Pyruvate Kinase M1 Suppresses Development and Progression of Prostate Adenocarcinoma

Prostate cancer is the second leading cancer-related cause of death in men and given time most men will develop prostate cancer (1). Loss of the tumor suppressive lipid phosphatase PTEN is associated with abnormal prostate growth in prostate cancer, with many prostate cancers exhibiting decreased PTEN expression due to mutation or epigenetic silencing (2–4). Loss of PTEN activity results in phosphatidylinositol (3, 4, 5)-triphosphate accumulation and activation of AKT signaling to drive uncontrolled proliferation and survival (5, 6). How these signaling events promote prostate cancer have been extensively studied (7), as have the ways in which growth factor signaling pathways can affect cell metabolism (8, 9). However, whether changes in metabolic enzyme expression affect prostate tumor initiation and progression is less well defined.

Changes in metabolism are necessary to sustain cancer cell proliferation (9, 10). Many cancer cells exhibit increased glucose uptake (10, 11). Increased tumor glucose consumption relative to normal tissues is exploited clinically to stage cancers through utilization of ¹⁸FDG-PET imaging (12); however, most prostate cancers grow at a slower rate than highly ¹⁸FDG-avid malignancies and ¹⁸FDG-PET is not often used in the clinical management of patients with prostate cancer, leading to the notion that these cancers rely less on glucose metabolism (13). Nevertheless, prostate cancer cells metabolize glucose in culture (14, 15), and ¹⁸FDG-PET avidity is observed in prostate cancer, including aggressive tumors (16, 17). Whether changes in glucose metabolism influence prostate cancer initiation and progression has not been extensively studied.

Induction of cellular senescence can suppress cancer development (18–20). Overcoming cellular senescence is thought to be particularly important in the pathogenesis of some prostate cancers, as Pten loss results in a p53-dependent senescence response by prostate epithelial cells in preclinical models (21), and can suppress tumor formation (18, 20). Changes in metabolism to increase glucose oxidation can promote senescence in some tissues (22) and senescent lesions can be hypermetabolic (23). Cellular senescence is also associated with decreased anaplerotic flux into the tricarboxylic acid cycle (24), and nucleotide deficiency can promote this phenotype by promoting DNA replication stress (25). Of note, loss of Pten in prostate epithelial cells can cause DNA replication stress, induce a DNA damage response, and contribute to senescence (26). Taken together, these studies suggest that changes in metabolism that affect nucleotides and DNA replication can contribute to the induction or maintenance of a quiescent or senescent state; however, the relationship between metabolism and tumor suppression due to cellular senescence is controversial (23, 27, 28), and whether changes in glucose metabolism influence senescence as a tumor suppressive mechanism in prostate cancer is not known.

Regulation of pyruvate kinase activity can influence the extent of glucose oxidation and nucleotide synthesis (29–31). Most human and murine tissues express an isoform of pyruvate kinase that is encoded by the PKM gene (32, 33). This gene produces an RNA product that is alternatively spliced to generate mRNAs encoding two different isoforms of the enzyme: PKM1 and PKM2 (34). The mRNA encoding PKM1 and PKM2 differ only in inclusion of either exon 9 for the PKM1 message or exon 10 for the PKM2 message. PKM1 is a constitutively active enzyme that promotes oxidative glucose metabolism (29, 32). PKM2 is allosterically regulated, and decreased pyruvate kinase activity associated with this isoform can promote anabolic metabolism including nucleotide synthesis (32, 35). Genetic deletion of the Pkm2-specific exon in proliferating primary mouse embryonic fibroblasts that normally express PKM2 results in PKM1 expression and an irreversible proliferation arrest (31). Moreover, deleting only one Pkm2 allele also results in PKM1 expression that leads to proliferation arrest despite continued expression of PKM2 at wild-type levels, indicating that expression of PKM1 rather than loss of PKM2 is responsible for the phenotype (31). Proliferation arrest caused by PKM1 expression can be prevented by addition of exogenous nucleotide bases (31), suggesting that high pyruvate kinase activity associated with PKM1 expression can limit nucleotide synthesis, but whether this results in tumor suppression is not known.

Although most tissues in mice express either the Pkm1 or Pkm2 isoform of pyruvate kinase (33), cancer cells preferentially express Pkm2 (32, 36). This is thought to be advantageous to cancer cells because PKM2 expression allows cells to adapt pyruvate kinase activity to different cell conditions (35, 37); however, why PKM2 is selected for in most cancers is controversial (33, 38). Pyruvate kinase is active as a homotetramer (30, 39, 40). Allosteric regulators that decrease PKM2 activity function by destabilizing the active tetramer. In contrast, PKM1 is constitutively active because residues encoded by the isoform-specific exon promote stable tetramer formation. Small molecule pyruvate kinase activators have been identified that stabilize the PKM2 tetramer to promote an enzyme state similar to PKM1 (30, 41–43). One such PKM2 activator, TEPP-46, is bioavailable when dosed orally in mice and can inhibit xenograft growth, phenocopying the effects of PKM1 expression (30). However, Pkm2 expression is not required for tumor growth in several mouse cancer models (33, 44–50), suggesting that loss of Pkm2 expression might limit the ability of pyruvate kinase activators to be effective cancer drugs.

Whether high pyruvate kinase activity due to PKM1 expression is a barrier to cancer initiation is not known. PKM1 expression is reported to provide a metabolic advantage to tumors in some contexts (38), and suppress tumor growth in others (29, 30, 45). Because PKM1 is constitutively active, understanding where PKM1 expression suppresses tumor growth could inform which tumor types might be sensitive to pyruvate kinase activating drugs. To study how PKM1 expression affects tumor formation, we generated mice harboring a conditional allele for the unique exon included in Pkm1. We crossed animals harboring this Pkm1-conditional allele, as well as mice harboring a Pkm2-conditional allele (45), to a Pten loss-driven mouse prostate cancer model (21, 51). We found that PKM isoform expression profoundly impacts prostate tumor initiation and progression. Deletion of both Pkm1 and Pten in the prostate results in the formation of aggressive prostate tumors that limit animal survival. In contrast, deletion of Pkm2 and Pten in the prostate results in high Pkm1 expression and suppresses tumor formation. Importantly, small molecule PKM2 activators can also suppress prostate tumor growth, and many human prostate cancers retain PKM2 expression, arguing that forcing pyruvate kinase into a high activity state might have a role in managing prostate cancer in patients.

The conditional allele for Pkm1 was generated using standard protocols to introduce loxP sites in the intronic region flanking exon 9 of the Pkm1 gene in a manner analogous to how the Pkm2 allele was generated (see ref. 45). For all experiments, the PbCre4 (MGI:2385927) allele was maintained in males due to previously observed germline recombination and mosaic expression of floxed alleles when the PbCre4 allele is transmitted through females (52). Males harboring PbCre4, Pten (MGI:2679886), Pkm1, or Pkm2 (MGI: 5547750) floxed alleles were crossed to females harboring Pten, Pkm1, or Pkm2 floxed alleles to generate prostate restricted deletion of these genes and splice products. All animals were maintained on a mixed background and littermates were used for direct comparisons.

Animals were fasted overnight before administration of 100 μCi of FDG ¹⁸F through a tail-vein catheter. Animals were kept warm using a heated water pad and placed under 2% anesthesia during the 1 hour uptake time to lower background signal. Because of the small size of the prostate and its proximity to the bladder in 7- to 11-week-old control and Pten^pc–/– animals, tissue was harvested after FDG administration and gamma counts used to assess FDG uptake in prostate and gastrocnemius muscle tissue. For studies involving 6-month-old mice, animals were imaged for 10 minutes by PET and 1 minute by CT using 720 projections at 50 kV and 200 μA using Sofie G8 PET/CT. Images were CT attenuation corrected and MLEM3D reconstructed. All images were decay corrected to the time of injection. The average signal intensity for three regions of each prostate was normalized to the average signal intensity for three regions in the heart of each animal.

Asp718 (Roche)-digested genomic DNA was analyzed by Southern blot using standard protocols and probe binding was visualized by autoradiography using an analogous strategy to what was described previously for the Pkm2 allele (45). Asp718 has the same restriction site specificity as Kpn1.

PCR genotyping for Pkm1 conditional mice was developed to detect and amplify the targeted Pkm genetic locus and performed using forward (5′-CACGCAACCATTCCAGGAGCATAT-3′) and reverse (5′-TGGTGACCTTGGCTGTCTTCCTGA-3′) primers. To genotype PbCre4, forward (5′-CTGAAGAATGGGACAGGCATTG-3′) and reverse (5′-CATCACTCGTTGCATCGACC-3′) primers were used as suggested by the NCI mouse repository. Genotyping of the Pten and Pkm2 alleles used in this study was performed as described previously (45, 53).

Western blots were performed using primary antibodies against PKM1 (Sigma-Aldrich, catalog no. SAB4200094, RRID:AB_10624711), PKM2 (Cell Signaling Technology, catalog no. 4053, RRID:AB_1904096), PKM (Cell Signaling Technology, catalog no. 3190, RRID:AB_2163695; Abcam, catalog no. ab6191, RRID:AB_2163678), and vinculin (Sigma-Aldrich, catalog no. V4505, RRID:AB_477617). For IHC, the following primary antibodies were used: 1:600 PKM1 (Cell Signaling Technology, catalog no. 7067, RRID:AB_2715534), 1:1,200 PKM2 (Cell Signaling Technology, catalog no. 4053, RRID:AB_1904096), 1:100 phospho-CHK1(ser317) (MyBioSource, catalog no. MBS9600779), 1:300 vimentin (Abcam, catalog no. ab24525, RRID:AB_778824), 1:200 pancytokeratin (Abcam, catalog no. ab9377, RRID:AB_307222); 1:200 synaptophysin (Thermo Fisher Scientific, catalog no. PA5–16417, RRID:AB_10989504), 1:16,000 PCNA (Cell Signaling Technology, catalog no. 2586, RRID:AB_2160343), 1:100 Ki-67 (BD Pharmingen, catalog no. 556003), 1:100 AR (Santa Cruz, catalog no. sc-816-G). IHC analysis of PKM1 and PKM2 in mouse tissue sections was performed as follows: 5-μmol/L-thick formalin-fixed paraffin-embedded (FFPE) tissue sections were deparaffinized and immediately underwent heat-mediated antigen retrieval in a pressure cooker at 125°C for 5 minutes in Citra pH 6.0 solution (Biogenex, catalog no. HK086). Endogenous peroxidase activity was quenched with BLOXALL (Vector Labs, catalog no. SP-6000) for 20 minutes. Sections were then blocked with Protein Block (Dako, catalog no. X0909) for 30 minutes, incubated overnight with primary antibody at 4°C, incubated with SignalStain Boost (Cell Signaling, catalog no. 8114) for 30 minutes, incubated with SignalStain DAB substrate (Cell Signaling, catalog no. 8059) for 2 minutes (PKM2) or 5 minutes (PKM1) at room temperature, and counterstained with hematoxylin. IHC analysis of pChk1 was performed as follows: 5-μmol/L-thick FFPE tissue sections were deparaffinized and immediately underwent heat-mediated antigen retrieval in a pressure cooker at 125°C for 5 minutes in Citra pH 6.0 solution (Biogenex, catalog no. HK086), Endogenous peroxidase activity was quenched with BLOXALL (Vector Labs, catalog no. SP-6000) for 20 minutes. Sections were then blocked for 30 minutes with 3% normal goat serum, incubated overnight with primary antibody at 4°C, incubated with avidin/biotin/HRP reagents per manufacturer recommended protocol (Vector Labs ABC-HRP Kit, catalog no. PK-4001), incubated with DAB substrate (Vector Labs, catalog no. SK-4100) for 5 minutes at room temperature, and counterstained with hematoxylin. For all other primary antibodies, incubation was at room temperature for 1 hour. For all other primary antibodies except vimentin, IHC analysis was performed as described above for pCHK1. Vimentin was detected with an AP-conjugated goat-anti-chicken (Thermo Fisher Scientific, catalog no. PA1–28799, RRID:AB_10984880) and stained with Vulcan Fast Red Kit 2 (Biocare, catalog no. FR805). Staining intensity for PKM1, PKM2, and pChk1 in mouse prostate tumors was scored using a 4-category system (negative, low, intermediate, high). Three high power fields per tumor were evaluated and each assigned to a staining intensity category based on the most prevalent staining intensity patterns of tumor cells within that field. The tumor was assigned the median category. Percent Ki-67 positive cells in anterior prostate tumors of Pten^pc–/– mice treated with TEPP-46 for 4 weeks (n = 2) and age matched controls (n = 4) was quantitated using an automated algorithm in QuPath (Version: 0.3.0; ref. 54). Ten high-power fields were selected per tumor. Within each high-power field, areas containing tumor cells were manually selected to generate a region of interest (ROI). Total number of tumor cells and Ki-67 positive cells per ROI were quantitated using the analyze → cell detection → positive cell detection function with the following parameters: setup parameters (hematoxylin OD, pixel size = 0.5 μmol/L), nucleus parameters (background radius = 8 μmol/L, median filter radius = 0 μmol/L, sigma = 1.5 μmol/L, min area = 10 μm², max area = 400 μm²), intensity parameters (threshold = 0.1, max background intensity = 2, split by shape selected), cell parameters (cell expansion = 5 μmol/L, include cell nucleus selected), general parameters (smooth boundaries selected, make measurements selected), intensity threshold parameters (nucleus DAB OD mean, single threshold). Percentage of Ki-67-positive cells equals Ki-67-positive cells divided by total number of tumor cells. All Ki-67 quantification was done by scoring sections in a blinded fashion.

For longitudinal measurements of tumor growth, WT, Pten^pc–/–, Pten;Pkm1^pc–/–, or Pten;Pkm2^pc–/– littermates were randomized into cohorts and prostate tissue size assessed biweekly using a Varian 7T MRI imaging system. Image sequences were acquired using the proton imaging FSEMS sequence (fast spin echo multiple slice) with TR: 4,000 milliseconds; TE: 12 milliseconds in the axial orientation. Additional settings were as follows: 256 × 256 data matrix; 45 × 45 mm region; 1-mm-thick slice; for 20 slices. OsiriX-Viewer was used for image analysis. MRI assessment of abnormal prostate growth was noted when the prostate tissue volume increased over at least two consecutive timepoints.

β-Galactosidase staining was conducted as reported previously (55). In brief, fresh frozen sections were cut to 8 μm thickness and briefly fixed in paraformaldehyde. A solution containing 1 mg of 5-bromo-4-chloro-3-indoyl β-d-galactoside (X-gal) per mL (diluted from a stock of 20 mg of dimethylformamide per mL) with 40 mmol/L citric acid/sodium phosphate pH 5.5, 5 mmol/L potassium ferricyanide in 150 mmol/L NaCl₂, and 2 mmol/L MgCl₂ was applied to the tissue. Sections were incubated in a CO₂ free incubator at 37°C for 12 to 16 hours and then visualized by conventional light microscopy.

For metabolite extraction, 10 to 40 mg of anterior prostate tissue was weighed and homogenized cryogenically (Retsch Cryomill) prior to extraction in chloroform:methanol:water (4:6:3). Samples were centrifuged to separate aqueous and organic layers, and polar metabolites were dried under nitrogen gas for subsequent analysis by mass spectrometry. For LC/MS, dried metabolites were resuspended in water based on tissue weight, and valine-D8 was used as an injection control (56). LC/MS analyses were conducted on a QExactive benchtop orbitrap mass spectrometer equipped with an Ion Max source and a HESI II probe, which was coupled to a Dionex UltiMate 3000 UPLC system (Thermo Fisher Scientific). External mass calibration was performed using the standard calibration mixture every 7 days. Sample was injected onto a ZIC-pHILIC 2.1 × 150 mm (5 μm particle size) column (EMD Millipore). Buffer A was 20 mmol/L ammonium carbonate, 0.1% ammonium hydroxide; buffer B was acetonitrile. The chromatographic gradient was run at a flow rate of 0.150 mL/min as follows: 0 to 20 minutes: linear gradient from 80% to 20% B; 20 to 20.5 minutes: linear gradient from 20% to 80% B; 20.5 to 28 minutes: hold at 80% B. The mass spectrometer was operated in full-scan, polarity switching mode with the spray voltage set to 3.0 kV, the heated capillary held at 275°C, and the HESI probe held at 350°C. The sheath gas flow was set to 40 units, the auxiliary gas flow was set to 15 units, and the sweep gas flow was set to 1 unit. The MS data acquisition was performed in a range of 70 to 1,000 m/z, with the resolution set at 70,000, the AGC target at 106, and the maximum injection time at 80 milliseconds. Relative quantitation of polar metabolites was performed with XCalibur QuanBrowser 2.2 (Thermo Fisher Scientific) using a 5 ppm mass tolerance and referencing an in-house library of chemical standards.

PCA analysis was performed using MetaboAnalyst 4.0 (McGill University). The input dataset included 111 metabolites that were detectable in primary prostate tumors of all three genotypes and wild-type prostate (four groups total). The dataset was filtered for metabolites with near-constant values by IQR. Missing values (0.7%) were replaced with a value equal to half of the minimum value detected (assumed to be the detection limit) for a given metabolite. Peak areas were log transformed and centered at the mean.

Six-month-old Pten^pc–/– animals were serially imaged using MRI until tumors were estimated to be >2 mm³, and then randomized to receive either PKM2 activator (TEPP-46) or a vehicle control delivered twice daily via oral gavage at a final dose of 50 mg/kg in a volume less than 250 μL for 4 weeks. The ages of mice treated varied from 8 to 18 months with cohorts selected on the basis of tumor size. TEPP-46 was formulated in 0.5% carboxymethyl cellulose with 0.1% v/v Tween 80 as reported previously (30).

PKM1 and PKM2 expression was determined by IHC in clinically annotated human prostate cancer tissue sections collected at Dana-Farber Cancer Institute during routine clinical care. Collection of tissue was approved by the Institutional Review Board of the Dana Farber Cancer Institute (Protocol 01–045) and Partners Healthcare (IRB 2006P000139). Briefly, FFPE prostate tissue from 345 patients (including three tumor cores and two matched normal cores per patient) were arrayed on seven panels. Tissue microarray (TMA) hematoxylin and eosin sections were reviewed by a board-certified genitourinary pathologist (ML) to confirm presence of tumor and normal prostate. Corresponding unstained TMA sections were stained with antibodies that detect either PKM1- or PKM2-specific epitopes as described above. In total, 304 patients for which adequate tumor or normal tissue was present were included in the final analysis. Each core was scored for PKM1 and PKM2 and assigned a categorical variable (negative, low, intermediate, or high) based on the intensity of staining. Each individual patient was assigned the median category of the three tumor cores (expression in tumor) and higher category of the two normal cores (expression in normal prostate).

Log-rank tests were performed to determine significance in survival or tumor incidence (SPSS Statistics). Two-tailed paired and unpaired Student t test were performed for all other experiments unless otherwise specified (GraphPad PRISM 7). Results for independent experiments are presented as mean ±SEM; results for technical replicates are presented as mean ±SD.

All animal experiments were approved by the MIT Committee on Animal Care. All tissue analyzed from patients with prostate cancer was obtained via protocols approved by the Institutional Review Board of the Dana Farber Cancer Institute (Protocol 01–045) and Partners Healthcare (IRB 2006P000139), with written informed consent obtained from all patients, and the tissue collected in accordance with ethical guidelines.

All data are included in the supplemental materials, or are available from the corresponding author upon reasonable request.

PTEN loss is sufficient to promote glucose uptake (57) even though silencing of Pten alone is not sufficient to transform prostate epithelial cells (58). PTEN is frequently lost in human prostate cancer (4), so to determine whether increased glucose uptake is an early consequence of Pten loss in the prostate, mice homozygous for a conditional Pten allele (Pten^fl) were crossed to mice with a PbCre4 allele that drives prostate-specific Cre-recombinase expression, enabling the generation of animals with prostate-restricted Pten deletion (Pten^pc–/–; refs. 21, 51, 59). These mice develop prostatic intraepithelial neoplasia (PIN) at approximately 3 months of age, which progresses to invasive cancer as the mice age. To assess the effect of Pten loss on glucose uptake prior to the development of invasive cancers, we measured ¹⁸FDG uptake in the prostate and muscle of Pten^pc–/– mice at 7 to 11 weeks of age. Glucose uptake was elevated in the prostate but not the muscle of Pten^pc–/– mice (Fig. 1A), even though this time point is prior to the onset of invasive cancer (21). These findings suggest that Pten loss is sufficient to increase glucose uptake in prostate tissue, and that increased glucose uptake can occur prior the development of histologically apparent neoplasia.

Because Pkm1 expression is sufficient to suppress proliferation in some cells despite high glucose uptake (31), we questioned whether changes in pyruvate kinase isoform expression might be associated with cancer formation in the prostate. The mouse prostate has three anatomically distinct lobes (60). Tumor formation in Pten^pc–/– mice is most prominent in the anterior prostate (AP), and growth of anterior prostate tumors are the major cause of mortality in this cancer model (21, 61), therefore we focused on histopathologic characterization of tumors arising in this lobe. PKM isoform-specific antibodies were used to evaluate PKM isoform expression by IHC. Non-prostate tissues with known pyruvate kinase isoform expression were stained to confirm the specificity of antibody staining by IHC, and in young wild-type mice, PKM2 expression was found to be highest in epithelial cells throughout the prostate and seminal vesicles, whereas PKM1 expression was found to be highest in the stromal compartment, with weaker expression observed in epithelial cells (Fig. 1B; Supplementary Figs. S1A and S1B). Prostate intraepithelial neoplasia (PIN) lesions in 3-month-old Pten^pc–/– mice were characterized by upregulation of PKM1 and PKM2 expression compared with normal prostate epithelium, and tumor associated stroma also showed increased PKM2 expression (Fig. 1B and C). Tumor progression from PIN to invasive disease in older mice was associated with increased expression of PKM2 and decreased expression of PKM1, and with androgen receptor expression (Fig. 1C and D; Supplementary Figs. S1C and S1D). Furthermore, we observed increased PKM2 and decreased PKM1 expression in prostate tumor regions with increased proliferation as assessed by PCNA staining (Fig. 1E). These data suggest that in Pten-driven prostate cancer, increased PKM2 expression and loss of Pkm1 expression is correlated with increased cell proliferation and tumor progression.

To generate a conditional allele that eliminates Pkm1 isoform expression in mouse tissues, we introduced loxP sites that flank the PKM1-isoform-specific exon 9 into the Pkm genomic locus of mouse embryonic stem (ES) cells using homologous recombination (Fig. 2A). Proper targeting of ES cells was confirmed by Southern blot analysis (Supplementary Fig. S2A). Targeted ES cells were used to generate chimeric mice, which were subsequently bred to achieve germline transmission of the conditional allele and then crossed to FLP recombinase transgenic mice to delete the Neo^r gene. Expected targeting of the Pkm genomic locus was confirmed in the animals by Southern blot analysis (Fig. 2B) and by a PCR-based approach developed for genotyping (Supplementary Fig. S2B). Intercrossing Pkm1 conditional mice yielded progeny born in the expected Mendelian ratios that display no overt phenotypes.

To determine the effect of Pkm1 deletion in the prostate, we crossed Pkm1 conditional mice to animals with a PbCre4 allele to achieve animals homozygous for the Pkm1^fl allele (Pkm1^fl/fl PbCre4, hereafter Pkm1^pc–/–). Examination of Pkm1 expression in the AP from these animals showed the expected decrease in Pkm1 mRNA transcript levels (Fig. 2C), and the absence of PKM1 protein expression by Western blot analysis in all prostate lobes, whereas PKM1 protein expression is retained in other PKM1-expressing tissues (Fig. 2D). Loss of PKM1 in the prostate also resulted in increased PKM2 expression in all three prostate lobes relative to other tissues (Fig. 2D). These results confirm the conditional allele functions as designed, and demonstrates that deletion of Pkm1 results in increased Pkm2 expression in mouse prostate tissue.

To determine the effect of Pkm1 deletion on prostate tumor initiation and progression, we crossed Pkm1^fl/fl mice to Pten^pc–/– mice (hereafter, Pkm1;Pten^pc–/–). We found that the survival of Pkm1;Pten^pc–/– animals was decreased compared with Pten^pc–/– animals (Fig. 3A). Tumors were never observed in Pkm1^pc–/– mice without Pten deletion when aged to 15 months, suggesting Pkm1 loss may accelerate cancer initiated by Pten deletion. Indeed, prostate tumors in Pkm1;Pten^pc–/– animals showed loss of Pkm1 expression and strong Pkm2 expression at earlier time points than in Pten^pc–/– animals (Figs. 1C and D and 3B and C), and histopathologic analysis showed Pkm1;Pten^pc–/– animals developed invasive adenocarcinoma more frequently in younger mice than Pten^pc–/– littermates (Supplementary Fig. S3A). Pkm1;Pten^pc–/– tumors stained positive for pancytokeratin and negative for vimentin consistent with prostate epithelial origin (Supplementary Fig. S3B). Like tumors in Pten^pc–/– mice (Supplementary Fig. S1D), they also stained positive for the androgen receptor (AR; Supplementary Fig. S3C). Focal synaptophysin expression was observed in a tumor harvested from a 15-month-old Pkm1;Pten^pc–/– animal (Supplementary Fig. S3B), suggesting that at late stages these tumors may undergo neuroendocrine differentiation as has been observed in younger animals with prostate cancer driven by Rb loss or N-myc overexpression in combination with Pten loss (62, 63). However, this phenotype was not investigated further because most Pkm1;Pten^pc–/– animals did not survive beyond 1 year (Fig. 3A).

Serial MRI showed that tumors in Pkm1;Pten^pc–/– animals arose earlier and grew faster than tumors in Pten^pc–/– animals (Fig. 3D and E). Analysis of prostate tissue from 6-month-old mice, a time where high-grade PIN and adenocarcinoma are observed in Pten^pc–/– mice (Supplementary Fig. S3A; ref. 21), demonstrated that Pkm1;Pten^pc–/– animals have larger tumors than Pten^pc–/– mice, and prostate weight in Pkm1;Pten^pc–/– mice is greater than that found in both Pten^pc–/– and wild-type (WT) control mice (Fig. 3F and G). Ki-67 staining indicated increased proliferation in Pkm1;Pten^pc–/– tumors relative to Pten^pc–/– tumors and WT prostate tissue (Fig. 3H and I). We found no evidence of macroscopic metastases in animals of either genotype at either 6 or 12 months of age. These data argue Pkm1 deletion accelerates growth of Pten-loss-driven prostate tumors.

To determine whether Pkm2 is required for prostate tumor initiation and/or progression, we crossed Pkm2-conditional mice (Pkm2^fl/fl; ref. 45) to Pten^pc–/– mice (hereafter, Pkm2;Pten^pc–/–). In contrast to both Pten^pc–/– and Pkm1;Pten^pc–/– mice, most Pkm2;Pten^pc–/– mice lived a normal lifespan. We therefore assessed prostate size by serial MRI and found a striking delay in abnormal prostate growth in Pkm2;Pten^pc–/– mice compared with Pten^pc–/– littermates (Fig. 4A), although some ultimately developed PIN and/or invasive cancer (Supplementary Fig. S3A). Analysis of prostate tissue in Pkm2;Pten^pc–/– mice younger than 12 months showed nearly normal appearing prostates in many animals even though all Pten^pc–/– littermates developed PIN lesions or invasive cancer by 6 months of age (Supplementary Fig. S3A). We confirmed loss of PKM2 expression in prostate tissue from Pkm2;Pten^pc–/– mice and observed strong expression of PKM1 (Fig. 4B and C). Deletion of Pkm2 had no effect on AR expression and lesions arising in these mice stained positive for pancytokeratin and negative for vimentin and synaptophysin (Supplementary Fig. S4). Delayed prostate tumor progression observed by MRI (Fig. 4A, D, and E) was confirmed at time of necropsy, where we found that prostates from 14- to 15-month-old Pkm2;Pten^pc–/– animals were smaller and weighed less than prostates from 6-month-old Pten^pc–/– littermates (Fig. 4F and G). Furthermore, Ki-67 staining suggested decreased proliferation in 6-month-old Pkm2;Pten^pc–/– prostates compared with Pten^pc–/– animals (Fig. 4H and I). These data suggest PKM2 rather than PKM1 expression may be necessary for tumorigenesis in this tissue, and when considered together with results from Pkm1;Pten^pc–/– mice, are consistent with PKM1 expression being tumor suppressive in the prostate.

To assess whether pyruvate kinase isoform expression affects glucose uptake in the prostate, 6-month-old WT, Pten^pc–/–, Pkm1;Pten^pc–/–, and Pkm2;Pten^pc–/– mice were imaged with FDG-PET. Consistent with PKM2 deletion leading to reduced neoplasia, prostates from Pkm2;Pten^pc–/– mice exhibited glucose uptake that was similar to WT animal prostates, and lower than that observed in prostates from Pten^pc–/– and Pkm1;Pten^pc–/– mice (Fig. 5A). These data suggest that a shift from PKM2 to PKM1 expression in prostates of Pkm2;Pten^pc–/– mice suppresses the increased glucose uptake associated with prostate neoplasia in Pten^pc–/– and Pkm1;Pten^pc–/– mice.

In mouse embryonic fibroblasts, PKM1-mediated suppression of cell proliferation is due to altered metabolism that results in nucleotide depletion and impaired DNA replication (31). To determine whether changes in metabolite levels are correlated with pyruvate kinase isoform expression and/or prostate tumor growth, we examined metabolite levels in the anterior prostates from 6-month-old WT, Pten^pc–/–, Pkm1;Pten^pc–/–, and Pkm2;Pten^pc–/– mice (Supplementary Data File S1). Principle component analysis found that metabolite levels in WT prostate tissue are distinct from Pten null prostate tissue regardless of Pkm genotype, consistent with an effect of Pten loss on metabolism (Fig. 5B). Compared with Pten^pc–/– and Pkm1;Pten^pc–/– prostate tissue, which clustered together, Pkm2;Pten^pc–/– prostate tissue clustered separately consistent with prostate neoplasia being suppressed in these mice (Fig. 5B). Interestingly, the majority of metabolites that were significantly different between Pkm2;Pten^pc–/– and Pten^pc–/– prostate tissue were related to nucleotide and redox metabolism (Supplementary Table S1; Fig. 5C). Specifically, when compared with Pten^pc–/– prostate tissue, Pkm2;Pten^pc–/– prostate tissue exhibited increased levels of several nucleosides and decreased levels of two nucleotide monophosphates as well as ribose-1-phosphate. Elevated nucleosides is suggestive of impaired salvage, which when coupled with decreased levels of nucleotide precursors may be indicative of impaired nucleotide metabolism in Pkm1 expressing prostate tissue. Prostate tissue from Pkm2;Pten^pc–/– mice also showed increased levels of 3-hydroxybutyrate (β-hydroxybutyrate) and lactate, potentially suggestive of a more reduced tissue redox state that would also be predicted to impair nucleotide synthesis (10, 64). Of note, none of these metabolites were significantly different when comparing tissue from Pkm1;Pten^pc–/– and Pten^pc–/– mice (Supplementary Table S1). Taken together, these data suggest that loss of PKM2 leading to PKM1 expression in the setting of Pten loss may affect nucleotide metabolism, possibly contributing to tumor suppression.

Impaired nucleotide synthesis can lead to DNA replication stress, which initiates a signaling response that leads to CHK1 phosphorylation and can be a barrier to tumor progression (65, 66). Consistent with published reports (26), we observed increased phospho-CHK1 staining indicative of DNA replication stress in prostate tissue from 6-month-old Pten-null mice (Supplementary Fig. S5A). Interestingly, prominent phospho-CHK1 staining persists in PKM1-expressing prostate tissue from Pkm2;Pten^pc–/– mice and is lost as tumors progress in Pkm1;Pten^pc–/– mice (Fig. 5D and E; Supplementary Fig. S5B). Nucleotide depletion can also underlie oncogene-induced senescence (25), and Pten deletion in the prostate initially results in cellular senescence that is overcome with time, or by Trp53 deletion (21). To evaluate whether tumor suppression by Pkm2 loss is associated with maintenance of senescence, we examined SA-β-gal staining as a marker of senescence in prostate tissue from 6-month-old WT, Pten^pc–/–, Pkm1;Pten^pc–/–, and Pkm2;Pten^pc–/– animals. We observed increased SA-β-gal staining specifically in prostate epithelial cells in tissue from 6-month-old Pten;Pkm2^pc–/– animals as compared with tissue from age-matched WT, Pten^pc–/–, and Pkm1;Pten^pc–/– mice (Fig. 5F). Taken together these data suggest that PKM1 deletion leading to PKM2 expression allows tumors induced by Pten loss to overcome replication stress, whereas PKM2 deletion leading to PKM1 expression promotes persistent Pten loss-induced replication stress and senescence.

Because Pkm1 encodes a constitutively active enzyme, and low pyruvate kinase activity can promote tumor growth (40, 45), we hypothesized that activating PKM2, thereby forcing it into a high-activity PKM1-like state might suppress tumor growth in Pten^pc–/– mice. To test this, we randomized Pten^pc–/– mice with established tumors to treatment with vehicle or TEPP-46, an orally bioavailable PKM2 activator (30), that was administered twice a day for 1 month. Tumor size was assessed at baseline and biweekly over the course of therapy using MRI. Most tumors from vehicle-treated Pten^pc–/– mice increased in size (as defined by >50% change in tumor volume) over the 1-month period of treatment, whereas fewer tumors from TEPP-46 treated animals grew over the same time period, with radiographic evidence of tumor shrinkage in some mice (Fig. 6A and B; Supplementary Fig. S6A). The proliferation index as assessed by Ki-67 staining also decreased in tumors treated for 1 month with TEPP-46 (Fig. 6C; Supplementary Fig. S6B). These data suggest that pharmacologic activation of PKM2 can suppress prostate tumor growth and supports the notion that high pyruvate kinase activity is tumor suppressive in this tissue.

Pkm2 expression is variable in many human cancers (33, 45, 48, 50) and knowing whether this is also the case in human prostate cancer is important to consider PKM2 activators as potential therapeutics. Thus, PKM2 expression was examined using IHC in sections from patients who underwent prostatectomy and in prostate cancer specimens on a prostate tumor microarray (TMA; Fig. 7). PKM2 expression was noted in some epithelial cells in normal prostate tissue, and was retained in cancer cells in both low- and high-grade tumors, whereas PKM1 expression was restricted to the stromal regions of both normal and malignant prostate tissue. Of note, we observed that most human prostate tumors exhibit some PKM2 expression, with intermediate to high PKM2 expression found in more than half of tumors including higher Gleason grade cancers and tumors from patients with more advanced disease (Fig. 7B and C). These data suggest that unlike some other human cancers, PKM2 expression is retained even in clinically aggressive prostate cancer, arguing that PKM2 activation may be effective in treating patients with this disease.

Modulation of pyruvate kinase isoform expression has large effects on Pten loss-driven prostate tumor initiation and growth. Despite the fact that PKM2 is the predominant isoform expressed in most human and mouse cancers, Pkm2 expression is dispensable for the formation and growth of multiple other cancer types. Deletion of Pkm2 in autochthonous models of breast cancer, acute myeloid leukemia, colon cancer, medulloblastoma, pancreatic cancer, hepatocellular carcinoma, and sarcoma has minimal effect on cancer growth, and in some cases accelerates cancer progression (33, 45–50). Thus, a tumor suppressive effect of Pkm2 deletion in prostate tissue appears to be the exception among mouse cancer models examined to date.

In many mouse cancer models where Pkm2 loss does not affect tumor growth, loss of Pkm2 is accompanied by minimal to undetectable PKM1 expression (33, 45–50). In prostate tissue, high PKM1 expression accompanies Pkm2 deletion. The relative correlation between PKM1 expression and tumor suppression in various autochthonous cancer models is consistent with a tumor suppressive role for high pyruvate kinase activity. Ectopic PKM1 expression has been shown to suppress both mouse and human tumor growth in mice, even in settings where PKM2 is not deleted and there is no selective pressure to retain PKM2 (29, 30, 45). Nevertheless, some tumors can grow despite retaining some PKM1 expression (46, 49) and a protumorigenic role for PKM1 has been reported in pulmonary neuroendocrine cancers (38). However, our finding that Pkm1 deletion in Pten null prostate tissue results in aggressive cancers is strongly supportive of a tumor suppressive role for PKM1 in this organ.

The mechanism by which PKM1 suppresses prostate tumor growth is not fully understood, although the fact that treating mice with PKM2 activators can phenocopy PKM1 expression suggests that high pyruvate kinase activity associated with PKM1 is involved. One mechanism by which high pyruvate kinase activity can suppress proliferation is by affecting nucleotide synthesis (31). Nucleotide depletion can promote cellular senescence (25), and an ability to overcome senescence is a barrier to prostate cancer initiation following Pten loss (21). Nucleotide depletion can also lead to DNA replication stress (25), and DNA replication stress downstream of Pten loss can lead to senescence of prostate epithelial cells (26). The observation that shifting from PKM1 to PKM2 expression during prostate cancer development impacts nucleotide metabolism and DNA replication stress signaling suggests that a change in pyruvate kinase isoform expression to enable less enzyme activity may support nucleotide synthesis and help cancer cells to overcome Pten loss-induced DNA replication stress and senescence.

Why high pyruvate kinase activity is particularly tumor suppressive in prostate tissue is not clear; however, it could be related to the distinct metabolic phenotype of this organ. Prostate tissue synthesizes both citrate and fructose for seminal fluid (67), and either PKM1 expression or PKM2 activation can promote oxidative glucose metabolism that may promote citrate synthesis (29, 30). A role for high pyruvate kinase to support normal prostate tissue metabolism may explain in part why PKM1 is expressed in some normal prostate epithelial cells, as well as explain why loss of both PKM1 and PKM2 expression is less prevalent in prostate cancer as compared with other malignancies.

The finding that PKM2 is dispensable for tumor growth in multiple model systems, and that some cancer cells appear to proliferate despite undetectable expression of either PKM1 or PKM2 (33, 44–50), suggests that the effectiveness of small molecule pyruvate kinase activators in treating cancer will be limited by loss of pyruvate kinase expression, even though molecules that activate pyruvate kinase appear to be well tolerated in both mice and humans (30, 68, 69). However, the finding that PKM2 is retained in the majority of human prostate cancers and that either genetic or pharmacological manipulation of pyruvate kinase appears to be tumor suppressive in an autochthonous mouse model suggests that prostate cancer might be an indication where pyruvate kinase activation could be effective for therapy. Further work is needed to test this hypothesis in additional prostate cancer models, as well as uncover how pyruvate kinase activation interacts with existing prostate cancer therapies, and inform how these agents should be tested in patients with prostate cancer.

W.J. Israelsen reports grants from NIH during the conduct of the study. E. Freinkman reports personal fees from Immunai outside the submitted work and also has a patent 20200033360 pending. K.D. Courtney reports grants from Astellas and personal fees from Exelixis outside the submitted work. A. Jha reports other support from Elucidata Corporations outside the submitted work. R.A. DePinho reports grants from NIH during the conduct of the study, personal fees and other support from Tvardi Therapeutics, Asylia Therapeutics, Stellanova Therapeutic, and Nirogy Therapeutics, and other support from Sporos Bioventures outside the submitted work. C.J. Thomas reports a patent that cover the lead PKM2 activators used in this study issued. L.C. Cantley reports grants from NCI R35 CA197588 during the conduct of the study; personal fees and other support from Novartis, Volastra, Larkspur, and Faeth,; grants, personal fees, and other support from Petra; and grants from Stand up for Cancer/AACR outside the submitted work; also has a patent for WO 2008019139A3 and WO 2009025781A1 pending, licensed, and with royalties paid from Agios. M.G. Vander Heiden reports personal fees from Agios Pharmaceuticals, iTeos Therapeutics, Droia Ventures, Sage Therapeutics, Faeth Therapeutics, and Auron Therapeutics outside the submitted work and also has a patent for Use of pyruvate kinase activators to treat cancer issued and licensed to Agios Pharmaceutics. No disclosures were reported by the other authors.

S.M. Davidson: Conceptualization, formal analysis, funding acquisition, investigation, methodology, writing–original draft, writing–review and editing. D.R. Schmidt: Formal analysis, investigation, methodology, writing–review and editing. J.E. Heyman: Methodology, writing–review and editing. J.P. O'Brien: Methodology, writing–review and editing. A.C. Liu: Methodology, writing–review and editing. W.J. Israelsen: Methodology. T.L. Dayton: Methodology. R. Sehgal: Methodology. R.T. Bronson: Formal analysis. E. Freinkman: Methodology. H.H. Mak: Methodology. G.N. Fanelli: Formal analysis. S. Malstrom: Methodology. G. Bellinger: Methodology. A. Carracedo: Methodology. P. Pandolfi: Methodology. K.D. Courtney: Methodology. A. Jha: Methodology. R.A. DePinho: Funding acquisition. J.W. Horner: Methodology. C.J. Thomas: Methodology. L.C. Cantley: Resources, funding acquisition. M. Loda: Methodology. M.G. Vander Heiden: Conceptualization, resources, supervision, funding acquisition, investigation, methodology, writing–original draft, writing–review and editing.

The authors thank the Swanson Biotechnology Center for tissue processing and members of the Vander Heiden Laboratory for thoughtful discussions. They also thank John Frangioni for assistance with FDG-PET studies and B. Bevis for help generating figures. S.M. Davidson was supported by an NSF Graduate Research Fellowship and T32GM007287. E. Freinkman acknowledges support from W81XWH-15-1-0337 from the Department of Defense. D.R. Schmidt acknowledges support by the Joint Center for Radiation Therapy Foundation and the Harvard University KL2/Catalyst Medical Research Investigator Training award (TR002542). C.J. Thomas acknowledges support from the Division of Preclinical Innovation, National Center for Advancing Translational Research and the Center for Cancer Research, NCI. L.C. Cantley acknowledges support from R35CA197588. M.G. Vander Heiden acknowledges support from the Ludwig Center at MIT, the Burroughs Wellcome Fund, the Damon Runyon Cancer Research Foundation, the MIT Center for Precision Cancer Medicine, a Stand Up To Cancer Innovative Research Grant (Grant Number SU2C-AACR-IRG-09-16), the Emerald Foundation, the NIH (P30CA1405141, R35CA242379, R01CA168653, K08CA136983, P50CA090381), and a faculty scholar award from the Howard Hughes Medical Institute. Stand Up To Cancer is a division of the Entertainment Industry Foundation. The indicated SU2C grant is administered by the American Association for Cancer Research, the scientific partner of SU2C.

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.