Metastatic Colorectal Cancer Rewrites Metabolic Program Through a Glut3-YAP-dependent Signaling Circuit.

Rationale: Cancer cells reprogram cellular metabolism to fulfill their needs for rapid growth and metastasis. However, the mechanism controlling this reprogramming is poorly understood. We searched for upregulated signaling in metastatic colorectal cancer and investigated the mechanism by which Glut3 promotes tumor metastasis.
Methods: We compared RNA levels and glycolytic capacity in primary and metastatic colon cancer. The expression and association of Glut3 with clinical prognosis in colon cancer tissues was determined by immunohistochemistry. Glut3 gain-of-function and loss-of-function were established using colon cancer HCT116, HT29, and metastatic 116-LM cells, and tumor invasiveness and stemness properties were evaluated. Metabolomic profiles were analyzed by GC/MS and CE-TOF/MS. The metastatic burden in mice fed a high-fat sucrose diet was assessed by intravenous inoculation with Glut3 knockdown 116-LM cells.
Results: Upregulation of glycolytic genes and glycolytic capacity was detected in metastatic colorectal cancer cells. Specifically, Glut3 overexpression was associated with metastasis and poor survival in colorectal cancer patients. Mechanistically, Glut3 promoted invasiveness and stemness in a Yes-associated protein (YAP)-dependent manner. Activation of YAP in turn transactivated Glut3 and regulated a group of glycolytic genes. Interestingly, the expression and phosphorylation of PKM2 were concomitantly upregulated in metastatic colorectal cancer, and it was found to interact with YAP and enhance the expression of Glut3. Importantly, a high-fat high-sucrose diet promoted tumor metastasis, whereas the inhibition of either Glut3 or YAP effectively reduced the metastatic burden.
Conclusion: Activation of the Glut3-YAP signaling pathway acts as a master activator to reprogram cancer metabolism and thereby promotes metastasis. Our findings reveal the importance of metabolic reprogramming in supporting cancer metastasis as well as possible therapeutic targets.

Introduction

Colorectal cancer (CRC) is the third most common cancer and the second most common fatal cancer worldwide. Approximately 25% of patients experience systemic metastases, and the most frequently metastatic sites are the liver and lung 1. Despite major research efforts and clinical progress, predictable biomarkers for colorectal cancer metastasis and therapeutics are not yet fully understood. Recently, alterations in cellular metabolism have been recognized as emerging hallmarks of cancer, and the most appreciated hallmark is aerobic glycolysis, also known as the Warburg effect 2. It is believed that cancer cells increase glucose consumption to meet the high energy and macromolecule demands required for rapid growth 3, and several studies have reported that oncogenic signals drive the malignant progression of cancer cells by modulating metabolic enzymes. For instance, the oncogene c-Myc transactivates most glycolytic genes, including glucose transporter 1 (Glut1) and lactate dehydrogenase A (LDHA) 4, 5. Hypoxia-inducing factor 1 (HIF-1) regulates LDHA and pyruvate dehydrogenase kinase isozyme 1 (PDK1) 6. Other recent evidence has shown that mitogen signaling promotes tumor growth by modulating pyruvate kinase M2 (PKM2) 7. These evidence indicate that the regulatory mechanisms underlying cellular metabolism in cancer cells are far more complex than previously believed and warrant investigation.

The glucose transporter (encoded by SLC2A) family comprises enzymes involved in the first step of glycolysis and facilitates glucose entry into cells. The Glut family includes 14 members that transport glucose or other substances in different tissues 8. Glut1 is ubiquitously expressed in normal tissues and is overexpressed in many cancer types. In contrast to Glut1, Glut3 is expressed in neuronal cells but is less expressed in other organs under normal conditions. Interestingly, recent studies have reported the elevated expression of Glut3 in glioma and hepatoblastoma 9. The induction of Glut3 facilitates metabolic adaptation to nutrient deprivation in brain tumor-initiating cells and is responsible for TGFβ-induced EMT in nonsmall cell lung cancer 10, 11. However, the regulation of Glut3 in CRC metastasis and the contribution of Glut3 to this process are largely unknown.

In an attempt to identify biomarkers for metastatic CRC, we investigated transcriptome in primary and metastatic CRC tissues. Interestingly, we found that Glut3 expression and glycolytic capacity were concomitantly increased in metastatic CRC. We further demonstrated that Glut3 operates through a positive regulatory loop by collaborating with the Hippo cascade transducer YAP. We show that Glut3 is responsible for the activation of YAP, which in turn transactivates glycolytic genes, including Glut3 itself. Moreover, oncogenic signal promotes the association of YAP and PKM2, thereby enhancing Glut3 expression and promoting tumor invasiveness. Blockade of Glut3 and YAP effectively represses CRC aggressiveness and metastasis.

Results

Overexpression of Glut3 is an independent marker for metastatic CRC

To identify genes associated with CRC metastasis, we analyzed the RNA level in 7 pairs of primary and liver metastatic CRC tissues. Interestingly, gene ontology annotation revealed that cellular metabolic processes were altered. In particular, we found that the expression of Glut3 was upregulated in metastatic tumor tissues (Figure ). Transcriptome sequencing data from TCGA revealed that Glut3 overexpression was closely associated with the aggressive clinicopathological features of CRC (Figure ). Accordingly, CRC patients with high Glut3 expression were associated with low overall and recurrence-free survival rates (Figure ). Immunohistochemistry analysis results confirmed that Glut3 protein was increased in metastatic CRC tissues compared with primary tumors in both paired and unpaired samples (Figure ). Intriguingly, data retrieved from TCGA revealed that the correlation between Glut3 and poor survival was found more often in colon and lung squamous cancers than in other cancer types (Figure ). Similar results from other independent cohorts also showed that Glut3 overexpression correlated with poor outcomes, particularly in colon and lung cancers (Figure ).

To better understand the involvement of the Glut family in CRC, we analyzed the expression of Glut isoforms 1-4. The results showed that the expression of Glut1 and Glut3, but not that of Glut2 and Glut4, was upregulated in tumors compared with normal tissues (Figure ). However, expression of Glut1 was associated only with tumor stage (Figure ). Overall and recurrence-free survival was not correlated with Glut1 expression (Figure ). These data indicate that Glut3 expression is an indispensable marker for CRC.

Aerobic glycolysis is upregulated in metastatic CRC

To gain insights into the role of cellular metabolism in CRC metastasis, we examined the bioenergetic profile in colorectal cancer HCT116 cells and their metastatic derivatives, 116-LM cells, which were obtained from pulmonary metastasis of HCT116 cells 12. The increased migration and invasion in 116-LM cells was confirmed by Transwell assay (Figure ). Importantly, 116-LM cells exhibited an increased glycolytic rate and glycolytic capacity (Figure ). Quantitative PCR analysis revealed that the expression of glycolysis-related genes was increased in 116-LM cells (Figure ). Notably, the expression of Glut3 rather than that of other isoforms was drastically elevated in 116-LM cells compared with the parental HCT116 cells (Figure ). Western blot results confirmed that Glut3 protein levels were upregulated in 116-LM cells (Figure ). In contrast, the protein level of Glut1 was similar to that of Glut3 to some extent in colon cancer cell lines, but the Glut1 protein levels between HCT116 and 116-LM cells were not changed (Figure ).

To further examine cellular metabolites in 116-LM cells, GC/MS was performed, and 43 metabolites were identified. The results showed that metastatic 116-LM cells increased glycolysis products, including glucose-6-phosphate, 3-phosphoglycerate, and lactate (Figure ), which is correlated with the upregulation of glycolytic capacity indicated by the ECAR (Figure ). Interestingly, the levels of amino acids and fatty acids were also upregulated in 116-LM cells, suggesting a demand for protein and macromolecule syntheses in metastatic cells (Figure ). To evaluate the contribution of upregulated glycolysis in 116-LM cells, cells were treated with high or low concentrations of glucose. Quantitative PCR analysis showed that high glucose promoted the expression of EMT and glycolytic genes (Figure ). In contrast, the inhibition of glycolysis by 2-deoxyglucose (2DG) suppressed EMT genes (Figure ). Moreover, the growth of 116-LM cells was more vulnerable to glucose restriction than the growth of HCT116 cells (Figure ). Low glucose treatment decreased the growth of 116-LM and HCT116 cells by 53% and 31% from their levels in high glucose, respectively (P<0.01). Together, these data suggest that metastatic CRC requires greater glucose utilization.

Glut3 promotes CRC aggressiveness and stemness through a YAP-dependent pathway

To determine the contribution of Glut3 to CRC metastasis, we generated Glut3-overexpressing HCT116 and Glut3-silencing 116-LM cells (Figure ). Glut3 overexpression increased the transcription of EMT genes and promoted cell migration and invasion (Figure ). Additionally, Glut3 overexpression increased the transcription of stemness-related genes (Figure ) and the formation of tumorspheres (Figure ), properties associated with cancer stem cells. However, this phenomenon was conversely reduced in Glut3 knockdown cells (Figure ). Knockdown of Glut3 also reduced the promoter activity of stemness-related transcription factors (Figure ). Moreover, knockout of Glut3 using the CRISPR/Cas9 systems significantly reduced cell migration and invasion as well as tumorsphere formation (Figure ). These data indicate that Glut3 is crucial for the invasiveness and stemness properties of metastatic CRC.

To study the mechanism underlying the promotion of CRC invasiveness and stemness by Glut3, we examined the Hippo pathway, as it has been reported to be regulated by cellular energy status and has drawn increasing attention for its role in tumor promotion. Moreover, we previously identified that the expression of YAP is upregulated and plays a role in the invasive properties of 116-LM cells 12. Thus, we investigated the relationship between Glut3 and Hippo signaling. Consistently, 116-LM cells exhibited increased levels of YAP compared with the parental HCT116 cells (Figure ), and treatment with glucose reduced YAP phosphorylation and increased protein level of YAP and its paralog TAZ (Figure ). In contrast, blocking ATP production by the addition of 2DG induced YAP phosphorylation, suppressed YAP/TAZ level (Figure ) and subsequently inhibited the expression of YAP downstream genes (Figure ). Moreover, low glucose concentrations (5 mM) robustly increased YAP protein levels in 116-LM cells, while YAP was equivalently induced by high glucose concentrations (25 mM) in the parental cells (Figure ). We hypothesized that metastatic CRC are more susceptible to glucose due to elevated levels of Glut3. Indeed, the release of glucose induced YAP upregulation in control knockdown 116-LM cells, while Glut3 silencing remarkably diminished YAP in either glucose starvation or glucose-releasing conditions (Figure ). Moreover, Glut3 silencing remarkably suppressed the invasiveness of 116-LM cells compared with HCT116 cells (Figure ), indicating that metastatic CRC is addicted to Glut3, possibly through the activation of YAP signaling.

To further determine whether Glut3 is responsible for YAP activation and therefore promotes CRC metastasis, we examined the expression of YAP target genes. Quantitative PCR assays revealed that the YAP target genes were upregulated in Glut3-overexpressing HCT116 cells (Figure ), while they were downregulated in Glut3-silenced 116-LM cells (Figure ). YAP facilitates the regulation of downstream gene expression by binding with TEAD transcription factors. We further found that Glut3 silencing repressed the activity of a TEAD-responsive luciferase reporter (Figure ). Additionally, the increased capacities for tumor migration/invasion and tumorsphere formation as a result of Glut3 overexpression were suppressed by YAP silencing (Figure ). Gene set enrichment analysis (GSEA) showed that expression of Glut3 was significantly associated with the YAP signature in CRC patients (Figure ). These findings indicate that the effects of Glut3 on CRC aggressiveness and stemness are dependent on YAP.

Glut3-YAP signaling loops are crucial for CRC malignancy

Our aforementioned data show that Glut3 regulates YAP signaling. Interestingly, a recent study reported that inhibition of YAP repressed Glut3 expression in brain tumors 13, suggesting that Glut3 might be a YAP downstream target. However, the detailed underlying mechanism is not yet understood. Thus, we further examined the relationship between Glut3 and YAP. Quantitative PCR assays showed that YAP silencing downregulated Glut3 mRNA levels in both 116-LM and HT29 cells (Figure ). We also found that knockdown of YAP decreased the expression of glycolysis genes (Figure ). Conversely, overexpression of YAP5SA, a constitutively active YAP mutant, induced the mRNA expression of Glut3 and glycolysis genes in HCT116 cells (Figure ). Accordingly, addition of a TEAD inhibitor, verteporfin, suppressed Glut3 production (Figure ). Moreover, ChIP assay revealed YAP binding in the -1164/-1155 region of the Glut3 promoter (Figure ), and YAP expression induced Glut3 promoter activity (Figure ). However, the deletion of the -1164/-1155 region diminished YAP-elicited transactivation of Glut3 (Figure ). Similarly, inhibition of YAP downregulated the promoter activity of Glut3 (Figure ), whereas overexpression of YAP5SA restored the production of Glut3 (Figure ). Furthermore, the reduced migration/invasion and tumorsphere formation in Glut3-silenced 116-LM cells were rescued upon reintroduction of YAP (Figure ). TCGA data further confirmed the positive correlation between Glut3 and YAP/TAZ mRNA in CRC patients (Figure ) and showed that patients who coexpressed Glut3 with YAP/TAZ had the worst survival probability (Figure ).

Because the inhibition of ATP production by 2DG suppressed YAP activity (Figure ), we wanted to determine whether inhibiting glycolysis impaired Glut3 expression. The results showed that inhibition of YAP by 2DG was accompanied by a reduction in Glut3 at both the mRNA and protein levels (Figure ). However, the inhibition of Glut3 by 2DG was rescued by YAP (Figure ). Altogether, these data support the observation of concomitant elevations in YAP and Glut3 in metastatic 116-LM cells (Figure ) and show that the reciprocal activation between Glut3 and YAP signaling acts as a feedback loop that fosters glucose-mediated CRC metastasis.

The association of YAP with PKM2 enhances Glut3 expression

Increasing evidence has shown that PKM2 acts as a key molecule in linking the reprogramming of cancer cell metabolism and oncogene-mediated tumorigenesis. Nonphosphorylated PKM2 exhibits pyruvate kinase activity, whereas phosphorylated PKM2 translocates to the nucleus and enhances the expression of oncogenic genes. Interestingly, it has been reported that the phosphorylation of PKM2 facilitates nuclear translocation through interaction with the WW domain of PIN1 protein 7. Because YAP possesses a WW domain, we therefore investigated the relationship between YAP and PKM2. We found that the total and phosphorylated PKM2 were increased in metastatic 116-LM cells compared with HCT116 cells (Figure ). Moreover, phosphorylation of PKM2 was increased in the nuclear fraction of 116-LM cells (Figure ); thus, we explored whether increased Glut3 and/or YAP has an effect on PKM2 in metastatic CRC. The results showed that Glut3 silencing reduced total PKM2, as well as PKM2 phosphorylation to some extent, compared with the YAP knockdown cells, which robustly suppressed the expression and phosphorylation of PKM2 (Figure ). Moreover, YAP suppression decreased the presence of nuclear PKM2, whereas the ectopic overexpression of YAP5SA induced the nuclear translocation of PKM2 (Figure ). We also found that suppression of YAP decreased PKM2 protein stability (Figure ).

To further study the mechanisms by which YAP regulates PKM2 in colorectal cancer, we examined the possibility that YAP interacts with PKM2. Interestingly, coimmunoprecipitation assays showed an interaction between endogenous YAP and PKM2 in 116-LM cells (Figure ). These findings were confirmed by ectopic overexpression of YAP and PKM2 (Figure ). Next, because the receptor tyrosine kinase signal regulates the phosphorylation of PKM2, we wanted to examine whether PKM2 phosphorylation promotes an interaction with YAP. 116-LM cells were treated with gefitinib, an epidermal growth factor receptor (EGFR) tyrosine kinase inhibitor that blocks PKM2 phosphorylation 14. The results showed that although total PKM2 was unaffected by gefitinib treatment, the levels of YAP and phosphor-PKM2 simultaneously decreased (Figure ). Importantly, the interaction of YAP and PKM2 was impeded in the presence of gefitinib (Figure ), suggesting that the dysregulation of mitogen signaling in metastatic colorectal cancer induces PKM2 phosphorylation and consequently facilitates its association with YAP.

To further explore the molecular basis of the interaction between YAP and PKM2, we constructed a series of deletion mutants of PKM2 and YAP. The results of the coimmunoprecipitation assay revealed that the Δ1 mutant of PKM2, which lacked the B, A2, and C domains, did not bind YAP. However, the Δ3 mutant of PKM2, lacking the C domain, did bind to YAP, suggesting that the B and A2 domains of PKM2 are responsible for the association with YAP (Figure ). Moreover, we found that deletion of the TEAD binding domain in YAP (YAP-M2) abrogated the binding with PKM2 (Figure ).

To better understand whether PKM2 participates in the regulation of Glut3, a luciferase reporter assay was employed, and the results showed that overexpression of PKM2 and YAP5SA increased the promoter activity of Glut3 by 1.7- and 8-fold, respectively. Importantly, cotransfection with YAP5SA and PKM2 significantly increased the promoter activity of Glut3 compared with that of cells transfected with YAP or PKM2 alone (Figure ). Similar results were observed with Glut3 protein (Figure ). Moreover, knockdown of PKM2 not only suppressed Glut3 but also impeded YAP (Figure ), and compared with shYAP or shPKM2 treatments, the inhibition of both YAP and PKM2 substantially repressed Glut3 expression and tumor migration/invasion (Figure ). These data together indicate that oncogenic signaling facilitates the association of PKM2 with YAP and consequently promotes Glut3 expression and the invasiveness of CRC.

YAP plays a crucial role in metabolic changes in metastatic CRC

To perceive the impacts of YAP on metabolic changes in metastatic CRC, quantitative analysis of cellular metabolites in HCT116, 116-LM, and 116-LM/shYAP cells was performed using CE-TOF/MS. A total of 101 metabolites were identified (Figure ). We evaluated metabolic parameters using absolutely quantified data obtained from CE-TOF/MS to investigate the physiological states of the cells (Table ). First, we found that the glucose 6-phosphate (G6P) to ribose 5-phosphate (R5P) ratio, a parameter that shows the significance of the activity of glycolysis and the pentose phosphate pathway, was increased over 2-fold in 116-LM cells compared with HCT116 cells (Figure ). The pentose phosphate pathway is a major source of purine metabolism. Our data showed that the levels of total adenylate and guanylate, which are related to de novo purine synthesis, were increased in 116-LM cells (Figure ). Likewise, 116-LM cells showed a higher glycerol 3-phosphate/dihydroxyacetone phosphate (DHAP) ratio, which is an indirect parameter reflecting the NADH/NAD+ ratio and energy status (Figure ). Indeed, ATP production was also higher in 116-LM cells (Figure ). In addition, amino acids, both essential and nonessential, were present at higher levels in 116-LM cells (Figure ). Specifically, the glucogenic amino acids, which are used in gluconeogenesis, were increased approximately 2-fold in 116-LM cells (Figure ). These data revealed that the synthesis of macromolecules was elevated in 116-LM cells and highlighted the importance of glucose utilization and energy demand in CRC metastasis.

Moreover, we found that reduced glutathione (GSH) was upregulated in 116-LM cells (Figure ). Glutathione removes and detoxifies carcinogens, which confers resistance to chemotherapeutic drugs and protects cancer cells. In contrast, the GSH to oxidized glutathione (GSSG) ratio, which reflects oxidative stress, was increased in YAP-knockdown 116-LM cells (Figure ). These data also revealed the indispensable role of YAP in drug resistance. However, these metabolic alterations were decreased in YAP-knockdown 116-LM cells (Figure ). Notably, the adenylate and guanylate energy charges were not affected in the three groups (Figure ). Energy charge is a cellular metabolic indicator that decreases under severe stress conditions. Together with the aforementioned metabolic parameters, these data demonstrated that YAP plays a crucial role in shaping metabolic reprogramming in CRC metastasis.

The Glut3/YAP axis is indispensable for high-fat/sucrose diet-mediated CRC metastasis and is a potential therapeutic target

To gain insights into the significance of high glucose-elicited tumor progression, we fed immunocompromised mice a diet containing 45% fat and 17% sucrose (HFSD), which reflects Western food-mediated metabolic changes and obesity in humans. Mice fed a HFSD showed an approximately 10% increase in body weight compared with mice receiving the standard chow diet (CD) (Figure ). However, fasting glucose was significantly elevated—by approximately 25%—in the HFSD-treated group (Figure ). Importantly, the metastatic burden was significantly increased in mice fed a HFSD compared with those fed a CD, whereas Glut3 suppression decreased the metastatic potential in both the CD- and HFSD-treated groups (Figure ). Silencing Glut3 had a greater impact on tumor metastasis in the HFSD group than in the CD group. These findings indicate that Glut3 indeed plays a crucial role in high sucrose-mediated CRC metastasis.

Since there is no treatment that currently targets Glut3, to further explore therapeutic strategies based on Glut3-YAP axis-mediated CRC metastasis, we investigated the potentiality of panobinostat, an FDA-approved histone deacetylase inhibitor (HDACi) whose inhibitory activity against colorectal and lung cancers we have previously identified by suppressing YAP and TAZ 12, 15. As expected, panobinostat treatment not only decreased YAP protein levels but also suppressed Glut3 in three CRC lines (Figure ). Moreover, total and phosphorylated PKM2 were decreased by panobinostat (Figure ). The results of Transwell assays also showed that panobinostat effectively suppressed the invasiveness of colorectal cancer cells (Figure ). Thus, we further examined the potential of panobinostat to act on HFSD-associated tumor development. Immunocompromised mice were fed an HFSD, and the average body weights were 20% greater than those of the CD group (data not shown). The results consistently showed that mice receiving HFSD exhibited greater metastatic burden; however, treatment with panobinostat significantly reduced metastatic burden in both the CD and HFSD groups (Figure ). Accordingly, the fasting glucose concentration was increased in the HFSD treatment (Figure ), but no significant difference was found between the vehicle- and panobinostat-treated groups (Figure ), suggesting that the inhibitory effect of panobinostat was independent of blood glucose. These data indicate that targeting YAP-mediated metabolic reprogramming might be therapeutically beneficial in the treatment of CRC metastasis.

Discussion

Emerging evidence indicates that the Hippo transducer YAP/TAZ regulates metabolic reprogramming in cancer cells. Certain metabolic enzymes involved in glucose and amino acid metabolism are targets of YAP/TAZ. YAP transcriptionally regulates hexokinase 2 (HK2) and PFKFB3 16, and YAP and TEAD1 work together to regulate Glut1 expression 17, 18. Moreover, a recent study has shown that YAP promotes Glut3 transcription in glioma 13; however, the detailed regulatory mechanism and its clinical relevance are unclear. In our study, we found that YAP and glycolytic genes were concomitantly elevated in metastatic CRC. We further identified the binding and transactivation of YAP on the Glut3 promoter region with a TEAD. Suppression of YAP downregulated the expression of glycolytic genes, including Glut3. Clinically, CRC patients show a highly positive correlation between YAP/TAZ and Glut3, and those whose tumors exhibit YAP and Glut3 coexpression have the poorest survival outcomes. Moreover, we showed that YAP plays an indispensable role in metabolic adaptation to metastatic colon cancer. Using metabolic parameters, we demonstrate that YAP regulates glycolysis, nucleotide synthesis, amino acid flux, and glutathione balance. Recently, Santinon and colleagues reported that the regulation of deoxynucleotide metabolism by YAP/TAZ is required for chemotherapeutic resistance 19. We found that YAP regulates nucleotide biosynthesis and cellular detoxification; thus, chemotherapeutics may take advantage of metabolic alterations in cells requiring YAP. Together, the metabolic adaptation and the increased aerobic glycolysis caused by YAP may be part of a novel strategy for the treatment of colon cancer.

Recent findings have implicated YAP/TAZ activity in the regulation of metabolic cues, including glucose levels 20, 21. In fact, glucose-induced upregulation of YAP/TAZ has been found not only by the inhibition of Hippo kinase via AMPK but also by interactions with the glycolytic enzyme PFK1 20, 22. Moreover, a recent study by Peng and colleagues showed that glucose-induced YAP glycosylation and activation promoted tumorigenesis 23. In our study, we found that the upregulation of YAP in 116-LM cells occurs at least partially through Glut3, which facilitates glucose uptake and therefore increases YAP/TAZ. Suppression of Glut3 significantly diminished YAP and its downstream targets. Because YAP/TAZ plays a pivotal role in tumor progression, we also found that YAP regulates the expression of glycolytic genes. Our data indicate that positive feedback by Glut3 and YAP is present in the promotion of glucose metabolism and tumor metastasis.

Increasing evidence has revealed the crucial role of oncogenic PKM2 in tumor growth and stemness, independent of its pyruvate kinase activity. Phosphorylation of PKM2 causes it to translocate to the nucleus and function as a transcriptional cofactor of HIF-1 to stimulate the expression of glycolytic genes, including LDHA, PDK1, and SLC2A1 (Glut1) 24. In addition, nuclear PKM2 associates with the oncoprotein c-Myc to regulate cell proliferation and metabolic reprogramming. Moreover, a recent study by Zhou and colleagues showed that PKM2 phosphorylation regulates YAP activity and thereby promotes the transformation of breast epithelial cells 25; however, the detailed regulatory mechanism is unclear. Herein, we found that metastatic CRC upregulates the expression of total PKM2 and phospho-PKM2, which facilitates the association with YAP. YAP stabilizes and promotes nuclear translocation of PKM2, thereby enhancing Glut3 expression. Interestingly, we also found that PKM2 silencing affected YAP expression. These data are concordant with previous findings and suggest that PKM2 and YAP may interplay to each other. PKM2 has been reported to interact with the nuclear transporter Pin1 by binding the WW domain of Pin1 7. Our data showed that the TEAD binding domain rather than the WW domain in YAP is responsible for its association with PKM2. Given that the TEAD-YAP complexes predominantly govern YAP downstream gene expression, nuclear PKM2 plays a crucial role in shaping YAP-promoted tumorigenesis. Intriguingly, PKM2 suppression drastically impeded Glut3 expression, suggesting that other factors such as HIF-1 may cooperate with PKM2 to regulate the expression of Glut3. Further study is necessary to determine the possible mechanism of PKM2 phosphorylation and the genetic regulation of cancer metabolism by YAP/PKM2.

Conclusion: Our study indicated that Glut3 is an independent tumor marker for metastatic CRC. We identified the role of Glut3 in the promotion of CRC metastasis through a positive regulatory loop with YAP and PKM2 and suggest that it is a potential therapeutic target. The results of our study highlights that metabolic reprogramming is a feasible entry point for the treatment of colon cancer.

Materials and Methods

Cell culture

Human colorectal cancer cell lines were purchased from the Bioresource Collection and Research Center (Hsinchu, Taiwan). 116-LM cells were obtained from lung metastatic tumor cells by tail-vein injection of HCT116 cells into nonobese diabetic/severe combined immunodeficiency (NOD/SCID) mice, as established previously 12. HCT15 and HT29 cells were cultured in RPMI. HCT116, 116-LM, SW480, and SW620 cells were cultured in DMEM. All culture media were supplemented with 7% fetal bovine serum and GlutaMAX (Gibco). All cell lines were authenticated through short tandem repeat profiling (GenePrint 10 System) to ensure that no culture contamination had occurred.

Reagents and plasmids

2-Deoxy-glucose (2DG) was purchased from Sigma. Panobinostat was obtained from Selleckchem. The 8xTGIIC luciferase reporter plasmid was provided by Dr. Stefano Piccolo (Addgene #34615). The pcDNA-flag-YAP plasmid was provided by Dr. Marius Sudol (Addgene #17793). The pCMV-flag-YAP5SA plasmid was provided by Dr. Kunliang Guan (Addgene #27371). The pEGFP-C1-PKM2 plasmid was provided by Dr. Axel Ullrich (Addgene #64698). Full-length human Glut3 in the pcDNA3/DYK vector was purchased from GenScript. The deletion mutants of YAP5SA and PKM2 were subcloned into the pcDNA3-HA vector. The promoter of Glut3 (-2100~+43) was constructed in the pGL4.1[luc2] reporter plasmid. The promoters of Oct4, Nanog, and c-Myc were previously constructed in pGL4.1[luc2] reporter plasmids 26.

Expression of the Glut family in human colon cancer patients

Gene expression patterns of the Glut family and the clinicopathological profile of human CRC were obtained from TCGA, Hong 27, and Skrzypczak 28 datasets (Oncomine). The survival prognosis of different cancer types as a function of Glut3 was determined using OncoLnc and PrognoScan databases. A high expression group was defined as greater than 30% of patients. The RNA expression value of 7 paired primary and liver metastatic colon cancer tissues was obtained from the GSE40367 dataset 29. Microarray data were analyzed with R version 3.3.1 software. GSEA was performed using the MSigDB. The correlation of Glut3 and the YAP signature (CORDENONSI_YAP_CONSERVED_SIGNATURE) in CRC patients were obtained from GSE41258 30, GSE68468, GSE40967 31, and GSE17536 32 datasets.

Measurement of glycolytic capacity

The real-time extracellular acidification rate (ECAR) was measured using a Seahorse XFe24 Flux Analyzer (Seahorse Bioscience) according to the manufacturer's instructions. Cells were seeded in a Seahorse plate overnight to reach 80% confluence. One hour before measurement, the culture medium was replaced with XF basal medium (Seahorse Bioscience). The ECAR was measured over time following the injection of 10 mM glucose, 10 μM oligomycin, and 50 mM 2-deoxyglucose (2-DG). Glycolytic capacity was normalized to the cell number.

Metabolite measurements

A total of 1x107 cells were washed with ice-cold phosphate-buffered saline (PBS), and their metabolism was quenched with 500 μl of methanol. Then, the sample was vortexed for 2 min and kept on ice for 5 min. Another 500 μl of cold water was added, and the treatment was repeated. The supernatant was collected by centrifuging the sample at 15,000 rcf for 5 min, and 100 μl of the extracted supernatant was dried with a SpeedVac. The dried residues were further derivatized with trimethylsilylation, and the samples were analyzed by a gas chromatography-mass spectrometer (GC-MS) (Agilent 7890A) coupled with a 5975C inert mass selective detector (MSD) for metabolic profiling. Heatmaps were generated by MetaboAnalyst 4.0.

Targeted quantitative analysis of cellular metabolites was conducted by the C-SCOPE package of HMT (Human Metabolome Technologies, Inc., HMT) using capillary electrophoresis time-of-flight mass spectrometry (CE-TOFMS) for cation analysis and CE-tandem mass spectrometry (CE-MS/MS) for anion analysis. Briefly, 3-5x106 cells were washed twice with 5% mannitol solution. The cells were then treated with methanol and then treated with Milli-Q water containing internal standards (H3304-1002, Human Metabolome Technologies (HMT), Tsuruoka, Yamagata, Japan). The extract was filtered through a Millipore 5-kDa cutoff filter (UltrafreeMC-PLHCC, HMT) to remove macromolecules. The filtrate was centrifugally concentrated and resuspended in Milli-Q water for metabolomic analysis at HMT. The metabolites were annotated based on the HMT metabolite database. Hierarchical cluster analysis (HCA) and principal component analysis (PCA) were performed by statistical analysis software (Human Metabolome Technologies, Inc.). The metabolic parameters were evaluated using the equation described in Table . All values were normalized to cell numbers.

Short hairpin RNA and lentiviral infection

Short hairpin RNAs (shRNAs) for human YAP (TRCN0000107265; TRCN0000107266) and PKM2 (TRCN0000195581; TRCN0000195588) were obtained from the National RNAi Core Facility (Academia Sinica, Taipei, Taiwan). Lentiviral preparation and virus infection were performed as previously described. In brief, HEK293T cells were cotransfected with pLKO.shRNA together with pCMV-∆R8.91 and pMDG plasmids. At 48-h posttransfection, virus-containing supernatants were incubated with target cells for another 48 h. Transduced cells were selected using puromycin (2 μg/ml). A control shRNA targeting red fluorescent protein (RFP) was used as a negative control. A pool of Glut3 lentiviral particles containing 3 target-specific shRNAs (sc-41218V) was obtained from Santa Cruz Biotechnology. For CRISPR/Cas9-mediated Glut3 knockout, 116LM/sgGlut3 cells were generated using the All-in-one sgRNA/Cas9 system (National RNAi core, Academia Sinica, Taipei, Taiwan). The shRNA sequences are shown in Table .

Cell migration and invasion assays

For cell migration, 1x105 cells in 200 μl of serum-free medium were seeded into the upper chamber of a Transwell insert. For cell invasion, 4x105 cells in 200 μl of serum-free medium were seeded into the upper chamber of a Matrigel (BD Biosciences)-coated Transwell insert. The lower wells were filled with complete medium and served as the chemo-attractant. After 24 h, the cells were fixed with methanol and stained with crystal violet. The Matrigel and unmigrated cells were removed using cotton swabs prior to analysis.

Tumorsphere formation assay

Cells (104) were seeded in six-well ultralow attachment plates (Corning Costar) and maintained in serum-free DMEM/F12 supplemented with 20 ng/ml epidermal growth factor (PeproTech), 25 ng/ml basic fibroblast growth factor (PeproTech), and B27 (Gibco) for 10 days. The formation of tumorspheres was observed, and they were counted under a light microscope.

Immunoprecipitation and Western blotting

Cells were lysed in ice-cold radioimmunoprecipitation assay (RIPA) buffer supplemented with a protease and phosphatase inhibitor cocktail (Roche). Whole-cell extracts were incubated with the indicated antibodies and protein A/G beads (Santa Cruz Biotechnology) overnight. The beads were washed with RIPA buffer three times and then boiled in SDS loading buffer. For Western blotting, cell extracts were separated by SDS-PAGE. After Western transfer, membranes were probed with antibodies against DDK (Origene ta50011), GFP (Santa Cruz biotechnology sc-101536), Glut3 (Santa Cruz biotechnology sc-74399), HA (Santa Cruz biotechnology sc-393579), PKM2 (Cell Signaling #3198), phospho-PKM2(Y705) (Cell Signaling #3827), YAP (Cell Signaling #12395), phospho-YAP(S127) (Cell Signaling #13008), actin (GeneTex GTX11003) and lamin C (GeneTex GTX101127). Bands were detected with an enhanced chemiluminescence system (Millipore). Western blotting was performed at least three times, and representative experiments are shown. Quantification was carried out using ImageJ software.

Luciferase reporter assay

Cells were seeded in 24-well plates and transiently transfected with a pGL4.1[luc2]-promoter reporter plasmid together with the Renilla luciferase (RL)-TK plasmid using PolyJET transfection reagent. Firefly and Renilla luciferase activities were measured using a Dual-Luciferase Assay System (Promega) according to the manufacturer's instructions.

Real-time PCR

Total RNA was extracted with an RNA extraction kit and reverse-transcribed with a high-capacity cDNA conversion kit (Invitrogen, Carlsbad, CA, USA). cDNA was amplified using EvaGreen Master Mix (Biotium, Hayward, CA, USA) using a StepOne Plus Real-Time PCR system (Applied Biosystems, Darmstadt, Germany) with specific primers. The results were calculated using the ΔΔCT equation and are expressed as multiples of change relative to a control sample.

Chromatin immunoprecipitation

The protein-DNA complexes were crosslinked using 1% formaldehyde, which was then quenched by adding glycine to a final concentration of 200 mM. Chromatin complexes were sonicated to an average size of 250 bp by a MISONIX Sonicator 3000. The chromatin was incubated with anti-YAP (Cell Signaling #14074) and protein A/G beads (EZ-Magna ChIP Kit, Millipore) overnight. The immunocomplexes were reverse crosslinked, and the purified DNA was subjected to PCR analysis. The PCR primers are listed in Table .

Immunohistochemistry

Colorectal cancer tissue microarrays were purchased from Biomax. Slides were deparaffinized, rehydrated and heated in an antigen-unmasking solution (Vector Laboratories). Slides were incubated first with anti-Glut3 primary antibody (Santa Cruz biotechnology) overnight at 4 °C and then with SignalStain Boost IHC detection reagent (Cell Signaling Technology), and they were then visualized by 3,3'-diaminobenzidine (DAB) peroxidase substrate (Vector Laboratories). The H-score was determined by estimating the percentage of tumor cells positively stained with weak, moderate, and strong intensity and calculation using the following formula: (low%) × 1 + (medium%) × 2 + (high%) × 3.

Animal study

All animal studies were performed according to the guidelines and approval of the Animal Care and Use Committee of Taipei Medical University. To study the effect of Glut3 on CRC metastasis, 5-week-old BALB/c nude mice were fed a diet containing 45% fat and 17% sucrose (high-fat/sucrose diet (HFSD)) (Research Diets) or a standard chow diet (CD). Meanwhile, mice were injected intravenously with 2x106 116-LM/shRFP and 116-LM/shGlut3 cells for 7 weeks. To study the effect of panobinostat on colorectal cancer metastasis, 5-week-old NOD/SCID mice were fed an HFSD or CD for 7 weeks, followed by intravenous injection with 116-LM cells. After one week of inoculation with tumor cells, mice were given an intraperitoneal injection with vehicle or panobinostat (8 mg/kg, every T and W, weekly) for an additional 6 weeks. All dissected lungs were embedded with paraffin, and tissue sections were stained with hematoxylin and eosin and scanned by TissueFax (TissueGnostics). The metastatic index was calculated as the ratio of the area of metastatic burden to the total area of lung. To measure fasting plasma glucose levels, mice were fasted for 6 or 9 h during the day. Food was removed, but mice were allowed free access to water. Blood glucose levels were measured using a blood glucometer (Roche Diagnostics).

Statistical analyses

Data are presented as the mean ± standard error (SE) of three independent experiments. Statistical significance was determined by an unpaired, two-tailed Student's t-test unless stated otherwise. * Indicates a p value <0.05; ** indicates a p value <0.01. Coexpression trends were analyzed by a chi-squared test. A correlation coefficient was analyzed by the Pearson test. The survival probability was analyzed by a log-rank (Mantel-Cox) statistical test. Statistical analysis was carried out using GraphPad Prism software.

Supplementary figures and tables.

Click here for additional data file.