mTORC2 responds to glutamine catabolite levels to modulate the hexosamine biosynthesis enzyme GFAT1

Metabolic pathways are defective in the absence of SIN1

As we have shown previously, mTORC2 integrity and function is disrupted in SIN1^−/− MEFs (Jacinto et al., 2006). In order to identify new mTORC2 targets or regulators, we performed microarray analysis of mRNA from SIN1^−/− MEFs in comparison to WT MEFs. Changes in expression of genes whose products are involved in metabolism accounted for >25% of the total (Figure 1A). Furthermore, genes involved in nitrogen-related metabolism comprised the majority (14%). By qRT-PCR, we verified the mRNA expression changes of some genes that are particularly involved in metabolism (Figure 1B and Table S1). We therefore performed metabolite analysis by mass spectrometry to gain insights on metabolic pathways that could be regulated by mTORC2. An overall decrease in metabolites from glycolysis, TCA cycle, and nucleotide metabolism occurred in SIN1^−/− MEFs relative to WT (Figure 1C). Furthermore, amino acids that could be synthesized from glutamine catabolism were also diminished. These cells displayed sluggish growth, consistent with decreased ATP and lactate production (Figure 1C and S1). Thus, there is a profound defect in metabolic pathways in the absence of mTORC2.

mTOR associates with GFAT1 and the hexosamine biosynthesis pathway is defective in the absence of SIN1

We also screened for possible mTORC2 targets by immunoprecipitation of mTOR from fractionated MEFs combined with mass spectrometry. One of the proteins that associated with mTOR specifically in low and medium density fractions in WT but not in SIN1^−/− was GFAT1 (Figure 2A and 2B). The mRNA encoding GFAT1 (Gfpt1) was also pronouncedly reduced in SIN1^−/− (Table S1 and Figure 1B). Immunoprecipitation of GFAT1 from fractionated extracts revealed that the majority of GFAT1 was present in the cytosol but was detectable in the membrane, particularly in the high-speed pellet fractions (Figure 2C). GFAT1 co-immunoprecipitated the mTORC2 components mTOR, rictor and SIN1 particularly in the high-speed fractions. Since one of GFAT1 substrates is glutamine, we examined if its interaction with mTORC2 could be sensitive to this amino acid. Prolonged (12 hr) glutamine withdrawal diminished GFAT1 in the membrane and the interaction with mTORC2 became more discernible in the cytosol fractions (Figure 2C). Hence, GFAT1 interacts with mTORC2 and this compartmental interaction is sensitive to the amounts of glutamine.

We then examined GFAT1 protein levels in SIN1^−/− MEFs and found that it was markedly diminished compared to WT (Figure 2D). This reduction can be rescued by overexpression of HA-SIN1 in SIN1^−/− (Figure 2E), indicating that the defective expression is specifically due to SIN1 loss. SIN1 deficiency also abrogated GFAT1 in the membrane, correlating with increased localization of rictor and mTOR in these cells (Figure 2F). Next, we compared the HBP (Figure 2G) in WT vs SIN1^−/− MEFs. By mass spectrometry, UDPGN was indeed considerably reduced in SIN1^−/− MEFs (Figure 2H). Interestingly, the immediate product of GFAT1, glucosamine-6-phosphate, was slightly elevated in these cells. This could be due to accumulation of this product owing to defects in the distal HBP reactions. Consistent with this notion, SIN1^−/− cells have decreased Uap1 (Figure 1B and Table S1), the enzyme catalyzing the terminal reaction in the HBP, along with diminished acetyl-CoA and nucleotides (Figure 1C), thus attenuating UDPGN production. Total cytosolic O-GlcNAc levels were also diminished in SIN1^−/−, further supporting defective HBP (Figure 2I). Moreover, as we have shown previously (Chou et al., 2014), CD147, a highly glycosylated cell surface protein involved in lactate transport (Le Floch et al., 2011), was present mainly as a low glycosylated form in mTORC2-disrupted cells. This form was sensitive to tunicamycin, which blocks N-glycosylation (Figure S2C). Using an in vitro assay, total cellular GFAT1 activity was lower when mTORC2 is disrupted, consistent with the reduced GFAT1 proteins in the SIN1^−/− MEFs (Figure 2J). Total GFAT1 activity in these cells was augmented by reconstitution with HA-SIN1β (Figure 2J). These results reveal that mTORC2 disruption leads to diminished GFAT1 expression and defective HBP.

Acute glutamine withdrawal activates mTORC2 but extended glutamine starvation diminishes mTORC2 activation and GFAT1 expression

To determine how mTORC2 is involved in the regulation of the HBP, we next addressed how mTORC2 regulates GFAT1 expression. We therefore examined if the nutrients that GFAT1 responds to could regulate its expression in an mTORC2-dependent manner. We incubated growing cells in fresh complete DMEM in the presence or absence of glutamine at the indicated hours. By 12 hrs in MEFs and 36 hrs in HeLa, there were no discernible changes in GFAT1 protein levels (Figure 3A) and cell viability (Figures S3A–B). However, when glutamine withdrawal was extended past these periods, the expression of several proteins including GFAT1 diminished (Figures 3B–C). Cell viability of WT but not SIN1^−/− MEFs plunged after 12 hr glutamine starvation (Figure S3A). In SIN1^−/− MEFs, GFAT1 expression remained lower than WT (Figure 3A). Interestingly, Akt Ser473 phosphorylation was immediately enhanced upon glutamine withdrawal and sustained for several hours but waned after extended starvation in both WT MEFs and HeLa (Figures 3A–C). Other mTORC2-regulated effectors such as NDRG1 and the constitutively phosphorylated Akt T450 and PKCα/βT638/641 displayed decreased phosphorylation upon prolonged glutamine deprivation (Figures 3B–C). Akt phosphorylation was abolished upon Torin1 treatment (Figure S4A) and remained absent in SIN1^−/− (Figure 3A), validating that this response is mTORC2-dependent. It was also reduced upon re-supplementation with glutamine, indicating that mTORC2 responds to glutamine or its catabolites (Figure S4B). While glutamine withdrawal consistently induced Akt Ser473 phosphorylation in WT MEFs, HeLa (Figures 3A–C) and primary thymocytes (Figure S4C), it simultaneously reduced mTORC1 signaling (S6K and S6 phosphorylation) only in HeLa (Figure 3C) but not in the other cell types (Figures 3A, S4C), arguing against the role of the mTORC1-induced negative feedback loop that purportedly promotes Akt phosphorylation under certain conditions (Harrington et al., 2005). In support of this, IRS-1 phosphorylation at the S6K-mediated sites only diminished upon extended glutamine starvation in both MEFs and HeLa (Figures S4D–F). Withdrawal of other amino acids such as cysteine or leucine also enhanced Akt Ser473 phosphorylation (Figure S4G–H), which was mitigated upon re-supplementation with glutamine (Figure S4I–J). Thus, while acute glutamine withdrawal enhances Akt Ser473 phosphorylation, prolonging glutamine starvation (>12 hr in MEFs and >36 hr in HeLa) decreased phosphorylation of other mTORC2 targets, reduced expression of GFAT1 and other growth promoting regulators such as Akt. These findings suggest that glutamine catabolite levels affect mTORC2 activation and GFAT1 expression.

To further examine whether GFAT1 expression and mTORC2 activation are indeed responsive to declining levels of glutamine or glutamine catabolites, we used the glutaminolysis inhibitor, BPTES, which specifically inhibits the glutaminase GLS. Addition of BPTES did not significantly affect GFAT1 expression until after 24 hrs whereas Akt phosphorylation was immediately enhanced (Figure 3D). In contrast, S6K phosphorylation was abolished. Thus, like glutamine withdrawal, blocking glutaminolysis activates mTORC2. Chronic inhibition of glutaminolysis, like prolonged glutamine withdrawal, eventually leads to downregulation of GFAT1 expression.

Since glutamine is also used as a substrate by amidotransferases in a number of biosynthetic pathways including the HBP, we examined how their inhibition could affect GFAT1 expression and Akt phosphorylation. We used the glutamine antagonist/analog, azaserine, which potently blocks amidotransferases including GFAT1 (Lyons et al., 1990). In growing untreated HeLa cells, GFAT1 protein levels remained comparable up to 24 hrs (Figure 3E). Even without glutamine withdrawal, Akt phosphorylation increased over time in HeLa (and PTEN^−/− lymphoma; Figure S4C), likely due to glutamine addiction of these cancer cells. Azaserine treatment led to a decline in GFAT1 levels while Akt phosphorylation intensified earlier (6 hr) (Figure 3E). When azaserine was combined with glutamine withdrawal, Akt phosphorylation was weaker and abolished earlier (by 12 hr) (Figure 3F) with concomitant decrease in GFAT1, along with Akt and S6K expression. These results further support that whereas declining glutamine catabolites potentiate mTORC2 activation, their depletion eventually downregulates expression of GFAT1 and other growth regulators such as Akt.

GFAT1 expression is enhanced and mTORC2 activation is sustained during prolonged glucose withdrawal

Since the glucose metabolite, fructose-6-P, is the other GFAT1 substrate, we next analyzed how glucose starvation could affect GFAT1 expression and Akt phosphorylation. Glucose withdrawal did not decrease GFAT1 protein levels (Figure 4A). In fact, an enhancement of its mRNA occurred by 9 hr starvation in WT MEFs, which did not occur in SIN1^−/− MEFs (Figure 4B). Viability of WT MEFs declined after 12 hr glucose starvation (Figure S3B). In HeLa, prolonged glucose starvation led to a striking increase in GFAT1 protein by 24 hrs and these cells remained viable for longer periods (Figures 4C and S3E). This enhancement was abolished by Torin1 treatment (Figure S5A). As previously reported, the expression of the ER chaperone BiP (Grp78) was concomitantly triggered upon prolonged glucose withdrawal (Figures 4A and 4C) (Munro and Pelham, 1986). On the other hand, Akt Ser473 phosphorylation was robustly increased upon glucose withdrawal and sustained throughout starvation (Figures 4A and 4C). This enhancement could be mitigated by the re-addition of glucose at the last hour of starvation (Figure S5B). The expression of other mTORC2 targets such as PKCα and NDRG1 was also maintained in WT MEFs and became elevated upon prolonged glucose withdrawal in HeLa (Figure S5C–D). No Akt phosphorylation occurred in SIN1^−/− cells even upon glucose withdrawal and despite robust phosphorylation of ERK1/2 (Figure 4A), which was previously linked to glucose stress-induced Akt phosphorylation (Shin et al., 2015). Phosphorylation of the mTORC1 target S6K was slightly enhanced in WT MEFS but abolished during prolonged glucose starvation in HeLa. IRS-1 phosphorylation at the S6K-mediated feedback-loop sites also diminished only upon prolonged glucose withdrawal (Figure S5C). Hence, the induction of Akt Ser473 phosphorylation at early time points of glucose withdrawal is likely not due to the mTORC1-mediated feedback loop. Together, these results reveal that similar to glutamine withdrawal, mTORC2 is activated upon glucose withdrawal. Unlike prolonged glutamine withdrawal, extended glucose starvation enhances GFAT1 expression.

Glutamine is required to maintain or enhance GFAT1 expression during glucose starvation

Since prolonged glutamine but not glucose starvation diminished GFAT1 expression, we then asked whether glutamine is in fact required for maintaining or enhancing GFAT1 expression during glucose deprivation. Combined withdrawal of both nutrients accelerated cell death such that we were unable to analyze GFAT1 expression past 6 hrs in MEFs (Figures 4D, S3C). At 6 hrs, GFAT1 levels appeared comparable to non-starved condition. Akt phosphorylation was strongly induced upon acute withdrawal of both nutrients but was abolished by 6 hrs (Figure 4D). Phosphorylation of other mTORC2 targets and effectors also declined at this point (Figure S5E). On the other hand, S6K phosphorylation was abrogated by 3 hrs of starvation. In SIN1^−/− MEFs, GFAT1 expression was further diminished, Akt phosphorylation remained absent and cell viability declined after 6 hrs (Figure S3C). Similar patterns were observed in HeLa wherein GFAT1 expression also remained comparable (Figure 4E). Cell viability of HeLa cells plunged abruptly after 12 hours (Figure S3F). Notably, phosphorylation of Akt during combined glucose and glutamine withdrawal was potentiated compared to withdrawal of either glucose or glutamine alone (Figure 4E). Interestingly, the enhancement of Akt phosphorylation was mitigated when excess glutamine was added back after the first hour of glucose starvation but was less apparent if added after prolonged starvation (Figure 4F). Protein and mRNA expression of GFAT1 in HeLa was also robustly enhanced upon prolonged glucose withdrawal and was further augmented when glutamine was supplemented at the last hour of starvation (Figure 4F–G). Hence, during glucose starvation, declining glutamine catabolites could induce mTORC2 activation. Moreover, glutamine and mTORC2 are required to maintain or enhance GFAT1 expression levels under glucose-deprived conditions.

To further investigate how glutamine catabolites are required for GFAT1 enhancement upon prolonged glucose withdrawal, we examined how BPTES could affect this response. BPTES incubation blunted the enhancement in GFAT1 expression upon glucose withdrawal (Figure 4H), indicating that glutaminolysis is required to allow robust enhancement of GFAT1 expression in HeLa. In contrast, Akt phosphorylation upon glucose withdrawal was further intensified when BPTES was present. However, at 36 hours starvation and BPTES incubation, Akt, along with GFAT1 expression became downregulated. When glucose-withdrawn cells were co-incubated with azaserine, GFAT1 expression was markedly reduced while Akt phosphorylation remained elevated (Figure S5F). Together, these findings support that glutamine catabolites are required for enhancement of GFAT1 expression and declining levels of these catabolites during glucose starvation activate mTORC2.

Glutamine is required for the mTORC2-dependent increase in GFAT1 expression via Xbp1s

GFAT1 expression is mediated by the unfolded protein response (UPR) effector, spliced Xbp1 (Xbp1s), which serves as a transcriptional regulator of GFAT1 (Wang et al., 2014). We therefore examined if Xbp1 could mediate GFAT1 expression by mTORC2. The microarray data, verified by qRT-PCR analysis, revealed that Xbp1 mRNA was significantly diminished in SIN1^−/− MEFs (Figure 5A and Table S1). Since Xbp1s translocates to the nucleus during UPR induction, we analyzed its protein levels in this compartment. While the unspliced Xbp1 (Xbp1u) is cytosolic, Xbp1s was exclusively found in the nuclear/membrane fractions and accumulated in this compartment by 12 hr glucose withdrawal (Figure 5B) but not during 12 hr glutamine starvation (Figure 5C). In SIN1^−/− cells, there were decreased levels of Xbp1u and the processed Xbp1s failed to accumulate in the nucleus. When glucose and glutamine were simultaneously withdrawn, Xbp1s also failed to accumulate in the nuclear fractions in both WT and SIN1^−/− (Figure 5D), despite robust activation of p38 (Figure S6A), which is implicated in Xbp1 regulation and poorly activated in rictor-deficient liver (Lamming et al., 2014). Likewise, in HeLa, Xbp1s accumulated in nuclear fractions upon glucose but not glutamine starvation (Figure 5E). Furthermore, when glucose-starved cells were supplemented with excess (8 mM) glutamine, nuclear accumulation of Xbp1s was augmented (Figure 5F). These findings corroborate the requirement for glutamine catabolites in enhancing GFAT1 expression.

TCA cycle metabolites are maintained during glucose withdrawal but become downregulated during glutamine starvation

To obtain clues on which intracellular glutamine catabolite(s) could affect mTORC2 activation and GFAT1 expression during glucose withdrawal, we performed metabolomics on MEFs subjected to acute (1 hr) vs prolonged glucose starvation (12 hr). Glycolytic metabolites were pronouncedly diminished by 1 hr and remained low at 12 hr (Figure 6A). UDPGN and metabolites from nucleotide synthesis (particularly UDP-glucose) were also diminished by 12 hr. In contrast, TCA cycle/glutaminolysis intermediates were augmented by 12 hr. Non-essential amino acids that can be generated from TCA intermediates were increased by 12 hr as well. These results are consistent with glutamine being funneled towards glutaminolysis to provide TCA cycle intermediates during glucose withdrawal.

Next, we examined the metabolite changes upon glutamine withdrawal. Whereas glycolytic, nucleotide synthesis metabolites and amino acids (except glutamine) remained relatively comparable at 1 hr starvation, the TCA cycle metabolites and UDPGN diminished (Figure 6B). By 20 hr of glutamine starvation, UDPGN recovered and glycolytic metabolites were slightly enhanced. Nucleotide levels remained comparable if not slightly enhanced. In contrast, TCA cycle intermediates and their amino acid derivatives further decreased. When both glucose and glutamine were simultaneously withdrawn, metabolites of glycolysis, TCA cycle and nucleotide synthesis further decreased (Figure 6C). These findings suggest that the presence of glutamine is required to maintain levels of TCA cycle/glutaminolysis metabolites and that these metabolites could particularly play a role in regulating GFAT1 expression.

mTORC2 is required for a robust response to glucose withdrawal by ensuring availability of sufficient glutamine catabolites

Because TCA cycle/glutaminolysis metabolites were maintained during glucose starvation but became attenuated by glutamine withdrawal (Figure 6A–B), we examined whether their levels could be defectively regulated in SIN1^−/− MEFs during starvation. Indeed, unlike the WT MEFs (Figure 6A), TCA cycle/glutaminolysis metabolites became downregulated in SIN1^−/− MEFs during glucose withdrawal (Figure 7A). These findings suggest that one or more metabolites from TCA cycle/glutaminolysis could play a role in maintaining or enhancing GFAT1 expression. We therefore examined how glutamate and α-ketoglutarate (α-KG), which can be generated by glutaminolysis, could enhance GFAT1 expression under conditions of starvation. In WT MEFs, addition of glutamate or dimethyl-α-ketoglutarate (DKG), a more cell-permeable precursor of α-KG, to media lacking both glucose and glutamine led to an increase in the levels of GFAT1 mRNA (Figure 7B). Furthermore, whereas GFAT1 protein levels decreased upon prolonged glutamine withdrawal, supplementation with DKG prevented this decrease (Figure 7C). It is noteworthy that enhanced GFAT1 were found in the low speed pellet fractions. Interestingly, Akt phosphorylation in the cytosol was also diminished effectively by addition of DKG. Thus, α-KG could serve as a key TCA cycle/glutaminolysis metabolite that prevents downregulation of GFAT1 expression.

We next examined if DKG could restore normal expression of GFAT1 in SIN1^−/− cells. However, its effect on GFAT1 was quite minimal (Figure 7D). Since Xbp1 expression was lower in SIN1^−/− MEFs, we then overexpressed Xbp1s to examine if this would boost GFAT1 expression. This was not sufficient by itself either (Figures 7E and S6B) but when combined with supplementation of DKG, GFAT1 protein levels became comparable to GFAT1 from WT cells by 6 hr incubation (Figure 7E). Thus, in addition to Xbp1s, sufficient α-KG levels are required for optimal GFAT1 protein expression.

Our findings have indicated that mTORC2 is required to respond to fluctuations in glutamine catabolites. In addition to regulating GFAT1, it could be involved in ensuring sufficient glutamine under starvation conditions in order to restore metabolic homeostasis. To address this, we took advantage of primary thymocytes that are deficient of rictor, raptor, or both (Figure S6C). In the thymus, the majority of the thymocytes consist of the quiescent Double Positive (DP) CD4⁺CD8⁺ population (~80%). However, there is a highly proliferating CD8-immature single positive (CD8-ISP) subset (Chou et al., 2014), which comprises only about 1% of the thymocyte population. We therefore compared how the highly proliferating CD8-ISP vs the quiescent DP would respond to glucose starvation by measuring the surface levels of CD98, a subunit of the glutamine transporter. In wild type CD8-ISP cultured ex vivo in complete media, CD98 increased fast but transiently over time and returned to basal levels by 20h (Figure 7F). Glucose withdrawal further upregulated CD98 fast and remained elevated up to 20 hrs. CD98 expression was consistently lower on rictor-deficient CD8-ISP thymocytes but glucose withdrawal augmented its expression to the level of wild type cells cultured in complete media. In contrast, raptor-deficient CD8-ISP thymocytes, which also had lower basal CD98 expression, gradually enhanced CD98 expression in absence of glucose to reach similar levels after 20h as observed in wild type cells upon glucose starvation. CD8-ISP thymocytes that are deficient of both rictor and raptor have the lowest CD98 expression and failed to upregulate its expression upon glucose withdrawal. These results reveal the involvement of both mTORC1 and mTORC2 in promoting glutamine metabolism. Without raptor, cells had a delayed response but without rictor, proliferating cells are unable to mount a robust response upon glucose starvation.

In the quiescent DP population, both wild type and rictor-deficient thymocytes remained insensitive to glucose withdrawal as neither upregulated CD98 expression (Figure 7G). However, when PTEN was deleted in the thymus, leading to lymphoma development of mice older than 11 weeks, the DP lymphoma cells became highly proliferative and sensitive to glucose withdrawal by upregulating CD98 expression (Figure 7G). Overall, these findings further support that highly proliferating cells require mTORC2 to mount a robust response to nutrient starvation.

Cell culture and stimulation

Wild type and SIN1^−/− MEFs (passaged no more than 4X) and HeLa were cultured in complete DMEM. After culturing for 24 hours reaching typically 60% confluency, cells were resuspended in either fresh complete media or starvation media (DMEM without glucose, without glutamine, without pyruvate), supplemented with dialyzed FBS, in the absence or presence of 25 mM glucose and/or 2 mM glutamine as indicated.

Cell Fractionation and co-immunoprecipitation

Cells were lysed in CHAPs lysis buffer and centrifuged at 14000 rpm unless otherwise noted. The soluble (cytosolic) lysates were subjected to SDS-PAGE and immunoblotting. Pellets, which contain membrane compartments, were further solubilized using RIPA and processed for immunoblotting as indicated. Cell fractionation using sucrose gradients or differential centrifugation and co-immunoprecipitation are described in detail in Supplemental Experimental Procedures.

Metabolic profiling and genomic microarray analysis

MEFs were harvested in triplicates or sextuplicates and intracellular metabolites were extracted by cold methanol extraction followed by LC-MS analysis. For microarray analysis, growing WT and SIN1^−/− MEFs were harvested and sample processing was done by Rutgers Univ. functional genomics facility using their standard protocol. Gene expression analyses were carried out using Affymetrix GeneChip Mouse Genome 430 2.0 Arrays. Pairwise differential expression analysis between the WT and SIN1^−/− MEFs was performed. Data normalization and expression values were transformed to Log, Student t-test with Benjamin and Hochberg multiple testing correction was performed. Relative mRNA expression levels of genes of interest were assessed by qRT-PCR.

Mice and Flow Cytometry

Homozygous C56BL/6 rictor^f/f, raptor^f/f, or PTEN^f/f (Jackson Lab) mice were crossed with C56BL/6 Lck-Cre mice (Taconic Farms, Germantown, NY), which generates T cell-specific rictor (rictor^T−/−), raptor (raptor^T−/−), or PTEN (PTEN^T−/−) knockout mice, respectively. All mice were genotyped by PCR using the respective primers described in Supplemental Information. Handling and experimentation protocols have been reviewed and used in accordance with Institutional Animal Care and Use Committee regulations of Rutgers Univ.