The metabolic waste ammonium regulates mTORC2 and mTORC1 signaling

Two structurally and functionally distinct mammalian TOR complexes control cell growth and metabolism in physiological and pathological contexts including cancer. Upregulated glutaminolysis is part of the metabolic reprogramming occurring in cancer, providing fuels for growth but also liberating ammonium, a potent neurotoxic waste product. Here, we identify ammonium as a novel dose-dependent signal mediating rapid mTORC2 activation and further regulating mTORC1. We show that ammonium induces rapid RICTOR-dependent phosphorylation of AKT-S473, a process requiring the PI3K pathway and further involving the Src-family kinase YES1, the FAK kinase and the ITGβ1 integrin. Release of calcium from the endoplasmic reticulum store triggers rapid mTORC2 activation, similar to ammonium-induced activation, the latter being conversely prevented by calcium chelation.Moreover, in analogy to growth factors, ammonium triggers the AKT-dependent phosphoinhibition of the TSC complex and of PRAS40, two negative regulators of mTORC1. Consistent with mTORC1 stimulation, ammonium induces the inhibitory phosphorylation of 4EBP1, a negative regulator of protein biogenesis. Ammonium however dually impacts on the phosphorylation of p70S6K1 triggering a transient AKT-independent decrease in the phosphorylation of this second mTORC1 readout. Finally, we reveal ammonium as a dose-dependent stimulator of proliferation. This study underscores an mTORC2 and mTORC1 response to the so-called ammonium waste.

TOR (target of rapamycin) is a serine/threonine kinase conserved in all eukaryotes that integrates extra and intracellular signals with cell growth 1 . Mammalian TOR is part of two distinct complexes. mTORC1 responds to nutrients, energy levels and growth factors and stimulates translation and anabolic metabolism while inhibiting autophagy. In contrast little is known about mTORC2 regulation. The latter complex responds mainly to growth factors via PI3K signaling, phosphorylates members of the AGC kinases family, including AKT, all linked to cancer and diabetes 2,3 . Though 'cancer' represents a puzzling number of at least 100 diseases, a set of common characteristics indicate important shared metabolic variations 4,5 . In many cancers, genetic alterations lead to constitutive activation of mTOR signaling, impacting on tumor cell metabolism 1,6 . One aspect of this metabolic reprogramming concerns the enhanced rate of glutamine transport and the upregulated catabolism of this amino acid, two processes currently targeted by therapeutic strategies 4,7,8 . In the mitochondria, glutamine is first deaminated by glutaminase (GLS) to glutamate that can in turn be converted to α -ketoglutarate, feeding the TCA cycle 5 . Increased glutaminolysis has been linked to oncogenic levels of Myc through a coordinated transcriptional program enhancing GLS expression 9 . High rate glutaminolysis is also expected to liberate high levels of ammonium. NH 4 + is in pH-dependent equilibrium with NH 3 and at pH7 for instance the ionic molecule largely predominates (> 98%) over the neutral form. Hereafter, the term 'ammonium' will refer to the sum of NH 4 + and NH 3 unless a molecular distinction is required and 'ammonia' will refer to NH 3 . Accordingly, the seminal work by Chance and collaborators reported elevated ammonium levels in the blood of tumor-bearing rats 10,11 . Ammonium concentrations of up to 5 mM were subsequently found in the interstitial fluids from human tumor xenografts 12 . Although the importance of glutamine as a tumor nutrient is recognized since the 1950's, the potential side-consequences of ammonium produced by glutaminolysis in tumors have been poorly documented. Ammonia was proposed to behave as an autocrine and paracrine diffusible factor stimulating autophagy in tumor masses through an

Results
Ammonium induces dose and time dependent AKT S473 phosphorylation. To highlight potential signaling pathways responding to ammonium, we probed a phospho-kinase array with lysates of MCF-7 breast cancer cells treated or not with 5 mM NH 4 Cl for 30 minutes (Fig. 1A). We used a 5 mM NH 4 Cl concentration in reference to the highest ammonium concentration measured in tumor xenografts, a concentration that is reported to induce the autophagic process 12 . Of note, the latter concentration is far below those often used in vitro to inhibit the lysosomal function [22][23][24] . In these conditions, we detected a prominent increase in the phosphorylation of AKT-S473. In addition, we also noted an increased ERK1/2 phosphorylation (Fig. 1A). The latter ERK phosphorylation is consistent with previous data obtained in a phosphoproteomic analysis 14 . We decided to focus on the analysis of the uncharacterized AKT-S473 phosphorylation.
We confirmed by western blot that 5 mM NH 4 Cl induces the S473 phosphorylation of the AKT/PKB readout in MCF-7 cells. We similarly observed AKT phosphorylation in U2OS osteosarcoma, HepG2 hepatoma and SW480 colon adenocarcinoma cells (Fig. 1B). Moreover, AKT-S473 phosphorylation was also observed in the BJ-1 fibroblast cell line (Fig. 1B). These data indicate that 5 mM NH 4 Cl mediates common signaling pathways activation in cancer and fibroblast cell lines. NH 4 Cl-induced AKT phosphorylation occurred rapidly, with a maximal immunodetection at the earliest time-point of 5 min treatment and tended to decrease at 240 min (Fig. 1C). AKT phosphorylation was immunodetected with NH 4 Cl concentrations far lower than 5 mM, already visible at 0.5 mM, suggesting the stimulation of a sensitive pathway rapidly responding to low NH 4 Cl concentrations, far below those inhibiting the lysosomal function (Fig. 1D). AKT phosphorylation was observed for NH 4 Cl concentrations higher than 5 mM as well (Fig. 1E). Eng and co-workers showed that ammonium supplied as either NH 4 Cl or NH 4 OH was efficient in stimulating autophagy, while other molecules as NaOH, CO 2 , NO, H 2 O 2 and lactic acid were inefficient 12 . Here, use of NH 4 OH, instead of NH 4 Cl, mediated a similar AKT-S473 phosphorylation both in terms of dose and time-dependencies ( Fig. 1F and G). These data indicate that ammonium supplied as either NH 4 Cl or NH 4 OH leads to the induction of AKT-S473 phosphorylation in a time and dose dependent manner.
Ammonium-induced AKT phosphorylation is dependent on mTORC2. We next dissected the upstream pathway leading to the rapid induction of AKT phosphorylation occurring upon NH 4 Cl supplementation. PI3K is largely described to transduce extracellular signals to activate AKT kinase 25 . Pretreatment of MCF-7 cells with LY294002 or Wortmannin to inhibit PI3K, impaired basal AKT phosphorylation level and prevented its induction by NH 4 Cl addition ( Fig. 2A). The mTORC2, composed of mTOR, RICTOR, SIN1, PROTOR and DEPTOR, has been shown to promote cell proliferation and survival by phosphorylation and activation of the AGC kinase family members AKT and SGK 26 . In contrast to mTORC1, little is known about the signals regulating mTORC2. The phosphorylation of AKT at Ser-473 is so far the best characterized readout for mTORC2 activity 26 . We show that upon siRNA-mediated knockdown of RICTOR expression, the basal and the NH 4 Cl-induced AKT-S473 phosphorylation were strongly reduced (Fig. 2B). Moreover, NH 4 Cl was found to induce T346 phosphorylation of NDRG1, a specific target of SGK1, another reliable readout of mTORC2 signaling 26 , and the latter phosphorylation was dependent on mTORC2 (Fig. 2C,D and E). These data together support that ammonium stimulates PI3K/mTORC2/AKT signaling.
Ammonium-induced mTORC2 activation is dependent on YES1 and FAK kinases and ITGβ1 integrin. We next addressed how ammonium activates mTORC2. It is reported that epidermal growth factor receptor (EGFR) stimulates mTORC2 kinase activity and signaling 27 . The use of pharmacological inhibition of EGFR or the downstream MEK activity with Gefitinib and U0126 respectively indicated that NH 4 Cl-induced AKT-S473 phosphorylation occurred independently of EGFR or MEK kinases ( Fig. 3A and B). In contrast, inhibiting SFK kinases by pretreatment of cells with PP2 reduced basal and NH 4 Cl-induced AKT-S473 phosphorylation ( Fig. 3C), revealing a SFK requirement for NH 4 Cl-induced mTORC2 activation. Specific siRNA-mediated knockdown of SRC kinase did not impair NH 4 Cl-induced AKT-S473 phosphorylation (Fig. 3D). In contrast, knockdown of YES1 kinase decreased basal and both NH 4 Cl-induced AKT-S473 and NDRG1-T346 phosphorylation (Fig. 3E). Moreover, pharmacological inhibition with PF573228 of the focal adhesion kinase (FAK), reported to interact with mTORC2 but not mTORC1 28 , reduced NH 4 Cl -induced AKT-S473 and NDRG1-T346 phosphorylation ( Fig. 3F and G). These results together indicate that YES1 and FAK kinases are involved in the NH 4 Cl-induced AKT phosphorylation. We also addressed whether integrins, known regulators of FAK signaling, could be involved in the NH 4 Cl-induced AKT-S473 phosphorylation 29 . We found that siRNA-mediated knockdown of integrin ITGβ 1, but not of ILK, decreased the basal and the NH 4 Cl-induced AKT-S473 phosphorylation ( Fig. 3H and I). Altogether our data indicate that ammonium-induced activation of mTORC2 involves ITGβ 1, FAK, YES1, and PI3K signaling.

Ammonium-induced mTORC2 activation is modulated by calcium.
Our data reveal that ammonium supplied as NH 4 Cl or NH 4 OH can trigger a rapid and relatively sensitive phosphorylation of AKT-S473. This raises the question of what is the molecular entity that is initially sensed as activator. Is NH 4 + , or NH 3, directly sensed, or is it a derivative or a consequence of NH 4 + /NH 3 entry in the cells? NH 4 Cl treatment has been shown to induce a rapid transient increase in intracellular Ca 2+ in cultured astrocytes 30 . We thus assessed the potential role of calcium in NH 4 Cl-induced mTORC2 activation. Pretreatment of cells with the cell-permeable calcium chelator BAPTA-AM decreased basal and NH 4 Cl-induced AKT-S473 and NDRG1-T346 phosphorylation ( Fig. 4A and B). In addition, treatment of cells with thapsigargin (TG), an inhibitor of endoplasmic reticulum Ca 2+ -ATPase known to trigger an increase in cytoplasmic calcium concentration 31 , induced a rapid increase in AKT-S473 phosphorylation (Fig. 4C). The latter TG-induced phosphorylation of AKT was dependent on mTORC2 activity  4D). These results together support that ammonium-induced mTORC2-dependent AKT phosphorylation is modulated by the mobilization of intracellular calcium stores.
Ammonium induces AKT-dependent activation of mTORC1 signaling. The impact of ammonium on the activity of mTORC1, whether neutral or inhibitory, remains unclear [12][13][14][15] . In line with a potential mTORC1 inhibition by ammonium, the kinase array (Fig. 1A) also revealed a decreased phosphorylation of p70S6K-T389, an mTORC1 readout, after 30 minutes treatment with 5 mM NH 4 Cl. Western blot analysis of the time-course impact of NH 4 Cl on pP70S6K-T389 confirmed the slight dephosphorylation (Fig. 5A). This occurred rapidly after NH 4 Cl addition, being detectable at the earliest time point of 5 minutes. However, the weak dephosphorylation of p70S6K-T389 appeared transient, phosphorylation being restored at time 240 min, suggesting that the rapid detected inhibition of mTORC1 was quickly compensated by a reactivation. Moreover, 5 mM NH 4 OH, used in place of NH 4 Cl, provoked a similar transient dephosphorylation of p70S6K-T389 (Fig. 5B). Although, we do not exclude an osmotic effect, these data suggest that ammonium induces a rapid but transient inhibition of mTORC1 activity.
We next wished to address the mechanism of mTORC1 activity recovery following the rapid inhibition. Activation of mTORC1 by growth factors is predominantly mediated through the PI3K-AKT pathway 2,32 . AKT activation has been shown to promote mTORC1 by inhibiting the TSC complex, a negative regulator of mTORC1 32 . Furthermore, AKT-mediated phosphorylation of another negative regulator of mTORC1, PRAS40, prevents its inhibitory role 32,33 . Given that ammonium promotes AKT signaling, we further investigated the effect of NH 4 Cl 5 mM on mTORC1 activation. We found that NH 4 Cl induced the rapid phosphorylation of TSC2-T1462, peaking during the first 30 minutes (Fig. 5A). It also induced rapid phosphorylation of PRAS40-T246 suggesting consequent rapid mTORC1 activation (Fig. 5A). The apparent mTORC1 activation however contrasts with the decreased phosphorylation of p70S6K-T389 that is rapidly observed after NH 4 Cl addition. We therefore tested another readout of mTORC1 activity, the phosphorylation of 4EBP1 (T37/46). Phosphorylation of 4EBP1 (T37/46) inhibits this negative regulator of translation thereby allowing protein biogenesis 34 . In this case, NH 4 Cl addition induced the rapid inhibitory phosphorylation of 4EBP1 (T37/46), consistent with mTORC1 activation (Fig. 5A). These data indicate that an additional regulatory process occurs upon NH 4 Cl addition that transiently counteracts the mTORC1-mediated stimulation of p70S6K-T389 phosphorylation.
Treatment with MK2206, a highly selective and potent inhibitor of AKT 35 , not only prevented the NH 4 Cl-induced phosphorylation of AKT, but also reduced the phosphorylation of TSC2, PRAS40 and 4EBP1, consistent with NH 4 Cl-induced activation of mTORC1 being AKT-dependent (Fig. 5C). Moreover, MK2206 inhibition of AKT did not impair the rapid NH 4 Cl-induced dephosphorylation of p70S6K-T389 (Fig. 5C). This data indicates that the additional regulatory process stimulated by ammonium, possibly corresponding to a phosphatase activation, is AKT-independent.
Collectively, these data are consistent with ammonium promoting AKT-dependent mTORC1 activation in addition to mTORC2 stimulation.
Ammonium stimulates cell proliferation in a dose-dependent manner. Finally, we evaluated the dose-dependent impact of NH 4 Cl on the proliferation of MCF-7 cells. Addition of 10 and 20 mM NH 4 Cl resulted in a significant cell growth inhibition after 3 and 5 days while NH 4 Cl concentrations below 5 mM had no negative impact on proliferation (Fig. 6A). In contrast, addition of 2, 3 or 5 mM NH 4 Cl resulted in slight but significant increase in cell proliferation. Moreover, the latter NH 4 Cl doses were able to rescue the growth defect of MCF-7 cells cultured in glutamine-free medium (Fig. 6B).
These data indicate that the so-called ammonium waste can act as a dose-dependent stimulator of proliferation.

Discussion
This study identifies ammonium as promoting both mTORC2 and mTORC1 signaling and brings insights into the molecular mechanism of the ammonium-mediated regulation (Fig. 7). In contrast to the many cues activating mTORC1, mTORC2 is mainly documented for being activated by growth factors 2,26 . However, recent advances, including observations made in the yeast model, indicate that mTORC2 can be activated by additional signals as well. Membrane tension has been reported to promote yeast TORC2 and mTORC2 activation, the latter enabling for instance a control of cell polarity and mobility 3,36-39 . On the other hand, methylglyoxal, an intermediate of glycolysis, was also recently reported to activate yeast TORC2 and mTORC2 40 . Our study now indicates that mTORC2 is further linked to the metabolic status by responding to ammonium, a byproduct of glutaminolysis. How mTORC2 activation is regulated is also not fully understood 3 . Growth factors were shown to promote, via PI3K signaling, association of mTORC2 to the ribosome and its activation 41 . Like for the activation of mTORC2 mediated by growth factors, the ammonium-induced activation turned to be dependent on PI3K, whereas it was independent of the growth factor receptor EGFR and MEK kinase. Of note, our study identifies YES1, a kinase of the Src family, as involved in the ammonium-induced phosphorylation of both AKT-S473 and NDRG1-T346 mTORC2 readouts. SFK kinases have been shown to participate to cancer cell proliferation and survival, YES1 being currently considered as a potential therapeutic target in a number of cancers 42 . We additionally found that the focal adhesion kinase FAK, a partner of mTOR2 but not of mTOR1, implicated in survival, migration and  invasion 43 , is also involved in the ammonium-induced phosphorylation of mTORC2 readouts. Interestingly, YES1 and FAK interaction has been reported in melanoma, suggesting that both proteins could work in a concerted manner 44 . The transmission of the ammonium signal leading to AKT-S473 phosphorylation also implicated the β 1-integrin, known to activate FAK signaling both in normal and cancer cells 45,46 .
Our data are thus consistent with ammonium-induced activation of AKT relying on β 1-integrin, FAK, YES1, PI3K and mTORC2 (Fig. 7). Interestingly, β 1-integrin/FAK/PI3K/AKT/mTOR signaling was recently implicated in the metabolic reprogramming favoring glycolysis in human mammary epithelial cells overexpressing Twist, a key transcription factor for epithelial-mesenchymal transition and thus invasion and metastasis 47 . Whether ammonium-mediated signalling to mTORC2 can lead to glycolysis activation remains to be evaluated. It is interesting to note that glutamine metabolism, as well as exogenous addition of ammonium, under normoxic conditions were reported to lead to the stabilization of the HIF1α transcription factor known to favor the transcription of several genes involved in glycolysis under hypoxic conditions 48,49 .
Our data also indicate that ammonium promotes mTORC1 signaling. Different pathways convey the signals enabling mTORC1 activation 1,32 . Conserved Rag GTPases for instance control the amino acids-dependent TORC1 signaling 50 . On the other hand, growth factors activate mTORC1 indirectly by suppressing the inhibitory effect of negative regulators, the TSC complex and PRAS40. Activated AKT phosphorylates and inhibits TSC2 in the TSC complex and PRAS40, relieving the TSC-mediated repression of Rheb and Raptor-PRAS40 interaction respectively, thereby increasing mTORC1 signaling 32 . We found that ammonium increased AKT-dependent phosphorylation of both TSC2 and PRAS40, consistent with mTORC1 activation. Indeed, ammonium induced the phosphorylation of the mTORC1 downstream target 4EBP1 suggesting that ammonium promoted protein biogenesis. Intriguingly, and as recently reported 15 , we observed an inhibitory impact of ammonium on the phosphorylation level of a second mTORC1 substrate, the p70S6K1 kinase. However, the weak dephosphorylation of p70S6K1 appeared to be transient and is AKT-independent. This observation could be explained by ammonium promoting, in addition to mTORC1, the activation of phosphatases favoring p70S6K1 dephosphorylation as a rapid response to ammonium. For instance, in the Drosophila model, S6K is dephosphorylated by PP2A phosphatase and this is implicated in the response to the nutritional conditions 51 . The PHLPP phosphatase is also reported to dephosphorylate p70S6K1 52 . Osmotic stress was recently shown to mediate a transient inactivation of mTORC1 53 . However, in this case the response was observed with NaCl concentration above 25 mM, thus well above the osmotic perturbation caused by 5 mM NH 4 Cl or NH 4 OH used in our experimental set-up. In addition, the osmotic stress affected both p70S6K1 and 4EPB1 mTORC1 substrates and it also transiently reduced the AKT-S473 and the ERK1/2 phosphorylation which contrasts with our observations with NH 4 Cl treatment. However, the decrease in p70S6K1 phosphorylation was found to be both TSC2/Rheb dependent Figure 7. Model of the multifaceted response to ammonium. Ammonium triggers mTORC2 activation which phosphorylates and activates AKT in integrinβ 1, YES1 (Src family kinase), and FAK (Focal adhesion kinase) dependent manner. In addition, activated AKT phosphorylates and inhibits TSC2 in the TSC complex and PRAS40, relieving the TSC-mediated repression of Rheb and Raptor-PRAS40 interaction respectively, thereby increasing mTORC1 signaling. Active mTORC1 phosphorylates and inhibits 4E-BP1. Once phosphorylated by mTORC1, 4E-BP1 releases eIF4E and stimulates translation initiation. We further propose that ammonium activates, yet to identify, phosphatases mediating the dephosphorylation of p70S6K1. In parallel to mTORC2 and mTORC1 stimulation, ammonium also promotes ERK signaling. and independent, and Calyculin-A-sensitive phosphatases were proposed to directly dephosphorylate p70S6K1 in response to the osmotic stress via an unknown pathway. Further investigations are needed to determine the mechanism of p70S6K1 dephosphorylation after ammonium addition.
We also found that calcium mobilization is required for ammonium-induction of mTORC2 signaling. Treatment with thapsigargin to enable the release of calcium from ER was sufficient to observe rapid mTORC2-dependent phosphorylation of AKT-S473, as similarly observed with ammonium. Growth hormones were shown to trigger the localization of mTORC2 at the mitochondrial-associated ER membrane where it controls aspects of the mitochondrial physiology, including metabolism and calcium transport 54 . Moreover, calcium was recently suggested to modulate both mTORC1 and mTORC2 in response to growth factors 55 . Our data in turn indicate that ammonium-induced activation of mTORC2 can be modulated by calcium. It is conceivable that ammonium affects the calcium stores via a pH-mediated effect. Ammonium is indeed reported to induce calcium release from the lysosome 30 . Ammonium is a widely used molecule to alter lysosomal function, by inducing a pH raise inhibiting proteases activity 56 . Ammonium has even been reported to inhibit rapamycin-induced autophagy in hepatocarcinoma HepG2 cells 57 . However, in these cases, ammonium concentrations used are above 10 mM, thus far above the levels found in tumor xenografts [10][11][12] . Ammonia was proposed to act as a diffusible signal triggering autophagy in distinct tumor regions 12 . The mechanism enabling ammonium-induced autophagy remains unclear 12,14,15 . As confirmed by our observations, ammonium was found to stimulate ERK1/2 phosphorylation 14 (Fig. 7). However, the ERK pathway was not highlighted as required to initiate ammonium-induced autophagy 14 . It is likely that in growing tumor where cells are engaged in active glutaminolysis, the progressive production and accumulation of ammonium will reach levels that stimulate autophagy and also activate pro-survival signaling pathways leading to stimulate proliferation. This could occur in an autocrine and paracrine fashion. It was recently reported that the glutamine required for the growth of glioblastoma tumors is provided by the glutamine synthetase activity of GS-positive glioma cells in the tumor or by normal astrocytes 58,59 . GS enabled glutamine synthesis from ammonium and glutamate, fueling nucleotide biosynthesis and growth of glioblastoma. Supplementation of both GS substrates rescued the growth of glutamine-deprived cells. The shuttle of ammonium between cells with complementary metabolic factories, or from the circulation to the glutamine-synthesizing cells, could play a key role in tumor growth. Interestingly, our data show that while high ammonium concentrations inhibit the proliferation of MCF-7 breast cancer cells, doses in the 2-5 mM range can stimulate growth both in the presence or absence of glutamine supplementation. Ammonium, likely used to feed glutamine synthesis in glutamine-starved cells, appears to further stimulate proliferation of glutamine-supplemented cells. High ammonium concentrations probably affect several signaling pathways, and inhibit autophagy in this case, overall ending to an inhibition of proliferation, at least in vitro.
While ammonium is largely documented as a toxic waste product in animals, it is a major nitrogen source in micro-organisms and plants 16 . It is a preferred nitrogen source in yeast activating TORC1 and sustaining fast growth and proliferation [19][20][21] . Ammonium participates to the general control of the expression of transporters and enzymes involved in the utilization of non-preferred nitrogen sources 60 and also triggers the inhibition of nutrient transport activity [61][62][63] . In addition, ammonium is a signal limiting yeast colonies overgrowth 64 , while ammonium limitation enables filamentous invasive growth 65 . In contrast, ammonium excess is also known to limit plant growth and to be toxic for yeast cells 66 . Hence, like further evidenced in this study, ammonium appears to have several dose-dependent effects in many domains of life.
Of note, although ammonium is widely assumed to be mainly transported across cell membranes by passive diffusion of ammonia, several transport proteins can mediate either specific or non-specific ammonium transport. For instance, a widely conserved protein family, comprising the human Rhesus factors, act as specific ammonium transport proteins [67][68][69] . We have recently shown that one of the Rhesus factors genes is upregulated in certain cancer cells 70 . Whether targeting ammonium transport proteins could be used as therapeutic strategy remains to be evaluated.
As part of their metabolic reprogramming, cancer cells liberate waste products, like lactate and protons derived from the upregulated glycolysis in the Warburg effect for instance 71,72 . Overall, our data reveal how another waste product of cancer cells, ammonium derived from upregulated glutaminolysis, could turn advantageous for proliferation by triggering key signaling pathways promoting growth.

Reagents
Phospho-Kinase array. The human phospho-kinase array kit (ARY003) was obtained from R&D. NH 4 Cl 5 mM was added for 30 minutes and the levels of phopsho-proteins in cell lysates were analyzed according to the manufacturer's instructions.
Cell proliferation assay. For cell proliferation assays, 2 × 10 4 cells were seeded in 24-well plates and cultured overnight. For glutamine-free conditions, cells grown with glutamine-containing medium were washed three times with pre-warmed glutamine-free medium and were then cultured in glutamine-free medium supplemented with NH 4 Cl at the indicated concentration for the indicated time. Cells were washed with PBS, fixed with 4% paraformaldehyde for 10 minutes and then stained with 0.1% crystal violet for 20 minutes at room temperature. Cells were washed 5 times with water and staining was extracted using 10% acetic acid. Absorbance was measured at 590 nm. All proliferation experiments were done in triplicates.
Statistical analysis. Data are expressed as means ± S.E.M. of a triplicate. Statistical comparisons are assessed by Student's t-tests using Graph Pad Prism version 5.00 software (Graph Pad Software). Differences are considered significant when the p value is below 0.05 (*p < 0.05, # p < 0.001), n = 3.