PDK4 Protein Promotes Tumorigenesis through Activation of cAMP-response Element-binding Protein (CREB)-Ras Homolog Enriched in Brain (RHEB)-mTORC1 Signaling Cascade*♦

Background: mTOR integrates various intracellular and extracellular signals to regulate cell growth, proliferation, and survival. Results: PDK4 promotes tumorigenesis through activation of the CREB-RHEB-mTORC1 signaling cascade. Conclusion: PDK4 is a novel activator of mTOR signaling. Significance: PDK4 activation of mTORC1 may be targeted for the treatment of mTOR-mediated cancer and other diseases. Mechanistic target of rapamycin (mTOR) integrates multiple extracellular and intracellular signals to regulate cell growth and survival. Hyperactivation of mTOR has been observed in various cancers. Regulation of mTOR activity is thus of importance in physiological processes and tumor development. Here, we present pyruvate dehydrogenase kinase 4 (PDK4) as a novel regulator of mTORC1 signaling. mTORC1 activity was augmented with PDK4 overexpression and reduced by PDK4 suppression in various cell lines. Furthermore, PDK4 bound to cAMP-response element-binding protein (CREB) and prevented its degradation. The enhanced CREB consequently transactivated the expression of Ras homolog enriched in brain (RHEB), a direct key activator of mTORC1, independent of AMP-activated protein kinase or tuberous sclerosis complex protein 2. PDK4 potentiated the mTORC1 effectors hypoxia-inducible factor 1α and pyruvate kinase isozymes M2 and promoted aerobic glycolysis (Warburg effect). Knockdown of PDK4 suppressed the tumor development of cancer cells with activated mTORC1. The abundance of PDK4 dictated the responsiveness of cells to the mTOR inhibitor, rapamycin. Combinatory suppression of mTOR and PDK4 exerted synergistic inhibition on cancer cell proliferation. Therefore, PDK4 promotes tumorigenesis through activation of the CREB-RHEB-mTORC1 signaling cascade.


Mechanistic target of rapamycin (mTOR) integrates multiple extracellular and intracellular signals to regulate cell growth and survival. Hyperactivation of mTOR has been observed in various cancers. Regulation of mTOR activity is thus of importance in physiological processes and tumor development.
Here, we present pyruvate dehydrogenase kinase 4 (PDK4) as a novel regulator of mTORC1 signaling. mTORC1 activity was augmented with PDK4 overexpression and reduced by PDK4 suppression in various cell lines. Furthermore, PDK4 bound to cAMP-response element-binding protein (CREB) and prevented its degradation. The enhanced CREB consequently transactivated the expression of Ras homolog enriched in brain (RHEB), a direct key activator of mTORC1, independent of AMP-activated protein kinase or tuberous sclerosis complex protein 2. PDK4 potentiated the mTORC1 effectors hypoxiainducible factor 1␣ and pyruvate kinase isozymes M2 and promoted aerobic glycolysis (Warburg effect). Knockdown of PDK4 suppressed the tumor development of cancer cells with activated mTORC1. The abundance of PDK4 dictated the responsiveness of cells to the mTOR inhibitor, rapamycin. Combinatory suppression of mTOR and PDK4 exerted synergistic inhibition on cancer cell proliferation. Therefore, PDK4 promotes tumorigenesis through activation of the CREB-RHEB-mTORC1 signaling cascade.
The mechanistic target of rapamycin (mTOR) 2 is an evolutionarily conserved serine/threonine kinase that integrates var-ious intracellular and extracellular signals and regulates cell survival, growth, proliferation, differentiation, and metabolism (1)(2)(3)(4). The receptor tyrosine kinase/PI3K/AKT pathway relays signaling from growth factors to mTOR by first suppressing the tuberous sclerosis complex 1 and 2 (TSC1 and TSC2) protein complex. The TSC1/2 protein complex is a negative regulator of the small GTPase Ras homolog enriched in brain (RHEB) as TSC2 exerts its GTPase-activating activity toward RHEB. The unleashed RHEB then activates mTOR by binding to its catalytic domain and stimulating mTOR phosphorylation (5,6). mTOR exists in rapamycin-sensitive mTOR complex 1 (mTORC1) and rapamycin-insensitive mTOR complex 2 (mTORC2). Activated mTORC1 exercises its cellular functions by modulating downstream effectors such as S6 kinase and 4E-binding protein 1 (4E-BP1), whereas mTORC2 phosphorylates AKT at Ser-473 (4). Conversely, hyperactive mTORC1 also induces feedback inhibition of AKT through up-regulation of S6 kinase, growth factor receptor-bound protein 10 (GRB10), and store-operated Ca 2ϩ entry and down-regulation of platelet-derived growth factor receptors and insulin receptor substrate (5)(6)(7)(8). Genetic mutations in human cancers that cause gain of function for proto-oncogenes or loss of function for tumor suppressors in this pathway lead to hyperactivation of mTORC1. Mutations of tumor suppressor TSC1 or TSC2 cause TSC, a multiorgan tumor syndrome, through activation of mTOR (9 -12). mTOR is also a sensor of energy status downstream of the liver kinase B1 (LKB1)/AMP-activated protein kinase (AMPK)-TSC1/2 signaling pathway. Deletion of either LKB1 or AMPK also activates mTOR (13). An important role of mTOR has been established in the regulation of cellular pro-* This work was supported by the National Science and Technology Major cesses such as metabolism, autophagy, and protein synthesis (14 -17). Hyperactive mTOR augments angiogenesis, aerobic glycolysis (Warburg effect), and cell proliferation in tumor development (18 -20). Therefore, the regulation of mTOR activity is of significance in normal cellular processes as well as tumor development.
Pyruvate is an intermediate pivot in glucose metabolism and can be decarboxylated to acetyl-CoA by pyruvate dehydrogenase (PDH) (21). PDH links glycolysis to the citric acid cycle and releases energy via NADH. Pyruvate dehydrogenase kinases (PDKs) inactivate PDH through phosphorylation of its E1a subunit. PDKs are mitochondrial proteins encoded by genomic genes, and their expressions vary in diverse tissues, indicating that they may have different functions. PDK4 is one of four isozymes of PDKs and is highly expressed in heart and skeletal muscle (22). The role of PDK4 in tumorigenesis is unclear. As an inhibitor of PDK4 and its isoforms, dichloroacetic acid (DCA) has been used to treat some mitochondrial diseases (23). DCA has antiproliferative and pro-apoptotic effects on cancer cells due to unknown mechanisms. Furthermore, DCA can suppress hypoxia-inducible factor 1␣ (HIF-1␣) (24), a downstream effector of mTOR signaling (25).
Here, we show that PDK4 activates mTORC1 through upregulation of the CREB-RHEB cascade. Overexpression of PDK4 sensitizes cells to mTOR inhibition, and suppression of PDK4 blunts tumorigenesis. Therefore, PDK4 may be targeted for the treatment of certain cancers.
Preparation of Plasmids and Viruses-The pLL3.7-shRNA vector was used for suppression of PDK4 expression. The target sequence for human PDK4 was 5Ј-GGACGTAAGAGATTCT-CAT-3Ј. The target of mouse Pdk4 was 5Ј-GGATTTGGTGG-AGTTCCAT-3Ј. Lentiviruses were generated by co-transfecting pLL3.7 and the packaging vectors (VSVG, REV, and pMDL) into HEK293T cells. After 48 h of transfection, viruses were collected by filtering through a 0.45-m filter. PC3 cells were infected by incubation with viruses for 2 days, and the expression of GFP protein was used to evaluate the infection rate.
Cell Proliferation Assay (MTT)-Cell proliferation was measured using an MTT assay kit (BioDev-Tech, Beijing, China). Cells were plated at 4 ϫ 10 3 cells per well, seeded in quintuplet in 96-well plates for 24 h, and then treated with rapamycin at various concentrations. Cells were incubated with 200 l of medium containing 20 l of MTT reagent at 37°C. After 2 h, the supernatant was removed, and 150 l of dimethyl sulfoxide was added. The plates were shaken under protection from light for 10 min, and the spectrometric absorbance at 490 nm was then recorded. Cell proliferation assays were replicated in at least two independent experiments.
Immunoprecipitation-The co-immunoprecipitation of CREB and PDK4 was performed as described previously (29). NIH/ 3T3 or HEK293T cells were cultured in 15-cm plates until they reached 80 -90% confluence and were then lysed in 1 ml of lysis buffer. The samples were centrifuged to remove insoluble debris, and the supernatant was split into 2 equal aliquots. Antiphospho-CREB antibody and control IgG antibody were added separately to each aliquot, and samples were incubated at 4°C overnight. After incubation, 100 l of a 50% slurry of protein G-agarose beads (Millipore) was added, and samples were rotated for 2 h. Immunoprecipitates were spun down and washed three times with lysis buffer. Immunocomplexes were then subjected to immunoblotting.
Chromatin Immunoprecipitation-Chromatin immunoprecipitation was performed as described previously (19). NIH/ 3T3 cells were cultured in 15-cm plates until they reached 80 -90% confluence. Cross-linking was achieved by incubation with 1% formaldehyde for 10 min and then stopped by the addition of glycine to a 0.125 M final concentration. Chromatin was sheared by sonication to fragment sizes between ϳ500 and 1000 bp and then immunoprecipitated with either anti-phospho-CREB antibody or normal IgG antibody at 4°C overnight. Sixty microliters of salmon sperm DNA/protein G-agarose (Cell Signaling Technology) were added to immunocomplexes, and samples were incubated at 4°C for 2 h. The immunocomplexes were sequentially washed once for 10 min in ChIP low salt wash buffer and once in ChIP high salt wash buffer and twice in ChIP LiCl buffer and TE buffer. DNA-protein complexes were eluted twice with elution buffer. Finally, the released DNA was extracted using phenol/chloroform followed by ethanol precipitation. DNA was resuspended in 30 l of MilliQ water and amplified by real time-PCR. The RHEB primers were designed by the computer software program PRIMER 6 within the predicted binding region (PBR) and 2 kb upstream of the transcription start site as no-binding control (NBR). The primer sequences were as follows: PBR forward, ACCTCTCCT-TGGCTCCACCCTT, and PBR reverse, TCCCACCTACTTCC-GCCGCTTT; NBR forward, AAGCTCCTCAAGGGACAAT-GGT, and NBR reverse, TGGCTCTCTCCAATGAGCATCC.
Measurements of Glucose and Lactate-Glucose and lactate in medium were measured as described previously (19). A total of 5 ϫ 10 4 cells per well were seeded in 12-well plates (n ϭ 3) for 12 h, and then cells were incubated with fresh medium with or without rapamycin for 48 h. Cell numbers were counted before measurement. The medium was collected, and the glucose and lactate concentrations were examined immediately using a glucose and lactate measuring instrument (EKF-Diagnostic GmbH, Magdeburg, Germany). The consumption of glucose and production of lactate were normalized by cell number. The assay data were replicated in three independent experiments.
Induction of Subcutaneous Tumors in Nude Mice-Subcutaneous tumors were established as described previously (2). Immunodeficient nude mice (BALB/c, 4 -6 weeks old) were obtained from the Institute of Laboratory Animal Sciences, Chinese Academy of Medical Sciences. Eight mice were used in each cohort.
Tumor growth and mouse survival were assessed following subcutaneous inoculation of 1 ϫ 10 6 PC3 cells in 100 l of DMEM into the flank region near the axilla. Mice were sacrificed when tumor size was 1000 mm 3 . The animal study protocol was approved by the Animal Center of the Institute of Basic Medical Sciences, Chinese Academy of Medical Sciences, and was in compliance with the regulations of the Beijing Administration Office of Laboratory Animals on the care of experimental animals. The data were replicated in two independent experiments.
Histology-Nude mouse tumor tissues were fixed in 4% buffered paraformaldehyde at room temperature and embedded in paraffin. Immunohistochemistry for detection of Ki67 and CD31 was conducted using a standard protocol. Anti-Ki67 and anti-CD31 antibodies were diluted at 1:500 and incubated in a humidified chamber overnight at 4°C. The secondary antibody used was horseradish peroxidase-goat anti-rabbit IgG (Zhongshan Jinqiao, Beijing, China). Diaminobenzidine was used to develop the signal. Sections were lightly counterstained with hematoxylin.
Statistics-The Kaplan-Meier log-rank test was used for analysis of mouse tumor development with GraphPad Prism software (GraphPad Software). Comparisons between two groups of data were conducted using the Student's t test. Drug combination effect was evaluated by calculating the combination index as described previously (19). Differences were considered significant when p Ͻ 0.05.

RESULTS
PDK4 Activates mTORC1-DCA, a PDK inhibitor, was previously shown to suppress HIF-1␣ expression. Because mTORC1 is a critical regulator of HIF-1␣, we treated cells with DCA and examined mTOR activity. As a substrate of PDKs, phosphorylated PDH (p-PDH) was used to indicate the action of DCA. DCA reduced the phosphorylation of S6, 4EBP1, and GRB10, three mTORC1 effectors, in mouse myoblast and adipoblast cell lines and various human cancer cells (Fig. 1a), and the inhibition was dose-dependent (Fig. 1b). However, we did not observe any difference in p-S6 and p-Grb10 between pdh knockdown cells and control cells (Fig. 1c), suggesting that DCA suppression of mTOR activity likely does not occur through PDH. To investigate the specificity of PDK4 regulation on mTORC1, we knocked down pdk4 and the other three pdk isozymes in NIH/3T3 cells. Knockdown of pdk4, but not pdk1, pdk2, or pdk3, reduced mTORC1 activity (Fig. 1, d and e). Taken together, our data demonstrate that DCA suppresses mTORC1 specifically through Pdk4 and independent of Pdh.
To further investigate PDK4 regulation of mTOR, we overexpressed PDK4 in NIH/3T3 cells. p-S6 and p-4EBP1 were elevated, and p-Akt was decreased in response to Pdk4 expression in a dose-dependent manner (Fig. 1f). In contrast, Pdk4 overexpression had no impact on p-Erk, a marker of the MAPK signaling pathway (Fig. 1f). Similarly, overexpression of Pdk4 increased phosphorylation of S6, 4EBP1, and Grb10 but not Erk upon serum stimulation after 24 h of starvation of NIH/3T3 cells (Fig. 1g), with a decrease of p-Akt. Consistently, inhibition of Pdk4 did not affect p-Erk ( Fig. 1b). We also observed an increase of p-Akt in DCA-treated cells in a dose-dependent manner (Fig. 1b) and an increase of p-Akt in PDK4 knocked down cells (Fig. 1d), which attributed to the feedback inhibition of mTORC1.
DCA-induced Inactivation of mTORC1 Is Independent of AMPK and PTEN-AKT-TSC2-After knocking down PDK4, suppression of mTOR activity by DCA was compromised (Fig.  2a), consistent with PDK4 regulation on mTOR. Both LKB1-AMPK and PTEN/AKT/TSC1/2 pathways are major mTOR regulatory networks. To examine the potential involvement of AMPK in PDK4 regulation of mTOR, AMPK ␣1/␣2 double knock-out MEFs were used. DCA-mediated p-S6 and p-4EBP1 reductions were not affected in Ampk Ϫ/Ϫ MEFs compared with wild-type MEFs. Furthermore, DCA did not alter p-Ampk levels in wild-type MEFs (Fig. 2b). Finally, overexpression of pdk4 had no effect on p-Ampk levels in NIH/3T3 cells (Fig. 2c). Taken together, these data indicate that PDK4 is unlikely upstream of AMPK.
PDK4 Stimulates RHEB Expression through CREB-To identify the link between PDK4 and RHEB, we analyzed the promoter region of rheb for potential protein-binding sites and identified binding sites for the transcription factor CREB (Fig.  4a). CREB is a cellular transcription factor. After phosphorylation at Ser-133, CREB binds to certain DNA sequences and increases or decreases the transcription of targeted genes (32,33). Chromatin immunoprecipitation assay was performed  were subjected to immunoblotting. c, total RNA was extracted from NIH/3T3 cells transfected with PDK4 siRNA or negative control siRNA for qRT-PCR. d, protein lysates were extracted from NIH/3T3 cells transfected with PDK4 siRNA or negative control siRNA and PC3 cells with shScr or shPDK4 for immunoblotting, respectively. e, relative expression of RHEB mRNA from NIH/3T3 cells transfected with siRNA for PDK1-3 or negative control siRNA was determined by qRT-PCR (n ϭ 3). nc indicates negative control. V indicates control plasmids. *, p Ͻ 0.05. with cross-linked chromatin from NIH/3T3 cells treated with p-CREB antibody or control IgG. CREB indeed bound to the predicted CREB-binding regions of the RHEB gene (Fig. 4b). To provide more direct evidence of CREB involvement in PDK4regulated mTORC1 activity, we treated CREB-knocked down NIH/3T3 cells with DCA. DCA-induced depression of mTOR activity was abolished in rheb knockdown cells (Fig. 4c, left  panel). Similarly, Pdk4 no longer activated mTORC1 in creb knockdown cells (Fig. 4c, right panel), which indicates that CREB is involved in PDK4 regulation of mTOR. A marked reduction of Rheb was observed in CREB knockdown cells compared with control cells (Fig. 4d), implicating PDK4 in the upstream of CREB and CREB in the stimulation of RHEB expression. Furthermore, Creb downstream target genes, such as c-fos, nurr-1, sst1, were enhanced in pdk4 overexpression cells (Fig. 4e, left and middle panels) and suppressed in pdk4 FIGURE 4. PDK4 stimulates RHEB expression through stabilization of CREB. a, schematic illustration of the promoter regions of the mouse RHEB gene. E1 represents RHEB exon 1; the dark rectangle denotes a PBR, and the open rectangle indicates an NBR in the RHEB promoter. The arrow above the gene indicates the transcription start site. b, NIH/3T3 cells were subjected to ChIP assay using an anti-phospho-CREB antibody; normal rabbit IgG antibody served as the negative control. c, lysates of NIH/3T3 cells transfected with CREB siRNA or negative control siRNA treated with or without 20 mM DCA for 24 h and transfected with PDK4 or control plasmids were subjected to immunoblotting. d, lysates of NIH/3T3 cells transfected with two independent siRNAs for CREB or negative control siRNA were subjected to immunoblotting. e, total RNAs were extracted from NIH/3T3 cells transfected with PDK4 or control plasmids and NIH/3T3 cells transfected with siRNA for PDK4 or control siRNA for qRT-PCR. f, CREB antibody-immunoprecipitated DNA from PDK4-overexpressing and control NIH/3T3 cells and NIH/3T3 cells treated with 20 mM DCA for 24 h were PCR-amplified for regions indicated in a. g and h, total protein lysates and RNA were extracted from NIH/3T3 cells transfected with PDK4 or control plasmids for immunoblotting and qRT-PCR, respectively. i, immunoblotting of CREB in PDK4-overexpressing or control NIH/3T3 treated with MG-132 (10 M). j, immunoblotting of CREB in NIH/3T3 with shPDK4 or shScr treated with CHX (25 M). k, co-immunoprecipitation of endogenous proteins from NIH/3T3 cells and HEK293T cells. Immunoprecipitation (IP) was performed with anti-phospho-CREB or control IgG, and complexes were examined by immunoblotting with either phospho-CREB or PDK4 antibody. Co-immunoprecipitation and ChIP assay data were replicated in two independent experiments. Data represent mean Ϯ S.E. of replicate real time PCR. nc indicates negative control. V indicates control plasmids. *, p Ͻ 0.05. **, p Ͻ 0.01; p Ͻ 0.001. knockdown cells (Fig. 4e, right panel). The interaction between CREB protein and the RHEB promoter was significantly higher in pdk4-overexpressing cells than in control cells (Fig. 4f, left  panel) and was suppressed in NIH/3T3 cells treated with DCA (Fig. 4f, right panel), indicating that CREB is the potential bridge between PDK4 and RHEB.
Next, we examined the regulatory mechanism of PDK4 on CREB. PDK4 overexpression increased both total Creb and p-Creb protein levels in NIH/3T3 cells (Fig. 4g). However, PDK4 did not affect the mRNA levels of Creb (Fig. 4h), suggesting that PDK4 regulates CREB post-transcriptionally. To test whether PDK4 stabilizes CREB, we established pdk4 knockdown NIH/3T3 cells (Fig. 4i, left panel). The synthesis rate of Creb protein, measured by inhibiting degradation with the proteasome inhibitor MG-132, was similar between shPDK4 NIH/ 3T3 cells and shScr control cells (Fig. 4i, right panel). To test whether Creb degradation was inhibited in pdk4-overexpressing cells (Fig. 4j, left panel), we treated cells with protein biosynthesis inhibitor cycloheximide (CHX). We found apparent suppression of Creb degradation in PDK4-overexpressing cells compared with control cells (Fig. 4j, right panel). In addition, endogenous PDK4 co-immunoprecipitated CREB (Fig. 4k), indicating that PDK4 physically interacts with CREB. Taken together, we speculated that PDK4 interacts with CREB and prevents CREB degradation. Therefore, PDK4 enhances the transactivation of CREB on RHEB.
PDK4 Potentiates HIF-1␣ and Its Target Pyruvate Kinase Isozymes M2 and Promotes Aerobic Glycolysis-Among the mTOR effectors, HIF-1␣ is one of the critical regulators in tumorigenesis. Elevated/attenuated HIF-1␣ and its downstream target pyruvate kinase isozymes M2 (PKM2) were observed in PDK4overexpressing/knockdown cells (Fig. 5a). Increased glucose consumption and lactate production are manifestations of aerobic glycolysis in mTOR-hyperactivated cells (19). The potentiated Warburg effect was reversed by the mTOR inhibitor rapamycin in PDK4-overexpressing NIH/3T3 cells (Fig. 5, b  and c). Glucose consumption (Fig. 5e) and lactate production (Fig. 5f) were reduced by PDK4 depletion in three human cancer cell lines, PC3, A549, and MCF7 (Fig. 5d). To clarify PDK4induced metabolic changes whether through mTORC1 or PDH, we knocked down PDH in shPDK4 and control cells. Glucose consumption and lactate production showed a slight increase in both cells (Fig. 5g). On the contrary, after knocking down PDK4, the inhibition of the Warburg effect by mTOR inhibitor rapamycin was significantly compromised (Fig. 5g). Thus mTOR indeed plays an important role in PDK4-related metabolic changes. These data also add additional lines of evidence to the PDK4 activation of mTORC1.
Inhibition of PDK4 Suppresses Cell Proliferation and Tumorigenesis-To investigate the role of PDK4 in tumor development of mTOR-activated cancer cells, we injected PTEN-mutated PC3 cells expressing control shRNA (shScr) or shPDK4 subcutaneously into nude mice. Reduction of PDK4 in PC3 cells dramatically retarded tumor initiation and development and extended the survival of tumor-bearing mice (Fig. 6, a and b). Additionally, the attenuated PDK4 expression with the marked decrease of RHEB expression and lower mTOR activity were detected in tumor tissues dissected from nude mice (Fig. 6c). Reduced Ki67 staining and microvessel density (CD31 staining) illustrated suppressed proliferation and angiogenesis, manifestations of blunted mTOR activity, in PC3 shPDK4 tumors (Fig.  6d). In addition, PDK4 highly expressing cells displayed greater sensitivity to rapamycin treatment (Fig. 7a). Consistently, knockdown of PDK4 rendered NIH/3T3 cells less sensitive to rapamycin (Fig. 7b) Furthermore, enhanced inhibition of proliferation was observed in PC3 cells treated with DCA and rapamycin (Fig. 7c). Consistently, the catalytic mTOR inhibitor Torin-1 also enhanced the inhibition of PC3 cell proliferation by DCA (Fig. 7d). DCA induced cleavage of caspase3 (Fig. 7e), a manifestation of cellular apoptosis. The MTT readout correlated with caspase activation, indicating a connection between MTT and viability.

DISCUSSION
mTOR is a central controller of multiple cellular processes. Here, we have identified PDK4 as a positive regulator of mTORC1 and demonstrated that PDK4 activates mTORC1 through CREB-mediated transcriptional regulation of RHEB (Fig. 7f). PDK4-CREB-RHEB-mTOR signaling cascade is critical for tumorigenesis. mTOR integrates intracellular and extracellular signals and modulates cell growth and proliferation. Environmental cues that affect mTOR include growth factors, stress, energy status, and amino acids (14). By coordinating these upstream inputs, mTORC1 controls a variety of cellular processes. Gain of function mutations of proto-oncogenes or loss of function mutations of tumor suppressors can augment mTORC1 signaling and result in tumorigenesis. Therefore, the regulation on mTOR is of considerable importance both in physiological and pathological states. Guo et al. (34) observed increased PDK4 and reduced mTOR signaling in H 2 O 2 -induced senescent cells in comparison with control cells. The authors proposed that PDK4 negatively regulated mTOR. In contrast, our findings from MEFs, human cancer cells, and a xenograft tumor model demonstrated that PDK4 is an activator of mTORC1 in various tissues (Figs. 1 and 6).
Both AMPK and TSC2 are negative regulators of mTOR (13,(35)(36)(37). Our data suggest that PDK4 regulation of mTOR is independent of AMPK or TSC2 (Fig. 2). We identified RHEB, an immediate activator of mTOR, as the connector between PDK4 and mTOR. TSC1/2, Bnip3, or p38-PRAK cascade inactivates mTORC1 through suppression of RHEB (27,29,38,39). The transcriptional regulation of Rheb was discovered in ATF6-Rheb-mTOR signaling (31). We found that the tran- scriptional activity of RHEB is significantly higher in PDK4overexpressing cells than in control cells, demonstrating that PDK4 activation of mTORC1 occurs through up-regulation of RHEB transcription. We then demonstrated that CREB serves as the transcriptional activator of RHEB. Furthermore, PDK4 post-transcriptionally regulates CREB by enhancing CREB accumulation (Fig. 4). Previous studies have shown that CREB is degraded by phosphatases or the ubiquitin-proteasome pathway (40,41). Because PDK4 physically interacts with CREB, PDK4 may bind to CREB and prevent its degradation. Taken together, we speculate that PDK4 stabilizes CREB and potentiates CREB activation of RHEB transcription.
Hyperactive mTOR promotes angiogenesis, potentiates aerobic glycolysis, and accelerates tumor growth (19,20). In this study, we identified PDK4-CREB-RHEB as the activating axis upstream of mTOR. This cascade may augment mTOR downstream events in tumorigenesis. Indeed, overexpression of PDK4 increased expression of mTOR effectors HIF-1␣ and PKM2 and promoted aerobic glycolysis. This critical role of PDK4 in aerobic glycolysis is mTOR-dependent as inhibition of mTOR with rapamycin reversed the glycolytic phenotype caused by PDK4 overexpression (Fig. 5). Moreover, the attenuated PDK4 expression suppressed cell proliferation and tumor angiogenesis (Fig. 6). The Warburg effect is characterized by an increase of lactate production with relative reduction of acetyl-CoA. This effect is partially achieved through mTOR-mediated HIF-1␣ up-regulation of glycolytic enzymes and PDK. Alternatively, the putative PDK4 suppression of PDH may prevent transformation of pyruvate to acetyl-CoA and consequently promote aerobic glycolysis.
As PDK4-CREB-RHEB signaling cascade activates mTORC1 and promotes tumorigenesis, components in this cascade may be potential targets for cancer therapy. Because PDK4-overexpressed cells displayed higher sensitivity to rapamycin, the level of PDK4 may be a sensitivity biomarker of cancer cells to mTOR inhibitors. In addition, combination of mTOR inhibitors and the PDK4 inhibitor exerted synergistic suppression on the proliferation of cancer cells with activated mTOR signaling (Fig. 7). This result is a proof of principle that combination therapy may be a superior approach for the management of cancers caused by the hyperactivated mTOR signaling cascade with fewer side effects. As the pan-inhibitor of PDKs, DCA has antiproliferative and pro-apoptotic effects on cancer cells, possibly through depolarization of mitochondria and induction of apoptosis of cancer cells (24). However, its potential to be translated into clinical treatment is debatable because of its broad side effects (42,43). Our data indicate that DCA inhibits mTOR activity, and this effect is specifically through PDK4, independent of PDH and the other three PDK isoforms (Fig. 1). Our study reveals an additional mechanism of DCA inhibition of cancer. To avoid the toxicity of DCA, it will be worthwhile to screen specific inhibitors of PDK4.
In summary, we have identified PDK4 regulation of mTOR through a newly identified PDK4-CREB-RHEB-mTOR signaling cascade. Activated PDK4 may potentiate mTOR and promote cell growth and tumorigenesis. Our study uncovers a novel therapeutic target that controls mTOR signaling, and the development of PDK4 inhibitors is thus warranted. In addition, the interplay between PDK4 and mTOR may display wider significance in physiological and pathological states, such as metabolic disorders, beyond the potential connection in cancers.