PI3K (phosphoinositide 3-kinase) signalling pathways regulate a large array of cell biological functions in normal and cancer cells. In the present study we investigated the involvement of PI3K in modulating small molecule metabolism. A LC (liquid chromatography)-MS screen in colorectal cancer cell lines isogenic for oncogenic PIK3CA mutations revealed an association between PI3K activation and the levels of polyamine pathway metabolites, including 5-methylthioadenosine, putrescine and spermidine. Pharmacological inhibition confirmed that the PI3K pathway controls polyamine production. Despite inducing a decrease in PKB (protein kinase B)/Akt phosphorylation, spermidine promoted cell survival and opposed the anti-proliferative effects of PI3K inhibitors. Conversely, polyamine depletion by an ornithine decarboxylase inhibitor enhanced PKB/Akt phosphorylation, but suppressed cell survival. These results suggest that spermidine mediates cell proliferation and survival downstream of PI3K/Akt and indicate that these two biochemical pathways control each other's activities, highlighting a mechanism by which small molecule metabolism feeds back to regulate kinase signalling. Consistent with this feedback loop having a functional role in these cell models, pharmacological inhibitors of PI3K and ornithine decarboxylase potentiated each other in inhibiting tumour growth in a xenograft model. The results of the present study support the notion that the modulation of spermidine concentrations may be a previously unrecognized mechanism by which PI3K sustains chronic proliferation of cancer cells.
- cell signalling
- metabolite profiling
- phosphoinositide 3-kinase (PI3K)
PI3K (phosphoinositide 3-kinase) signalling is involved in the control of essential cellular functions downstream of RTKs (receptor tyrosine kinases), G-protein-coupled receptors and Ras [1–3]. Upon receptor stimulation, heterodimeric class I PI3K subunits are recruited to the plasma membrane where they phosphorylate PtdIns(4,5)P2 to produce PtdIns(3,4,5)P3, a lipid second messenger that nucleates a large array of downstream effectors, resulting in the activation of cascades involving other lipid kinases and phosphatases, GTPases, and protein kinases such as PKB (protein kinase B)/Akt [4–7]; these in turn regulate a range of biochemical and cell biological pathways, including those controlling glucose metabolism, survival, cell cycle progression and motility .
In addition to its numerous functions in normal physiology, PI3K signalling is also a target for cancer treatment. Overactivation of the PI3K/Akt pathway in cancer cells can occur through activating mutations in PIK3CA, the gene encoding the p110α catalytic isoform of class IA PI3K. These mutations are common in colorectal and breast cancers among others  and have been shown to be oncogenic [10,11]. The PI3K pathway can also be overactive in cancer as a result of overexpression or activating mutations in PKB/Akt  or by inactivation of PTEN (phosphatase and tensin homologue deleted on chromosome 10) , the lipid phosphatase that dephosphorylates PtdIns(3,4,5)P3 to PtdIns(4,5)P2, thus terminating the PI3K signal. PTEN is one of the most frequently deregulated tumour suppressors in cancer . Since the pathway operates downstream of RTKs, PI3K/Akt signalling activity is often also increased in cancers driven by RTKs, which can occur as a result of activating mutations or overexpression of RTKs themselves or by the presence of high concentrations of growth factors in the tumour microenvironment . In addition, the transformation potential of Ras can be dependent upon its interaction with p110α .
The PTEN–PI3K–Akt axis contributes to several of the biochemical and cell biological mechanisms that underlie cancer biology. Indeed, by acting upstream of the translation and transcription factors that promote cell survival, protein synthesis and cell cycle progression, an overactive PI3K pathway plays a key role in the ability of cancer cells to evade cell death and to maintain chronic proliferation, two fundamental traits of cancer . In addition, p110α acts downstream of VEGFR (vascular endothelial growth factor receptor) and has a key role in physiological angiogenesis ; thus this PI3K isoform could also promote metastasis. Another potential role of PI3K in contributing to the hallmarks of cancer involves the modulation of energy metabolism. In insulin-responsive tissues, the PI3K–Akt axis regulates energy homoeostasis by acting downstream of the insulin receptor to control the activity of enzymes involved in gluconeogenesis, glycolysis and lipid metabolism . Cancer cells promote the use of glucose through the glycolytic pathway even in the presence of oxygen (the Warburg effect) and thus PI3K could have a role in regulating fluxes of energetic metabolic pathways in cancer cells. Deregulation of energy metabolism is being rediscovered as being a key hallmark of cancer . However, although PI3K has been implicated in the regulation of energetic processes in insulin-responsive tissues , the extent of PI3K implication in the regulation of small molecule metabolism in cancer is not yet fully defined.
The aim of the present study was to investigate the involvement of PI3K in the control of small molecule metabolism, without a preconception of the biochemical pathways that may be downstream of this enzyme. Metabolites were profiled by LC (liquid chromatography)-MS across isogenic colorectal cancer cell lines bearing either an oncogenic mutation in PI3KCA or a WT (wild-type) allele. We found that molecules associated with polyamine metabolism were more abundant in cells bearing the oncogenic PIK3CA mutation, compared with cells with the WT allele or with cells treated with PI3K inhibitors. Interestingly, the polyamine spermidine opposed the effects of PI3K inhibitors in arresting cell proliferation, indicating that the proliferative effects of the PI3K/Akt pathway are mediated, at least in part, by spermidine. In addition, modulation of polyamine levels by small molecule inhibitors showed that spermidine modulates PI3K/Akt signalling activity, revealing the existence of a negative feedback loop in which polyamine and PI3K pathways control each other's activity. Consistent with a functional role for this feedback loop in cancer cell proliferation, a combination of inhibitors against PI3K and enzymes of polyamine synthesis decreased cell proliferation and tumour growth to a greater extent than treatment with either inhibitor alone. These results therefore indicate that the modulation of polyamine metabolism by PI3K may be a mechanism by which this oncogenic signalling pathway promotes cancer cell proliferation.
MATERIALS AND METHODS
Antibodies against phospho-PKB/Akt (Ser473), PKB/Akt, MTAP (S-methyl-5′-thiadenosine phosphorylase), cleaved PARP [poly(ADP-ribose) polymerase] and vinculin were from Cell Signaling Technology; anti-LC3-2G6 antibody was from Nanotools and anti-α-tubulin antibody was from Sigma. The ODC (ornithine decarboxylase) inhibitors ODC-I and DFMO (α-difluoromethylornithine) were from Merck and from Tocris respectively. PI-103 was purchased from Tocris and GDC-0941 was from Chemdea. LC-MS grade solvents (water, acetonitrile and methanol) were from LGC Promochem and formic acid was from Fisher Scientific. Guava viacount reagent was from Millipore. Tris/HCl was from Severn Bio Tech and PVDF membrane was from Millipore. Metabolite standards, including ornithine, putrescine, spermidine, spermine and MTA (5-methylthioadenosine) were from Sigma. Other reagents were from Sigma unless otherwise stated.
The isogenic DLD-1 cell lines with WT or mutant PI3KCA were provided by Dr Victor E. Velculescu (Ludwig Center for Cancer Genetics and Howard Hughes Medical Institutions, Johns Hopkins Kimmel Cancer Center, Baltimore, MD, U.S.A.) . The colorectal cell lines LS123, LS174T and SW948 were purchased from A.T.C.C. (supplied by LGC Standards, Teddington, U.K.). DLD-1 human colon cancer cell lines were cultured at 37°C in a 5% CO2 incubator in DMEM (Dulbecco's modified Eagle's medium) containing 0.5% FBS (fetal bovine serum). The LS123  and LS174T [22,23] cell lines were cultured in EMEM (Eagle's minimal essential medum) and SW948 cells were cultured in Leibovitz's L-15 medium [22,24].
Cells were homogenized in lysis buffer (1% Triton X-100, 50 mM Tris/HCl, pH 7.5, 150 mM NaCl and 1 mM EDTA) supplemented with protease and phosphatase inhibitors. Proteins were separated by SDS/PAGE (10% gels), transferred on to PVDF membranes and then probed with antibodies against phospho-Akt (Ser473), PKB/Akt, MTAP, cleaved PARP or LC3.
Cells were seeded in 96-well plates (4×103 cells per well). After culturing for the indicated time periods, 20 μl of MTS solution was added to each well and absorbance was recorded at 570 nm.
Cell viability and cell count were determined utilizing a Guava PCA (principal component analysis) cell analyser (Guava Technologies) using the Guava ViaCount Reagent (Millipore).
Direct cell counts
Cells were counted using a haemocytometer. Cell proliferation was determined by subtracting the number of cells obtained after treatment to the number of cells seeded.
Extraction of metabolites from cell lysates and media
Metabolites were extracted with 60 μl of ice-cold methanol (50%) containing internal standard in media, vortex-mixed and sonicated on ice for 20 min. The tubes were centrifuged at 13200 g for 10 min at 4°C and supernatants were dried under vacuum.
Metabolite analysis by nanoflow LC-MS/MS (tandem MS)
Metabolites were analysed in a LTQ-Orbitrap XL mass spectrometer (Thermo Fisher Scientific) connected online to a nanoflow ultra-HPLC (LC, nanoAQUITY; Waters). Chromatographic separations were performed on a 100 μm×100 mm ACQUITY™ 1.7 μm C18 BEH column (Waters) at a flow rate of 600 nl/min. A 15 min gradient (5–30% B) was employed with mobile phase A of 0.1% formic acid in water and mobile phase B of 0.1% formic acid in acetonitrile. Full scan survey spectra (m/z 50–1000) were acquired in the Orbitrap with a resolution of 60000. Progenesis LC-MS, Sieve and Pescal were used for data analysis. The putative identities of metabolite masses were determined by inputting m/z values in the human metabolome database (http://www.hmdb.ca/) allowing an m/z window of 5 p.p.m.
Cells were lysed in a urea-based lysis buffer and proteins were digested using trypsin as reported previously . Phosphopeptides were enriched from total peptides by TiO2 chromatography essentially as described previously [26,27]. Enriched phosphopeptides were analysed by LC-MS/MS in the LTQ-Orbitrap-nanoAcquity system described above with a gradient consisting of 1–35% mobile phase B over 100 min followed by 80% mobile phase B for 10 min. Peptide identification was performed by searching against the SwissProt database using the Mascot search engine with mass tolerance of 5 p.p.m. and 600 mmu (milli-mass units) for parent and fragment ions respectively. Allowed variable modifications were methionine oxidation, pyroglutamate at the N-terminus and phosphorylation of serine, threonine and tyrosine residues. Quantification was done using Pescal as described previously [25,28].
Metabolite anlaysis by SRM (single reaction monitoring) in LC triple quadrupole MS
Metabolites were analysed on a 2.1 mm×150 mm C18, 2.7 μm fused core technology column. A 12 min gradient was employed with mobile phase A of 0.1% formic acid in water and mobile phase B 0.1% formic acid in methanol at a flow rate of 400 μl/min (10–80% mobile phase B over 6 min). MS analysis was performed on a TSQ Vantage LC-MS/MS System (Thermo Scientific). Quantification was carried out using SRM to monitor the transitions m/z 298.1–136.2 for MTA, 146.2–72.2 for spermidine, 203.3–112.3 for spermine, 133.2–70.2 for ornithine and 89.3–72.3 for putrescine. Xcalibur software (Thermo Scientific) was used for data acquisition and analysis.
DLD-1 xenografts in mice
The present study was carried out in accordance with the regulations of the Animals (Scientific Procedures) Act 1986. The protocol was approved by the local Ethical Review Committee and by the U.K. Home Office. MUT (E545K mutant allele of PIK3CA) cells (3×106) were injected subcutaneously into the flanks of 8-week-old female Fox Chase SCID® Mice (Charles River Laboratories). After 7 days, mice with tumours greater than 75 mm3 were randomly divided into four groups. A first group was treated with GDC-0941 (25 mg/kg of body mass) in 0.5% methylcellulose and 0.2% polysorbate 80 (Tween 80) in de-ionized water (MCP buffer), administrated by oral gavage. A second group was treated with 0.1% DFMO in drinking water, changed daily. A third group was treated with GDC-0941 (25 mg/kg of body mass) and 0.1% DFMO in drinking water. The control group was treated with MCP buffer according to the same dose schedule. Mice were anaesthetized with pentobarbital and killed after 7 days of treatment when the control tumours reached ~400 mm3. Tumours were removed, weighed and snap frozen in liquid nitrogen until further analysis.
Statistical significance was assessed using the Student's t test calculated using Microsoft Excel 2007 after log2 transformation of the data (for phosphoproteomics) and by Mann–Whitney test (using Prism). P values of the phosphoprotemics data were corrected for multiple testing in Microsoft Excel 2007 using the Benjamini–Hochberg procedure .
Metabolite profiles downstream of PI3K in DLD-1 colorectal cancer cells
We used LC-MS to compare metabolite profiles in isogenic DLD-1 colorectal cancer cell lines bearing either a WT or an E545K mutant allele of PIK3CA  (these cell lines are hereafter referred to as WT and MUT respectively). Metabolite amounts were also profiled in these cell lines after treatment with PI-103, a class I PI3K inhibitor  with some degree of selectivity towards p110α at low doses . These analyses were performed with cells growing at 0.5% FBS because differences in the phosphorylation of PKB/Akt at Ser473, a key marker of PI3K pathway activation, were only apparent between the MUT and WT cells at this concentration of serum (Figure 1A). These data are consistent with previously published results on these cell lines  and indicate that PI3K is constitutively active in MUT cells in the near absence of upstream signals.
A total of 3736 different ion masses were detected by LC-MS across the present study. The intensities of the whole dataset produced two groups by PCA, one containing untreated MUT cells, with the other containing MUT cells treated with PI-103, WT cells treated with vehicle or with PI-103 (Figure 1B). These results indicate that significant changes in global metabolite profiles occur upon disruption of PI3K signalling. A total of 294 m/z values were significantly different between WT and MUT cells by at least 2-fold and P<0.05 (by Benjamini–Hochberg corrected t test, n=3). Based on accurate mass (±5 p.p.m.), 63 metabolites could be putatively identified; these belong to diverse chemical classes, including amino acids, carbohydrates, nucleic acids and other organic acids (Figure 1C and Supplementary Table S1 at http://www.biochemj.org/bj/450/bj4500619add.htm). It should be noted that these metabolite identities were based on accurate mass only and therefore these identifications remain putative.
Validation of MTA as a metabolite produced downstream of PI3K
Among the different metabolites found to be potentially regulated by the PI3K pathway in WT and MUT cells, the ion at m/z 298.097 was of particular interest because, on the basis of accurate mass, it could in principle be derived from MTA, the substrate of MTAP. MTAP expression is frequently lost in cancer and this enzyme is thought to be a tumour suppressor in certain cancer types [32,33]. The ion at m/z 298.097 was one of the masses more significantly inhibited upon PI3K pathway modulation by PI-103 in both WT and MUT cells, decreasing its ionic intensities ~3-fold in both cell types (Figure 1D; P<0.01). Moreover, under the basal state, the intensity of this ion was >2-fold greater in MUT relative to WT cells (P=0.01, n=3; Figure 1E).
In order to confirm that the compound at m/z 298.097 was MTA, we compared the spectra derived from this ion with that of an MTA standard obtained from a commercial source. The MTA standard co-eluted with the endogenous compound at m/z 298.097 by LC-MS and spiking the MTA standard in the sample produced a single peak, thus confirming this co-elution (Supplementary Figure S1 at http://www.biochemj.org/bj/450/bj4500619add.htm). The MS/MS of the MTA standard, the spiked MTA in the sample and the endogenous cellular compound produced the same fragment ions at equivalent intensity ratios (Supplementary Figure S1), consistent with the notion that the MTA standard and the endogenous ion at m/z 298.097 were the same compound, and confirming its identity as MTA.
MTA, a byproduct of polyamine metabolism, is metabolized to methylthioribose-1-phosphate and adenine by MTAP. We hypothesized that an accumulation of MTA in MUT cells was due to a modulation of MTAP expression and/or activity downstream of PI3K. Consistent with this hypothesis, MTAP expression was lower in MUT cells relative to the WT as revealed by immunoblotting for MTAP (blue arrows in Figure 1F, and Supplementary Figure S2 at http://www.biochemj.org/bj/450/bj4500619add.htm), with PI-103 treatment resulting in an increase of MTAP expression in MUT cells (red arrows in Figure 1F, and Supplementary Figure S2). These results indicate that MTAP expression is modulated, at least in part, by PI3K and that this results in an accumulation of MTA in cells with oncogenic mutations on PIK3CA.
We next investigated three other colorectal cancer cell lines, namely LS123, LS174T and SW948, all of which had high basal phosphorylation of PKB/Akt on Ser473, a site known to be downstream of PI3K. Of these, only LS123 showed a reduction in PKB/Akt phosphorylation upon prolonged (24 h) treatment with PI-103 (Figure 2A). This reduction in PKB/Akt phosphorylation was associated with an increase in MTAP levels (Figure 2A), a decrease in MTA levels (Figure 2B) and a decrease in cell viability (Figure 2C). Thus reduction of MTA levels, concomitant with an increase in MTAP as a result of prolonged PI3K inhibition, only occurred in the cell line that responded to PI-103, but not in the cell lines resistant to this inhibitor. These results provide further support to the notion that PI3K modulates MTA levels in colorectal cancer cell lines.
Targeted analysis by LC-MS identifies polyamine metabolism as a pathway downstream of PI3K in DLD-1 cells
Since MTA is a byproduct of polyamine metabolism (; Figure 3A), we wondered whether other steps in this pathway were also downstream of PI3K signalling. Unfortunately, the polyamines were not detected in our original analysis by high-resolution MS because these small molecules were not amenable to analysis by our nanoflow LC-MS system, most likely due to their basicity and because very small molecules are not always well separated by capillary LC using standard conditions .
We therefore designed specific MS-based assays to target the analysis of the polyamines ornithine, putrescine, spermidine and spermine, which together constitute the ‘classical’ polyamine metabolic pathway (Figure 3A). These analyses were performed by SRM. MTA was also analysed by these targeted SRM assays. Whereas the levels of ornithine and spermine (the first and last metabolite in the pathway respectively) were not significantly different in MUT and WT cells and were unaffected by 24 h treatment with PI-103, differences were found in the other pathway components. These include putrescine, spermidine and MTA, whose concentrations were higher in MUT cells compared with WT cells (Figures 3D–3F). Treatment of cells with PI-103 significantly reduced the concentration of putrescine in MUT cells (the signals of which could not be detected in WT cells) and of spermidine and MTA in both WT and MUT cell lines (Figure 3). These results indicate that PI3K signalling controls the levels of the polyamines spermidine and putrescine in these cell models.
Polyamines modulate PI3K/Akt signalling in DLD-1 cells
The results of the present study indicated that PI3K signalling regulates polyamine production. Polyamines have pro-survival effects  and modulate PKB/Akt phosphorylation in neuroblastoma and epithelial cells [37,38]. Consistent with previous studies , the exogenously added polyamines spermidine and spermine, which are cell-permeable , promoted cell viability at concentrations of up to 1 μM (above which they reduced viability, Supplementary Figure S3 at http://www.biochemj.org/bj/450/bj4500619add.htm). However, given that PI3K modulates the levels of spermidine, but not spermine (Figure 3), we focused on spermidine for subsequent experiments.
In MUT cells, the addition of spermidine (1 μM) to the medium reduced Ser473 phosphorylation on PKB/Akt approximately 3-fold (Figure 4A). We also modulated endogenous polyamines in cells using ODC-I , a commercial inhibitor of ODC, the first and rate-limiting enzyme of the polyamine pathway . As expected, addition of ODC-I to cells decreased the levels of cellular polyamines (Supplementary Figure S4 at http://www.biochemj.org/bj/450/bj4500619add.htm). ODC-I also increased PKB/Akt phosphorylation 1.5–2-fold in both WT and MUT cells (Figure 4B). The spermidine (1 μM)-induced reduction in PKB/Akt phosphorylation could be reversed by co-treatment with ODC-I (Figure 4C). Taken together, these results indicate that spermidine levels regulate Akt phosphorylation in MUT and WT cells.
In order to investigate whether spermidine affects kinase signalling other than PI3K/Akt, we performed an untargeted LC-MS/MS-based phosphoproteomic analysis in MUT cells treated with spermidine. We identified and quantified 1009 phosphopeptides, of which 69 and 432 showed significantly (P<0.05; >±2-fold) altered intensities upon 1 and 16 h treatment with spermidine respectively (Figure 4D). Proteins whose phosphorylation was modulated by spermidine belonged to different ontologic classes (Supplementary Table S1). Ontologies increased upon 16 h of spermidine treatment included those related to the ribosome and translation, whereas ontologies related to the cytoskeleton, serine/threonine protein kinase activity, and proliferation were enriched in proteins whose phosphorylation was decreased after 16 h of spermidine treatment (Supplementary Table S1). Supplementary Figure S5 (at http://www.biochemj.org/bj/450/bj4500619add.htm) shows the identity of proteins that belong to ontologies significantly enriched in our analysis and Figure 4(E) shows phosphorylation sites on serine/threonine protein kinases that were modulated by spermidine. Consistent with spermidine decreasing the activity of the PI3K/Akt pathway (Figures 4A–4C), sites on PAK [p21 protein (Cdc42/Rac)-activated kinase] 1 and PAK2 were reduced upon treatment of cells with this polyamine (Figure 4E). Depending on the system under investigation, PAK isoforms can be downstream or upstream of PI3K/Akt or MAPK (mitogen-associated protein kinase) signalling pathways [40,41] and are involved in the onset and progression of different cancers including colon . Taken together, the phosphoproteomics data indicate that spermidine has pleiotropic effects on kinase signalling, with PI3K/Akt being one of the pathways modulated by this polyamine.
PI3K promotes proliferation of DLD-1 colon cancer cells by regulating spermidine production
Both PI3K/Akt signalling and the polyamine metabolic pathway are established regulators of cancer cell proliferation [8,36]. Therefore our next question asked whether spermidine may mediate PI3K-dependent cell proliferation and survival. This possibility was investigated by assessing whether exogenously added spermidine, which readily enters cells , would affect the proliferative/survival effects of PI3K. As expected, incubation of cells with PI-103 reduced the viability of both MUT and WT cells, as measured by MTS (Figure 5A) and cell counting (Figure 5B). Interestingly, these effects of PI3K were partially reversed by the addition of 1 μM spermidine (relevant data points are marked with arrows in Figures 5A and 5B).
PI3K inhibition also induced cell death in both WT and MUT cells, as measured by increased PARP cleavage and a Guava assay, an effect that was also partially reversed by spermidine in MUT cells (Figure 6A), but not in WT cells, possibly due to the lower endogenous levels of polyamines in the latter compared with MUT cells (Figure 3). PI3K inhibition also increased autophagy, but this effect was not reversed by spermidine (Supplementary Figure S6 at http://www.biochemj.org/bj/450/bj4500619add.htm). Collectively, these results are consistent with the notion that the control of cell proliferation and survival exerted by PI3K are mediated, at least in part, by spermidine.
Inhibitors of PI3K and ODC potentiate each other in reducing DLD-1 colon cancer cell proliferation
The results of the present study indicate that the PI3K/Akt pathway in MUT and WT cells controls the production of spermidine, which in turn promotes cell proliferation and survival (Figures 5 and 6). This is despite an inhibitory effect of spermidine on PKB/Akt phosphorylation (Figure 4), which would be expected to result in an inhibition of proliferation/survival . These data suggest the existence of a negative feedback loop by which PI3K/Akt and polyamine biochemical pathways control each other's activities (Figure 6B). In line with this hypothesis, co-treatment of cells with PI-103 and ODC-I, at concentrations that by themselves did not reduce cell viability, resulted in a 30% decrease in cell viability of MUT cells as measured by MTS, but not WT cells (Figure 7A). Measuring cell proliferation by direct cell counting revealed that PI3K and ODC inhibition had an additive effect (Figure 7B). As expected, inhibition of either PI3K or ODC induced apoptosis (as measured by PARP cleavage) in both MUT and WT cells, an effect that could be enhanced by co-treatment with both ODC-I and PI3K inhibitors, to a greater extent in WT cells than in MUT cells (Figure 8A). The differences in the behaviour of MUT and WT cells may be attributed to the fact that WT cells have lower levels of endogenous spermidine than MUT cells under basal conditions (Figure 3), and due to differences in PI3K activation between the two cell lines.
In order to test whether polyamine and PI3K pathways co-operate in tumour growth in vivo, xenografts of MUT cells in nude mice were treated for 7 days with sub-optimal concentrations of GDC-0941 (pan-class I PI3K inhibitor), DMFO (ODC-I inhibitor) or both. These compounds were used because they have better in vivo pharmacological properties than PI-103 and ODC-I. The masses of the tumours after treatment and post mortem excision were not significantly reduced by treatment by both GDC-0941 and DMFO (Supplementary Figure S7A at http://www.biochemj.org/bj/450/bj4500619add.htm) and no differences in PKB/Akt phosphorylation were observed between treatments (Supplementary Figure S7B), indicating that by the end of treatment, GDC-0941 and DMFO were no longer efficient at pathway inhibition. However, although each tested compound in isolation slowed tumour growth relative to vehicle-treated mice (Figures 8B–8D), co-treatment with both inhibitors effectively reduced tumour volume in half of the cohort (Figure 8E).
PI3K signalling and polyamine metabolism are two biochemical pathways with pleiotropic cellular functions. The expression of ODC, the rate-limiting enzyme in polyamine production, was found to be regulated by the proto-oncogene c-Myc [42–44], and to be indispensable for mammalian development  and cancer cell proliferation [46–48]. Similarly, PI3K is one of the most frequently deregulated pathways in several cancer types [9,10,49]. In the present study we have found a previously unrecognized function of PI3K signalling in modulating polyamine metabolism.
The connection between polyamine synthesis and PI3K was uncovered in the present study using an unbiased LC-MS approach that identified MTA as a metabolite downstream of PI3K. Since MTA is a byproduct of polyamine metabolism, these results led us to speculate that other molecules in the polyamine biosynthetic pathway would also be downstream of PI3K signalling. We thus designed SRM assays to quantify the polyamines in WT and MUT cells treated with a PI3K inhibitor. These results confirmed that the polyamines spermidine and putrescine, but not ornithine or spermine, were downstream of PI3K. These observations are consistent with published results showing that a reduction in spermidine and putrescine as a result of inhibiting ODC in mice is not associated with a reduction in spermine levels  and with the predictions of mathematical modelling of polyamine metabolism showing that ODC inhibition does not significantly affect spermine production .
Spermidine had pro-survival and proliferative effects, and partly reversed the inhibition of proliferation and the induction of apoptosis caused by PI3K inhibition (Figures 5 and 6), despite the observation that this polyamine reduced the activity of the PI3K/Akt pathway (Figure 4). These observations support the hypothesis that spermidine acts downstream of PI3K to induce cell proliferation and survival.
Our data also point to the existence of a feedback loop by which spermidine and PI3K control their reciprocal activities (Figure 6B) and suggest a mechanism for the homoeostasis of spermidine levels and cell proliferation. In said mechanism, spermidine would promote cell proliferation and survival downstream of PI3K (Figure 5 and 6) by its known roles in modulating cellular processes such as translation . However, when its concentration increases above a certain threshold, spermidine abrogates kinase signalling pathways (Figure 4), leading to a reduction in spermidine levels in cells (Figure 3D), and a concomitant reduction in cell proliferation. When spermidine levels fall below the proposed threshold, kinase signalling would be activated, promoting spermidine production again and thus maintaining its cellular concentrations, and the ability of cells to proliferate, within a constant range. This mechanism explains the observation that relatively low concentrations of exogenous spermidine have proliferative effects, whereas it is cytotoxic at high concentrations ([8,36,52], and Supplementary Figure S3, i.e. at concentrations in which kinase signalling is completely inhibited). The proposed mechanism is also consistent with the results of the present study showing global modulation of protein phosphorylation by spermidine (Figure 4) and by reports documenting that polyamines modulate the activity of specific kinase pathways [37,53].
The existence of a PI3K/Akt/polyamine feedback mechanism led us to hypothesize that a functional consequence of inhibiting both PI3K and ODC would be a reduction in cell viability to a greater extent than if either pathway was inhibited in isolation. Consistent with this hypothesis, PI3K and ODC inhibitors had additive effects in reducing in vitro cell proliferation (Figure 7 and 8) and potentiated each other in reducing in vivo tumour growth (Figure 8). Indeed, although GDC-941 and DMFO slowed down tumour growth, the combination of both compounds had a dramatic effect on the reduction in tumour volume in the first 7 days of treatment, with 50% of tumours showing a regression in volume as a result of co-treatment. These results suggest that a drug combination that targets both PI3K and polyamine pathways may be more effective than inhibiting either pathway by itself. These data, together with the observation that spermidine has an impact on ~40% of the phosphoproteome (Figure 4D), also indicate that the extent by which metabolic pathways feedback to modulate oncogenic signalling may be greater than previously appreciated.
Vinothini Rajeeve performed the experiments, analysed the data, prepared the Figures and contributed to writing the paper. Wayne Pearce performed mouse experiments. Marta Cascante contributed to obtaining funding, interpreted the results and edited the paper prior to submission. Bart Vanhaesebroeck contributed to obtaining funding and edited the paper prior to submission. Pedro Cutillas conceived the study, contributed to obtaining funding, designed the experiments, analysed the data, interpreted the results, prepared the Figures and wrote the paper.
This project was funded by Cancer Research U.K. [grant number C27327/A9914] and Barts and the London Charity [grant number 297/997]. W.P. and B.V. received support from Cancer Research U.K. [grant number C23338/A10200]. M.C. was funded by Spanish Government and Feder Funds [grant numbers SAF2011-25726 and ISCIII-RTICC, RD06/0020/0046] and an Icrea Academia award 2010.
We thank Dr Victor Velculescu for isogenic cell lines, Ed Wilkes and Dr Pedro Casado for feedback, Alex Montoya and Dr Essam Ghazaly for technical assistance, and Dr Imma Berenjeno and Dr Khaled Ali for advice on the in vivo experiments. B.V. is an advisor to GlaxoSmithKline (Stevenage, U.K.) and Activiomics (London, U.K.). P.R.C. is an advisor to Activiomics (London, U.K.).
Abbreviations: DFMO, α-difluoromethylornithine; FBS, fetal bovine serum; LC, liquid chromatography; MS/MS, tandem MS; MTA, 5-methylthioadenosine; MTAP, S-methyl-5′-thiadenosine phosphorylase; MTS, 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium; ODC, ornithine decarboxylase; PAK, p21 protein (Cdc42/Rac)-activated kinase; PARP, poly(ADP-ribose) polymerase; PCA, principal component analysis; PI3K, phosphoinositide 3-kinase; PKB, protein kinase B; PTEN, phosphatase and tensin homologue deleted on chromosome 10; RTK, receptor tyrosine kinase; SRM, single reaction monitoring; WT, wild-type
- © The Authors Journal compilation © 2013 Biochemical Society