 6.0) at 125°C for 10 minutes. Slides were blocked with PeroxAbolish, avidin, biotin, 3% BSA in PBS and Rodent Blocking M for xenografts. Slides were incubated at 4°C overnight with primary antibodies indicated in figures (sources of antibodies are listed in Supplementary Table S3), followed by biotinylated secondary antibody and streptavidin-conjugated horseradish peroxide. Slides were stained in DAB chromogen for 1 minute and then counterstained with hematoxylin. Tissue microarray (TMA) was scored and analyzed by two investigators (H. Yang and Z. Lu) independently.

MCF7 and HIM3 breast cancer cells were infected with SMARTvector Tet-inducible shPFKFB2 lentivirus (Clone ID V3SH7669-224728719) and subsequently selected using 1 μg/mL of puromycin according to the manufacturer's protocol (Dharmacon). Clonal populations were generated by limiting dilution under puromycin selection. Clones under best knockdown efficiency were selected by Western blotting under 1 μg/mL of doxycycline for 48 hours. MCF7- and HIM3-inducible cells were maintained in DMEM (Corning) supplemented with 10% FBS and 0.5 μg/mL of puromycin (Sigma).

Experiments with Hsd:Athymic nude-Foxn1^nu homozygous mice (Envigo) and NOD/SCID mice (Charles River Laboratory) were reviewed and approved by the Institutional Animal Care and Use Committee of M.D. Anderson Cancer Center (IACUC 00001052; Houston, TX). For ovarian cancer xenografts, 60 female nude mice were injected intraperitoneally with 1 × 10⁶ A2780 cells or 1 × 10⁶ HeyA8 cells. After 7-day inoculation, mice were randomly assigned to four treatment groups (n = 10) as indicated. In the MCF7 breast cancer xenograft study, mice were implanted with 17β-estradiol pellets (0.5 mg per pellet, 60-day release; Innovative Research of America) under 2% isofluorane. A total of 3 × 10⁶ MCF-7 cells suspended in 100 μL of equal part of PBS and Matrigel with reduced growth factor (Corning) were injected into the fourth mammary fat pads 3 days after estradiol pellet implantation. When xenografts reached 50 mm³, tumor-bearing mice were randomly divided into four groups as indicated (n = 7). In the HIM3 xenograft study, xenograft models were established according to published protocols (28). A total of 2.5 × 10⁵ irradiated immortalized human mammary stromal fibroblasts were mixed with 2.5 × 10⁵ nonirradiated cells. Fibroblasts were then mixed with 1 × 10⁶ tumor cells and one third volume of Matrigel. This mixture was injected into the fourth mammary fat pads of 4-week-old NOD/SCID mice. After xenografts reached 50 mm³, tumor-bearing mice were randomly divided into four groups as indicated (n = 13; 8 for tumor growth and 5 for IHC). For IHC analysis, mice were sacrificed and tumors were obtained after 1 week of doxycycline treatment. Tumor sizes were measured every week in two dimensions using a caliper, and tumor volume was calculated with the following formula: tumor volume (mm³) = 0.5 × ab² (a and b being the longest and the shortest diameters of the tumor, respectively). Mice were monitored until tumor burden reached 1,000 mm³ for euthanasia.

Data are represented as means ± SDs unless specified otherwise. Statistical significance was determined using an independent sample Mann–Whitney U test. Survival was analyzed by the Kaplan–Meier method and compared using log-rank (Mantel–Cox) test. All experiments were performed in triplicate and repeated at least 2–3 time. The minimal level of significance was P = 0.05. * or #, P <0.05; ** or ##, P <0.01. If not specified, the analysis is not significant.

We previously performed siRNA kinome screens to identify kinases that modulate paclitaxel resistance in ovarian cancer cells (23). To identify targetable molecules that alter paclitaxel sensitivity in ovarian and breast cancer cells, we conducted a confirmatory screen in multiple cancer cell lines, knocking down the 12 positive hits from our initial screen that were known to participate in glucose and energy metabolism. Each of the 12 positive hits was knocked down in four ovarian and two breast cancer cell lines to determine whether decreased expression of each kinase would enhance paclitaxel sensitivity (z-score < −0.5) in at least three of the cancer cell lines. Of the 12 kinases screened, knockdown of four kinases (CKMT1B, CKMT2, PFKFB2, and PFKFB4) with siRNAs were shown to have significant effects in at least three of six ovarian and breast cancer cell lines. Knockdown of these four kinases preferentially enhanced paclitaxel sensitivity in ovarian and breast cancer cell lines that preserve wtTP53 (HeyA8, A2780, and ZR-75-1) compared with cell lines that contained mutant or null TP53 (SKOv3ip, OC316, and BT-20; Supplementary Fig. S1A; Supplementary Table S4).

Gene expression of the 12 kinases was examined using the Cancer Genome Atlas (TCGA) database. Only PFKFB2 was overexpressed in a fraction of both ovarian (10%) and breast cancers (26%; refs. 29–31; Supplementary Fig. S1B). Of the 10% ovarian cancer samples, 4.2% of samples are wtTP53 and of the 26% breast cancer samples, 49.4% of samples are wtTP53. PFKFB2 is a bifunctional glycolytic enzyme that regulates the level of fructose-2,6-bisphosphate (F2,6-BP), a strong allosteric activator of phosphofructokinase 1 (PFK1; refs. 32, 33), one of the most important regulatory enzyme of glycolysis. Consequently, PFKFB2 was chosen for further investigation. To assess the prevalence of PFKFB2 overexpression in ovarian and breast cancers, PFKFB2 protein expression was measured by IHC analysis, in ovarian and breast cancer TMAs prepared from specimens obtained at the time of initial surgery. The expression of PFKFB2 was significantly upregulated in cancer cells compared with normal ovarian/fallopian (P = 0.026) and breast epithelial cells (P < 0.0001; Fig. 1A; Supplementary Fig. S2; Supplementary Tables S5 and S6). Intense staining (3–4+) was observed in 25% of ovarian cancers and 25% of breast cancers, but in 0% of normal ovarian/fallopian epithelium and 4.3% of normal breast epithelial cells. In addition, differential expression of PFKFB2 was shown between ovarian and breast cancers (intense staining) and all other normal organ tissues (Supplementary Fig. S3).

To validate the siRNA screen results and address whether high expression of PFKFB2 is one of the important factors contributing to tumorigenic phenotype in ovarian and breast cancers, we performed clonogenic assays after PFKFB2 siRNA or control siRNA transfection. PFKFB2 siRNA significantly inhibited cell growth, when compared with control siRNA. Notably, the most marked effects of silencing PFKFB2 on cell proliferation were seen in HeyA8, A2780, and MCF-7 ovarian and breast cancer cell lines that express wtTP53, with more modest inhibition observed in the OC316, SKOv3ip, OVCAR8, and MDA-MB-231 cells that express mutant or null p53 (Fig. 1B; Supplementary Fig. S4A). Moreover, the ability of PFKFB2 knockdown to sensitize ovarian and breast cancer cells to paclitaxel was greater in HeyA8, A2780, MCF-7, and ZR-75-1 than in OC316, OVCAR8, SKOv3ip, and MDA-MB-231 cells. When the concentrations of paclitaxel required to inhibit cancer cell growth by 50% (IC₅₀) were compared with and without PFKFB2 knockdown, silencing PFKFB2 using two individual PFKFB2 siRNAs in the HeyA8, A2780, MCF-7, and ZR-75-1 cell lines decreased cell viability and reduced the average paclitaxel IC₅₀ by 2.5-fold (range = 2.3–2.6-fold; Fig. 1C; Supplementary Fig. S4B). However, when the IC₅₀ of paclitaxel was measured in OC316, OVCAR8, SKOv3ip, and MDA-MB-231 cells with mutant or null TP53, silencing PFKFB2 with a single siRNA reduced the average paclitaxel IC₅₀ by only 0.95-fold (range = 0.8–1 fold; Fig. 1C; Supplementary Fig. S4B). Similar knockdown efficiencies were observed across both sets of cell lines. (Fig. 1D). In addition, we used paired isogenic patient-derived xenograft (PDX) lines differing only in TP53 (ER-PR-HER2-TP53 wild-type and a TP53-deficient subline; ref. 28), to evaluate the cancer cell growth and paclitaxel sensitivity in cell culture. Consistent with the above data, we observed significant growth inhibition (Fig. 1E) and paclitaxel sensitization (Fig. 1F) with the wtTP53 parental cell line HIM3 TP53 WT, but not with the TP53-deficent subline HIM3 TP53 knockout (KO) by addition of doxycycline to silence PFKFB2 in cultured cells. Conversely, overexpression of PFKFB2 increased cell growth and reduced sensitivity to paclitaxel, increasing the IC₅₀ of paclitaxel (P < 0.05 in HeyA8 and P < 0.01 in MCF-7), when HeyA8 and MCF-7 cells were transiently transfected with a PFKFB2 plasmid and selected with G418 (Supplementary Fig. S4C). Thus, these observations support the hypothesis that high expression of PFKFB2 can facilitate cell proliferation and paclitaxel resistance in ovarian and breast cancer cells that express wtTP53.

Increased glucose metabolism in cancer cells is required to meet their high energetic and biosynthetic demands. Many glycolytic enzymes, including PFKFB2, are induced in malignant cells (34) and PFKFB2 is thought to play an important role in regulating carbon flux through the glycolytic pathway and PPP (Fig. 2A; refs. 35, 36). To understand the metabolic basis for the therapeutic effect of silencing PFKFB2 on tumor cell growth and sensitivity to paclitaxel, we have studied the mechanism(s) by which silencing PFKFB2 alters glycolysis and the PPP.

Initially, we measured the concentration of F2,6-BP in two ovarian cancer cell lines and two breast cancer cell lines by LC/MS with or without silencing PFKFB2. Knockdown of PFKFB2 increased F2,6-BP and fructose-1,6-bisphosphate (F1,6BP; siPFKFB2/siControl > 1) in all the cell lines tested (Fig. 2B), consistent with increased glycolysis. The relative concentrations of F1,6BP + F2,6BP were considerably higher in the A2780 and MCF-7 ovarian and breast cancer cell lines that express wtTP53 than those in the SKOv3ip and MDA-MB-231 cell lines that express mutant TP53 (Fig. 2B). Consistent with these results, studies with the Seahorse XF glycolysis stress assay demonstrated a higher glycolytic flux in HeyA8, A2780, and MCF-7 cells than in OC316, SKOv3ip, and MDA-MB-231 cells (Fig. 2C; Supplementary Fig. S5A). Moreover, pyruvate, the end product of glycolysis, was significantly higher in wtTP53 cell lines than that in TP53-mutant or -null cell lines after depletion of PFKFB2 (Fig. 2D; Supplementary Fig. S5B). In contrast, we measured two oxidative PPP products by LC/MS (6-phosphoglyconate, 6PG and ribulose 5-phosphate, Ru5P) and found that silencing PFKFB2 increased the concentrations of both (Fig. 2E) in SKOv3ip and MDA-MB-231 cells compared with A2780 and MCF-7 cell lines, suggesting that knockdown of PFKFB2 resulted in the shunting of glycolytic intermediates to the oxidative PPP in cancer cells with mutant TP53.

One possible consequence of increasing fructose-2,6-bisphophate (F1,6BP/F2,6BP) would be an increase in PFK1 activity. It would be predicted to increase glycolytic flux and decrease the amount of glucose-6-phosphate (G6P) entering the PPP, which would reduce the biosynthesis of NADPH. On the other hand, higher levels of 6PG and Ru5P would imply higher oxidative PPP flux and increased production of NADPH. To test this, we measured NADPH and NADP⁺ using cancer cell lysates and found that silencing PFKFB2 decreased the ratio of NADPH/NADP⁺ (Fig. 2F; Supplementary Fig. S5C) in A2780, HeyA8, and MCF-7 cells with wtTP53. In contrast, silencing PFKFB2 produced minimal changes in the NADPH to NADP⁺ ratio in OC316, SKOv3ip, OVCAR8, and MDA-MB-231 cancer cells with mutant TP53. Thus, oxidative PPP flux was significantly upregulated in TP53-mutant cell lines, however it was downregulated in wtTP53 cell lines after depletion of PFKFB2 by siRNA.

To document differences in response to depletion of PFKFB2 in wtTP53 and TP53-mutant cells, the basal activity of G6PD was measured in HeyA8, A2780, MCF-7, OC316, SKOv3ip, OVCAR8, and MDA-MB-231 cells. We found that basal activity of G6PD was much lower in A2780, HeyA8, and MCF-7 cancer cells with wtTP53 (Fig. 3A) than in OC316, SKOv3ip, OVCAR8, and MDA-MB-231 with mutant TP53 (Fig. 3A). This low basal activity was significantly increased after knockdown of TP53 in wtTP53 cell lines (Fig. 3B; Supplementary Fig. S6A). In line with these results, decreased NADPH was observed in wtTP53 cancer cells following knockdown of PFKFB2, and the reduction in NADPH after PFKFB2 knockdown was rescued by knockdown of TP53 (Fig. 3C; Supplementary Fig. S6B) and decreased NADPH was also observed in wtTP53 cancer cells following knockdown of G6PD (Fig. 3D; Supplementary Fig. S6C). These results are consistent with our observation that silencing PFKFB2 in wtTP53 cell lines resulted in increased glycolysis (Fig. 2D) and decreased PPP flux (Fig. 2E). Taken together, these results suggest that tumor-associated TP53 mutants in high grade ovarian cancer which lack G6PD inhibitory activity may promote tumor cell proliferation by directing glucose toward PPP flux and NADPH production. In contrast, ovarian cancers with wtTP53 may attenuate PPP activity and instead channel glucose through glycolysis to generate ATP and pyruvate for tricarboxylic acid cycle and OxPhos metabolism (Supplementary Fig. S6D).

To explore whether decreased levels of NADPH following PFKFB2 silencing result in ROS accumulation and decreased viability in ovarian and breast cancer cells, ROS was measured using two fluorescent dyes, CM-H2DCFDA and dihydroethidium. Fluorescence intensity of the two dyes was enhanced by silencing PFKFB2, indicating the accumulation of ROS. However, more significant accumulation of ROS was observed in HeyA8, A2780, and MCF-7 cells with wtTP53 (Fig. 3E) than that in OC316, SKOv3ip, and MDA-MB-231 cells with mutant TP53 (Supplementary Fig. S7A). These results are consistent with the hypothesis that ROS accumulation may result from diminished pools of intracellular NADPH in wtTP53 ovarian and breast cancer cell lines. To measure the effect of increased ROS on ovarian cancer growth, we performed clonogenic assays in the presence and absence of N-acetyl-l-cysteine (NAC), an ROS inhibitor. NAC rescued PFKFB2-knockdown–mediated growth inhibition in HeyA8 and A2780 cancer cells (Fig. 3F; Supplementary Fig. S7B). In contrast, NAC had little effect on OC316, SKOv3ip, and OVCAR8 cells (Supplementary Fig. S7C). These studies support the interpretation that PFKFB2 knockdown affects the proliferation of ovarian cancer cells with wtTP53 in an ROS-dependent manner.

Previous studies indicate that oxidative stress can trigger the JNK pathway and the phosphorylation of p53 (37, 38). Therefore, we examined phosphorylation of JNK and phosphorylation of p53 by Western blot analysis in HeyA8, A2780, and MCF-7 cells using siRNA to silence PFKFB2. We found that silencing PFKFB2 induced expression of p53 and, more importantly, significantly increased the phosphorylation of both JNK and p53, suggesting that ROS induced-JNK phosphorylation led to p53 activation (Fig. 4A). To confirm that p53 activation is triggered by ROS-mediated JNK phosphorylation, we investigated the effect of ROS and JNK inhibitors on p53 phosphorylation (Fig. 4B and C). We observed that NAC (Fig. 4B) and JNK inhibitor (Fig. 4C) substantially impeded JNK and p53 phosphorylation after PFKFB2 silencing in HeyA8, A2780, and MCF-7 cells. This result confirms that JNK is a mediator of p53 activation in response to oxidative stress resulting from downregulation of PFKFB2. Collectively, our data suggest that p53 is a crucial node downstream of ROS/JNK molecular pathway, inhibiting growth in ovarian and breast cancers.

To identify possible mechanism(s) by which PFKFB2 knockdown enhances the cytotoxic activity of paclitaxel in ovarian and breast cancer cells, we first examined the effect of PFKFB2 siRNA on apoptosis in the presence and absence of paclitaxel. Flow cytometry revealed that PFKFB2 siRNA induced a higher percentage of annexin V–positive cells, and the percentage was further increased when combined with paclitaxel (Fig. 5A). We then tested whether p53 is a critical mediator in response to PFKFB2-knockdown–induced apoptosis. We measured the expression of Puma and p21, which are downstream targets of TP53 in HeyA8, A2780, and MCF-7 cells. The mRNA and protein levels of Puma and p21 were increased by silencing PFKFB2 (Fig. 5B and C; Supplementary Fig. S8), implying that PFKFB2 siRNA–induced apoptosis in HeyA8, A2780, and MCF-7 cells is mediated by p53 activation. Moreover, TP53 knockdown decreased the mRNA and protein levels of Puma and p21 (Fig. 5B and C; Supplementary Fig. S8) and rescued cells from apoptosis induced by PFKFB2 knockdown (Fig. 5D), confirming that p53, Puma, and p21 are critical mediators of ROS/JNK induced apoptosis.

Given the outcome of cell culture studies, we asked whether depletion of PFKFB2 affects the ability of wtTP53 ovarian and breast cancer cells to form tumors in immunodeficient mice. Orthotopic human ovarian and breast xenograft models were used to determine the potential of PFKFB2 as a target for clinical therapy. We used a well-characterized DOPC (1,2-dioleoyl-sn-glycero-3-phosphatidylcholine) nanoliposomal system for in vivo siRNA delivery in mouse ovarian cancer models (20). For each cancer cell line, four groups of 10 mice were treated for 4 weeks as follows: (i) control siRNA-DOPC, (ii) PFKFB2 siRNA-DOPC, (iii) a combination of control siRNA-DOPC plus paclitaxel, and (iv) a combination of PFKFB2 siRNA-DOPC plus paclitaxel. Liposome-encapsulated PFKFB2 siRNA significantly inhibited tumor growth in both ovarian cancer xenograft models (HeyA8 and A2780). Greater growth inhibition was observed when PFKFB2 siRNA was combined with paclitaxel than that with either single treatment (P < 0.05; Fig. 6A). In a murine human breast cancer xenograft model, MCF-7, a lentiviral vector–mediated doxycycline (+doxycycline)-inducible shRNA expression system was used to silence PFKFB2. After addition of doxycycline to drinking water, PFKFB2 was silenced and the inhibitory effect on tumor growth as a single agent or in combination with paclitaxel treatment was examined. shPFKFB2 (+doxycycline) or paclitaxel alone produced significant inhibition of tumor growth (P < 0.01), compared with control (−doxycycline), moreover, shPFKFB2 plus paclitaxel produced greater inhibition of tumor growth than either single agent in this orthotopic breast cancer mouse model (P < 0.01), which is consistent with the ovarian cancer data (Fig. 6B). To further evaluate the role of TP53 in PFKFB2 depletion–mediated growth inhibition in vivo, we used paired isogenic PDX lines and assessed the outcome in xenograft models by inducing shPFKFB2 expression using doxycycline. Doxycycline induction (silencing PFKFB2) plus paclitaxel significantly inhibited tumor growth, and prolonged survival of mice bearing HIM3 wtTP53 tumor (P = 0.008), compared with mice bearing HIM3 TP53 KO tumor (Fig. 6C). Moreover, the concentration of paclitaxel used in the experiments remained identical (5 mg/kg) in MCF-7 and HIM3 breast cancer xenograft studies. In the MCF-7 model, paclitaxel alone showed statistically significant inhibition of tumor growth as described above. However in the HIM3 wtTP53 model, which is a triple-negative breast cancer, paclitaxel alone did not inhibit tumor growth; but when combined with PFKPB2 shRNA (+doxycycline), tumor growth was significantly inhibited with prolonged survival. As expected, p-JNK and p-p53 increased in HIM3 wtTP53 xenografts with depletion of PFKFB2 (+doxycycline) by IHC (Fig. 6D). In addition, HIM3 TP53 KO xenografts showed minimal changes in p-JNK and p-p53 with PFKFB2 depletion (Fig. 6D). Taken together, these results demonstrate an important effect of knockdown of PFKFB2 and identify a potential therapeutic target to enhance paclitaxel sensitivity in breast cancers that retain functional TP53.

In this article we document that the PFK-2/FBPase-2 isoenzyme (PFKFB2) drives growth of ovarian cancer cells with wtTP53 and regulates paclitaxel sensitivity. While wtTP53 is found primarily in the small subset of low grade serous ovarian cancers, we found that PFKFB2 plays a similar role in the more prevalent wtTP53 breast cancers, making it an attractive target for drug development. Growth inhibition and cell death induced by silencing PFKFB2 have been shown to depend on dysregulation of the oxidative PPP, failure to detoxify ROS, activation of JNK and p53, upregulation of p21, Puma, and the induction of apoptosis in wtTP53 cancer cells.

Cancer cells express altered levels of the different PFK-2/FBPase-2 isoenzymes and even modulate their relative kinase and bisphosphatase activities according to metabolic needs in a spatial and/or temporal manner (36). The high biosynthetic demand of rapidly proliferating tumor cells also requires the production of large amounts of NADPH and the bisphosphatase activity may be indispensable to limiting glycolytic flux and diverting G6Pinto the oxidative PPP for NADPH production (39). It is therefore possible that cancer cells tightly regulate the activity of different PFK-2/FBPase-2 enzymes to meet their metabolic demands to maintain viability. In response to depletion of PFKFB2, the levels of NADPH and 6PG were decreased in cancer cells that harbored wtTP53, suggesting that PFKFB2 is acting primarily as a phosphatase in those cells. Consequently, silencing PFKFB2 results in ROS accumulation, oxidative stress, and apoptosis in ovarian and breast cancers. A recent publication from Schulze and colleagues reported that PFKFB4 isoform of PFK2 plays a role in prostate cancer cell survival (35). PFKFB4 was found to be required to maintain redox balance and to support tumor growth, highlighting the importance of metabolic regulation of biosynthesis and antioxidant production in prostate cancer, which is consistent with PFKFB2 function that we have observed in ovarian and breast cancers.

Our data have shown that silencing PFKFB2 only inhibits growth and increases paclitaxel sensitivity in cancer cell lines and xenografts that contain wtTP53, consistent with the context that p53 plays an important role in PPP flux and redox balance. TP53 is frequently mutated in many human malignancies including ovarian and breast cancers. Low grade ovarian cancers generally have wtTP53, but 99% of the more common high grade serous ovarian cancers exhibit TP53 mutations. More than half of breast cancers retain wtTP53, which decreases activity of the oxidative PPP by binding to G6PD, the first and rate-limiting enzyme of the PPP (40). Wtp53 binding to G6PD prevents the formation of the active dimer and decreasing the production of NADPH for ROS detoxification and biosynthesis. Therefore, cancer cells with mutant TP53 that lack G6PD inhibitory activity are prone to enhanced PPP glucose flux and NADPH production, resulting in reduced oxidative stress and decreased apoptosis. Silencing PFKFB2 in ovarian and breast cancers that harbor wtTP53 causes ROS/JNK-mediated activation of p53, implying the presence of a positive feedback loop, which can block further PPP through induced expression of p53.

Elevated levels of ROS have been detected in almost all cancers, where they promote many aspects of tumor development and progression (41). However, tumor cells also express increased levels of antioxidant proteins to detoxify ROS (41, 42) suggesting that careful regulation of intracellular ROS levels is required for cancer cell survival. Our study revealed that the ROS accumulation in ovarian and breast cancer cells in response to PFKFB2 depletion induces p53-dependent apoptosis and causes irreversible cellular damage, suggesting the importance of ROS metabolism in mutant TP53–driven cell proliferation and cancer development. Importantly, because metabolic reprogramming in cancer drives tumor cells to be highly dependent on ROS metabolism, disrupting the production of antioxidants could provide an effective treatment strategy (43). Chemotherapeutic agents that induce oxidative stress could be combined with strategies to block NADPH production for achieving synergistic effects. Our study confirmed that PFKFB2 modulates ROS production and activity of p53, resulting in growth inhibition and increased sensitivity to paclitaxel. It is, however, difficult to conclude that TP53 has a direct role as a determinant or biomarker of cell sensitivity to paclitaxel, particularly considering the conflicting reports on this theme in the literature (44–48).

In summary, PFKFB2 silencing upregulated glycolysis and attenuated PPP flux in cancer cells with wtTP53 function. Silencing PFKFB2 induces apoptotic cell death and paclitaxel sensitivity in ovarian cancer and breast cancer cells that harbored wtTP53 in vitro and in vivo. Our data support the importance of PFKFB2 as a potential predictive biomarker and target for molecular therapy. As PFKFB2 inhibitors become available, trials can be conducted to enhance the initial response of type I low grade ovarian cancers to paclitaxel therapy. A majority of breast cancers have wtTP53 and even a small fraction of triple-negative breast cancers retain p53 function (Supplementary Fig. S9A and S9B) making this approach applicable for several distinct cancer types. Although it remains to be determined if a particular subtype of breast cancer (e.g., luminal A, luminal B, and basal/triple negative) would be particularly susceptible to the combination treatment, our data suggest that those tumors with elevated PFKFB2 expression and functional p53 would likely benefit from knockdown of PFKFB2 to enhance paclitaxel sensitivity. Furthermore the data are consistent with TCGA analysis, which indicates an inverse correlation between high expression of PFKFB2 and poor overall survival (P = 0.02389) in luminal A tumors where the majority are wtTP53 (Supplementary Fig. S9B and S9C). Finally, it will be interesting to evaluate whether depletion of PFKFB2 will provide antitumor efficacy and enhance chemosensitivity in other types of cancer that have retained wtTP53.

A.K. Sood has ownership interests (including patents) at BioPath, is a consultant/advisory board member for Merck and Kiyatec, and reports receiving commercial research grants from M-Trap. No potential conflicts of interest were disclosed by the other authors.

Conception and design: H. Yang, Z. Shu, G. Lopez-Berestein, X.-F. Le, R.C. Bast, Z. Lu

Development of methodology: H. Yang, Z. Shu, S.L. Jeter-Jones, A. Ornelas, C. Rodriguez-Aguayo, N.M. Zacharias, A.K. Sood

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H. Yang, Z. Shu, Y. Jiang, W. Mao, L. Pang, A. Redwood, S.L. Jeter-Jones, A. Ornelas, C. Rodriguez-Aguayo, G. Bartholomeusz, S.W. Millward, H. Piwnica-Worms, A.K. Sood

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H. Yang, Z. Shu, Y. Jiang, W. Mao, L. Pang, A. Redwood, A. Ornelas, S.W. Millward, G. Lopez-Berestein, A.A. Ahmed, H. Piwnica-Worms, A.K. Sood

Writing, review, and/or revision of the manuscript: H. Yang, Z. Shu, A. Ornelas, N.M. Zacharias, S.W. Millward, A.A. Ahmed, H. Piwnica-Worms, A.K. Sood, R.C. Bast, Z. Lu

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H. Yang, Z. Shu, N.B. Jennings, J. Zhou

Study supervision: H. Yang, R.C. Bast, Z. Lu

Other (conducted the high throughput siRNA screens): L.R. Iles

This work was supported by the Cancer Prevention and Research Institute of Texas RP110595-P1, the MD Anderson SPOREs in Ovarian Cancer NCI P50 CA 083639 and CA 217685, the Shared Resources of the MD Anderson CCSG grant NCI P30 CA016672, R35 CA209904, the American Cancer Society Research Professor Award, the Frank McGraw Memorial Chair in Cancer Research, The National Foundation for Cancer Research, the philanthropic support from generous donations from Stuart and Gaye-Lynn Zarrow, the Mossy Foundation, and the Roberson Endowment. A.A. Ahmed is supported by grants from the NIHR National Institute for Health Research Oxford Biomedical Research Centre and Ovarian Cancer Action. We also acknowledge grants R21CA181994 and R25T CA557730.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.