Research ArticleCell Biology

Ammonia Derived from Glutaminolysis Is a Diffusible Regulator of Autophagy

See allHide authors and affiliations

Science Signaling  27 Apr 2010:
Vol. 3, Issue 119, pp. ra31
DOI: 10.1126/scisignal.2000911


Autophagy is a tightly regulated catabolic process that plays key roles in normal cellular homeostasis and survival during periods of extracellular nutrient limitation and stress. The environmental signals that regulate autophagic activity are only partially understood. Here, we report a direct link between glutamine (Gln) metabolism and autophagic activity in both transformed and nontransformed human cells. Cells cultured for more than 2 days in Gln-containing medium showed increases in autophagy that were not attributable to nutrient depletion or to inhibition of mammalian target of rapamycin. Conditioned medium from these cells contained a volatile factor that triggered autophagy in secondary cell cultures. We identified this factor as ammonia derived from the deamination of Gln by glutaminolysis. Gln-dependent ammonia production supported basal autophagy and protected cells from tumor necrosis factor–α (TNF-α)–induced cell death. Thus, Gln metabolism not only fuels cell growth but also generates an autocrine- and paracrine-acting regulator of autophagic flux in proliferating cells.


Cancer cells exhibit several notable metabolic alterations that confer a growth advantage in the stressful tumor microenvironment. Foremost among these metabolic abnormalities is the reliance on glycolysis for energy production under both aerobic and hypoxic conditions (commonly termed the Warburg effect) (1, 2). Metabolism of glucose to lactate supplies most of the bioenergetic needs of cancer cells and reduces the requirement for mitochondrial oxidation of glucose-derived carbon to generate metabolic energy in the form of adenosine triphosphate (ATP). During oxidative metabolism of glucose, pyruvate is converted to acetyl–coenzyme A (CoA) instead of lactate, and acetyl-CoA donates its carbons to the tricarboxylic acid (TCA) cycle in the mitochondria. In addition to its role in supporting mitochondrial oxidative phosphorylation, the TCA cycle supplies critical intermediates for the biosynthesis of metabolic building blocks (for example, nucleic acids and fatty acids) as well as NADPH (reduced form of nicotinamide adenine dinucleotide phosphate) (3). Whereas the requirement for mitochondrial ATP production is reduced in glycolytic tumor cells, the demand for TCA cycle–derived biosynthetic precursors and NADPH is unchanged or even increased. Under these conditions, carbon influx into the TCA cycle comes from a source other than glucose, typically the nonessential amino acid, glutamine (Gln) (3, 4). Compared to normal cells, many cancer cells exhibit high rates of Gln transport (5) and metabolism (6), in part because of high glutaminase activity in mitochondria (7, 8). In mitochondria, Gln is sequentially deaminated to glutamate and ultimately α-ketoglutarate (α-KG), which enters the TCA cycle. A by-product of glutaminolysis is ammonia, a potentially toxic metabolite that is removed by diffusion or transport out of the cell or through incorporation into α-keto acids, such as pyruvate, to generate the corresponding amino acid (alanine in this example) (9).

Autophagy supports cell survival during metabolic stress by packaging intracellular proteins and organelles into autophagosomes for delivery to and degradation by lysosomes, effectively recycling macromolecules into metabolic precursors that support processes essential for cell survival. In addition to its function in the starvation response, autophagy contributes to the maintenance of cellular homeostasis by facilitating the turnover of damaged proteins and organelles (10). Hence, basal autophagy promotes cell and organismal fitness, and alterations in autophagic activity have been implicated in the pathogenesis of many diseases, including cancer and neurodegenerative diseases (11). In this report, we describe a previously unknown function for Gln-derived ammonia as a diffusible regulator of autophagy in human cells.


Production of a diffusible, autophagy-inducing factor

We initially observed that U2OS cells showed increasing numbers of autophagosomes with increasing time in culture without medium replacement, as measured with an ectopically expressed green fluorescent protein (GFP)–LC3 reporter protein (Fig. 1A). The cell-conditioned medium (CCM) from these primary cultures, but not fresh, unconditioned medium, also triggered an autophagic response in secondary cultures of freshly plated U2OS cells, as measured by GFP-LC3 puncta counts (Fig. 1, B and C) or by the generation of cleaved, lipid-modified LC3 (LC3-II) (Fig. 1D). Similar results were obtained whether the primary cultures were incubated in fetal bovine serum–supplemented or serum-free medium (Fig. 1, C and D).

Fig. 1

Prolonged cell culture induces autophagy. (A) GFP-LC3–expressing U2OS cells were plated on day 0, and autophagy was monitored by GFP-LC3 puncta formation on days 1 to 4. Scale bar, 50 μm. (B) CCM from U2OS cells cultured for 4 days in complete medium (10% FBS in DMEM) or serum-free medium (SFM; DMEM alone) was collected and transferred to secondary, day-old cultures of U2OS GFP-LC3 cells. GFP-LC3 puncta were visualized after 24 hours. Scale bar, 50 μm. (C) Treated cells from (B) were imaged for GFP-LC3 and the number of cells positive for GFP-LC3 puncta was assessed. Data represent means ± SEM from three independent experiments (>150 cells per experiment). *P < 0.05. (D) Treated cells from (B) were lysed and analyzed for GFP-LC3 processing and endogenous LC3 lipidation. (E) U2OS GFP-LC3 cells were treated with serum-free CCM for the indicated amounts of time. Autophagic activity was determined by immunoblotting for free GFP and endogenous p62 protein.

A steady-state increase in the number of GFP-LC3 puncta or the amount of lipid-modified LC3 does not necessarily reflect a true stimulation of autophagic activity (12). For example, cellular treatment with bafilomycin A1, an agent that blocks autophagosome fusion with lysosomes, also provoked the accumulation of GFP-LC3 puncta and LC3-II (fig. S1, A and B). Consequently, we used two additional assays to ensure that we were observing an actual increase in autophagic flux in the U2OS cell cultures. The first assay monitors the processing of GFP-LC3 to free GFP in autolysosomes (1215). Fusion of GFP-LC3–containing autophagosomes with the lysosome generates the autolysosome, which mediates the proteolysis of the GFP-LC3 reporter, as well as that of endogenous LC3. However, the free GFP moiety generated during GFP-LC3 degradation is relatively resistant to proteolysis; hence, the appearance of free GFP reflects cargo delivery from autophagosome to lysosome. Increased numbers of GFP-LC3 puncta, combined with an increase in the abundance of free GFP, provide a reliable indicator of increased autophagic flux (14). The assay was validated with established modulators of autophagic flux. Stimulation of autophagy by the mammalian target of rapamycin (mTOR) inhibitor CCI-779 (16) increased free GFP in the U2OS reporter line. In contrast, inhibition of the autophagic pathway at an early stage [by depletion of Atg5 or Ulk1, two proteins that are required for autophagy (17)] or late stage [by bafilomycin A1 treatment, which inhibits autolysosome formation and function (18, 19)] suppressed the accumulation of free GFP in these cells (fig. S1, B and C). A second assay that reflects autophagic flux involves measuring the abundance of p62 protein, which is degraded mainly through autophagic clearance. Thus, induction of autophagy leads to decreased p62 abundance, whereas downstream suppression of autophagic flux causes the accumulation of this protein (fig. S1B) (12, 20). U2OS CCM stimulated both increased GFP-LC3 processing (Fig. 1, D and E) and decreased p62 abundance (Fig. 1E), confirming that the higher numbers of GFP-LC3 puncta observed in these cells was attributable to the stimulation of autophagy, rather than to a more distal block in autolysosome formation or function.

An obvious explanation for the autophagy-inducing activity observed in U2OS CCM is that nutrient depletion by the proliferating cells triggers a starvation response, leading to autophagy. Therefore, we replenished the U2OS CCM with amino acids, glucose, pyruvate, or vitamins at concentrations found in fresh medium and applied the reconstituted CCM to secondary U2OS cell cultures. Surprisingly, replenishment of these nutrients, individually or in combination, did not prevent the induction of autophagy by CCM (Fig. 2A). To confirm this finding, we mixed fresh medium or CCM with an equal volume of twofold concentrated fresh medium. This manipulation raised the concentrations of all nutrients to at most 1.5-fold higher than that found in fresh medium. The mixture containing fresh medium supported both normal cell viability and the stimulation of autophagy by CCI-779. In contrast, the CCM–fresh medium mixture directly stimulated autophagy in secondary U2OS cell cultures (Fig. 2B). In addition to the depletion of nutrients, CCM also contained reduced bicarbonate concentrations (Fig. 2C). However, replenishment of bicarbonate to the original concentration in fresh medium (44 mM) did not prevent induction of autophagy by CCM (Fig. 2D). Furthermore, a change in the medium buffering system from bicarbonate-based to Hepes-based did not impair the autophagy-inducing activity of CCM (Fig. 2E). Collectively, these results indicate that autophagy stimulation by CCM is not a consequence of nutrient depletion or alterations in pH during prolonged cell culture. Furthermore, CCM generated by various transformed and nontransformed human cell lines triggered variable but detectable autophagy, both in the host cells themselves and in the U2OS reporter line (Fig. 2, F and G).

Fig. 2

Autophagy induction by CCM is not caused by nutrient depletion. (A) U2OS GFP-LC3 cells were treated with fresh serum-free medium or CCM. Amino acids, glucose, pyruvate, and vitamins were replenished to concentrations found in fresh medium, and autophagy was determined after 24 hours. Panels depict nonadjacent lanes from the same blot. (B) U2OS GFP-LC3 cells were treated with fresh medium, CCM, or CCI-779, alone or diluted 1:1 with fresh serum-free medium at twice the standard concentration (final concentration = 1.5×). Autophagy was determined after 24 hours by GFP-LC3 processing and endogenous LC3 lipidation. (C) The concentration of CO2-NaHCO3 was measured in fresh and CCM samples. Data represent means ± SEM from three independent experiments. (D) U2OS GFP-LC3 cells were treated with fresh medium or CCM, untreated or adjusted to 44 mM final bicarbonate concentration. Autophagy was determined after 24 hours by GFP-LC3 processing. (E) CCM was generated with 10 mM Hepes (pH 7.4) as a buffer instead of sodium bicarbonate. U2OS GFP-LC3 cells were then treated with Hepes-buffered fresh medium or CCM. All incubations were performed in the absence of CO2. (F) Cell lines were cultured for 2 days in serum-free DMEM. CCM was collected, and all the samples were adjusted to a final concentration of 30 mM sodium bicarbonate. Secondary cultures of U2OS GFP-LC3 cells were treated for 24 hours with the indicated CCM samples. (G) MDA-MB-361 (breast adenocarcinoma) and H4 (neuroglioma) cells stably expressing GFP-LC3 were treated with fresh medium or host-CCM. Autophagy was determined after 24 hours by GFP-LC3 processing.

An alternative hypothesis to explain these findings posits that CCM contains a diffusible activator of autophagy. Initial efforts to characterize this putative autophagy-inducing factor revealed that the activity was stable after both heating and multiple freeze-thaw cycles (fig. S2A). In addition, the factor passed through a 1-kD cutoff membrane during ultrafiltration (fig. S2B), was retained on a cation exchange column (fig. S2C), and was lost during efforts to concentrate the CCM by lyophilization or evaporation (fig. S2, D and E). The latter finding suggested that the factor might be volatile, and, indeed, we found that rotary evaporation and distillation of CCM led to the transfer of the autophagy-inducing factor into the distillate (Fig. 3, A to C).

Fig. 3

Ammonia is the autophagy-inducing factor in CCM. (A) Fresh serum-free medium or CCM from U2OS cells was subjected to rotary evaporation (Rotovap). U2OS GFP-LC3 cells were treated for 24 hours with the nonvolatile evaporate fraction (after resolubilization with water) or the liquid distillate fraction (after addition of one-tenth volume of 10× DMEM) with fresh FBS. Cells were imaged for GFP-LC3 puncta. Scale bar, 50 μm. (B) Treated cells from (A) were imaged for GFP-LC3 and the number of cells positive for GFP-LC3 puncta was assessed. Data represent means ± SEM from three independent experiments (>150 cells per experiment). *P < 0.05. (C) Treated cells from (A) were lysed and assayed for GFP-LC3 processing and endogenous LC3 lipidation. (D) Samples from (A) were assayed for ammonia concentrations. The accumulation of ammonia parallels the autophagy-inducing activity of the samples. Data represent means ± SEM from three independent experiments. *P < 0.05. (E) U2OS GFP-LC3 cells were treated for 24 hours with increasing concentrations of NH4OH, NaOH, or NH4Cl. Autophagic activity was assayed by GFP-LC3 processing and endogenous LC3 lipidation. (F) U2OS cells were treated with fresh medium, CCM, 100 nM CCI-779, or 2 mM NH4OH. Cells were imaged for GFP-LC3 puncta. Scale bar, 50 μm. (G) Treated cells from (F) were imaged for GFP-LC3 and the number of cells positive for GFP-LC3 puncta was assessed. Data represent means ± SEM from three independent experiments (>150 cells per experiment). *P < 0.001 by ANOVA relative to control cells cultured in fresh medium. (H) U2OS GFP-LC3 cells were treated with 0.75 mM NH4OH in serum-free medium for the indicated times, and autophagic activity was determined by immunoblotting for free GFP and p62. (I) Circulating mouse plasma and tumor interstitial fluids from subcutaneous xenografts (HCT116, H1975, Colo 205, BT-474, and HT-29) were assayed for ammonia concentrations. Data represent means ± SEM from tumor tissues derived from five independent xenografts for each cell line. *P < 0.01 relative to plasma samples from the same host mice.

One metabolite whose physicochemical properties were consistent with all of the characteristics described above is ammonia. Both CCM and the CCM-derived distillate contained millimolar concentrations of ammonia, whereas the nonvolatile residue (evaporate) remaining after rotary evaporation was ammonia-free and devoid of autophagy-inducing activity (Fig. 3, A to D). Direct treatment of cells with ammonia [as ammonium hydroxide (NH4OH)] induced autophagy (Fig. 3E) with kinetics and magnitude similar to that observed with CCM (Fig. 3, F to H, compare to Fig. 1E). The concentrations of ammonia that increased autophagic activity were similar to those found in CCM (Fig. 3, D and E). To determine whether the ammonia concentrations found in CCM were attainable in tumor tissues in vivo, we collected interstitial fluid from multiple human cancer cell line xenografts and found ammonia concentrations (2 to 5 mM) comparable to those seen in CCM, whereas plasma ammonia concentrations in the same mice were consistently ≤0.5 mM (Fig. 3I). We also tested other potentially volatile metabolites (CO2, NO, H2O2, and lactic acid) for autophagy-inducing activity, with uniformly negative results (fig. S3, A and B). Induction of autophagy by NH4OH was not an indirect consequence of altered pH, because NaOH showed modest activity in this assay, whereas the ammonium salt, NH4Cl, induced autophagy to a similar degree as that provoked by NH4OH treatment (Fig. 3E). At a higher concentration (20 mM), both NH4OH and NH4Cl provoked the accumulation of LC3-II, but reduced the cleavage of GFP-LC3, consistent with their known inhibitory effects on lysosomal function at these higher concentrations (21, 22).

Regulation of basal autophagy by glutaminolysis-derived ammonia

Ammonia is generated during the conversion of Gln to α-KG in mitochondria. The anaplerotic function of glutaminolysis can be circumvented by direct addition of the TCA cycle intermediate, α-KG (as a cell-permeable dimethyl ester), to the culture medium (23). We observed that the addition of increasing concentrations of α-KG to complete medium led to a progressive decline in the production of ammonia in primary U2OS cell cultures, consistent with a reduction in the rate of glutaminolysis (Fig. 4A). Furthermore, the addition of α-KG over the same concentration range to the primary cultures suppressed the appearance of autophagy-inducing activity in CCM (Fig. 4B). In contrast, addition of α-KG to previously generated CCM failed to suppress the autophagy-inducing activity in this medium (Fig. 4B). Finally, we generated CCM from U2OS cells cultured in medium lacking Gln, minus or plus α-KG, to determine whether Gln was the primary source of the autophagy-inducing ammonia. Gln-free CCM generated in the absence or presence of α-KG was devoid of autophagy-inducing activity (Fig. 4C). Moreover, add-back of fresh Gln to either batch of CCM failed to restore the autophagy-inducing activity, indicating that Gln was the source, and not simply supportive, of the autophagy-inducing activity in CCM (Fig. 4C). In the absence of all other amino acids, Gln alone was sufficient for the production of active CCM (Fig. 4D). Thus, ammonia derived from glutaminolysis appears to be necessary and sufficient for the activity of the diffusible autophagy-inducing factor in CCM.

Fig. 4

Glutaminolysis is responsible for ammonia-induced and basal autophagy. (A) Serum-free CCM containing Gln was generated in the presence of the indicated concentrations of dimethyl α-KG before measurement of ammonia concentrations. Data represent means ± SEM from three independent experiments. *P < 0.001 by ANOVA compared to CCM alone. (B) U2OS GFP-LC3 cells were treated with Gln-containing CCM generated in the presence of α-KG, or with fresh medium or CCM containing fresh α-KG. (C) U2OS GFP-LC3 cells were treated with CCM produced in the presence or absence of Gln or α-KG. CCM generated in the absence of Gln does not promote autophagy. (D) Cells were treated with CCM produced in amino acid–free EBSS, with or without Gln. Addition of Gln is sufficient for generation of autophagy-inducing CCM. (E) Cells were incubated for 8 hours with or without Gln in 10% dialyzed FBS. The culture medium was replaced with fresh medium of the same type, with added bafilomycin A1 (Baf A1; 200 nM) and NH4OH (2 mM), and cells were cultured for an additional 16 hours.

Although stimulation of autophagy is commonly induced by metabolic stress, cells typically exhibit basal autophagy, which likely contributes to normal cellular homeostasis. A previous study reported up-regulation of basal autophagy with increasing Gln concentrations (24). Basal autophagy can be detected by treatment of cells with bafilomycin A1, which results in the accumulation of cleaved, lipid-modified LC3-II as a result of interference with autolysosome formation. Removal of Gln from the cell culture medium substantially reduced this basal autophagic activity, as demonstrated by the reduced accumulation of LC3-II after bafilomycin A1 treatment (Fig. 4E). Addition of ammonia to the Gln-deprived cells partially restored basal autophagy (Fig. 4E), suggesting that glutaminolysis-derived ammonia supports this activity in metabolically active cells.

Nutrient limitation induces autophagy through inhibition of the mTOR complex 1 (mTORC1) (25). Phosphorylation of S6 kinase 1 (S6K1) and 4E binding protein 1 (4E-BP1), which are markers of mTORC1 activity (26), or Akt at Ser473, which is a readout for mTORC2 activity (27), was unchanged after cellular treatment with CCM or ammonia (Fig. 5A), suggesting that ammonia-induced autophagy was independent of mTOR signaling pathways. To confirm this conclusion, U2OS cells were depleted of mTOR complex components and then treated with CCM. Inhibition of mTORC1 by knockdown of either mTOR or Raptor increased basal autophagy; however, addition of CCM to these cells further increased autophagy in an additive fashion (Fig. 5B). Conversely, activation of mTORC1 through TSC2 depletion decreased basal autophagy but had no notable effect on CCM-induced autophagy (Fig. 5B). Reduction of mTORC2 activity by Rictor knockdown had no effect on either basal or CCM-induced autophagy, consistent with a previous report that autophagic activity is not modulated by mTORC2 signaling (28). We concluded that the autophagy-inducing activity of CCM is not mediated through suppression of mTORC1 signaling functions.

Fig. 5

Induction of autophagy by CCM is independent of mTOR but requires Ulk1. (A) U2OS GFP-LC3 cells were treated for 24 hours with CCI-779, CCM, or NH4OH, and lysates were immunoblotted to detect changes in mTOR substrate phosphorylation. (B) U2OS GFP-LC3 cells were transfected twice at 24-hour intervals with siRNA duplexes specific for the indicated mTOR complex components. Seventy-two hours after the first transfection, cells were treated for 24 hours with CCM. Panels depict nonadjacent lanes from the same blot. (C) U2OS GFP-LC3 cells were transfected twice at 24-hour intervals with siRNA duplexes targeting Ulk1. Seventy-two hours after the first transfection, cells were treated for 24 hours with fresh medium, CCM, or 2 mM NH4OH. (D) Wild-type U2OS cells were treated with fresh medium, CCM, 2 mM NH4OH, or CCI-779 for 16 hours, or 1 μM staurosporine (Staur) for 2 hours. The percentage of cells with depolarized mitochondria was determined on the basis of JC-1 fluorescence with the Guava MitoPotential assay (Millipore). *P < 0.001 by ANOVA compared to cells cultured in fresh medium.

Autophagy mediated by mTOR inhibition occurs through activation of the kinase Ulk1, an ortholog of the kinase Atg1 in budding yeast (2932). Depletion of Ulk1 with multiple small interfering RNAs (siRNAs) inhibited autophagy induction by both CCM and ammonia (Fig. 5C). Therefore, ammonia induces autophagy in an mTOR-independent but Ulk1-dependent manner. Because some mTOR-independent mechanisms of autophagy occur through calcium signaling (33, 34), we treated cells with BAPTA-AM [1,2-bis(2-aminophenoxy)ethane-N,N,N′,N′-tetraacetic acid tetraacetoxymethyl ester], a membrane-permeable calcium chelator, to examine the role of calcium in ammonia-induced autophagy. Consistent with previous results (35), BAPTA-AM suppressed basal as well as CCI-779–induced autophagy; however, ammonia- or CCM-stimulated autophagy was more resistant to this compound (fig. S4A). Treatment of cells with actinomycin D or cycloheximide, inhibitors of messenger RNA (mRNA) transcription or translation, respectively, also reduced CCI-779–induced autophagy, but had minimal effects on ammonia or CCM-induced autophagy (fig. S4B).

In astrocytes, increased ammonia concentrations increase mitochondrial membrane permeability, leading to dissipation of the transmembrane potential gradient (36), which has been implicated in the stimulation of autophagy (37). Treatment of U2OS cells with CCM or ammonia did not perturb mitochondrial membrane potential as shown by the lack of change in JC-1 fluorescence, whereas treatment with a mitochondria-damaging agent, staurosporine (38, 39), resulted in mitochondrial membrane depolarization (Fig. 5D). Thus, ammonia or CCM-induced autophagy does not result from gross alterations in mitochondrial membrane permeability.

Cytoprotective effects of glutaminolysis-derived ammonia

In addition to enhancing survival during nutrient limitation, autophagy protects cells from killing by certain cytotoxic agents, such as TNF-α (40), tunicamycin, and proteasome inhibitors (41). To further examine the interplay between Gln metabolism and autophagy, we used previously characterized atg5+/+ and atg5−/− immortalized baby mouse kidney (iBMK) cell lines (41, 42). We confirmed that loss of basal autophagy, whether because of genetic deletion of atg5 (Fig. 6A) or siRNA-mediated depletion of Atg5 or Ulk1 (fig. S5), enhanced cell killing by TNF-α. Removal of Gln from the culture medium also increased TNF-α–induced cytotoxicity to a similar degree as that stemming from Atg5 deficiency. The cytoprotective effect of Gln is partially explained by anaplerotic fueling of the TCA cycle, because replacement of Gln with dimethyl α-KG enhanced cell survival. Notably, addition of ammonia to the α-KG–containing medium further decreased TNF-α–induced cell death to the extent seen in cells cultured in Gln-containing medium (Fig. 6A). This ammonia-induced restoration of cell viability is autophagy-dependent, because Atg5-deficient cells were refractory to rescue by ammonia addition.

Fig. 6

Cytoprotective effects of ammonia-induced autophagy. (A) Wild-type (6.1B11) and atg5−/− (7.1B4) iBMK cells were cultured for 6 hours in medium with or without Gln, α-KG, or 2 mM NH4OH as indicated. Fresh medium of same type was then re-added to the cells with or without TNF-α (100 ng/ml). After 16 hours, cell viability was assayed with the Guava Viacount assay. Background cell death (<7%) was subtracted to depict cell death specifically induced by TNF-α. Data represent means ± SEM from five independent experiments. *P < 0.001. (B) Wild-type (6.1B11) and atg5−/− (7.1B4) iBMK cells were treated for 24 hours with CCM, evaporated CCM (reconstituted with distilled water), or evaporated CCM supplemented with 2 mM NH4OH. Autophagic activity was assayed by immunoblotting for endogenous LC3. (C) Wild-type (6.1B11) and atg5−/− (7.1B4) iBMK cells were incubated under a hypoxic atmosphere (1% O2) in CCM, evaporated CCM, or evaporated CCM supplemented with 2 mM NH4OH. Medium was changed daily; after 3 days, cell growth and proliferation were assayed by SRB staining. Cell proliferation in CCM in 1% O2 was normalized to 100%. Data represent means ± SEM from four independent experiments. *P < 0.001 compared to CCM-treated cells by ANOVA of the raw data.

During metabolic stress, autophagy enhances cell survival at the expense of cell growth and proliferation (43). To determine whether ammonia suppresses cell proliferation during nutrient limitation, we cultured wild-type and Atg5-null iBMK cells in CCM under hypoxic conditions. Removal of ammonia from CCM by evaporation prevented autophagy induction (Fig. 6B) and significantly enhanced cell growth and proliferation in wild-type iBMK cells, but not their autophagy-defective, Atg5-null counterparts. This effect was reversed by re-addition of ammonia to the evaporated CCM (Fig. 6C). Thus, the ammonia released from metabolically active cells suppresses cell growth and proliferation in a paracrine fashion through an autophagy-dependent mechanism.


Reprogramming of cell metabolism in diseases such as cancer has been a matter of scientific interest for many decades. Much of the attention in this field has focused on glucose metabolism, based in part on Warburg’s seminal observations that many cancer cells metabolize glucose primarily through the extramitochondrial glycolytic pathway, which yields lactate as an excretable end product (1, 2). Warburg proposed that the shift away from mitochondrial oxidative phosphorylation was a default response to the accumulation of dysfunctional mitochondria in cancer cells. However, subsequent findings argued against the hypothesis that mitochondrial dysfunction is a hallmark feature of transformed cells. Instead, the current model posits that the mitochondria in highly glycolytic cells are reprogrammed from generating metabolic energy (in the form of ATP) to producing key metabolic intermediates that support numerous biosynthetic reactions and combat oxidative stress. In this regard, cellular uptake and mitochondrial metabolism of Gln have taken center stage as the processes that fuel the TCA cycle and enable mitochondria to support anabolic metabolism in the setting of glycolysis, which diverts carbon away from the TCA cycle (4, 9). We note that these metabolic reprogramming events are not restricted to tumor cells, because normal cells, such as activated T lymphocytes, also rely on glycolysis and Gln metabolism when undergoing explosive proliferation (44, 45).

The present report adds yet another dimension to the diverse array of metabolic events that stem from the cellular uptake and metabolism of Gln in cycling cells. In mitochondria, Gln deamination by glutaminase yields ammonia, which is actively or passively exported from the cell (9). The resulting glutamate is deaminated to α-KG, either by transaminases, which generate amino acids such as alanine or asparagine, or by glutamate dehydrogenase, which generates a second ammonia molecule as a by-product. Here, we report that that the ammonia liberated during glutaminolysis exerts a previously unrecognized influence on cell metabolism through the stimulation of autophagy. Thus, in addition to its diverse contributions to anabolic metabolism, glutaminolysis generates a by-product that, under appropriate conditions, accumulates to concentrations sufficient to trigger macromolecule turnover by autophagy. The present study focused mainly on cultured cell lines; however, we observed that interstitial fluids from human tumor xenografts contained ammonia at concentrations comparable to those found to trigger autophagy in vitro (Fig. 3I). The latter finding raises the possibility that ammonia is not simply a toxic metabolite as thought previously, but also a biologically active molecule that stimulates autophagy, a major stress-response pathway in tumor cells. Gln supports the generation of reducing equivalents, in the form of glutathione and NADPH, which protect cells from oxidative stress (9). Our results indicate that glutaminolysis exerts an additional stress-protective function through the release of ammonia, a diffusible stimulator of autophagic flux in actively proliferating cells.

We can envisage at least two mutually nonexclusive functions for glutaminolysis-dependent autophagy, particularly in the pathological setting of solid tumor tissues. First, we have shown that Gln metabolism to ammonia supports basal autophagy, which, as mentioned above, promotes stress resistance through removal of damaged macromolecules (for example, misfolded proteins) and organelles. Similarly, autophagy represents a cytoprotective response that potentially limits tumor cell killing by cytotoxic chemotherapy (46). Second, Gln-derived ammonia might represent a diffusible messenger that allows cancer cells in nutrient-replete regions of the tumor tissue to support the viability of neighboring cancer cells residing at greater distances from a vascular supply, and hence coping with nutrient deprivation conditions. The latter scenario is reminiscent of recent findings that lactate released from glycolytic tumor cells residing in hypoxic regions of tumor tissue is imported and metabolized by cancer cells existing in more well-oxygenated regions of the tumor tissue (47). A definitive test of this model awaits the development of imaging technologies that will allow intravital assessments of autophagic flux in solid tumors.

The exact mechanism through which ammonia stimulates autophagic flux remains unclear. Our results suggest that ammonia derived from Gln metabolism supports basal autophagy in metabolically active cells. Perhaps the most well-studied mechanism of autophagy regulation involves mTORC1 (25, 48). mTORC1 is responsive to both extracellular and internal stimuli that affect cell growth and biosynthetic pathways, and conditions such as nutrient deprivation or ATP depletion that lead to suppression of mTORC1 activity stimulate autophagy through activation of the Ulk1 complex (32, 49). The results presented in this report indicate that CCM stimulates autophagy through an mTORC1-independent pathway. However, we found that both CCM- and ammonia-induced autophagy were sensitive to depletion of Ulk1 by RNA interference. One explanation for this result is that ammonia intersects with the autophagy machinery at a point downstream from mTORC1 and upstream of Ulk1. An equally viable alternative model is that Ulk1, like ammonia, supports basal autophagy, but that Ulk1 plays a “push-forward” role in the autophagic pathway (through stimulation of autophagosome formation), whereas ammonia serves a “pull-through” function at a more distal point in the pathway (for example, through stimulation of autophagosome-lysosome fusion). Clearly, considerably more work is needed to fully understand the interplay between Gln metabolism and autophagy in normal and transformed cells.

The discovery that Gln metabolism generates a diffusible regulator of autophagy has some important implications for tumor biology. Expression of the c-Myc protooncogene is often deregulated in cancer, and the Myc oncoprotein directly increases the abundance of Gln transport and catabolizing enzymes in transformed cells (8, 23). In addition, Myc-transformed cells undergo rapid apoptosis upon withdrawal of Gln from the culture medium (50). We speculate that the liberation of ammonia from tumor cells engaged in glutaminolysis provides an autocrine- and paracrine-acting signal that promotes autophagy and, in turn, protects cells in different regions of the tumor from internally generated or environmental stress. A correlate of this model may be found in budding yeast cells, which use ammonia as a diffusible signal to mediate local transmission of growth-suppressive signals from overgrown, nutrient-deficient yeast colonies to neighboring colonies (51). Whether this intriguing correlation represents a true evolutionary connection is unclear, as published studies have not defined a link between ammonia signaling and autophagy in yeast cells. Nonetheless, the current findings add to the growing body of evidence that Gln exerts a pervasive influence on cancer biology, and that the therapeutic manipulation of Gln uptake or metabolism may represent a promising approach for the treatment of cancer and other proliferative diseases.

Materials and Methods

Cell culture

U2OS osteosarcoma cells were maintained in a humidified, 5% CO2 atmosphere in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% heat-inactivated fetal bovine serum (FBS), penicillin-streptomycin, and 6 mM l-Gln. The iBMK cell lines 6.1B11 (atg5+/+) and 7.1B4 (atg5−/−) were previously described (41, 42) and cultured in DMEM with 10% FBS, penicillin-streptomycin, and l-Gln. U2OS GFP-LC3 cells were generated by transfection of U2OS cells with pEGFP-LC3 and selection of stable transfectants with G418 (500 μg/ml).

siRNA transfections

U2OS GFP-LC3 cells were plated at low density and transfected twice with siRNA at 24-hour intervals with Lipofectamine RNAi max (Invitrogen) and 40 nM siRNA duplexes according to the manufacturer’s protocol. Autophagy was typically induced at 72 hours after the first transfection. SMARTpool siRNA duplexes (Dharmacon) were used for knockdown of human Atg5, Ulk1, mTOR, Raptor, Rictor, and TSC2. Single siRNA duplexes for Ulk1 knockdown were also obtained from Dharmacon (catalog nos. J-005049-08 and J-081408-12).

Autophagy induction

U2OS cells stably expressing GFP-LC3 were plated 1 day before treatment. Cells were typically treated for 24 hours to induce autophagy, unless otherwise indicated. Concentrations of drugs used were 100 nM CCI-779 (a rapamycin analog), 200 nM bafilomycin A1, or CCM produced from 70 to 90% confluent wild-type U2OS cells cultured for 2 to 4 days in serum-free DMEM.

Cell growth and viability assays

Cell growth and proliferation were assayed by staining with sulforhodamine B (SRB). Briefly, cells were fixed with 10% TCA in serum-free medium for 1 hour at 4°C. Cells were stained with 0.057% SRB for 10 min, and SRB absorbance in 10 mM tris (unbuffered) was determined at 540 nm. For cell viability determination, total cell populations were stained with the Guava Viacount Flex Reagent (Millipore) and analyzed on a Guava PCA-96 instrument.


Adherent cells were washed once on ice with phosphate-buffered saline (PBS), and cell lysates were prepared by scraping in NuPAGE LDS buffer and water bath sonication. Protein concentrations were determined with the BioRad RC/DC Protein assay, and equal amounts of protein were separated with NuPAGE 4 to 12% bis-tris gels with MES running buffer. After transfer to nitrocellulose and primary antibody incubation overnight at 4°C, proteins were detected with the Odyssey Infrared Imaging system (Licor) using IRDye680 and IRDye800CW conjugated secondary antibodies. The following antibodies were used for Western blotting: LC3 (NB100-2220, Novus Biologicals), p62 (sc-28359, Santa Cruz Biotechnology), Ulk1 (sc-33182, Santa Cruz Biotechnology), and actin (MAB1501, Chemicon). Antibodies for GFP (#2555), Atg5 (#2630), S6K1 phospho-Thr389 (#9205), Akt phospho-Ser473 (#4501), 4E-BP1 phospho-Thr70 (#9455), mTOR (#2983), Raptor (#2280), Rictor (#2114), and TSC2 (#3635) were obtained from Cell Signaling Technologies.


GFP-LC3–expressing U2OS cells were imaged live on an inverted Olympus IX51 fluorescent microscope at 40× magnification. For quantification, cells displaying >20 brightly fluorescent GFP-LC3 puncta were counted as positive.

Mitochondrial potential

Wild-type U2OS cells were treated with autophagy-inducing agents for 16 hours or with 1 μM staurosporine for 2 hours. Cells were collected and analyzed for JC-1 fluorescence with the Guava EasyCyte MitoPotential Kit (Millipore) according to the manufacturer’s instructions.

Biochemical assays and measurements

Serum-free CCM was filtered through a regenerated cellulose membrane with a 1-kD cutoff in a nitrogen-pressured stirred cell (Millipore). The solid material remaining after filtration (wash) was redissolved in an equal volume of serum-free medium before cell treatment. Cation exchange chromatography was carried out with Pierce Cation Exchange Maxi spin columns. Fresh medium or CCM was diluted 1:13.5 in water (25 mosM, final) before column application. The column was washed with water, and fractions were eluted in increasing concentrations of Earle’s balanced salt solution (EBSS). Cells were treated at 1× EBSS supplemented with MEM amino acids, pyruvate, vitamins, glucose, and Gln. For rotary evaporation, serum-free fresh medium or CCM was rotated under vacuum over a 37°C water bath, and distilled material was collected after condensation in a dry ice–isopropanol cold trap. Ammonia concentrations were measured with the Ammonia Assay Kit from Sigma (catalog # AA0100). For quantification of ammonia in mouse plasma and human tumor xenograft–derived interstitial fluid, samples were first deproteinized with 20% TCA and neutralized with 0.45 M NaOH before determination of ammonia concentration. Bicarbonate concentrations were measured with the CO2 assay kit from Equal Diagnostics (catalog # E598100).

Statistical analysis

Data from three or more independent experiments were analyzed with a two-tailed, paired Student’s t test, or with analysis of variance (ANOVA) with a posttest LSD (least significant differences) analysis where indicated. P < 0.05 was considered statistically significant.


Acknowledgments: We thank L. Toral-Barza, C. Shi, and W.-G. Zhang for excellent technical support; E. Rosfjord and T. Zhou for statistical assistance; M. L. Mercado and R. Mathew for reagent provision; and E. Graziani for helpful discussions. Funding: This work was supported by Wyeth Oncology Discovery, the Pfizer Center for Integrative Biology and Biotherapeutics (CIBB) and the Pfizer Postdoctoral Program, and NIH R01 grant CA130893 to E.W. Author contributions: C.H.E. performed the experiments; J.L. generated and collected the tumor interstitial fluid samples, E.W. provided transgenic iBMK cell lines and commented on the manuscript; C.H.E., K.Y., and R.T.A. analyzed data and interpreted experimental results; and C.H.E. and R.T.A. designed the experiments and wrote the paper.

Supplementary Materials

Fig. S1. Measurement of autophagic flux.

Fig. S2. Characterization of CCM.

Fig. S3. CO2, H2O2, NO, and lactate do not promote autophagy.

Fig. S4. Induction of autophagy is unaffected by intracellular Ca2+ buffering or inhibitors of mRNA transcription and translation.

Fig. S5. Atg5 or Ulk1 knockdown enhances TNF-α sensitivity.

References and Notes

View Abstract

Stay Connected to Science Signaling

Navigate This Article