Pharmacologic Activation of PKM2 Slows Lung Tumor Xenograft Growth Free

Proliferation of cancer cells requires the accumulation of biomass—sufficient biosynthetic building blocks to replicate each nucleic acid, protein, and lipid in the cell. As a tumor grows, the need for nutrients and oxygen may exceed the capacity of poor vascularity that characterizes most tumors. Faced with such challenges, cancer cells must adjust their metabolic pathways to proliferate.

Glucose provides cancer cells with the means to generate biomass through the generation of glycolytic pathway intermediates (1, 2). In cancer cells, the M2 isozyme of pyruvate kinase (PKM2) determines whether glucose is used for biomass generation or energy production (1). PKM2 is allosterically regulated by fructose-1,6-bisphosphate (FBP), an early glycolytic intermediate that binds and converts PKM2 from a less active dimeric form with low affinity for its substrate, phosphoenolpyruvate (PEP), to an active tetrameric form with high PEP affinity (3, 4). When glucose is abundant, FBP levels increase and PKM2 is activated, leading to high glycolytic flux. When glucose is limiting, FBP levels decrease and PKM2 is inactivated, allowing glycolytic intermediates to accumulate and be diverted into biosynthetic pathways (5).

PKM2 has been shown recently to have a critical function in regulating serine biosynthesis. Serine is synthesized de novo from the glycolytic intermediate 3-phosphoglycerate, and serine itself is used in the synthesis of nucleotides, proteins, lipids, and glutathione (6). When serine is absent from the growth media, PKM2-expressing cells reduce glycolytic flux (presumably through PKM2 inactivation) and accumulate glycolytic intermediates such as 3-phosphoglycerate. This allows PKM2-expressing cells to proliferate in serine-depleted media to a significantly greater degree than cells expressing PKM1 (7). Serine, like FBP, is an allosteric activator of PKM2 (8, 9). Perhaps when cellular serine levels are sufficiently high, PKM2 is converted to the active tetrameric form, restoring glycolytic flux to lactate.

PKM2 is upregulated in cancer cells (10), and has been shown to increase tumorigenicity compared with the alternatively spliced and constitutively active PKM1 isoform (11, 12). Specific RNAi knockdown of PKM2 also has been shown to regress established tumors in some xenograft models (12), while having no effect in others (13). These findings indicate that control of glycolytic flux through PKM2 activation/inactivation may be important for tumor growth in some indications. Cancer cells inactivate PKM2 through multiple mechanisms, including oncoprotein binding (14, 15), tyrosine phosphorylation (16–18), lysine acetylation (19), cysteine oxidation (20), and prolyl hydroxylation (21). In each case, decreased PKM2 activity correlates with increased tumorigenicity. Concordantly, PKM2 mutations that inhibit tetramerization also increase tumorigenicity (22). In light of such evidence, efforts have focused on the discovery and development of small-molecule PKM2 activators. There are several examples of small molecules that activate PKM2 in biochemical assays (23–25), and recent reports have shown that PKM2 activators affect cancer cell growth in vitro and tumor growth in vivo (26, 27).

Here, we describe a novel series of small molecules that activate PKM2 effectively in biochemical and cell-based assays. We find that PKM2 activators potently inhibit proliferation of cancer cells in media lacking serine, highlighting the importance of PKM2 regulation in serine biosynthesis. Finally, we show that PKM2 activators slow the growth of aggressive A549 lung adenocarcinoma xenografts. These findings support PKM2 activation as a potential cancer therapy.

Full length, human PKM2 (hPKM2) enzyme was obtained from Promab. All reagents, unless otherwise noted, were obtained from Sigma. Cell lines were obtained from American Type Culture Collection. No authentication of cell lines was conducted. Media were obtained from Invitrogen.

Compounds were preincubated with 2 nmol/L PKM2 enzyme in reaction buffer [50 mmol/L Tris-HCl, pH 8.0, 200 mmol/L KCl, 30 mmol/L MgCl₂, 2 mmol/L dithiothreitol (DTT), 5% dimethyl sulfoxide (DMSO] for 30 minutes at 25°C. ADP and PEP were then added to final concentrations of 75 μmol/L and 15 μmol/L, respectively. After 30 minutes, ATP formation was measured by Kinase Glo from Promega, and Concentration at half-maximal activation (AC₅₀) values were determined using Prism GraphPad Software.

A publicly available hPKM2 expression construct was obtained from the Structural Genomics Consortium (SGC). His6-hPKM2 was purified using NiNTA affinity capture and Hiload Superdex 16/60 S75 size exclusion chromatography. hPKM2 was crystallized using hanging drop vapor diffusion. Protein solution [20 mg/mL, 25 mmol/L Tris/HCl, pH 7.5, 0.1 mol/L KCL, 5 mmol/L MgCl2, 10% (volume for volume) glycerol] was mixed in a 1:1 ratio with reservoir solution containing 0.1 mol/L KCl, 0.1 mol/L ammonium tartrate, 25% (w/v) PEG3350. Crystals were soaked overnight in a solution containing 2 mmol/L SGI-9380, cryoprotected, and flash frozen in liquid N2. X-ray diffraction data were collected from a single crystal at 100K at Beamline-ID29 at the ESRF. Diffraction data were processed using XDS AutoPROC from Global Phasing and SCALA (CCP4; ref. 28). Molecular replacement was conducted using model 3H6O (SGC) in CSEARCH (29), and maximum likelihood refinement was carried out using a mixture of automated (30) and manual refinement protocols using Refmac (CCP4). Ligand fitting was conducted using Autosolve (29) and manual rebuilding. Each of the 4 PKM2 monomers and 4 activator ligands, comprising the tetramer in the asymmetric unit, were refined as independent entities.

A549 or NCI-H1299 lung cancer cells were plated at 20,000 cells per well (96-well plate) in minimal essential medium (MEM) media plus 10% FBS, with no additional glutamine or pyruvate. Following overnight incubation, cells were washed with PBS followed by 4-hour incubation in MEM media. Compounds were added to the cells in 1% final concentration DMSO. After 30 minutes, cells were lysed, and pyruvate kinase activity in lysates was determined by Pyruvate Kinase Activity Assay (BioVision). Maximum velocity values were calculated from the kinetic data, and AC₅₀ values were determined using Prism GraphPad Software. Compound washout experiments were as described above, except after a 30-minute incubation with SGI-9380 (469 nmol/L), the cells were washed with PBS followed by the addition of MEM media for the indicated time period, lysed, and pyruvate kinase activity determined.

HEK-293 cells were stably transfected with FLAG-PKM2 (Origene) and grown in RPMI-1640 media. When the cells were 80% confluent in a 6-well plate, the cells were washed with PBS followed by a 3-hour incubation in MEM without pyruvate. Compounds were added in 1% final concentration DMSO. Following 3 hours of incubation with compounds, the cells were washed, lysed in 1% Triton X-100 buffer, and immunoprecipitated with FLAG-M2-agarose beads (Sigma). Eluted proteins from the immunoprecipitation were separated by SDS-gel electrophoresis followed by Western blotting using anti-PKM2 antibody (Cell Signaling Technology).

Cells were seeded at 150,000 cells per well in 6-well plates, in RPMI or Basal Medium Eagle (BME) media and, after 3 days of incubation, washed and lysed in 1% Triton X-100 buffer with 1× phosphatase inhibitor cocktails. Clarified lysates were separated by SDS-gel electrophoresis followed by Western blot analysis and probed with a phospho-PKM2 (Tyr205) antibody (Cell Signaling Technology), stripped and reprobed with PKM2 antibody (Cell Signaling Technology), then stripped and reprobed with actin antibody (Cell Signaling Technology).

Cells were seeded at 5,000 cells per well (in 96-well plate) in BME media lacking nonessential amino acids + 5% dialyzed serum. After 18 hours, DMSO or compound in 0.1% final concentration DMSO was added. After 72 hours, cell viability was determined by ATPlite assay and half maximal effective concentration (EC₅₀) values were determined using Prism GraphPad Software. Serine rescue experiments were as described above, with or without 30 μmol/L serine included in the growth media.

Female athymic Nu/Nu mice were implanted with 10⁷ A549-luc-C8 cells in 1:1 volume with Matrigel into the right hind flank on day 1, followed by bioluminescent measurement and randomization for each group (n = 18 for xenograft study in Fig. 5). Mice were dosed intraperitoneally starting on day 1 with either vehicle (5% ethanol/7.5% DMSO/25% PEG400/12.5% cremophor EL/50% D5 water) or vehicle plus compound at a dose of 50 mg/kg, once daily, 5 days on, 2 days off, for 5 weeks. Tumor volumes were determined by caliper measurement starting on day 7 after implantation. Statistical significance at day 31 was determined by Student t test.

We screened a library of fragment-like small molecules (1,767 compounds), with molecular weights ranging from 150 to 300 Daltons, for PKM2 activators using a coupledbiochemical assay with a luminescent readout (23). A number of PKM2 activators were identified with AC₅₀ values ranging from 20 to 100 μmol/L, and maximal activation levels similar to FBP. We focused our medicinal chemistry efforts around XC-409, an N-benzyl-4-bromo-1H-pyrrole-2-carboxamide (Supplemental Fig. S1A), that activated PKM2 with an AC₅₀ value of 36 μmol/L (Supplementary Fig. S1B). Optimization of this scaffold yielded potent tool compounds SGI-9380 and SGI-10067 (Fig. 1A), with AC₅₀ values of 62 nmol/L and 11 nmol/L, respectively, and maximal activation levels similar to FBP (Fig. 1B and Table 1). SGI-9380 and SGI-10067, like FBP, increased the affinity of PKM2 for PEP (Fig. 1C), but not ADP (Fig. 1D).

A cocrystal structure of SGI-9380 bound to tetrameric PKM2 at 2.03 Å resolution was obtained (Fig. 2). SGI-9380 did not bind to the FBP site, but instead bound to an allosteric site of unknown physiologic significance at the dimer–dimer interface. SGI-9380 bound to the allosteric site with occupancy of 2 molecules per dimer (Fig. 2), whereas other PKM2 activators that bind to the same site are reported to bind with occupancy of one molecule per dimer (27). We were unable to obtain a crystal structure of SGI-10067 bound to PKM2. However, based on structural similarities between SGI-9380 and SGI-10067, we reason that the interactions of SGI-10067 with PKM2 are similar to those observed with SGI-9380.

Having shown that the PKM2 activators are biochemically active and bind directly to PKM2, we next tested the ability of SGI-9380 and SGI-10067 to stimulate pyruvate kinase activity in cells. The lung adenocarcinoma cell line A549 was selected because of previously reported high levels of phospho-PKM2 at Tyr105 (16), independently confirmed in our experiments (Supplementary Fig. S2). Phosphorylation of PKM2 at Tyr105 renders the enzyme insensitive to FBP activation (16). However, we found that SGI-9380 and SGI-10067 potently activated pyruvate kinase activity in A549 cells (Fig. 3A and Table 2). These results showed that SGI-9380 and SGI-10067 effectively permeate the cell membrane to activate PKM2, despite high levels of p-Tyr105, and suggest that the allosteric site bound by SGI-9380/SGI-10067 is not greatly affected by PKM2 phosphorylation at Tyr105.

On the basis of the tetrameric form of PKM2 in the SGI-9380 cocrystal, and the requirement of PKM2 tetramer formation for full activation (31), we reasoned that the PKM2 activators induce tetramer formation in cells. To investigate this possibility, we used a previously described (20) indirect assay for PKM2 subunit association. Briefly, HEK-293 cells expressing both endogenous PKM2 and exogenous FLAG-PKM2 were treated with PKM2 activators, followed by FLAG immunoprecipitation of cell lysates. We found that PKM2 activators increased the amount of endogenous PKM2 that coimmunoprecipitated with FLAG-PKM2 (Fig. 3B). This result suggested an increase in cellular association between FLAG-PKM2 and endogenous PKM2 consistent with tetramer formation (although an increase in dimer complexes cannot be ruled out). SGI-10067 was more effective than SGI-9380 in stimulating coimmunoprecipitation of endogenous PKM2, possibly reflecting its increased affinity for PKM2 (Tables 1 and 2). PKM2 activation in cells was relatively stable, exhibited by the slow kinetics of PKM2 inactivation following compound washout (Fig. 3C).

Having shown that the PKM2 activators were able to stimulate PKM2 activity in cells, we next addressed their effects on cancer cell proliferation. Previous reports showed that other classes of small-molecule PKM2 activators have little or no effect on the viability of cancer cells when grown under normal conditions (26, 27). We observed similar results with SGI-9380 and SGI-10067 against a number of adherent and suspension cell lines (data not shown). However, consistent with the report by Kung and colleagues (26), we found that SGI-9380 and SGI-10067 potently inhibited growth of A549 cells (EC₅₀ values of 210 nmol/L and 89 nmol/L, respectively) when grown in BME media that lack nonessential amino acids (Fig. 4A). The effects of SGI-9380 and SGI-10067 on viability in BME media were rescued by the addition of serine (30 μmol/L) to the media (Fig. 4B), similar to the observation of Kung and colleagues (26), indicating that PKM2 activation interferes with de novo serine biosynthesis. These results are consistent with additional reports that PKM2 inhibition promotes de novo serine synthesis in cancer cells (7, 13), and that PKM2 is allosterically activated by serine (9).

To determine whether sensitivity to PKM2 activators in the absence of serine is a general phenomenon, we evaluated the effects of SGI-10067 on proliferation against 7 additional lung cancer cell lines when grown in BME media lacking nonessential amino acids. Two of the cell lines tested (NCI-H23 and NCI-H522) were unable to proliferate in BME media, even upon addition of 100 μmol/L serine (Supplementary Fig. S3). Proliferation of NCI-H1299 cells was affected only modestly by the switch to BME media (Supplementary Fig. S3), suggesting that this cell line may have robust de novo serine biosynthesis. The remaining 4 cell lines were similar to A549 in which proliferation was slowed significantly in BME media, and largely rescued by the addition of serine (Supplementary Fig. S3). Of the cell lines that proliferated (slowly or well) in BME media, half were sensitive to SGI-10067 (IC₅₀ values 0.1–1 μmol/L) and half were relatively insensitive when tested with up to 30 μmol/L SGI-10067 (IC₅₀ values 10 μmol/L or greater; Fig. 4C). One of the SGI-10067–sensitive cell lines, NCI-H1299, has been studied in various genetic models in which pyruvate kinase activation inhibited tumor growth (7, 11, 16), and was recently shown to be sensitive to other classes of PKM2 activators in vitro and in vivo (26, 27).

Cell lines from additional indications other than lung cancer (including breast, colon, pancreas, and prostate) were also examined and found to have varying sensitivities to both serine deprivation (Supplementary Fig. S4A) and SGI-10067 treatment (Supplementary Fig. S4B). We hypothesized that cell line sensitivity to SGI-10067 in the absence of serine might correlate with PKM2 phosphorylation at Tyr105, either in normal or serine-free media. However, interrogation of PKM2 phosphorylation status in a subset of cell lines (Supplementary Fig. S2) showed that PKM2 phosphorylation did not predict sensitivity to SGI-10067 in media lacking serine. As an example, both A549 and DU-145 cells showed high levels of phosphorylated PKM2 (Supplementary Fig. S2), as previously reported (16). However, unlike A549 cells, DU-145 cells were insensitive to SGI-10067 when grown in BME media (IC₅₀ > 30 μmol/L; Supplementary Fig. S4B). Conversely, HCT-116 cells displayed only modest levels of phosphorylated PKM2 (Supplementary Fig. S2), yet were sensitive to SGI-10067 in BME media (IC₅₀ = 110 nmol/L; Supplementary Fig. S4B). The phosphorylation status of PKM2 was similar for cells grown in normal media and media lacking serine (Supplementary Fig. S2), indicating that phosphorylation of PKM2 at Tyr105 is not stimulated by serine deprivation.

Given recent studies showing the importance of de novo serine biosynthesis in breast cancer (32) and melanoma (33), we were interested in determining the effects of PKM2 activators on cell lines with PHDGDH gene amplification. PHGDH encodes 3-phosphoglycerate dehydrogenase, the enzyme responsible for the first step in de novo serine biosynthesis. PHGDH amplification has been shown to be important for proliferation of breast cancer (32) and melanoma (33) cells. We found that MDA-MB-468 and Malme-3M, PHGDH-amplified cell lines from breast cancer (32) and melanoma (33), respectively, were able to proliferate well in BME media lacking serine (Supplementary Fig. S4A), and were insensitive to SGI-10067 (Supplementary Fig. S4B). These results suggest that the ability of cancer cells with PHGDH amplification to divert glycolytic intermediates into the de novo serine biosynthetic pathway may be independent of the activation state of PKM2.

Overall, our results suggested that SGI-9380 and SGI-10067 potently activated PKM2 in vitro. To investigate whether the PKM2 activators would show efficacy in vivo, we used a subcutaneous A549 xenograft tumor prevention model. Compounds were administered intraperitoneally, beginning before the presence of established tumors, at a daily dose of 50 mg/kg. We observed no significant body weight loss or other gross toxicity in the cohorts receiving SGI-9380 or SGI-10067, suggesting few on- or off-target toxic effects. Both SGI-9380 and SGI-10067 significantly slowed tumor growth (Fig. 5). These results are comparable with a recent report showing inhibition of tumor growth in a NCI-H1299 lung adenocarcinoma xenograft model by a different chemical class of PKM2 activator (27). Our findings are also consistent with reports in PKM2 genetic models that correlate PKM2 activation with decreased tumorigenicity (11, 16). Our results provide additional, independent evidence that pharmacologic activation of PKM2 slows tumor growth in vivo.

Targeting tumor-specific metabolic pathways is a promising anticancer strategy (34–37). In this report, we show that activation of PKM2 with small molecules inhibited cancer cell growth in vitro and slowed tumor formation in vivo. These promising results indicate that PKM2 activation may be a useful anticancer strategy in the clinic.

PKM2 activation slowed the growth of subcutaneous A549 lung adenocarcinoma xenografts. This result is significant as it confirms that pharmacologic activation of PKM2 affects tumor growth in vivo (36). The effects we observed on tumor growth were statistically significant, yet modest in impact. Potentially, more potent PKM2 activators or more optimal dosing schedules should result in greater antitumor effects. Alternatively, the tumor growth inhibition we observed could be the maximal effect obtainable for PKM2 activation in a subcutaneous A549 xenograft, whereas orthotopic implantation of this model may be more sensitive, or other xenograft models including patient-derived primary explants may be more sensitive to PKM2 activation. Significant inhibition of an NCI-H1299 lung cancer xenograft was shown with similarly potent PKM2 activators (27). Future investigation is needed to determine whether the in vitro effects of PKM2 activation, which seem to be directly related to serine biosynthesis (9, 26, 27), are correlative with activity in vivo. The reduction in tumor growth in the A549 xenografts may have been due to defects in serine biosynthesis or, alternatively, due to impairment of other functions of PKM2, such as the nonmetabolic role of dimeric PKM2 in nuclear signaling (21, 38, 39).

Preclinical studies will benefit from increased focus on combinations of PKM2 activators with other anticancer agents. PKM2 activation theoretically should not be toxic to normal cells, and we observed no toxicity with PKM2 activators at high doses in vivo, potentially enabling combination studies that may yield regimens with high therapeutic indices.

S.B. Kanner has a honoraria from speakers' bureau from the University of Washington. No potential conflicts of interest were disclosed by the other authors.

Conception and design: K.M Parnell, J.M Foulks, R.N. Nix, B. Luo, M. Saunders, X.-H. Liu, M.V. McCullar, K.-K. Ho, S.B. Kanner

Development of methodology: K.M Parnell, J.M Foulks, R.N. Nix, B. Luo, A. Senina, D. Vollmer, J. Liu, V. McCarthy, Y. Xu, M. Saunders, X.-H. Liu, S. Pearce, K.-K. Ho

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K.M Parnell, A. Clifford, J. Bullough, A. Senina, D. Vollmer, J. Liu, V. McCarthy, M. O'Reilly, K.-K. Ho

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K.M Parnell, J.M Foulks, R.N. Nix, A. Clifford, J. Bullough, B. Luo, A. Senina, D. Vollmer, J. Liu, V. McCarthy, M. Saunders, X.-H. Liu, M. O'Reilly, K.-K. Ho, S.B. Kanner

Writing, review, and/or revision of the manuscript: K.M Parnell, J.M Foulks, B. Luo, A. Senina, D. Vollmer, Y. Xu, X.-H. Liu, M.V. McCullar, K.-K. Ho, S.B. Kanner

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Clifford, J. Bullough, A. Senina, D. Vollmer, K. Wright, M.V. McCullar, K.-K. Ho, S.B. Kanner

Study supervision: J.M Foulks, B. Luo, A. Senina, D. Vollmer, M.V. McCullar, S.B. Kanner

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.