Nutrient withdrawal rescues growth factor-deprived cells from mTOR-dependent damage

Nutrient restriction protects 293T Phoenix cells from death by serum deprivation

Most immortalized cell lines undergo mitotic catastrophe and cell death with morphological and biochemical features of apoptosis when deprived of fetal calf serum or growth factor supply [20]. Upon serum withdrawal, 293-T Phoenix cells, a retrovirus packaging line derived from E1A-transformed embryonic human kidney cells (HEK-293) carrying a temperature sensitive T antigen, displayed severe and time-dependent loss of viability, as revealed by a Propidium Iodide uptake assay (Figure 1A). Nearly 100% of cells appeared dead by day 4 of culture (96 hours) (Figure 1A). Remarkably, removal from the culture medium of either Glucose or Aminoacid Supplement (Glutamine + DMEM Non Essential Aminoacids), the two main energy fuels for most cultured transformed cells [21], resulted in a drastic protection from cell death. Typically we detected a maximum of mortality of up to 30% at day 4 under glucose deprivation, and below 10%, comparable to average mortality in the presence of serum (Figure 1B), for aminoacid-starved cultures.Reduction of glucose from high (4.5 g/l) to low (1 g/l) concentration had no significant effect on cell viability, indicating that even physiological concentrations of glucose promote death of Phoenix cells in the absence of serum. Simultaneousremoval of glucose and aminoacid supplement from the culture medium resulted in rapid (12 hours) loss of viability, in a fashion which could not be prevented by addition of Pyruvate, Dimethyl-Succinate or Free Fatty Acids (not shown); this confirms that glucose and glutamine account for most of the energy supply for these cells, at least in the tested experimental conditions.

Live cell count revealed that Phoenix cells continue proliferating robustly in the absence of serum, and are therefore, at least in part, self-sufficient for mitogenic stimulation. Cell proliferation and death appear to occur concomitantly (Figures 1A and 1C), and are likely to be mechanistically linked [22]. Proliferation also occurred, although to a lesser extent, in nutrient deprived cultures, yet associated with no or minimal cell loss (Figures 1A and 1C).

Beneficial effect of nutrient restriction on cell viability prompted us to evaluate the consequence of pharmacological interference with cellular metabolism. As expected, the glycolysis inhibitor 2-deoxyglucose fully rescued cells from death in the presence of glucose, to an even larger extent than glucose deprivation (Figure 1B). Similarly, significant protection was obtained by interference with mitochondrial respiration: in fact, both complex I inhibitor Rotenone and complex II inhibitor 3-Nitropropionic acid (NPA) drastically reduced death of serum-deprived cultures. Also the uncoupling agent 2,4-dinitrophenol(2,4-DNP), at non toxic concentration, had the same protective effect as mitochondrial inhibitors on cell survival in 2 g/l glucose (Figure 1B); noteworthy, both DNP and electron transport chain (ETC) blockers rapidly killed Phoenix cells in the absence of glucose (not shown), indicating that mitochondria are functional in this cell line and support energy demand when glycolysis is prevented.

In order to evaluate the impact of nutrient restriction on the energy balance of Phoenix cells, ATP content was measured 48 hours after cell transfer to the different culture media. As expected based on survival data, no drastic reductions in cellular ATP levels were observed upon nutrient withdrawal (Figure 1D). Glucose deprivation led to a modest (about 20%) decrease of cellular ATP, and aminoacid removal to no reduction at all, compared to standard growth medium (2 g/l glucose and aminoacid supplement). ATP reduction was more pronounced (about 50%) in cells treated with Rotenone (Figure 1D), indicating that mitochondria contribute significantly to ATP generation in this tumor cell line.

Thus, survival of Phoenix cells in serum free medium is clearly subdued to a metabolic regulation by nutrient availability, that operates independently from severe changes in cellular energy levels.

Nutrient toxicity in serum-deprived Phoenix cells is not mediated by ROS

Cell death by serum withdrawal is associated with the formation of harmful reactive oxygen species (ROS) [20], and nutrients may generate ROS through their oxidation in mitochondria [23]. Since nutrient restriction or mitochondrial blockade rescued Phoenix cells from serum deprivation, we tested possibility that cell protection might be mediated by an attenuation of cellular oxidative stress. To this end, Phoenix cells were transiently transfected with a redox-sensitive variant of the yellow fluorescent protein (rxYFP) and the intracellular redox state evaluated by confocal microscopy and fluorescence ratiometric analysis, 24 hours after serum or serum and glucose deprivation. RxYFP consistently appeared more reduced (as indicated by higher values of the Ratiometric Index R) in glucose-fed than in glucose-starved cells, revealing significantly higher levels of ROS in the latter cell population (Figures 2A, a and b). This finding was further supported by evidence of higher content of reduced NAD(P)H in glucose-fed cultures, as determined by cell microfluorimetry (Figure 2A, c). No significant redox changes were observed in cells deprived of Glutamine and NEAA or exposed to the mTOR inhibitor Rapamycin. As expected, addition of FCS further reduced the intracellular environment in glucose-fed cells (Figure 2A, b).

Based on these findings, excess oxidative stress unlikely accounts for impaired cell viability by nutrients. In keeping with this conclusion, no major changes in cell viability were induced, in the presence or absence of glucose, by saturating concentration of the ROS scavenger and glutatione precursor N-acethyl-cysteine (NAC, 10 mM) (Figure 2B, a). Similarly, overexpression of the ROS scavengers Catalase (Figure 2B, a) and SOD2 (Figure 2C and 2D, b) did not provide glucose-fed cells protection from death, nor affected cell viability in glucose free-medium. Notably, overexpression of Catalase effectively increased cell antioxidant capacity, as revealed by flow cytometry of cells loaded with the redox-sensitive dye Dichlorofluoresceine Diacetate (H2-DCF-DA) and exposed to a bolus of exogenous hydrogen peroxide (Figure 2B, b). Finally, overexpression of the class III deacetylase Sirt-1, a molecule linking, in model organisms and in mammalian cells, nutrient restriction to increased resistance to oxidative stress [24], did not rescue cells from glucose-induced death in serum-free medium (Figures 2C and 2D, a). Thus, collectively, these data suggest that generation of ROS and oxidative stress do not mediate the effects of glucose on cell viability in our experimental model. Additionally, failure of Sirtuin-1 to prevent or attenuate glucose-induced cell death indicates that this major nutrient sensor and regulator of cell survival is unlikely involved in the protective response of Phoenix cells to nutrient restriction.

Blockade of mTOR prevents nutrient-induced cell death

Since glucose and aminoacid withdrawal provided comparable protection to serum-starved Phoenix cells, in spite of having different effects on cell energy (Figure 1D) and redox balance (Figures 2A and B and data not shown), we reasoned that a common signaling mechanism might underlie the antiapoptotic action of the two starvation modes. The mTOR/S6K signaling cascade, which is modulated by both glucose and aminoacids and regulates cell proliferation and survival [6], was therefore evaluated as a potential candidate.

Even in the absence of exogenous growth factors, mTOR activity remained remarkably elevated in glucose-fed cells 24 hours after serum withdrawal, as revealed by the phosphorylation patterns of the major mTOR effectors S6 kinase and 4E-BP1, and of the downstream substrate S6 (Figure 3A, lane 1). Note that in this analysis phospho-site specific antibodies often recognize multiple bands, the uppermost, slowest-migrating one generally representing the most heavily phosphorylated form of the protein (see arrows) [25]. Based on this criterion, we observed a marked reduction of mTOR activity in glucose-starved, and to an even larger extent, in aminoacid-starved cells (Figure 3A, lanes 2 and 3). A drastic reduction in S6 kinase phosphorylation was also observed in glucose-fed cells treated with mitochondrial inhibitors or with the uncoupler 2,4-DNP, in keeping with the starvation-mimicking effects of these treatments on cell survival (Figures 3B and 1C). In glucose-starved cells we also observed a small increase in the phosphorylation of the AMP-activated protein kinase (AMPK-α) (Figure 3A), the putative negative regulator of mTOR in this experimental condition [8]. This modest, although detectable biochemical change, which reflects the small reduction in cellular ATP content reported in figure 1 D, likely accounts for reduced mTOR signaling (Figure 1B) [8] in Phoenix cells grown in the absence of glucose. Thus, collectively, these observations confirm that mTOR is responsive to glucose and aminoacids, and that its activity positively correlates with nutrient availability and extent of cell death in serum-deprived Phoenix cells.

In order to address the role of the mTOR/S6K cascade in nutrient-dependent death of Phoenix cells, we evaluated the effect of mTOR blockade on cell viability in standard and glucose-depleted medium. Rapamycin, a macrolide antibiotic widely used as an immunosuppressive drug, directly inhibits mTOR activity within the nutrient sensitive TORC1 complex, by complexing with the cellular protein FKBP12; another drug, 5-aminoimidazole-4-carboxamide ribo- nucleoside (AICAR), indirectly suppresses mTOR signaling through AMP kinase, by mimicking cell de-energization and accumulation of adenosine mono-phosphate (AMP) [26]. As expected, both drugs drastically decreased the phosphorylation of the mTOR substrate S6 kinase in cells grown in the presence of both glucose and aminoacids (Figure 4A). More importantly, both Rapamycin and AICAR dramatically reduced cell death in nutrient repleted medium (Figure 4B).

In order to rule out potential non-specific effects of drug compounds, in a parallel series of experiments mTOR expression in Phoenix cells was genetically inactivated by shRNA technology. As displayed in figure 3C, lentiviral expression of a mTOR specific shRNA resulted in a substantial reduction of the mTOR expression level, and in a reduced phosphorylation of its downstream targets S6K, S6 and 4E-BP1, in nutrient-rich samples (Figure 4C, compare lanes 1 and 4 in each panel). In keeping with evidence of cell protection by Rapamycin and AICAR, mTOR inactivation allowed a higher percentage of cells (about 50%) to survive in nutrient repleted medium with respect to mock-infected cells, and nearly abolished protection by glucose withdrawal (Figure 4D). It should be noted, however, that here and in general in experiments involving genetic manipulation of Phoenix cells, mortality in the absence of nutrients was often higher than the usual (compare Figure 1A with 4D and S2,a), possibly due to cellular distress from the experimental procedure. Notwithstanding this limitation, survival data with mTOR-silenced cells confirm the observations made with chemical inhibitors, demonstrating that activation of the mTOR cascade is instrumental to nutrient-triggered cell death in the cell line under study, and that protection by nutrient restriction is conceivably mediated by the inhibition of this cascade.

Akt is activated by mTOR inhibition but does not account for cell protection While in most cancer-related models the mTOR cascade exerts antiapoptotic functions downstream of the PI3 kinase/AkT PKB signaling axis [27], few examples of cell protection by inhibition of mTOR/S6K have been reported [28-31]. It is also known that hyperactivation of the mTOR cascade can downregulate survival signaling by the AkT/PKB kinase [15,16]. In search for a molecular mechanism linking nutrient-dependent mTOR signaling to massive cell death of serum-deprived Phoenix cells, we sought to evaluate the phosphorylation of AkT at serine 437, a biochemical correlate of AkT kinase activity. Consistent with previous reports, we found increased levels of AkT phosphorylation/activity in cells deprived of glucose or treated with Rapamycin, in a fashion which inversely correlated with activation of the mTOR effector S6 kinase (Figure 5 A, a). mTOR-silenced cells also displayed increased phosphorylation of AkT in nutrient rich medium, although to a lower extent compared to control cells treated with Rapamycin (Figures 5A, a and b); moreover, transfection of rat mTOR cDNA in cells deprived of human mTOR rescued mTOR expression and activity (as assessed by phosphorylation of S6) and in parallel decreased the phosphorylation of AkT (Supplementary Figure 1). Thus, taken together, these observations confirmed that, in serum-deprived Phoenix cells, nutrients downregulate AkT phosphorylation/activity through the mTOR cascade. This raises the possibility that increased AkT function might be responsible, at least in part, for the dramatic protection provided by restriction of glucose or aminoacid supply in this experimental model.

To further investigate the mechanistic role of the PI3K-AkT cascade in cell survival induced by nutrient deprivation, cells were treated with the PI3 kinase specific inhibitor LY294002, and exposed to normal or glucose-deprived media in the absence of serum. Surprisingly, the PI3K inhibitor failed to reverse cell protection by glucose withdrawal, but attenuated cell death in nutrient-repleted medium (Figure 6A, a), suggesting that residual AkT activity may be detrimental rather than protective in this culture condition. Interestingly, biochemical studies revealed that, while AkT/PKB phosphorylation was, as expected, decreased, also the serine phosphorylation of S6 was strongly downregulated. This is an evidence that inhibition of the mTOR cascade, a downstream target of AkT [10], accompanied PI3K blockade (Figure 6A, b).

In a complementary series of experiments, over-expression of a constitutively active form of AkT (myrAkT) in Phoenix cells failed to prevent cell death in nutrient rich medium, while slightly decreasing cell protection by glutamine withdrawal (Supplementary Figure 2).

Collectively, these data do not support a role for AkT in cell survival by nutrient restriction in our cell model, but rather indicate that protection operates also in the context of PI3K (and AkT) inhibition, provided that the mTOR cascade is also blocked. Conversely, AkT appears to increase cell death in the presence of abundant nutrients, and to negatively interfere with cell protection by Glutamine deprivation.

mTOR inhibition attenuates hyperglycemic damage in primary endothelial cells

In an attempt to verify that mTOR-dependent nutrient toxicity is not restricted to one single transformed cell line, we cultivated primary human endothelial cells (HUVEC) in high (4.5 g/L) ambient glucose, a well established model of endothelial hyperglycemic damage. Specific endothelial growth factors (EGF, FGF-B, VEGF and IGF-1), normally required for the optimal propagation of these cells, were omitted from the culture medium, while FBS (5%) was included to limit cellular stress. In these harsh conditions, a majority of cells detached from the plate and appeared dead after 48-60 hours of incubation, the percentage of live cells (quantified by flow cytometry as the percentage of cells with high forward scatter, low side scatter profile) ranging from 10 to 15%. Cell survival, however was significantly improved (nearly doubled) by Rapamycin, to an extent even larger than by cultivation in normal (0.9 g/L) glucose (Figure 6A, a). Importantly, these differences matched the phosphorylation level of S6, an index of mTOR activity (Figure 6B). Likewise, mTOR knock-down by lentivirus-delivered shRNA consistently increased cell survival by about 20%, in accordance with the evident although incomplete inhibitory effect on mTOR signaling (Figures 6A, a and 6B).

Additionally, accumulation of O-GlcNacylated proteins, a biochemical hallmark of endothelial damage by high glucose [2,32], was drastically reduced by Rapamycin and, although to a lesser extent, by mTOR knock-down (Figure 6B).

Thus, inhibition of the mTOR cascade partially rescues primary human endothelial cells from hyperglycemic damage under growth factor restriction, confirming and extending analogous findings obtained in Phoenix cells.

R= (F-F_ox)/(F_red-F_ox)

and ranging between 0 (complete oxidation) and 1 (complete reduction), by means of a dedicated software generated through the Labview 7.1 interface [23]. For image quantitation, average R values were determined within multiple Regions of Interest (ROIs, single cells or small cell clusters) for each sample, and their Mean±SD (n = 7 to 9) determined and utilized for further statistical analysis (Student t-test). In some experiments, after initial cell imaging in nutrient-deficient medium, nutrients were added back and cell redox responses monitored for 30 minutes or longer.

Intracellular NAD(P)H was measured, in the same experimental settings as above, by two-photon confocal analysis of cell green autofluorescence after two-photon excitation at 366 nm, as described by Patterson et al. [44].

Biochemical studies. For protein phosphorylation studies Phoenix cells were plated at 1.5 x 10⁵/well in 12-well plate and incubated for 16 to 24 hours in complete medium. The day after, cells were switched to DMEM without glucose, glutamine/NEAA and FCS, and these components were added back where necessary. Serum-free samples were given BSA 20 mg/ml (stock) at the same dilution as FCS (typically 10%, i.e. 2 mg/ml final concentration). Antioxidants and chemical inhibitors were also added at this stage of the experiment (see above, viability assay). After 24 hours supernatants were removed and cells lysed in 100 microliters of ice-cold lysis buffer (NaCl 150 mM, Tris-hcl 50 mM pH 8; 2 mM EDTA) containing 1% v/v Triton X-100, 0.1 % v/v SDS, 1:1000 Protease Inhibitor cocktail (Sigma), 1 mM Sodium Orthovanadate, 1 mM NaF; 2 mM β-glycerophosphate. After 15 minutes on ice with occasional vortexing cells were spun down at 14,000 rpm, 4°C to remove debris and unlysed cells, and supernatant quantified for protein content (DC Protein Assay, BIORAD), resuspended in 6X Laemmli buffer, boiled for 2 minutes and stored at -80°C or directly loaded onto denaturing discontinuous polyacrylamide gels for SDS-PAGE. Proteins were then electroblotted onto nitrocellulose membrane (PROTRAN®, Whatman, Dassel, Germany). After reversible Ponceau S staining to confirm protein transfer and equal loading throughout the lanes, membranes were blocked in TBS-T containing 5% skim milk. Antisera were added in 3% milk at the appropriate dilution and incubated for 16 hours on a rotating plate at 4°C. After extensive wash in TBS-T, immuno-complexes were visualized by incubation with HRP-conjugated secondary reagents (BIORAD) followed by enhanced chemoluminescence (ECL, GE Healthcare, Milan, Italy) and autoradiography. In some experiments autoradiograms were digitalized and band intensity (band volume, i.e. area x mean pixel intensity) quantified with a dedicated software (Quantity One, BIORAD). Quantitation was normally not performed when differences displayed were immediately evident. Occasionally membranes were stripped in 2% SDS at 60°C, washed, blocked and subdued to a second round of hybridization.

For protein O-glycation and phosphorylation studies on HUVEC, cells were handled as for viability assays, except that after 24 hours of incubation supernatants and floating cells were removed and adherent cells lysed as described above.

Accumulation of GlcNAcylated proteins was determined by immunoblotting using a specific anti O-GlcNAc monoclonal antibody (CTD 110.6, COVANCE [45]).

Catalase assay. As an indirect assessment of intracellular catalase activity, cells were switched to serum-free medium (without BSA), loaded for 30 minutes with the redox sensitive fluorescent dye H2-Dichlorofluorescein Diacetate (H2-DCF-DA) and challenged with 1 mM hydrogen peroxide for 15 minutes. Cells were then quickly transferred to tubes for flow cytometry and green fluorescence (Fl-1) quantified. Resistance to H₂O₂-induced cell oxidation was assumed to correlate with cell capacity to degrade hydrogen peroxide. Extracellular catalase and the catalase inhibitor Aminotriazol were used to validate this procedure.

Determination of intracellular ATP . ATP was quantified by chemiluminescence using a dedicated kit (ENLITEN^® ATP Assay, PROMEGA, Milan, Italy) according to the manufacturer's recommendations. For each sample luminescent emission was normalized for total protein content, determined as described above.

Statistics. Data sets (usually triplicate culture wells) were compared by the two-tailed Student's t-test for independent samples. Threshold for statistical significance was set at p<0.05