Two parallel pathways connect glutamine metabolism and mTORC1 activity to regulate glutamoptosis

ARTICLE Two parallel pathways connect glutamine metabolism and mTORC1 activity to regulate glutamoptosis Clément Bodineau 1,2, Mercedes Tomé1, Sarah Courtois 3, Ana S. H. Costa 4,5, Marco Sciacovelli 4, Benoit Rousseau6, Elodie Richard7, Pierre Vacher 7, Carlos Parejo-Pérez8, Emilie Bessede3, Christine Varon3, Pierre Soubeyran7, Christian Frezza 4, Piedad del Socorro Murdoch1,9, Victor H. Villar 10 & Raúl V. Durán 1,2,7✉ Glutamoptosis is the induction of apoptotic cell death as a consequence of the aberrant activation of glutaminolysis and mTORC1 signaling during nutritional imbalance in proliferating cells. The role of the bioenergetic sensor AMPK during glutamoptosis is not defined yet. Here, we show that AMPK reactivation blocks both the glutamine-dependent activation of mTORC1 and glutamoptosis in vitro and in vivo. We also show that glutamine is used for asparagine synthesis and the GABA shunt to produce ATP and to inhibit AMPK, independently of glutaminolysis. Overall, our results indicate that glutamine metabolism is connected with mTORC1 activation through two parallel pathways: an acute alpha-ketoglutaratedependent pathway; and a secondary ATP/AMPK-dependent pathway. This dual metabolic connection between glutamine and mTORC1 must be considered for the future design of therapeutic strategies to prevent cell growth in diseases such as cancer. https://doi.org/10.1038/s41467-021-25079-4 OPEN 1Centro Andaluz de Biología Molecular y Medicina Regenerativa—CABIMER, Consejo Superior de Investigaciones Científicas, Universidad de Sevilla, Universidad Pablo de Olavide, Seville, Spain. 2Institut Européen de Chimie et Biologie, INSERM U1218, Université de Bordeaux, Pessac, France. 3Bordeaux Research in Translational Oncology, INSERM U1053, Université de Bordeaux, Bordeaux cedex, France. 4Medical Research Council Cancer Unit, Hutchison/ MRC Research Centre, Box 197, Cambridge Biomedical Campus, University of Cambridge, Cambridge, UK. 5Cold Spring Harbor Laboratory, Cold Spring Harbor, NY, USA. 6Service Commun des Animaleries, Animalerie A2, University of Bordeaux, Bordeaux, France. 7INSERM U1218, Institut Bergonié, Bordeaux, France. 8Instituto de Bioquímica Vegetal y Fotosíntesis, Consejo Superior de Investigaciones Científicas, Universidad de Sevilla, Seville, Spain. 9Departamento de Bioquímica Vegetal y Biología Molecular, Universidad de Sevilla, Seville, Spain. 10 CRUK Beatson Institute, Glasgow, UK. ✉email: [email protected] NATURE COMMUNICATIONS | (2021) 12:4814 | https://doi.org/10.1038/s41467-021-25079-4 | www.nature.com/naturecommunications 1 1234567890():,; Among all amino acids, glutamine is the most abundant in the blood1–3. The importance of glutamine for cancer cells growth has been extensively described and linked to its role as a precursor for α-ketoglutarate (αKG) to sustain the tricarboxylic acid (TCA) cycle. Indeed, glutamine is mainly metabolized through glutaminolysis in a two-step reaction: first the enzyme glutaminase (GLS) catalyzes the hydrolysis of glutamine to glutamate; and then glutamate is converted to αKG through an oxidative deamination reaction catalyzed by the enzyme glutamate dehydrogenase (GDH). Another amino acid, leucine, acts as an allosteric activator of the second enzyme, being consequently necessary for glutaminolysis induction4,5. Although glutamine can participate in the synthesis of different amino acids through the production of glutamate, only asparagine requires glutamine for de novo synthesis. Asparagine synthetase (ASNS) is the enzyme that converts glutamine and aspartate into glutamate and asparagine in an ATP-dependent manner6. ASNS expression is upregulated during amino acid starvation through the activation of activating transcription factor 4 (ATF4)7. Asparagine has been identified as an exchange amino acid factor that regulates the serine/threonine kinase mammalian target of rapamycin complex 1 (mTORC1), nucleotide biosynthesis and proliferation7. The role of GABA shunt is a bypass for two steps of the TCA cycle, wherein GABA is synthesized from glutamate by a GAD1encoded decarboxylase, transaminated into succinic semialdehyde, and metabolized into the TCA cycle intermediate succinate8. In addition to its role as a neurotransmitter, GABA has been detected in a wide range of peripheral tissues9and recently linked to castration-resistant prostate cancer progression through the regulation of nuclear androgen receptor by GABA10. That study reported that GABA shunt is upregulated in response to the activating phosphorylation of GAD65 by the PI3K pathway. In a different cancer type, the increased uptake and metabolism of GABA in breast to brain metastatic cells was linked to an increase of NADH levels in the microenvironment conferring a proliferative advantage to the tumor11. Previously, our results demonstrated that glutamine, and most particularly its catabolic conversion to αKG through glutaminolysis, activates mTORC1 at the short and long-term5,12. mTORC1 is a major signaling hub that integrates different inputs, such as nutrients, oxygen, energy, and growth factors, to regulate the metabolic pathways controlling cell growth and proliferation13. As mTORC1 is over-activated in 80% of solid tumors14, different analogs of its inhibitor rapamycin have been developed to inhibit mTORC1 in patients. At a clinical level, the results remain modest and a clear need for co-therapies has been described15–17. Our previous work determined that the activation of mTORC1 by glutaminolysis during nutritional imbalance lead to the anomalous inhibition of autophagy and a subsequent form of apoptosis named glutamoptosis12,18. But still, the impact of glutamine metabolism on the bioenergetic status of the cells during longterm mTORC1 activation and glutamoptosis induction remains to be defined. Eukaryotes have developed a system to sense low ATP levels via another serine/threonine kinase, the AMP-activated protein kinase (AMPK) complex. Under low intracellular ATP levels, AMP or ADP can directly bind to the γregulatory subunit of AMPK, leading to a conformational change that allows the activating phosphorylation of AMPK13. Once activated, AMPK redirects the metabolism towards the increase of catabolism and the decrease in anabolism through phosphorylation of downstream key protein pathways, ultimately leading to mTORC1 inhibition13. Herein, we report that the amino acid glutamine is sufficient to sustain the production of ATP in absence of any other amino acid, following a glutaminolysis-independent mechanism. During glutamine sufficiency, ASNS and GABA shunt are responsible for metabolizing glutamine to generate ATP and to inhibit AMPK. Thus, our results indicate that glutamine activates mTORC1 following a two branches mechanism: a short-term mechanism involving αKG production, and a long-term mechanism involving ASNS, GABA, and AMPK. Results Glutaminolysis sustains the production of ATP to inhibit AMPK and to activate mTORC1. Previously, we showed that the addition of leucine and glutamine (“LQ treatment”) to amino acid-starved cells is sufficient to activate glutaminolysis and subsequently mTORC15,19. To better understand the role of the bioenergetic status of the cell in this pathway, we first investigated if LQ treatment impacted the production of ATP in the cell and the activation of AMPK. Confirming our previous observations, short-term (2 h) amino acid starvation had no effect in the levels of ATP in cultured cells. However, we observed that, at longer times, amino acid withdrawal significantly decreased ATP/ADP ratio in U2OS and HCT116 cellular models (Fig. 1A, B and Supplementary 1A, B). LQ addition in amino acid-starved cells sustained the levels of ATP at 4, 8, 12, 24, 48, and 72 h (Fig. 1C and Supplementary 1C–E). By contrast, the addition of methionine or arginine, two amino acids which have been reported to activate mTORC1 at short-term, did not sustain the production of ATP (Supplementary 1F, G). Measurements of glutamine and leucine levels in the medium indicated that these amino acids remained fully available even after 72 h of LQ treatment (Supplementary 1H–I). In agreement to these previous observations, we also observed that LQ treatment increased the basal respiration of U2OS cells compared to amino acid starvation (Fig. 1D). The maximal oxygen consumption rate (OCR) was also increased in the presence of LQ, which resulted in an increase of ATP production linked to respiration (Fig. 1E). Thus, we concluded that long-term glutaminolysis, even in the absence of any other amino acid, was sufficient to increase the respiratory capacity and the production of ATP in the cell. We next investigated if the observed production of ATP in LQtreated cells had any impact on the phosphorylation of AMPK. The removal of amino acids from the culture medium led to the inhibition of mTORC1 in U2OS cells, as broadly reported before20, but concomitantly it led also to the phosphorylation of AMPK at threonine 172 (Fig. 1F), in agreement with the observed decrease in ATP levels. Accordingly, the phosphorylation of AMPK was reverted by the addition of LQ, which again correlated with the increase in ATP levels in this condition (Fig. 1F). The LQ-induced inhibition of AMPK was observed at 24, 48, and 72 h (Fig. 1G). Long-term activation of mTORC1 downstream of AMPK during LQ treatment was confirmed by the phosphorylation of its target proteins S6K, S6, and 4EBP1 (Fig. 1F, G), and by its translocation at the surface of the lysosome using LAMP2 (Fig. 1H, I) or CD63 (Supplementary 1J, K) as lysosomal markers. To confirm a mechanistic role of AMPK during LQ-mediated mTORC1 activation at long-term, we investigated the status of mTORC1 by the addition of the AMPK activator AICAR in LQ-treated U2OS and HCT116 cells. As observed in Fig. 1J and Supplementary 1L, reactivation of AMPK using AICAR was sufficient to prevent the LQ-mediated activation of mTORC1. Similar results were obtained by transfecting a MYC-tagged constitutively active form of AMPK (CA-AMPK, kindly provided by Prof. Benoit Viollet, Paris, France) (Fig. 1K) or using other AMPK activators, such as metformin or A769662 (Fig. 1L). Surprisingly, the ablation of AMPK using AMPK−/−MEFs (again, kindly provided by Prof. ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-25079-4 2NATURE COMMUNICATIONS | (2021) 12:4814 | https://d oi.org/10.1038/s41467-021-25079-4 | www.nature.com/naturecommunications Benoit Viollet) also prevented the LQ-mediated activation of mTORC1 (Supplementary 1M). This result suggested that, in addition to acting as a negative regulator of mTORC1, the presence of AMPK is necessary for the connection between glutamine metabolism and mTORC1, underscoring the complexity of the metabolic adaptations in these circumstances. We also confirmed that this AMPK-regulated activation of mTORC1 during long-term LQ treatment had a direct physiological impact in the cell in terms of autophagy activation, a key process during glutamoptosis induction12. As observed in Fig. 1M, N, LQ-treated U2OS cells stably expressing the autophagic reporter GFP-LC3 displayed an increase of GFP-LC3 aggregates upon AMPK reactivation using AICAR, metformin, or A769662. Consistently, reactivation of AMPK using these different activators also decreased the levels of endogenous p62 and LC3I, increasing LC3II formation (Fig. 1L), well-known autophagy markers. We also evaluated the autophagic flux using bafilomycin A1 (Baf A1) to trap the formation of autophagosomes. The inhibition of autophagy using Baf A1 led to a greater accumulation of p62 and LC3I in cells treated with LQ. However, 35 35 45 70 55 55 70 70 70 35 25 25 55 70 15 10 55 55 55 70 55 15 10 70 55 70 55 35 25 25 35 35 35 * 0 5 10 15 20 OCR (pmol/min/ g protein) -LQ+ 0 20 40 60 80 100 % GFP-LC3 aggregates -LQ AICAR Metformin A769662 + LQ - - - + - - - + - - - + * * 0 2 4 8 12 24 0.5 1.0 1.5 0 ATP/ADP ratio * hours -AA N.S 0.5 1.0 1.5 0 ATP/ADP ratio +-+-AA 2h 72h C * AB FG -LQ AICAR Metformin A769662-- +LQ -AA K H +AA -AA +LQ LAMP2 mTORC1 merged LQ Actin S6-pS235/236 S6K S6 24h S6K-pT389 AMPK-pT172 +-+-+- 48h 72h -AA AMPK E - -AICAR LQ -AA + - - + + + AMPK-pT172 AMPK S6K-pT389 S6K S6-pS235/236 S6 Actin IJ L Actin LC3I/II p62 AICAR Metformin A769662 LQ - - - + - - - + - - - + - - - S6-pS235/236 S6 AMPK-pT172 AMPK S6K-pT389 S6K LQ AICAR Baf A1 + - - - - - p62 LC3I/II Actin + + - - + - + - + - - + + + + - + + -0.2 0.0 0.2 0.4 0.6 Pearson correlation coefficient Colocalisation between LAMP2 and mTORC1 + - - - - + AA LQ ** M LQ AA + - - - - + Raptor-pS792 S6 S6-pS235/236 AMPK-pT172 S6K-pT389 S6K 4EBP1-pT37/46 4EBP1 AMPK Actin Raptor 0 102030405060708090100 0 10 20 30 40 Time (minutes) OCR (pmol/min/ gof protein) -AA +LQ Oligo FCCP Rot + Ant A *** *** CD O AA LQ CA-AMPK - - - + + - - - - + + - EV Actin S6 CA-AMPK S6-pS235/236 N -++++ -AA 70 70 55 70 70 35 35 0.5 1.0 1.5 2.0 2.5 0 ATP/ADP ratio -+ LQ 24h -+ 48h -+ 72h -AA *** 135 55 35 135 70 70 55 70 70 35 25 35 55 35 48 15 15 NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-25079-4 ARTICLE NATURE COMMUNICATIONS | (2021) 12:4814 | https://doi.org/10.1038/s41467-021-25079-4 | www.nature.com/naturecommunications 3 reactivation of AMPK using AICAR concomitantly to Baf A1 and LQ did not accumulate p62 at the same level (Fig. 1O), and restored the levels of LC3II. Altogether, these results showed that long-term LQ treatment was sufficient to increase the levels of ATP, leading to the inhibition of AMPK. AMPK inhibition in LQ-treated cells led to the subsequent activation of mTORC1 and the inhibition of autophagy, two major processes controlling glutamoptosis. AMPK inhibition is necessary for glutamoptosis both in vitro and in vivo. Next, we assessed the role of AMPK in the induction of apoptosis following the activation of mTORC1 by glutaminolysis during nutritional imbalance, i.e., glutamoptosis. In agreement with our previous results12, long-term LQ treatment in absence of any other amino acid decreased cell viability (Fig. 2A, B). The LQ-induced increase in cell death was prevented by AMPK reactivation using AICAR, metformin, or A769662 (Fig. 2A, B). Further, a clonogenic assay confirmed that the pharmacological reactivation of AMPK prevented LQ-mediated cell death induction (Fig. 2C, D). AMPK reactivation was also sufficient to prevent apoptosis induced by LQ, as determined by the decrease in the double annexin V/PI staining observed by flow cytometry (Fig. 2E, F), and by the decrease in the pro-apoptotic markers cleaved PARP and cleaved caspase 3 (Fig. 2G). On the contrary, AMPK ablation did not increase the induction of cell death during amino acid withdrawal, as mTORC1 inhibition and subsequent autophagy activation under these conditions prevented glutamoptosis (Supplementary 2A). This result further positioned mTORC1/autophagy downstream of AMPK during glutamoptosis induction. These results were confirmed using ATG5−/−MEFs in which the impairment of autophagy completely abolished the capacity of AICAR to promote cell survival during amino acids starvation (Fig. 2H, I) as autophagy is necessary as survival mechanism. Thus, AMPK inhibition upstream of mTORC1/autophagy was necessary for glutamoptosis in vitro. The participation of AMPK in glutamoptosis was also investigated in vivo using implanted tumors in xenograft mouse models. Two different cohorts of mice were implanted in the right dorsal flank with HCT116 cells. Mice were treated with either vehicle or with a cell-permeable αKG derivative, dimethyl-αKG (DMKG) to induce glutamoptosis in vivo. Mice were then cotreated with the mTORC1 inhibitor temsirolimus (a rapamycin derivative) or with metformin. First, the induction of mTORC1 by DMKG was assessed by the analysis of S6 phosphorylation in these tumors. As expected, the injection of DMKG in mice induced the activation of mTORC1. A decrease of S6 phosphorylation and thus mTORC1 activation was observed in mice cotreated with temsirolimus or metformin (Fig. 2J, K and Supplementary 2B). Second, the in vivo induction of glutamoptosis was assessed by immunohistochemical analysis of cleaved caspase 3. We confirmed that DMKG treatment induced an increase in apoptotic cell death in xenograft tumors in vivo, as assessed by caspase 3 cleavage (Fig. 2L–M and Supplementary 2C). This is a validation of the physiological relevance of glutamoptosis in vivo, confirming that glutamoptosis can be used as a potential tool to induce tumoral cell death in vivo. Furthermore, as shown in Fig. 2L–M and Supplementary 2C, we also confirmed that the inhibition of mTORC1 using temsirolimus prevented glutamoptosis in vivo, similar to what we observed in cultured cells12. This result validated the physiological model by which the aberrant activation of mTORC1 during nutritional imbalance induces apoptotic cell death in animal models. Finally, and of note, AMPK reactivation using metformin which was confirmed by IHC on T172 AMPK phosphorylation (Supplementary 2D, E) was similarly sufficient to prevent glutamoptosis in vivo (Fig. 2L–M and Supplementary 2C). Altogether, these data confirmed the necessity of AMPK inhibition for glutamoptosis induction both in vitro and in vivo. Glutaminolysis is not necessary for LQ-mediated ATP production. After confirming the capacity of LQ to sustain ATP levels, we investigated the possible role of glutaminolysis, the main pathway to catabolise glutamine, in the production of ATP in response to LQ. For this purpose, and following a loss-offunction approach, we inhibited GLS either pharmacologically or genetically (using two different inhibitors, BPTES and DON, or RNA-interference mediated knockdown). We then investigated the capacity of LQ treatment to induce ATP levels in GLSinhibited cells. Surprisingly, the inhibition of GLS did not prevent the production of ATP in response to LQ in U2OS, HCT116 or HEK293 cells (Fig. 3A, B and Supplementary 3A, B). Confirming Fig. 1 Glutaminolysis sustains the production of ATP to inhibit AMPK and to activate mTORC1. A ATP/ADP ratio of U2OS cells incubated in the presence or the absence of all amino acids for 2 or 72 h. Fed cells (C) are used as control. BATP/ADP ratio of U2OS cells incubated in absence of amino acid for the indicated time. CATP/ADP ratio of amino acid-starved U2OS cells incubated in the presence or absence of LQ during the indicated times. DOCR analysis by Seahorse®technology of amino acid-starved U2OS cells incubated in the presence (blue) or absence (purple) of LQ during 72 h. OCR was measured either in basal conditions or after the injection of oligomycin, FCCP, and rotenone/antimycin A. Data are mean ± SEM of three biologically independent experiments performed with five replicates. EBasal respiration used to drive ATP production as determined by OCR quantification of data obtained in (D). FImmunoblot of mTORC1 activity markers (S6K, S6, and 4EBP1 phosphorylation) and AMPK phosphorylation of U2OS cells incubated with or without amino acids, in the presence or absence of LQ during 72 h. GImmunoblot of mTORC1 activity markers (S6K and S6 phosphorylation) and AMPK phosphorylation of amino acid-starved U2OS cells incubated in the presence or absence of LQ during 24, 48, or 72 h. HImmunofluorescence microscopy captions of U2OS cells incubated with or without amino acids, in the presence or absence of LQ during 72 h. Cells were stained against LAMP2 (lysosomal marker, red), mTORC1 (green) and DAPI (blue). Scale bar represents 10 µm. IQuantification of the colocalization between LAMP2 and mTORC1 as shown in (H). Person’s R value was evaluated using ImageJ coloc2 plugin on 25 ROI in three biologically independent experiments (75 ROI in total per condition). JImmunoblot of mTORC1 activity markers (S6K and S6 phosphorylation) and AMPK phosphorylation of amino acid-starved U2OS cells incubated in the presence or absence of LQ, with or without AICAR, during 72 h. KImmunoblot analysis of mTORC1 activity marker (S6 phosphorylation) of U2OS cells expressing a myc-tagged, constitutively active AMPK mutant in the presence or absence of amino acids and LQ as indicated. LImmunoblot of autophagy (p62 and LC3-I/II) and mTORC1 (S6K and S6 phosphorylation) markers of amino acid-starved U2OS cells incubated in absence or presence of LQ, AICAR, metformin or A769662 for 72 h, as indicated. MFluorescence microscopy captions of GFP-LC3 expressing amino acid-starved U2OS cells incubated in the presence or absence of LQ, with or without AICAR, metformin or A769662 for 72 h. Autophagosome formation upon GFP-LC3 aggregation was assayed using confocal microscopy. The scale bar represents 10 µm. NQuantification of the number of GFP-LC3 dots per cell of captions obtained in (M). >100 cells were counted per experiment. OImmunoblot analysis of autophagy markers (p62 and LC3I/II) of U2OS cells treated with LQ, AICAR and/or Bafilomycin A1 as indicated for 72 h. Graphs show mean values ± SEM (n=3 biologically independent experiments). *p< 0.05 (ANOVA analysis followed by a post hoc Bonferroni test). Source data are provided as a Source Data file. ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-25079-4 4NATURE COMMUNICATIONS | (2021) 12:4814 | https://d oi.org/10.1038/s41467-021-25079-4 | www.nature.com/naturecommunications this result, silencing GDH did not impair the production of ATP in LQ-treated cells either (Fig. 3C). Thus, glutaminolysis did not show a necessary role for the capacity of LQ to sustain ATP levels. Further confirming this conclusion, we also observed that the addition of DMKG to amino acid-starved cells did not sustain the production of ATP to a similar extent than LQ treatment (Fig. 3D and Supplementary 3C, D). Hence, glutaminolysis was neither necessary nor sufficient to sustain ATP production. The lack of decrease on ATP levels upon glutaminolysis inhibition correlated with an absence of AMPK reactivation (Fig. 3E, F), confirming that glutaminolysis did not mediate the connection between LQ and the ATP/AMPK axis. 0 10 20 30 40 50 Cell death % + -- -AA AICAR - ++ -- -- + MEF ATG5-/- * MEF ATG5+/+ * 0 20 40 60 80 100 Viability % - -+ -- +LQ AICAR + + * MET TEM Vehicle -DMKG +DMKG S6-pS235/236 MET TEM Vehicle -DMKG +DMKG cCas3 +LQ-LQ AICAR Metformin A769662-- A - - + - - + + + LQ AICAR C +LQ -LQ AICAR Metformin A769662 - F H I - + + + -AA - - LQ AICAR S6 cPARP cCasp 3 S6K-pT389 S6K S6-pS235/236 AMPK-pT172 AMPK Actin G B DE -LQ AICAR Metformin A769662 +LQ - - - + - - + - - - + - - - L M 1.84%1.84% 12.04%12.04% BAX 0 1 2 3 4 5 S6-pS235/236 IHC visual score - - - - + - DMKG Temsirolimus Metformin - - + + - - + + - + - + ** 0 1 2 3 4 5 cCaspase 3 IHC visual score - - - - + - DMKG Temsirolimus Metformin - - + + - - + + - + - + ** cPARP AA AICAR + - cCasp 3 ATG5-ATG12 Actin - - - + + - MEF ATG5+/+ - - - + MEF ATG5-/- J K Raptor-pS792 Raptor -AA -AA 0 10 20 30 40 50 % Late apoptotic cells - * * 100 15 20 70 70 55 135 135 70 70 25 25 55 70 15 25 35 55 0 50 100 150 200 250 Number of colonies - -+ -- ++ +LQ AICAR * * 82.34%82.34% 3.78%3.78% 4.81%4.81% 29.93%29.93% 64.29%64.29% 0.97%0.97% 0.90%0.90% 3.41%3.41% 95.04%95.04% 0.65%0.65% 1.44%1.44% 5.21%5.21% 1.00%1.00%92.35%92.35% 0.78%0.78% 4.90%4.90% 93.04%93.04% 1.28%1.28% NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-25079-4 ARTICLE NATURE COMMUNICATIONS | (2021) 12:4814 | https://doi.org/10.1038/s41467-021-25079-4 | www.nature.com/naturecommunications 5 Glutamine is sufficient to sustain ATP levels, but not to activate mTORC1. During LQ treatment, leucine addition plays an allosteric activation role for GDH, necessary for αKG production5,21. As we concluded that glutaminolysis did not affect the ATP/AMPK pathway, we then assessed whether leucine had any role in ATP levels during LQ treatment. For this purpose, we treated amino acid-starved U2OS, HCT116 and HEK293A cells with glutamine or leucine, and subsequently measured ATP levels. Unlike leucine, glutamine alone was sufficient to sustain ATP levels in all three cell lines (Fig. 4Aand Supplementary 4A, B). Confirming this result, we observed that basal respiration of the cells in amino acid starvation did not change in response to leucine addition (Fig. 4B). However, cells in the presence of glutamine showed a significant increase of the respiratory activity, independently of the presence or the absence of leucine (Fig. 4B). Similarly, glutamine sufficiency induced a clear increase of the OCR linked to ATP production by the mitochondria to meet the energetic needs of the cell (Fig. 4C). This increase of ATP levels by glutamine alone also correlated with an increase of endoplasmic reticulum calcium pool due to the activity of the sarcoendoplasmic reticulum Ca2+ ATPase (SERCA) (Supplementary 4C). In a similar manner, and correlating with ATP levels, a strong decrease of AMPK phosphorylation was observed upon glutamine addition to otherwise amino acid-depleted cells, while the addition of leucine did not reduce the phosphorylation status of AMPK (Fig. 4D and Supplementary 4D). Fig. 2 AMPK inhibition is necessary for glutamoptosis both in vitro and in vivo. A Representative microscopy images of amino acid-starved U2OS cells incubated in absence or presence of LQ, AICAR, metformin, or A769662 for 72 h, as indicated. Scale bar represents 100 µm. BCell viability as estimated by a trypan blue exclusion assay of amino acid-starved U2OS cells incubated in the presence or absence of LQ and AICAR during 72 h. CRepresentative images of clonogenic assay of U2OS cells treated as in (B). DColony quantification of images obtained in (C). EQuantification of late apoptosis population of three biologically independent experiments (double positive, annexin V and PI) as obtained in (F) for the indicated condition. FFlow cytometry analysis of annexin V/PI staining of amino acid-starved U2OS cells incubated with or without LQ in combination with AICAR, metformin, or A769662 during 72 h. GImmunoblot of the pro-apoptotic markers (BAX, cleaved caspase 3, and cleaved PARP), mTORC1 activity markers (Raptor, S6K, and S6 phosphorylation), and AMPK phosphorylation of amino acid-starved U2OS cells incubated in the presence or absence of LQ, with or without AICAR, during 72 h. HCell viability as estimated by a trypan blue exclusion assay of ATG5+/+and ATG5−/−MEFs incubated in the presence or absence of amino acids (AA) and AICAR during 72 h. IImmunoblot analysis of pro-apoptotic markers of ATG5+/+and ATG5−/−MEFs incubated as in (H). J–LRepresentative immunohistochemistry microscopy pictures (×40 magnification) of xenograft tumors of mice treated as indicated (TEM: Temsirolimus; MET: Metformin). Samples were stained against S6-pS235/236 (J) and cleaved caspase 3 (L). Scale bars represent 100 µm. K–MIHC visual score of S6-pS235/236 (K) and caspase 3 (M) of images from (J) and (L), respectively. The upper and lower limits of the boxes represent quartiles, with the line within the boxes indicating the median and the whiskers showing the extremes (n≥10 images per treatment). Graphs show mean values ± SEM (n=3 biologically independent experiments). *p< 0.05 (ANOVA analysis followed by a post hoc Bonferroni test). Source data are provided as a Source Data file. 0.5 1.0 1.5 2.0 0 ATP/ADP ratio - + - - - + + + - LQ siCtl siGDH + - + N.S 0.5 1.0 1.5 2.0 0 ATP/ADP ratio - + - - - + + + - LQ siCtl siGLS1 + - + N.S 0.5 1.0 1.5 2.0 0 ATP/ADP ratio - - - + - - + + - LQ BPTES DON + - + * - - LQ BPTES Actin S6 S6K AMPK AMPK-pT172 S6-pS235/236 S6K-pT389 -AA + - + +DON LQ -AA - - Actin AMPK AMPK-pT172 S6-pS235/236 S6 S6K-pT389 S6K + - + + A BC DE F Actin GLS1 siCtl siGLS1 Actin GDH siCtl siGDH 55 55 35 35 70 70 55 70 70 25 25 55 70 70 55 70 70 35 25 25 45 0.5 1.0 1.5 2.0 0 ATP/ADP ratio - - + - - + LQ DMKG ** Fig. 3 Glutamine is sufficient to sustain ATP levels, but not to activate mTORC1. A ATP/ADP ratio of amino acid-starved U2OS cells incubated in the presence or absence of LQ with or without BPTES or DON during 72 h. B,CGLS or GDH expressions were knocked down using small interfering RNA (siRNA) in U2OS cells for 48 h. Cells were then treated with LQ for 72 h and the ATP/ADP ratio was measured. Scramble non-targeting siRNA was used as a control. Immunoblots of GLS or GDH levels are presented as a control of the knockdown. DATP/ADP ratio of amino acid-starved U2OS cells treated with LQ or DMKG for 72 h. E,FImmunoblot of amino acid-starved U2OS cells treated with or without LQ in combination with BPTES (E) or DON (F) for 72 h. Activity markers of AMPK and mTORC1 were analysed. Graphs show mean values ± SEM (n=3 biologically independent experiments). *p< 0.05 (ANOVA analysis followed by a post hoc Bonferroni test). Source data are provided as a Source Data file. ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-25079-4 6NATURE COMMUNICATIONS | (2021) 12:4814 | https://d oi.org/10.1038/s41467-021-25079-4 | www.nature.com/naturecommunications Similar to what was observed for LQ treatment (Fig. 3B, C), glutaminolysis inhibition by GLS1 or GDH knockdown did not affect the capacity of glutamine sufficiency to induce ATP levels in U2OS and HCT116, confirming that glutaminolysis was not necessary for glutamine-dependent ATP production (Supplementary 4E–H). It is worth noting that neither glutamine alone nor leucine alone (in absence of any other amino acid) were sufficient to activate mTORC1 after 72 h in either U2OS or HCT116 cells (Fig. 4D and Supplementary 4D). Consistently, mTORC1 localization at the surface of the lysosome was only observed in the presence of both leucine and glutamine, but it was not observed if only one of these amino acids were added (Fig. 4E, F and Supplementary 4I, J). In a similar manner 0 5 10 15 20 OCR (pmol/min/ gprotein) - - + - L Q - + + + * RagB WT Q RagB GTP Actin S6 S6K S6-pS235/236 S6K-pT389 -AA - + - - - + + + - + - + L Q Actin S6 S6K-pT389 S6K - - + - - + + + AMPK-pT172 AMPK -AA S6-pS235/236 A D C E G F JH 0 102030405060708090100 0 10 20 30 40 Time (minutes) OCR (pmol/min/μg of protein) St L Q LQ Oligo FCCP Rot + Ant A B *- - Q L + - + + Actin LC3I/II p62 I - - + - + + Q L -0.2 0.0 0.2 0.4 0.6 Pearson correlation coefficient Colocalisation between LAMP2 and mTORC1 - - + - - + L Q + + * L Q - LQ LAMP2 mTORC1 merged * 0.5 1.0 1.5 2.0 2.5 3.0 0 ATP/ADP ratio - - + - - + L Q + + 0 20 40 60 80 100 120 % GFP-LC3 aggregates - - + - + + Q L 55 70 55 70 70 70 25 35 25 55 70 15 55 70 70 35 35 25 35 Fig. 4 Glutamine metabolism activates mTORC1 following two parallel, necessary branches. A ATP/ADP ratio of amino acid-starved U2OS cells incubated with leucine and/or glutamine during 72 h. BOCR analysis by Seahorse®technology of amino acid-starved (purple) U2OS cells incubated with leucine (red) and/or glutamine (Q green, LQ blue) during 72 h. OCR was measured either in basal conditions or after the injection of oligomycin, FCCP, and rotenone/antimycin A. Data are mean ± SEM of three biologically independent experiments performed with five replicates. CBasal respiration used to drive ATP production as determined by OCR quantification of data obtained in (B). DImmunoblot of mTORC1 activity markers (S6K and S6 phosphorylation) and AMPK phosphorylation of amino acid-starved U2OS cells incubated with leucine and/or glutamine during 72 h. EImmunofluorescence microscopy captions of U2OS cells incubated with leucine and/or glutamine during 72 h. Cells were stained against LAMP2 (lysosomal marker, red), mTORC1 (green) and DAPI (blue). Scale bar represents 10 µm. FQuantification of the colocalization between LAMP2 and mTORC1 as shown in (E). Person’s R value was evaluated using ImageJ coloc2 plugin on 25 ROI in three biologically independent experiments (75 ROI in total per condition). GFluorescence microscopy captions of GFP-LC3 expressing amino acid-starved U2OS cells incubated in the presence of glutamine and/or leucine during 72 h. Autophagosome formation upon GFP-LC3 aggregation was assayed using confocal microscopy. The scale bar represents 10 µm. HQuantification of the number of GFP-LC3 dots per cell of captions obtained in (G). >100 cells were counted per experiment. IImmunoblot of autophagy (p62 and LC3-I/II) markers of amino acidstarved U2OS cells incubated in the presence of glutamine and/or leucine during 72 h. JU2OS cells were transfected with RagB WT plasmid or RagB 54 L plasmid. Amino acid-starved cells were then incubated with or without glutamine for 72 h. Downstream targets of mTORC1 (S6K and S6 phosphorylation) were assessed by immunoblot. Graphs show mean values ± SEM (n=3 biologically independent experiments). *p< 0.05 (ANOVA analysis followed by a post hoc Bonferroni test). Source data are provided as a Source Data file. NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-25079-4 ARTICLE NATURE COMMUNICATIONS | (2021) 12:4814 | https://doi.org/10.1038/s41467-021-25079-4 | www.nature.com/naturecommunications 7 than for LQ, glutamine levels were measured in the cell culture medium during glutamine sufficiency condition to verify the availability of this amino acid even after 3 days of treatment (Supplementary 4K). Autophagy analysis downstream of mTORC1 also confirmed the inability of glutamine sufficiency to activate mTORC1. Thus, long-term glutamine addition did not inhibit autophagy activation during amino acid starvation, as assessed by the autophagy reporter GFP-LC3 by confocal microscopy, and by immunoblot analysis of the autophagic endogenous markers p62 and LC3I/II. In contrast, and as shown previously, the addition of both leucine and glutamine strongly decreased the number of GFP-LC3 aggregates and increased p62 levels in U2OS cells (Fig. 4G–I). 0.5 1.0 1.5 2.0 0 ATP/ADP ratio Q BPTES siCtl siASNS - - - - + - + - + + + + N.S * GABA Asparagine Glutamate Aspartate Citrate NAA Alanine Glutamine Malate Pyruvate Serine Isoleucine Ornithine Glucose Palmitate Stearate Lactate Sorbitol Fructose αKG Leucine KMV/KIC GSH Glycine Thymine Lysine Threonine Methionine Phenylalanine Valine Arginine Aconitate 2 1 0 -1 -2 Row Z-score AMPK-pT172 Q BPTES - - + - + + - - + - + + AMPK ASNS Actin siCtl siASNS + - - - - - AA L Q - + - - - + - + + siCtl ASNS Actin siASNS AB C DE F G HI + L + [U13C]Q+AA + [U 13C]Q + [U13C]Q +AA + [15N2]Q + [15N2]Q +L + [15N2]Q AA BPTES S6 Actin S6-pS235/236 ASNS - - - + + - + + - - - + + - + + siCtl siASNS AMPK AMPK-pT172 S6K S6K-pT389 4EBP1 4EBP1-pT37/46 0 1106 2106 3106 Ion counts Glu m+5 (17) (30) (22) ** 0 110 5 210 5 Ion counts GABA m+4 (7) (45) (32) ** 0 2105 4105 6105 8105 Asn m+4 (34) (28) (33) * * 0 1104 2104 3104 4104 5104 Asp m+4 (14) (21) (15) ** 0 2104 4104 6104 8104 Citrate m+4 (7) (20) (3) ** 1104 2104 0 Malate m+4 (8) (21) (8) * 0 1106 2106 3106 4106 Ion counts (35) (59) (51) Glu 15N ** 0 2104 4104 6104 3105 4105 5105 (33) (59) (35) Asp 15N ** 0 2103 4103 1105 2105 3105 4105 (17) (41) (23) Asn 15N2 ** 0 1104 2104 3104 1105 2105 (39) (21) (24) Asn 15N * 0.5 1.0 1.5 2.0 2.5 0 ATP/ADP ratio - - + - - + Q DMKG + + * * 55 35 70 55 55 55 55 35 70 100 70 35 25 25 25 55 55 70 55 55 0 5 10 15 20 ASNS mRNA levels (RU) N.S * * - - - - + - - - + + - - + + + - - - - + + - - + Q L BPTES DMKG ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-25079-4 8NATURE COMMUNICATIONS | (2021) 12:4814 | https://d oi.org/10.1038/s41467-021-25079-4 | www.nature.com/naturecommunications It is well documented that the re-addition of leucine alone for 10–30 m after a short-term amino acid starvation (1–2h) is sufficient to activate mTORC122–25. However, our data showed that this reactivation of mTORC1 disappeared after longer starvation periods (>4 h of amino acid starvation) (Supplementary 4L), which correlated with a strong decrease of glutamine intracellular levels following 4 h of amino acid starvation (Supplementary 4M). In contrast, re-addition of both leucine and glutamine activated mTORC1 even after longer periods of amino acid starvation (Supplementary 4N). These results support a model by which leucine alone activates mTORC1 after shortterm amino acid starvation thanks to the remaining pool of intracellular glutamine. It has been proposed elsewhere that intracellular glutamine is required for leucine uptake26. However, our data showed that, in our setup, intracellular levels of leucine were not significantly affected by the co-addition of glutamine in starved cells (Supplementary 4O), suggesting that intracellular glutamine was not necessary for leucine uptake. Rather, our data sustained a model by which leucine only activates mTORC1 by cooperating with glutamine to activate glutaminolysis. Glutamine metabolism activates mTORC1 following two parallel, necessary branches. In light of our results, we investigated the possibility that two parallel pathways connect glutamine and mTORC1: a previously reported glutaminolysis-dependent branch; and an ATP-dependent but GLS/GDH-independent branch. We previously showed that glutaminolysis activates mTORC1 via RagB. Thus, αKG generation by glutaminolysis increases GTP loading of RagB, leading to mTORC1 translocation to the surface of the lysosome5. It is documented that the overexpression of a constitutively GTP-bound mutant variant of RagB (RagB-GTP mutant) is sufficient to force the translocation of mTORC1 to the surface of the lysosome even in amino acidstarved cells27. The expression of this RagB-GTP mutant is sufficient to activate mTORC1 in amino acid-starved cells at short times (1–2h) 5,27. However, we observed that RagB-GTP mutant expression did not activate mTORC1 in amino acid-starved cells at longer times (72 h, Fig. 4J and Supplementary 4P). Strikingly, the combination of both RagB-GTP expression and glutamine addition induced the full activation of mTORC1 at 72 h, as determined by the phosphorylation of the downstream targets S6K and S6 (Fig. 4J). Similar results were obtained using a RagDGDP mutant, which also upregulates the lysosomal translocation of mTORC1 (Supplementary 4P). Opposite to what we observed with glutamine, the combination of both RagB-GTP and leucine did not induce the activation of mTORC1 signaling (Supplementary 4Q). These results further sustained a model by which glutaminolysis (activated by both glutamine and leucine) is necessary to produce αKG and to induce the Rag-mediated translocation of mTORC1 at the surface of the lysosome. On the other hand, glutamine, independently of GLS and GDH, sustains the production of ATP and the inhibition of AMPK, necessary for the full activation of mTORC1 at the surface of the lysosome, mediated by the mTORC1 coactivator Rheb28. ASNS and GABA shunt are alternative pathways to metabolize glutamine. So far, our results indicated that glutamine sufficiency produces ATP through mitochondrial activity, but glutaminolysis was not necessary for this. To uncover the pathways involved in glutamine-induced ATP production, we performed a metabolomics analysis (Fig. 5A). We observed that, during amino acid starvation (72 h), the levels of a large number of metabolites dropped. Addition of leucine did not restore the vast majority of them, again confirming that leucine has mostly an allosteric role, not having a significant impact on the metabolism of the cell by itself. In contrast, the addition of glutamine changed completely the pattern, allowing us to identify a group of metabolites which levels raised very high, particularly GABA, asparagine, glutamate and aspartate (Fig. 5A and Supplementary 5A–F). This result indicated an increased synthesis of asparagine, mediated by the activity of ASNS, responsible for the production of asparagine and glutamate from glutamine and aspartate (Supplementary 5G). To experimentally test this possibility, we traced glutamine during glutamine sufficiency in U2OS cells using [U13C]- glutamine. As expected, we observed an accumulation of both glutamate m +5 and asparagine m +4 in the intracellular extract of U2OS cells after 72 h of treatment. Further, glutamine sufficiency also produced m +4 labeled aspartate (Fig. 5B), indicating that aspartate used by ASNS was indeed originated (recycled) from the glutamate through oxaloacetate generation at the TCA cycle. Glutamate produced by ASNS was further converted into aspartate likely through the activity of the enzyme aspartate transaminase (GOT1) (Supplementary 5G). In parallel, we also incubated amino acid-starved cells in the presence of [15N2]- glutamine for 72 h. In agreement with our previous observations, we observed a large accumulation of 15N-asparagine, confirming the implication of ASNS in the metabolization of glutamine during glutamine sufficiency (Fig. 5C). The production of both 15N-aspartate and 15N-asparagine in this condition finally Fig. 5 ASNS and GABA shunt are alternative pathways to metabolize glutamine. A Heatmap representation of metabolite levels, as determined by LC–MS analysis, in amino acid-starved U2OS cells incubated with glutamine and/or leucine. The heatmap was created with MetaboAnalyst3.0 with the total pools of the detected metabolites. B13C-labeled metabolite levels, as determined by LC–MS analysis, in U2OS cells incubated with or without all amino acids or leucine alone as indicated, in the presence of (U)-13C-glutamine during 72 h. Total ion counts of glutamate m +5, asparagine m +4, aspartate m +4, GABA m +4, citrate m +4, and malate m +4 are graphed. The percentage of labeling with respect to total metabolite levels is shown in parenthesis for each metabolite. C15N-labeled metabolite levels, as determined by LC–MS analysis, in U2OS cells incubated with or without all amino acids or leucine alone as indicated, in the presence of 15N2-glutamine during 72 h. Ion counts of 15N -glutamate, 15N -asparagine, 15N aspartate, and 15N2asparagine are plotted. The percentage of labeling relative to the total metabolite is shown in parenthesis for each metabolite. DASNS expression was knocked down using small interfering RNA (siRNA) in U2OS cells during 48 h. Cells were then treated with glutamine and BPTES as indicated for 72 h and the ATP/ADP ratio was measured. Scramble non-targeting siRNA was used as a control. Immunoblot of ASNS levels is presented as a control of the knockdown. EImmunoblot analysis of AMPK phosphorylation in U2OS cells treated as in (D). FRelative mRNA expression levels of ASNS as determined by qPCR in amino acid-starved U2OS cells incubated as indicated during 72 h. GATP/ADP ratio of amino acid-starved U2OS cells incubated with glutamine and/or DMKG during 72 h. HImmunoblot analysis of mTORC1 downstream target (S6K, S6, and 4EBP1 phosphorylation) and AMPK phosphorylation in U2OS cells incubated during 24 h in the presence of all amino acids with dual inhibition of GLS and/or ASNS. ASNS expression was knocked down using small interfering RNA (siRNA) in U2OS cells during 48 h. Scramble non-targeting siRNA was used as a control. Cells were then incubated in the presence of absence of all amino acids and/or BPTES as indicated during 24 h. ISchematic representation of the two branches model connecting glutamine metabolism and mTORC1 signaling. Graphs show mean values ± SEM (n=3 biologically independent experiments). *p< 0.05 (ANOVA analysis followed by a post hoc Bonferroni test). Source data are provided as a Source Data file. NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-25079-4 ARTICLE NATURE COMMUNICATIONS | (2021) 12:4814 | https://doi.org/10.1038/s41467-021-25079-4 | www.nature.com/naturecommunications 9