Glutamine and asparagine activate mTORC1 independently of Rag GTPases

Nutrient sensing by cells is crucial, and when this sensing mechanism is disturbed, human disease can occur. mTOR complex 1 (mTORC1) senses amino acids to control cell growth, metabolism, and autophagy. Leucine, arginine, and methionine signal to mTORC1 through the well-characterized Rag GTPase signaling pathway. In contrast, glutamine activates mTORC1 through a Rag GTPase–independent mechanism that requires ADP-ribosylation factor 1 (Arf1). Here, using several biochemical and genetic approaches, we show that eight amino acids filter through the Rag GTPase pathway. Like glutamine, asparagine signals to mTORC1 through Arf1 in the absence of the Rag GTPases. Both the Rag-dependent and Rag-independent pathways required the lysosome and lysosomal function for mTORC1 activation. Our results show that mTORC1 is differentially regulated by amino acids through two distinct pathways.

activation. Growth factors alone cannot attain maximal mTORC1 activity without amino acid supplementation (5). Increased amino acid concentrations promote mTORC1 lysosomal localization and subsequent activation. Downstream of growth factors, the small G-protein Rheb binds to and activates mTORC1 at the lysosome. Rheb resides at the lysosomal surface via the last 15 amino acids, including a C-terminal CAAX box (6,7). GTP-bound Rheb directly interacts with mTORC1, and structural analysis of this interaction reveals a conformational change in the mTOR active site and allosteric activation of the kinase (8). Tuberous sclerosis complex (TSC) dissociates away from the lysosome and Rheb, downstream of growth factors, resulting in mTORC1 activation (6). TSC is a GTPaseactivating protein for Rheb and promotes inactive Rheb bound to GDP (9 -11). Mutations in TSC elevate mTORC1 activity and cause tuberous sclerosis and lymphangioleiomyomatosis (12)(13)(14). Thus, amino acids and growth factors intersect at the lysosome to promote mTORC1 activation.
The Rag GTPases link amino acid signaling to mTORC1 activation at the lysosome (5,15). RagA or RagB forms a heterodimer with RagC or RagD, and dimerization is essential for Rag GTPase protein stability and mTORC1 activation (16). In mammals, there are four Rag proteins: RagA and RagB are high in sequence similarity and functionally redundant; likewise, RagC and RagD are also highly related in sequence and redundant. The guanine nucleotide loading of the Rag GTPases is important for their physiological function, where GTP-bound RagA or RagB interacts with the mTORC1 component Raptor at the lysosome. RagC or RagD GDP-bound forms a heterodimer with the GTP-bound RagA or RagB. Other components have been reported to be involved in the Rag GTPase signaling cascade to mTORC1. For example, the Ragulator complex (7,17), the vacuolar H ϩ -ATPase (v-ATPase) (18), GATOR complexes (referred to as GATOR1 and GATOR2) (19), FLCN-FNIP complex (20), KICSTOR complex (21), and SLC38A9 (also referred to as SNAT9) (22)(23)(24)(25) were shown to regulate mTORC1 in a Rag-dependent manner.
In human serum, amino acid concentrations range from 6 to 650 M, depending on diet, age, and disease (43,44). Therefore, to determine the kinetics of AA mTORC1 -induced activation of mTORC1, we starved MEF and HEK293A cells of all amino acids and performed a dose response, utilizing concentrations of each individual amino acid ranging from 100 nM to 1 mM and analyzed mTORC1 activity ( Fig. 1B and Fig. S1 (A and B)). Concentrations in which individual amino acids activated mTORC1 ranged from 10 M to 1 mM (except His in MEF), comparable with the concentrations of amino acids found in human plasma and cell culturing medium. AA mTORC1 could induce the activation of mTORC1 between 15 min and 1 h (Fig.  1C). To exclude the possibility that amino acids that take awhile to activate mTORC1 (e.g. Gln and Asn), indirectly signal to mTORC1 through the degradation of serum proteins, we performed amino acid starvation and stimulation experiments in the absence of serum. Gln and Asn can still signal to mTORC1 (Fig. S1C). Furthermore, because growth factors and amino acids merge at the lysosome to achieve optimal mTORC1 activation, the addition of the growth factor insulin further increased mTORC1 signaling (Fig. S1C).
Amino acid withdrawal experiments were performed, where each individual amino acid was removed from the Dulbecco's modified Eagle's medium (DMEM), and mTORC1 activity was analyzed (Fig. S1D). HEK293A cells were maintained in DMEM supplemented with Ala and Asn (amino acids absent in the DMEM) for several weeks prior to withdrawal experiments. Short-term withdrawal (4 h) of Ala, Met, and Arg decreased mTORC1 activity (Fig. S1D, top). Also, Asn and Gln withdrawal significantly inhibited mTORC1 after 4 h. Long-term withdrawal (48 h) of Ala, Leu, Asn, Gln, Arg, and Ser inhibited mTORC1 (Fig. S1D, bottom). Starvation of cells of all amino acids or Met for 48 h resulted in cell death. We did not observe a dramatic change of mTORC1 activity when starving cells of His, Thr, or Val for 4 or 48 h. Because there are multiple mTORC1 sensors (23,24,29,30,33), withdrawal of only one amino acid may not significantly alter mTORC1 activity. Consistent with the amino acid stimulation experiments (Fig. 1), the withdrawal of seven amino acids (Ala, Met, Arg, Leu, Asn, Gln, and Ser) appears to regulate mTORC1.
To exclude the possibility that phosphorylation of S6K1 at threonine 389 was due to a compensating kinase or inactivation of a phosphatase, we analyzed other mTORC1 substrates, such as the phosphorylation of ULK1 at serine 758 and the electrophoretic mobility shift of 4EBP1. Phosphorylation of S6K1 and 4EBP1 promotes protein synthesis, whereas phosphorylation of ULK1 inhibits autophagy (45,46). Stimulation of cells with AA mTORC1 resulted in the phosphorylation of 4EBP1 (as measured by electrophoretic mobility shift) and ULK1 (Fig. S1E). Furthermore, cells pretreated with the mTORC1 inhibitor rapamycin prior to the addition of each individual amino acid blocked AA mTORC1 signaling to mTORC1 (Fig. 1D). Thus, 10 amino acids activate mTORC1 in both MEF and HEK293A cells. In agreement with previous work, Leu, Arg, Gln, and Met modulate mTORC1 activity (5, 26 -32, 42).

Individual amino acids promote mTORC1 lysosomal localization and require lysosomal function
The lysosome and lysosome machinery are essential for mTORC1 regulation by amino acids (16,18,47,48). Amino acid availability promotes mTORC1 lysosomal localization, where it is subsequently activated by Rheb downstream of growth factors (45,46). Stimulation of cells with only Leu or Gln is sufficient to promote mTORC1 lysosomal localization (5,16). Because our data show that 10 amino acids activate mTORC1 in MEF and HEK293A cells, we investigated whether mTORC1 lysosomal localization was required. Amino acid stimulation promoted the translocation of mTOR (green) to LAMP2-positive lysosomal membranes (red) in MEF cells (Fig. 2, A (left column, first row) and B) (6,7,16). In contrast, significantly less mTOR localized to LAMP2-positive lysosomal membranes in MEF cells starved of amino acids (Fig. 2, A (right column, first row) and B). Each of the 10 amino acids (AA mTORC1 ) alone was able to induce the co-localization of mTOR and LAMP2-positive lysosomal membranes (Fig. 2, A (left column, rows 2-7; right column, rows 2-5) and B). However, amino acids such as Lys, Phe, and Trp, which do not activate mTORC1 in MEF and HEK293A cells (Fig. S1, A and B), were unable to induce the co-localization of mTOR and LAMP2-positive lysosomal membranes (Fig. 2, A (right column, row 6) and B). Corresponding immunoblots were performed in parallel to the imaging experiments as controls and showed that AA mTORC1 induced mTORC1 activation, whereas the other amino acids did not (Fig. 2C). Similar patterns of mTOR localization were also confirmed in HEK293A cells stimulated with Leu, Asn, and Ser (Fig. S2, A and B). Based on our results, the same 10 amino acids (AA mTORC1 ) that regulate mTORC1 activity also promote its lysosomal localization (Figs. 1 and 2 and Figs. S1 and S2 (A and B)).

EDITORS' PICK: Glutamine and asparagine sensing by mTORC1
Both Rag-dependent and Rag-independent activation of mTORC1 require the v-ATPase, responsible for acidifying the lysosome and maintaining lysosomal function (16,18,49). For example, Leu was suggested to accumulate inside the lysosome and activate mTORC1 through an "inside-out" model, whereas v-ATPase serves as a critical component (18). However, it is unclear whether v-ATPase and lysosomal function is a general requirement for all amino acids to activate mTORC1. To test whether v-ATPase function is involved in AA mTORC1 signaling to mTORC1, we treated cells with the v-ATPase inhibitor bafilomycin A (50). Pretreatment of cells with bafilomycin A inhibited mTORC1 activation, even after the addition of individual A, table summarizing the amino acids that activate mTORC1. B, MEF or HEK293A cells were starved of amino acids (ϪAA) for 1-2 h, followed by the addition of each amino acid at the indicated concentration for 30 min. mTORC1 activity was analyzed by immunoblotting for the phosphorylation status of S6K1 (pS6K1) at threonine 389. S6K1 was used as loading control. C, MEF or HEK293A cells were starved of amino acids for 1-2 h, and each amino acid (500 M to 1 mM) was added for the indicated time. D, MEF cells were starved of amino acids for 1-2 h and pretreated with or without 100 nM rapamycin for 30 min, and then amino acids (ϩAA) or individual amino acids (500 M to 1 mM) were added for 30 min. mTORC1 activity was analyzed as in B. NC, normal conditions. EDITORS' PICK: Glutamine and asparagine sensing by mTORC1 amino acids (AA mTORC1 ) or amino acids combined (Fig. S2C). Choloroquine, a v-ATPase-independent inhibitor of the lysosomal pH gradient, also inhibited AA mTORC1 signaling to mTORC1 (Fig. S2D). Collectively, these results show that the v-ATPase and lysosomal function are required for the 10 amino acids (AA mTORC1 ) to activate mTORC1.

Glutamine and asparagine activate mTORC1 independently of the Rag GTPase signaling pathway
We previously discovered that mTORC1 is activated in response to amino acid stimulation in the absence of the Rag GTPases (16). As described previously, there are four Rag proteins in mammals: RagA, RagB, RagC, and RagD (5,15). Rag Because the concentrations of each amino acid in DMEM and MEM nonessential amino acid solution can vary significantly, we tested the 10 amino acids (AA mTORC1 ) at similar concentrations (500 M or 1 mM) (Fig. 1C). We starved MEF or HEK293A cells of amino acids and replenished cells with either amino acids or individual amino acids (Fig. 3 (A (left) and B (left)) and Fig. S3 (A (left) and B (left)). Interestingly, Asn like Gln activates mTORC1 in the absence of the Rag GTPases (Fig. 3A (middle) and Fig. S3A (right)). Gln and Asn activation of mTORC1 in the Rag GTPase KO cells is comparable with that in WT cells (Fig.  3A, right). Likewise, Asn and Gln signaled to mTORC1 without other crucial components in the Rag GTPase signaling pathway, such as GATOR2 (Mios) (Fig. 3B (right) and Fig. S3B). Ser, Thr, and Ala slightly activate mTORC1 in RagA/B and Mios KO cells (Fig. 3, A and B). These amino acids can be metabolized into pyruvate, and it has previously been revealed that pyruvate can regulate mTORC1 through the TTT-RUVBL1/2 complex (42). As mentioned previously, the activity of the Rag GTPases is determined by their guanine nucleotide status, and the overexpression of a constitutively inactive Rag GTPase complex (RagA/B GDP-bound and RagC/D GTP-bound) inhibits amino acid signaling through the Rag GTPases (5,15). Gln and Asn can still activate mTORC1 in WT and RagA/B KO cells expressing RagA/B GDP -RagC/D GTP (Fig. S3C). As mentioned above, amino acids promote lysosomal localization and activation of mTORC1 (Figs. 1 and 2 and Fig. S1 and S2 (A and  B)). Even in the absence of Rag GTPases, Gln still induces lysosomal localization of mTORC1 (16). In addition, like Gln, Asn also promotes mTOR localizing to lysosomes (Fig. S3, D and E). Collectively, these data suggest that Asn and Gln can activate mTORC1, independently of the Rag GTPase signaling pathway (Fig. 3 (A-C) and Fig. S3).
Stimulation of HEK293A cells with saturating concentrations of either Gln or Asn activated mTORC1. However, the simultaneous addition of saturating concentrations of both amino acids further increased mTORC1 activity, suggesting synergy between the amino acids (Fig. 3D). These results are consistent with different sensors or modes of action used by Gln and Asn to activate mTORC1. As expected, the combination of Leu and Arg also synergistically enhanced mTORC1 activity in HEK293A cells. Leu and Arg activate mTORC1 by binding to Sestrin2 and CASTOR1, respectively (29,(33)(34)(35). Taken together, eight amino acids (Ala, Arg, His, Leu, Met, Ser, Thr, and Val) primarily utilize the Rag GTPase-dependent pathway to activate mTORC1, whereas Gln and Asn signal to mTORC1 in the absence of the Rag GTPases. VPS34 and Pib2 have been shown to play a role in the Rag GTPase-independent TORC1 signaling pathway in yeast (38 -40). To test whether VPS34, a class III phosphatidylinositol 3-kinase, plays a role in Gln and Asn signaling to mTORC1, we inhibited VPS34 (VPS34-IN1) in RagA/B KO HEK293A cells. mTORC1 activity was decreased with treatment of VPS34-IN1 under normal culturing conditions. However, Gln-or Asn-induced mTORC1 activation was not significantly changed when cells were treated with VPS34-IN1 (Fig. S4A). Pib2 is a vacuolar membrane-associated phosphatidylinositol 3-phosphatebinding protein with no orthologs in mammals. The Pib2 FYVE domain shares high sequence similarity with LARPF/phafin1 (gene PLEKHF1), whereas the domain on Pib2 (motif E) that interacts with TORC1 is similar to a region in R3H and coiledcoil domain-containing protein 1 (gene R3HCC1). Knockdown of PLEKHF1 or R3HCC1 failed to alter Gln-or Asninduced mTORC1 activation (Fig. S4, B and C).
Gln (31,42,52,53) and Asn withdrawal decreases mTORC1 activity (Fig. S1D). Gln and Asn are both nonessential amino acids that can be made by the cell via Gln and Asn synthetases (GLUL and ASNS, respectively) (54). Glutamine synthesis does not appear to compensate for acute withdrawal of exogenous glutamine, and we did not observe further decrease of mTORC1 when depleting glutamine from GLUL stable knockdown cells (Fig. 3E). In addition, we performed asparagine deprivation through acute ASNS knockdown for cells growing in regular Asn-free DMEM and observed impaired mTORC1 activity (Fig. 3F), similar to a previous report (52). These data

Glutamine and asparagine signaling to mTORC1 requires Arf1
Gln signaling to mTORC1 in the absence of the Rag GTPases requires Arf1 (16). We previously demonstrated that Arf1 guanine nucleotide cycling is crucial for Gln, but not Leu or the Rag GTPase pathway, to induce mTORC1 activation and lysosomal localization. Because Asn, like Gln, can activate mTORC1 independently of the Rag GTPases (Fig. 3, A-C), we tested whether Asn signaling to mTORC1 also requires Arf1. Consistent with our previous findings, treatment of WT MEF cells with brefeldin A (BFA), an Arf1 guanine exchange factor inhibitor (55), modestly reduced mTORC1 activation in response to amino EDITORS' PICK: Glutamine and asparagine sensing by mTORC1 acids (Fig. 4A). BFA inhibited both Gln-and Asn-induced mTORC1 activation in both WT and Rag A/B KO MEFs (Fig. 4,  A-C). Moreover, BFA treatment in WT MEF cells dramatically impaired lysosomal localization of mTORC1 induced by Gln and Asn (Fig. 4, D and E).
In addition to BFA treatment, we also utilized siRNA to knock down Arf1 in WT or RagA/B KO HEK293A cells and analyzed the ability of Gln and Asn to regulate mTORC1. Gln and Asn stimulated mTORC1 activation in WT and Rag A/B KO cells treated with a control siRNA; however, Gln and Asn failed to signal to mTORC1 in cells that were depleted of Arf1 (Fig. 4, F and G). Thus, Arf1 is involved in Gln and Asn signaling to mTORC1, independent of the Rag GTPase pathway.
In summary, we show that eight amino acids filter through the well-studied Rag GTPase pathway (Fig. 4H, left). Whereas the sensors of Leu, Arg, and Met have been identified (29,30,(33)(34)(35)(36), the mechanisms by which Ala, His, Ser, Thr, and Val signal to mTORC1 are still unclear. Importantly, in addition to Gln (16), we discovered that Asn also activates mTORC1 in a Rag GTPase-independent manner and requires Arf1 (Fig. 4H, right). Our results show that mTORC1 is differentially regulated by amino acids through two distinct pathways.

Amino acid starvation and stimulation of cells
Amino acid-free medium was made following the Sigma (#D5796) high-glucose DMEM recipe with the exception that all amino acids were omitted. All experiments with amino acid starvation and stimulation contained 10% dialyzed FBS (#F0392 from Sigma) instead of regular FBS (#F2442 from Sigma) unless otherwise indicated. Amino acid starvation was performed by replacing regular medium with amino acid-free medium for ϳ1-2 h prior to amino acid stimulation unless otherwise indicated. For the confocal experiments, cells were starved of amino acids for 4 h before the addition of amino acids. Glutamine-free DMEM (#D5671 from Sigma) containing 10% dialyzed fetal bovine serum (#F0392 from Sigma) were used in glutamine starvation experiments. For all amino acid stimulation experiments, amino acids were used with the indicated concentration and time points.
Detailed procedures of amino acid stimulation of cells, plasmids, cDNA transfection, RNAi, RNA extraction, reverse transcription, real-time PCR, Western blotting, immunofluorescence microscopy, generation of stable cell lines, generation of Mios knockout cells using CRISPR/Cas9 genome editing, mTOR localization analysis, and statistical analysis are available in the supporting information. . Arf1 is required for glutamine-and asparagine-induced mTORC1 activation. A, WT MEF cells were starved of amino acids (ϪAA) for 2 h and pretreated with or without 10 M BFA for 1 h, and then amino acids (ϩAA), Gln, or Asn was added for 2 h at 4 mM. mTORC1 activity was analyzed by immunoblotting for the phosphorylation status of S6K1 (pS6K1) at threonine 389. S6K1 was used as loading control. B, RagA/B KO MEF cells were starved of amino acids for 2 h and pretreated with or without 10 M BFA for 1 h, and then amino acids, Gln, or Asn was added for 2 h at 4 mM. mTORC1 activity was analyzed as in A. C, immunofluorescence (IF) analysis depicting mTORC1 activity by staining for phospho-S6 (yellow). RagA/B KO MEF cells were starved of amino acids for 2 h and pretreated with or without 10 M BFA for 1 h, and then Gln or Asn was added for 2 h at 4 mM. 4Ј,6-Diamidino-2-phenylindole is shown in blue. D, immunofluorescence analysis of mTOR (green) and the lysosomal marker LAMP2 (red) in WT MEF cells. Cells were amino acid-starved, BFA-pretreated, and Glnor Asn-stimulated as described in B. Higher-magnification images of the depicted area are shown on the right. E, corresponding Western blots were performed in parallel to the imaging (D). mTORC1 activity was analyzed as in A. F, WT HEK293A cells were transfected with control siRNA (Ϫ) or siRNA against Arf1 (ϩ) for 48 h and starved of amino acids for 2 h, and then Gln or Asn was added for 2 h at 4 mM. mTORC1 activity was analyzed as in A. G, RagA/B KO HEK293A cells were transfected with control siRNA (Ϫ) or siRNA against Arf1 (ϩ) for 48 h and starved of amino acids for 2 h, and then Gln or Asn was added for 2 h at 4 mM. mTORC1 activity was analyzed as in A. H, model of Rag GTPase-dependent and -independent mTORC1 pathways. Left, eight amino acids (Ala, Arg, His, Leu, Met, Ser, Thr, and Val) activate mTORC1 through the Rag GTPase pathway. Right, two amino acids (Gln and Asn) activate mTORC1 independently of the Rag GTPases, whereas the activity of Arf1 is required through an unknown mechanism. Both pathways require the v-ATPase, the lysosome, and growth factor signaling (through TSC-Rheb) to achieve optical mTORC1 activation.