Multiple Metabolic Alterations Exist in Mutant PI3K Cancers, but Only Glucose Is Essential as a Nutrient Source

Targeting tumour metabolism is becoming a major new area of pharmaceutical endeavour. Consequently, a systematic search to define whether there are specific energy source dependencies in tumours, and how these might be dictated by upstream driving genetic mutations, is required. The PI3K-AKT-mTOR signalling pathway has a seminal role in regulating diverse cellular processes including cell proliferation and survival, but has also been associated with metabolic dysregulation. In this study, we sought to define how mutations within PI3KCA may affect the metabolic dependency of a cancer cell, using precisely engineered isogenic cell lines. Studies revealed gene expression signatures in PIK3CA mutant cells indicative of a consistent up-regulation of glycolysis. Interestingly, the genes up- and down-regulated varied between isogenic models suggesting that the primary node of regulation is not the same between models. Additional gene expression changes were also observed, suggesting that metabolic pathways other than glycolysis, such as glutaminolysis, were also affected. Nutrient dependency studies revealed that growth of PIK3CA mutant cells is highly dependent on glucose, whereas glutamine dependency is independent of PIK3CA status. In addition, the glucose dependency exhibited by PIK3CA mutant cells could not be overridden by supplementation with other nutrients. This specific dependence on glucose for growth was further illustrated by studies evaluating the effects of targeted disruption of the glycolytic pathway using siRNA and was also found to be present across a wider panel of cancer cell lines harbouring endogenous PIK3CA mutations. In conclusion, we have found that PIK3CA mutations lead to a shift towards a highly glycolytic phenotype, and that despite suggestions that cancer cells are adept at utilising alternative nutrient sources, PIK3CA mutant cells are not able to compensate for glucose withdrawal. Understanding the metabolic dependencies of PIK3CA mutant cancers will provide critical information for the design of effective therapies and tumour visualisation strategies.


Introduction
The PI3K-AKT-mTOR pathway is a key oncogenic signalling pathway and as such has a central role in regulating cell proliferation, cell survival, cancer cell invasion and metastasis [1][2][3]. Hyper-activation of the pathway is common in human cancers and can be achieved in a number of ways, including mutation of PIK3CA. PIK3CA, the gene encoding the alpha catalytic subunit of the kinase, is mutated in approximately 15% of colorectal tumours and approximately 30% of breast cancers, with most mutations occurring at three hotspots, E542K, E545K and H1047R [4][5][6][7][8][9]. The first two hotspot mutations, E542K and E545K, are found in exon 9, which lies within the helical domain, while H1047R, the most prevalent mutation in breast cancer, is found in exon 20, which lies within the kinase domain. Mutations within both regions lead to constitutive activation of PI3K-AKT-mTOR signalling [3,5,10].
The PI3K-AKT-mTOR pathway not only regulates cell proliferation and cell survival, but is an important pathway in the control and regulation of cancer metabolism [2,[11][12][13][14][15]. The integration of metabolic changes with proliferation signalling is crucial to conferring a competitive advantage upon a cancer cell, resulting in oncogenesis and metastasis.
Metabolic adaption of cancer cells is a well known phenomenon, with the studies by Otto Warburg that demonstrated cancer cells consume glucose at an elevated rate even in the presence of oxygen, remaining seminal to the field (reviewed [16][17][18]). Metabolic alterations do not simply result in changes in energy production, but are key for regulating the production of macromolecular building blocks required by a cell and for the maintenance of redox balance [16,[19][20][21][22]. Although aerobic glycolysis (the Warburg effect) is the most widely documented metabolic activity in tumours and cancer cell lines, additional pathways such as mitochondrial glutamine metabolism and fatty acid synthesis cooperate to produce the macromolecules needed to support continued cell growth. Indeed, glutamine contributes to many core metabolic tasks of proliferating tumour cells and is often only considered secondary to glucose in terms of importance in tumour cell metabolism. Glutamine is the most abundant amino acid in human plasma, and cancer cells metabolise glutamine in excess of any other amino acid [23][24][25]. Notwithstanding, a number of published studies have linked other amino acid dependencies with metabolic effects and survival of cancer cells [26][27][28]. These nutrient studies highlight the complex and somewhat flexible nature of tumour metabolism, since cancer cells can not only adapt and switch between different metabolic pathways, but also have simultaneous access to multiple nutrients.
The PI3K signalling pathway, largely through its effects on AKT and mTOR, has been associated with regulation of a number of metabolic effects, including glucose uptake, stimulation of the Warburg effect, and enhanced synthesis of lipids and proteins [11,29,30]. AKT activation for example has been shown to stimulate glycolysis by increasing expression and membrane translocation of the glucose transporter GLUT4. Active AKT also acts via GSK3b to regulate glycogen synthase and promote the association of Hexokinase 2 with the mitochondrial membrane, where Hexokinase 2 can more readily phosphorylate glucose [11,13,[29][30][31][32][33].
While historical studies have demonstrated connections between the PI3K-AKT-mTOR pathway and metabolism, the model systems used have often relied on comparisons between cell lines that have multiple genetic differences or used over-expression strategies that are not reflective of patient tumour genetics. The predominant focus of research has centred on the influence of tumour genetics on glucose metabolism. However, genetic alterations can also dictate alternative nutrient and metabolic dependencies.
In order to explore how endogenous PIK3CA mutations specifically alter metabolic pathways and to better understand whether any of these potential changes establish therapeutically targetable cellular metabolic dependencies, we have performed focused metabolic gene expression analysis, nutrient switching and siRNA experiments using isogenic cell line models that are genetically identical, apart from the mutation status of the endogenous PIK3CA gene.

Growth Dependency Assays
Proliferation of cells over a 120 hour period was evaluated using the sulphorhodamine B (SRB) assay [35]. Cells were seeded into 96-well plates in triplicate in glucose and glutamine free DMEM (PAA) supplemented with the required concentration of glucose (Sigma) and glutamine (PAA). MCF10A isogenic cell lines were additionally supplemented with 5% horse serum, 10 mg/ml insulin, 0.5 mg/ml hydrocortisone, 0.1 mg/ml cholera toxin and 0.2 ng/ml hEGF. All other cell lines were supplemented with 10% foetal bovine serum only. Where indicated media was also supplemented with 0.5 mM fructose, 10 mM galactose, 0.5 ml/l fatty acid cell culture supplement or 0.1 mM aspartic acid (Sigma).

Gene Expression Studies
MCF10A isogenic cell lines were seeded into 25 cm 2 flasks in DMEM/F12 media supplemented with 5% horse serum, 10 mg/ ml insulin, 0.5 mg/ml hydrocortisone, 0.1 mg/ml cholera toxin and 0.2 ng/ml hEGF. HCT116 isogenic cell lines were seeded into 25 cm 2 flasks in McCoy's 5A media supplemented with 10% foetal bovine serum. After 48 hours the cells were harvested by trypsinisation and RNA prepared using the RNeasy Kit (Qiagen), according to the manufacturer's instructions.
Total RNA was reverse transcribed using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems). The transcription levels of 45 genes involved in glycolysis, glutaminolysis, pentose phosphate pathway, oxidative phosphorylation, and lipid metabolism were quantified by real-time PCR along with three normalization genes (NUP214, PPIG and SLU7). 5 ng of cDNA were used per PCR. Amplifications were performed with DZO5 DNA polymerase in a buffer containing 20 mM Tris pH7.85, 30 mM KCl, 3 mM MgCl 2 , 100 mM dA,G,CTP, 200 mM dUTP, 0.8units uracil N-glycosylase, 6% glycerol, 1X ROX, and 0.2X SYBR green. The reactions were amplified on the Prism 7900 using the following cycling parameters: 50uC for 2 minutes, 95uC for 12 minutes, followed by 45 cycles of 95uC for 20 seconds and 60uC for 1 minute. A reference pool (Universal Human Reference RNA, Stratagene) at 2.5 ng was amplified with each gene. All amplifications were performed in duplicate.
Gene expression was determined using the 2(-DDCt) method described by Livak and Schmittgen [36]. The expression of a gene in PIK3CA mutant cells relative to PIK3CA wild type cells was determined using the formula 2 (DDCt PI3K mutated-DD Ct PI3K WT) .

Statistical Analysis
Statistical analysis was performed using 2-way ANOVA tests followed by Bonferroni's multiple comparison test. For simplicity, p values are represented as p,0.05 (*) or p,0.01 (**) only.

The Presence of a PIK3CA Activating Mutation Results in Gene Expression Changes Associated with Increased Glycolytic Dependency and Additional Metabolic Pathway Alterations
To fully characterise the influence of a PIK3CA mutation on metabolic alterations as a whole, a targeted gene set was first compiled, representing 45 genes from different metabolic pathways such as glycolysis, oxidative phosphorylation, glutaminolysis, pentose phosphate pathway and fatty acid metabolism. Gene expression was then analysed across replicate RNA samples, prepared from two isogenic model systems. The first cell model system was a non-tumorigenic cancer cell line pair comprised of MCF10A parental cells in which PI3Ka is wild type (WT), PI3Ka (+/+) and MCF10A cells into which the H1047R activating mutation had been introduced, PI3Ka (H1047R/+). The second was a colon cancer cell line HCT116, which is heterozygous for the H1047R PIK3CA mutation, PI3Ka (H1047R/+), paired with cells in which the mutant allele has been knocked-out to render the cells functionally WT, PI3Ka (+/2). Analysis of gene expression for the PIK3CA mutant cell samples relative to PIK3CA WT cells revealed some striking alterations in metabolic gene expression levels. While the precise genes up-and down-regulated in the PIK3CA mutant cells compared to PIK3CA WT differed between the two cell model systems, the overall changes in gene expression were indicative of an increased aerobic glycolytic phenotype within PIK3CA mutant cells. Introduction of an activating mutation within PIK3CA in MCF10A cells led to an increase in expression of the key glycolytic enzyme Hexokinase 2 (HK2), and a concurrent decrease in expression of the fructose transporter, GLUT5 ( Figure 1A). In the HCT116 colon cancer cell background, the presence of a mutated PIK3CA resulted in elevated expression of the glucose transporter, GLUT1 ( Figure 1B).
Other gene expression changes as a consequence of the presence of a PIK3CA mutation suggested that additional metabolic pathway alterations may exist. In MCF10A PI3Ka (H1047R/+) cells, the fatty acid transporter SLC27A1 showed reduced expression, with several genes involved in oxidative phosphorylation also showing reduced expression ( Figure 1A). Increases in the expression of genes such as asparagine synthetase (ASNS) and glutaminase (GLS) were consistently seen in PI3Ka (H1047R/+) cells, suggestive of elevated glutaminolysis ( Figure 1A and 1B).
Changes in mRNA expression were then confirmed at the protein level, demonstrating elevated levels of key glycolytic enzymes such as HK2 and GLUT1, and additionally revealing upregulation of the lactate transporter MCT4 in the MCF10A PI3Ka (H1047R/+) cells (Figure 2). Activation of the PI3K pathway was confirmed in both isogenic models as shown by AKT phosphorylation.

Growth of PIK3CA Mutant Cells is Selectively Dependent on Glucose While Glutamine is Essential in Both WT and PIK3CA Mutant Cells
Results from the gene expression profiling were suggestive of a number of alterations in cellular metabolism as a result of mutated PIK3CA, specifically increased glycolysis and altered glutaminolysis. To explore the possibility that cells may be more dependent not only on specific pathways but also demonstrate nutrient source dependency, we investigated the effect of nutrient depletion on cellular growth, focusing on cellular dependency on glucose and glutamine. Over 120 hours, complete withdrawal of glucose (in the presence of glutamine) negatively impacted the ability of MCF10A PI3Ka (H1047R/+) to proliferate and grow, while having no effect on the growth of MCF10A PI3Ka (+/+) cells ( Figure 3A). This differential impact was also true for a second MCF10A cell line containing an alternative PIK3CA mutation, MCF10A PI3Ka (E545K/+) ( Figure 3A). Likewise, growth of HCT116 PI3Ka (H1047R/+) was more severely impacted following glucose depletion and withdrawal than HCT116 PI3Ka (+/2) cells ( Figure 3B). This observation suggested that the PIK3CA WT cells were better able to tolerate glucose withdrawal than PIK3CA mutant cells, albeit in the presence of glutamine. To better understand the involvement of glutamine, we performed a matrix of studies, evaluating glucose depletion and withdrawal in conjunction with glutamine depletion and withdrawal. The growth of MCF10A cells, irrespective of PIK3CA mutation status was severely compromised by complete withdrawal of glutamine, even in the presence of glucose ( Figure 3C). Partial depletion of glutamine also negatively impacted growth of the cells, but the extent of dependency did not appear to correlate to PIK3CA mutation status. Similarly, complete glutamine withdrawal severely compromised growth of HCT116 cells. In contrast to MCF10A cells, HCT116 cells were better able to tolerate an intermediate level of glutamine withdrawal, with no growth impairment seen in either PIK3CA WT or mutant cells ( Figure 3D). The pattern of glucose sensitivity was maintained with this intermediate level of glutamine ( Figure S1).

The Glucose Dependency Exhibited by PIK3CA Mutant Cells Cannot be Overridden by Supplementation with Other Nutrients
In order to further characterise the glycolytic phenotype and apparent glucose dependency of PIK3CA mutant cells, we used the MCF10A isogenic model system to evaluate whether alternative nutrient sources could compensate for glucose and override the dependency. Switching the nutrient source to an alternative monosaccharide (fructose or galactose) was unable to compensate for glucose withdrawal in MCF10A PI3Ka (H1047R/+) cells, while growth of MCF10A PI3Ka (+/+) cells remained unaffected by removal of glucose or nutrient switching (Figure 4). Since gene expression data suggested that PIK3CA mutant cells may have alterations in metabolic pathways other than glycolysis, we further investigated whether the glucose growth dependency could be overridden by supplementing with nutrients that may promote these alternative pathways. Based on gene expression changes in  asparagine synthetase and fatty acid transporters, we supplemented cells with aspartic acid or a mixture of fatty acids. Neither supplement was able to override the effects of withdrawing glucose on the growth of MCF10A PI3Ka (H1047R/+) cells (Figure 4). Similar results were seen when the MCF10A PI3Ka (E545K/+) and HCT116 PI3Ka (H1047R/+) cells were profiled ( Figure S2).

Disruption of the Glycolytic Pathway Demonstrates Antiproliferative Effects in PIK3CA Mutant Cells
The MCF10A PI3Ka (H1047R/+) cells show a high level of glucose dependency and are unable to compensate for the absence of glucose by utilizing other nutrients. In conjunction with our data demonstrating that PIK3CA mutant cells show increased expression of the key glycolysis enzyme HK2, we hypothesised that disruption of the glycolytic pathway should be sufficient, even in the presence of glucose, to inhibit growth of the mutant cells. We therefore evaluated the effects of targeted disruption of glycolysis using two siRNA molecules to knockdown levels of HK2 in the MCF10A isogenic cells. Using a non-targeted siRNA molecule as a control, we found that specific HK2 knockdown does indeed result in anti-proliferative effects, selective for the PIK3CA mutant cells ( Figure 5A). Knockdown of HK2 protein levels in both PIK3CA WT and mutant cells lines was confirmed by Western blotting ( Figure 5B), with specificity of knockdown confirmed by monitoring HK1 expression.

Glucose Dependency is Seen in Other Cancer Cell Lines Harbouring Endogenous PIK3CA Mutations
Our investigations have used a precise and genetically defined system to investigate the effects of a PIK3CA mutation on cellular metabolism in isolation. We extended our studies to understand whether these findings were true across a broader panel of cancer cell lines. We therefore determined the glucose growth dependency of six colon and breast cancer cell lines, all containing endogenous PIK3CA mutations (Table 1). However, these lines also have the caveat of having a number of other genetic differences, which may modulate or bypass the putative glucose dependency determined in isogenic cell-systems. The extent to which growth was affected varied between the lines, with 3 of the 6 cell lines (T47D, BT20 and DLD-1) demonstrating reduced growth following partial glucose depletion ( Figure 6). However, growth of all 6 cell lines was compromised by complete withdrawal of glucose even in the presence of glutamine, supporting the concept that PIK3CA mutant cancers are heavily reliant on glucose to support cell growth.

Discussion
Interest in studying cancer metabolism and developing metabolic targeted agents as therapies has grown over recent years, and so an improved understanding of the unique metabolism of cancer cells is imperative for effective drug design and the rational treatment of patients, as well as labelled nutrient visualisation and treatment response monitoring. Despite activation of the PI3K-AKT-mTOR pathway being a common feature of many cancers, and activation of this pathway being associated with regulation of a number of metabolic effects, comprehensive profiling of the consequence of PIK3CA mutations on metabolic dependencies has not been studied. By using genetically defined isogenic cell models, we have been able to precisely define how PIK3CA mutations can influence metabolic pathways within a tumour.
Regulation of metabolic processes and nutrient utilisation are complex processes. Gene expression analysis in this study has revealed that the overall consequence of PIK3CA mutations is a shift towards an increased glycolytic phenotype. However, the exact way a cell effects this shift may vary. Rather than a single rate limiting molecule, several steps creating a 'regulatory pathway' may share the homeostatic control of energy metabolism. Multi-step control was highlighted by the differential gene up-regulation of key glycolysis enzymes in the different isogenic models. Though differences were seen, the net effect is suggestive of up-regulation of glycolysis in both cases. This link between patient relevant PIK3CA mutations and resulting glycolytic phenotype is consistent with previous studies that have associated activation of the AKT signalling pathway with stimulating the Warburg effect [11,13]. This concept is built on and strengthened by having revealed a clear dependence of PIK3CA mutant cells on glucose for continued growth. Despite suggestions that cancer cells are adept at utilizing other nutrient sources, this study has demonstrated that PIK3CA mutant cells are not readily able to compensate for glucose withdrawal, either through use of glutamine, alternative monosaccharides or other nutrients. Dependence on glucose for growth was also shown across a wider panel of cancer cell lines harbouring endogenous PIK3CA mutations. The results of this study could provide critical information for designing therapeutic strategies, since they have highlighted critical nodes within metabolic pathways that might be key to target. Indeed, knockdown studies confirmed that targeting the glycolytic pathway for disruption results in anti-proliferative effects, specific to PIK3CA mutant cells, suggesting that this could be a therapeutic strategy for specifically targeting patients with PIK3CA mutant tumours. The overt glucose dependence of cells with this mutation also offers routes for non-invasive imaging of PI3K-pathway activated tumours.
Through characterisation of the metabolic phenotype of genetically matched isogenic cells, we have been able to define the metabolic consequences of a single engineered genetic alteration, in this case an activating mutation within the PIK3CA gene. Whilst confirming that activation of the PI3K-AKT-mTOR pathway results in an increased glycolytic phenotype, we have also demonstrated a specific dependence for PIK3CA mutated cells on glucose and illustrated that targeting this pathway offers therapeutic potential and tumour imaging insight. Figure S1 Glucose sensitivity of HCT116 PI3Ka (H1047R/+) cells was maintained when grown in 0.5 mM glutamine media. HCT116 isogenic cells (+/2 and H1047R/+) were grown for 120 hours in media containing 0.5 mM glutamine and the indicated concentrations of glucose. Cell growth was assessed using SRB staining. The growth of each cell line is expressed relative to growth in media containing 0.5 mM glutamine and 25 mM glucose. (TIF) Figure S2 The glucose dependency of PIK3CA mutant cells cannot be overridden by supplementation with alternative nutrients. MCF10A PI3Ka isogenic cells (+/+ and E545K/+) and HCT116 isogenic cells (+/2 and H1047R/+) were grown for 120 hours in media containing 2 mM glutamine and the indicated concentrations of glucose (A and B respectively). Additionally, cells grown without glucose were supplemented with either fatty acid cell culture supplement (+FA) or 0.1 mM aspartic acid (+AA). Cell growth was assessed by SRB staining. (TIF)