Ammonia Metabolism in Cancer 2017 Science

RESEARCH

METABOLISM identified 15N-isotopologs of proline, aspartate, branched-chain amino acids (BCAAs), and glutamate, which have no previous biosynthetic con- nection to the amide nitrogen on glutamine (Fig. Metabolic recycling of ammonia via 1C and fig. S3B). The labeled nitrogen was liberated as ammonia before production of these glutamate dehydrogenase supports metabolites, which suggests that an ammonia- recycling pathway may synthesize the other glu- breast cancer biomass tamine derivatives detected. To test whether ammonia released during glutaminolysis was necessary for production of Jessica B. Spinelli,1,2 Haejin Yoon,1 Alison E. Ringel,1 Sarah Jeanfavre,2 these unanticipated amide-nitrogen glutamine Clary B. Clish,2 Marcia C. Haigis1* derivatives, we treated cells with the glutaminase inhibitor bis-2-(5-phenylacetamido-1,3,4- Ammonia is a ubiquitous by-product of cellular metabolism; however, the biological thiadiazol-2-yl)ethyl sulfide (BPTES), which at consequences of ammonia production are not fully understood, especially in cancer. 1 mM is not cytotoxic or cytostatic in T47D and We found that ammonia is not merely a toxic waste product but is recycled into central MCF7 human breast cancer cell lines (16)(fig.S4, amino acid metabolism to maximize nitrogen utilization. In our experiments, human breast A to C). BPTES treatment significantly decreased cancer cells primarily assimilated ammonia through reductive amination catalyzed by 15N-isotopologs of glutamate, proline, and as- glutamate dehydrogenase (GDH); secondary reactions enabled other amino acids, such as partate, whereas metabolites involved in direct proline and aspartate, to directly acquire this nitrogen. Metabolic recycling of ammonia glutamine metabolism, such as nucleotides and accelerated proliferation of breast cancer. In mice, ammonia accumulated in the tumor Downloaded from asparagine, remained labeled (Fig. 1D and fig. microenvironment and was used directly to generate amino acids through GDH activity. S4D). Addition of ammonia to BPTES-treated These data show that ammonia is not only a secreted waste product but also a fundamental cells restored metabolites depleted by glutamin- nitrogen source that can support tumor biomass. ase inhibition, demonstrating the specific contri- bution of ammonia (fig. S4E). This is consistent ncreased nutrient consumption can supply assimilation: (i) carbamoyl phosphate synthe- with findings that ammonia partially rescues pro-

carbon, nitrogen, oxygen, and sulfur to ac- tase I (CPS1), the adenosine triphosphate (ATP)– liferative defects in glutamine-deprived breast http://science.sciencemag.org/ commodate the extensive bioenergetic, bio- dependent, rate-limiting step of the urea cycle; cancer cells (17). I synthetic, and prosurvival requirements of (ii) glutamate dehydrogenase (GDH), a NADPH We examined the potential mechanisms under- rapidly proliferating cells (1–3). As a conse- (reduced nicotinamide adenine dinucleotide lying assimilation of ammonia liberated during quence, such cells generate an excess of metabolic phosphate)–dependent enzyme that catalyzes glutaminolysis. Because [15N]amide-glutamine waste products, which are cleared in mammals the reductive amination of a-ketoglutarate; and did not elicit any isotopes of four urea cycle in- through the excretory system. However, in the tu- (iii) glutamine synthetase (GS), which catalyzes termediates (ornithine, citrulline, argininosuc- mor microenvironment, metabolic waste products the ATP-dependent amination of glutamate to cinate, and arginine), we ruled out the activity such as lactate and ammonia accumulate (4, 5). generate glutamine (12, 13) (fig. S1A). Analysis of CPS1 as a mechanism for ammonia assimi- Although lactate is well studied in cancer, little is of transcriptomic data from The Cancer Genome lation (fig. S2). Instead, our data indicated that known about the mechanisms by which cancer Atlas for the ammonia-assimilating enzymes in GDH was the primary point of ammonia as-

cells manage increased amounts of ammonia healthy and cancerous tissues revealed that ex- similation because glutamate is upstream of on June 17, 2019 (NH3) generated by glutamine and asparagine pression of GS and GDH mRNA is significantly proline, aspartate, glutamine, and BCAA syn- catabolism, de novo cysteine synthesis through the increased across many cancer subtypes, whereas thesis. However, the Michaelis-Menten constant transsulfuration pathway, and salvage nucleotide CPS1 mRNA is increased only in the colon (Fig. 1B). (Km) of GDH for ammonia is high (approximate- metabolism (6). Ammonia has been considered a Amonghealthytissues,GSandGDHareubiqui- ly 9 mM) and GDH reportedly favors oxidative toxic by-product that must be exported from cells tously expressed and CPS1 is expressed only in the deamination over reductive amination in can- and is subsequently cleared through urea cycle liver (fig. S1B). Breast cancers display increased cer cells (18–21). By contrast, GDH-catalyzed re- activity in the liver (7–9). expression of both GS and GDH. Specifically, es- ductive amination is prevalent in the liver, where Glutamine has been called a “nitrogen reser- trogen receptor–positive (ER+) breast cancers have there is a sufficient concentration of ammonia voir” for cancer cells because of its anabolic role in increased expression of GS and GDH relative to to enable catalysis in this direction (22). We nucleotide synthesis (6, 10). However, the role of that in other subtypes (14). Therefore, we used wondered whether increased concentrations glutamine as a nitrogen reservoir is contradicted ER+ breast cancer as a representative model to of ammonia in the tumor microenvironment in catabolic glutamine metabolism, because nitro- probe for ammonia assimilation. might also permit GDH-catalyzed reductive gen is liberated as the by-product ammonia (11). To investigate the fate of glutamine-derived amination. The fate of ammonia in the metabolism of pro- ammonia, we performed a metabolic tracing anal- To determine whether GDH assimilates ammo- liferating cells and tumors remains unclear. We ysis with hydrophilic interaction liquid chroma- nia generated by glutamine catabolism, we used hypothesized that ammonia might be reassimi- tography tandem mass spectrometry (HILIC-MS) short hairpin RNA (shRNA) to deplete cells of lated into central metabolism to maximize the and assessed the fate of [15N]amide-glutamine, GDH, cultured them with [15N]amide-glutamine, 15 efficiency of nitrogen utilization. In this study, we which liberates [ N]NH3 through glutaminase and then subjected them to nitrogen metabolome sought to clarify the role of ammonia as either a activity (15). To identify the metabolic derivatives scanning (fig. S4F). MCF7 and T47D cell lines toxic waste product or a biosynthetic metabolite of [15N]amide-glutamine in an unbiased manner, express both GDH1 and GDH2 isoforms, and (Fig. 1A). we developed a method to screen the nitrogen shRNAs targeted both isoforms (fig. S4G). The Mammals have three enzymes that can over- metabolome, which contained 211 15N-isotopologs abundance of 15N-isotopologs of glutamate and come the thermodynamic hurdles of ammonia (table S1). The majority of the nitrogen meta- downstream metabolites (proline, aspartate) bolome did not acquire 15N labeling; of 211 15N- was significantly decreased in cells depleted of isotopologs, only 33 metabolites were labeled (fig. GDH (Fig. 1E). Urea cycle intermediates (citrul- 1 Department of Cell Biology, Harvard Medical School, Boston, S2). Consistent with previous studies, [15N]amide- line, argininosuccinate) remained unlabeled in MA 02115, USA. 2Broad Institute of MIT and Harvard, Cambridge, MA 02142, USA. glutamine was incorporated into asparagine and cells lacking GDH, underscoring the lack of *Corresponding author. Email: [email protected] nucleotides (10) (Fig. 1C and fig. S3A). We also CPS1-mediated ammonia assimilation in breast

Spinelli et al., Science 358, 941–946 (2017) 17 November 2017 1of6 RESEARCH | REPORT cancer cells (Fig. 1E). Reexpression of shRNA- from [15N]amide-glutamine (23) to measure the (fig. S6, A to C). Because ~2% of the glutamate pool 15 insensitive GDH1 rescued labeling onto glutamate amount of [ N]NH3 generated after 8 hours of acquired this label in a glutaminase-dependent and downstream metabolites (fig. S5). treatment with [15N]amide-glutamine. In MCF7 manner (fig. S3B), we hypothesized that ammo- Next, we used a liquid chromatography–mass and T47D cells, we found that only 3.5% of the nia recycling from glutaminolysis is highly effi- 15 spectrometry method for the detection of [ N]NH3 total ammonia pool derived from glutaminolysis cient. We quantified the efficiency of ammonia

Assimilation Waste Fold-Change Expression Compared to Healthy Tissue AABCDEFGHI JK LM GS 1.50 1.16 1.12 1.16 1.15 1.06 -1.15 1.10 1.27 1.11 ? GDH 1.34 1.32 1.25 1.22 1.03 -2.17 NH 3 CPS1 1.36 1.05 OO OO Sample Size 594 102 60 69 331 471 360 38 173 542 18 138 58 - (Patients) H N O -O O- 2 + NH + 3 NH3 Glutamine Significantly Not Significantly Significantly Glutamate Decreased Altered Increased

-O O O O O O

HN T47D MCF7 Downloaded from P O O O - - N O O H2N N O OH H 2.0 (-) BPTES O- 3.0 OH N-Carbamoyl Aspartate **** (+) BPTES HO NH Proline (M+1) **** 1.5 UMP (M+1) 2.0 **** **** O O 1.0 O P **** **** - 1.0 O H2N O O O- O- O 0.5

H N Isotopologue 2 - - O carbamoyl phosphate Abundance (%) O http://science.sciencemag.org/ N O- O NH + 0.0 3 + 0.0 Asparagine (M+1) O NH3 NH3 Aspartate (M+1) Glu Pro Asp Glu Pro Asp OO O O

- -O O- α H2N O -ketoglutarate + GLS + NH NH3 T47D Aspartate 3 Glutamate (M+1) Glutamine O 1.5 shControl - **** HO O O- **** shGDH #1 P O NH2 PRPP O NH + **** shGDH #2 FGAR O 3 **** Leucine (M+1) OH 1.0 **** Gln HO **** - N HO O H H2N - N O O OH 0.5 P O O N H N O O P on June 17, 2019

O Isotopologue NH N O OH N Abundance (%) HO HO NDNDND NDND ND OH 0.0 AMP (M+2) Unexpected Derivatives Glu Pro Asp Cit Asa 15NH 3 Ammonia Recycling Direct Pathways

OO O O O O

15 O- - Nucleotides H2 N HO O -O O- GLS 15 GDH + + NH3 NH3 O Glutamine Glutamate α-Ketoglutarate

Asparagine

Proline Aspartate BCAAs

Fig. 1. Glutamine-derived ammonia is recycled. (A) Schematic of melanoma. (C) Schematic of 15N-isotopologs after treatment with [15N]amide- fates of ammonia in cancer. (B) mRNA expression of ammonia-assimilating glutamine. UMP, uridine monophosphate; PRPP, phosphoribosyl enzymes from The Cancer Genome Atlas in cancerous versus normal pyrophosphate; FGAR, 5´-phosphoribosyl-N-formylglycineamide; AMP, tissue. GS (glutamine synthetase), GDH1 (glutamate dehydrogenase), and adenosine monophosphate. (D) Isotopolog abundance of unexpected CPS1 (carbamoyl phosphate synthetase 1) RNA levels were assessed [15N]amide-glutamine derivatives ± 1 mM BPTES in T47D and MCF7 cell using Oncomine.org; values are the mean of fold change (cancer/normal, lines. Values are means ± SEM; n = 4 per condition. (E) Isotope abundance where normal = 1.00) measured across the sample size shown at the bottom. of [15N]amide-glutamine–derived metabolites in control cells and cells A, ovarian serous cystadenocarcinoma; B, colon adenocarcinoma; C, rectal depleted of GDH (shGDH #1 and shGDH #2). Glu, glutamate M+1; Pro, adenocarcinoma; D, lobular and ductal breast carcinoma; E, lung adenocarci- proline M+1; Asp, aspartate M+1; Cit, citrulline M+1; Asa, argininosuccinate noma; F, squamous lung cell carcinoma; G, endometrial adenocarcinoma; M+1; ND, 15N-isotopolog not detected. Values are means ± SEM; n =4 H, bladder urothelial carcinoma; I, gastric adenocarcinoma; J, glioblastoma; per condition. (F) Schematic of ammonia recycling. GLS, glutaminase. K, pancreatic adenocarcinoma; L, hepatocellular carcinoma; M, cutaneous ****P < 0.0001 (two-tailed t test for all comparisons).

Spinelli et al., Science 358, 941–946 (2017) 17 November 2017 2of6 RESEARCH | REPORT recycling from glutamine catabolism by incu- of the total amount of glutamate (glutamate ammonia recycling (fig. S6F). Because both 15 13 bating MCF7 cells with [ N2- C5]glutamine. M+6 and glutamate M+1) to glutamate directly processes are mitochondrial, localization may 15 13 Glutaminolysis generated [ N- C5]glutamate generated in glutaminolysis (glutamate M+6) support this high efficiency (24). GDH is a bi- (glutamate M+6), and ammonia recycling was (fig. S6E). In total, 1.57 molecules of gluta- directional enzyme, so we also tested whether measured by detection of [15N]glutamate (glu- mate were generated from a single reaction the catalytic activity of oxidative deamination tamate M+1) (fig. S6D). We calculated the ratio of glutaminolysis, indicating a 57% efficiency of or reductive amination was prevalent. In GDH- depleted cells, ammonia recycling (glutamate M+1) was decreased, but a-ketoglutarate (M+5) was unchanged, suggesting a net activity of re- Cytotoxicity Assay Steady-State Metabolite Levels ductive amination in this system (fig. S6, G and Control vs Ammonia H). In sum, these data indicate that ammonia **** derived from glutaminolysis is recycled by re- MCF7 T47D 0.5 100 100 *** *** *** ductive amination catalyzed by GDH to support 80 80 * the synthesis of glutamate and downstream me- 60 60 0.0 tabolites (Fig. 1F). 40 Because numerous reactions other than gluta- 40 * 20 20 *** minolysis generate ammonia, we investigated (Ammonia/ Control) -0.5 * 0 0 *** whether free ammonia could be assimilated into log2 Fold Change Peak Area % Dead (PI Uptake) 0 1 5 0 1 5 metabolic pathways. To optimize NH Cl for tracing 0.1 0.5 10 20 50 05 0.1 0.5 10 20 50 4 0.05 0.75 0. 0.75 studies, we investigated whether exposure to in- [NH Cl] (mM) valine 4 alanine leucine creased concentrations of NH4Cl was toxic to tumor aspartateglutamate Downloaded from oxaloacetate cells. Physiological concentrations of ammonia in α-ketoglutarate α plasma are 0 to 50 mMinhealthyhumanadults, α-ketoisocaproate-ketoisovalerateKeto-acid Amino Acid 50 to 150 mM in newborns, and up to 1.0 mM in 15 25 N-Labeled Metabolites patients with hyperammonemia ( ). Supraphy- siological concentrations of ammonia are toxic to 25 MCF7 25 T47D neurons and are sometimes assumed to be toxic

20 20 to tumor cells (7, 26, 27). However, NH4Cl was not http://science.sciencemag.org/ toxic to tumor cells, even at concentrations that 15 15 were toxic to primary human astrocytes (Fig. 2A 10 10 and fig. S7A). Previous reports have shown that high concentrations of ammonia induce autoph- 5 5 agy in tumor cells (5). In MCF7 and T47D cell ND ND NDNDNDND ND ND ND ND ND ND lines, light chain 3 II (LC3II) lipidation was not 0 0

Isotopologue Abundance (%) induced until 10 mM ammonia was added to Ile GluPro Ala Leu GluProAsp Ala Leu Ile Cit Asp IMP Orn CitAsa IMP Orn Asa — GSH UMP AMP GSH UMP AMP media a concentration exceeding levels of ammo-

Gln (M+1) Asn (M+1)Gln (M+2) Gln (M+1) Asn Gln(M+1) (M+2) nia reported in the tumor microenvironment (fig. S7B). Moreover, ammonia concentrations of 0

Glutamate M+1 Metabolites Downstream of Glutamate to 10 mM did not alter uptake of glucose or on June 17, 2019 T47D MCF7 glutamine, nor basal respiration (fig. S7, C to **** **** E). Expression of the ammonia-assimilating en- **** **** **** zymes GS, GDH, and CPS1 was not affected by 50 **** 60 25 80 **** **** **** increasing ammonia concentration (fig. S7, F to **** **** 20 40 60 **** I), nor did 10 mM ammonia alter the pH of the 40 **** culture media (fig. S7J). These data indicate that 30 15 **** 40 **** supraphysiological concentrations of ammonia 20 10 **** did not induce toxicity or metabolic stress in 20 20 breast cancer cells. 10 5 We also examined ammonia uptake by cells. 0 0 0 0 When breast cancer cells were cultured in low Isotopologue Abundance (%) T47D MCF7 Isotopologue Abundance (%) Proline Alanine Aspartate Proline Alanine Aspartate concentrations of ammonia (0 to 1.0 mM), we ob- (M+1) (M+1) (M+1) (M+1) (M+1) (M+1) shControl shGDH #1 shGDH #2 served a net output of ammonia, which reverted to net uptake as the extracellular concentration Fig. 2. Ammonia is assimilated by GDH to generate amino acids. (A) Propidium iodide (PI) staining of NH4Cl increased above 1 mM (fig. S7K). At of cells treated with a dose of NH4Cl for 48 hours. Values are means ± SEM from a representative approximately 1.0 mM NH4Cl, ammonia was taken experiment of three replicates; n =3.(B) Abundance of keto- and amino acids involved in up from the medium, such that ammonia entry transaminase reactions in T47D cells treated with 0.75 mM NH4Cl. Values are means ± SEM from a may be regulated by diffusion. In agreement with representative experiment of two replicates; n =4.(C)Abundanceof15N-isotopologs in MCF7 and this result, the characterized mechanism of am- 15 + T47D cells after 8 hours of treatment with 0.75 mM [ N]NH4Cl. (M+1) and (M+2) indicate labeling with monium (NH4 ) import and export is through one or two nitrogens, respectively. Values are scaled to account for total intracellular ammonia and facilitated diffusion with rhesus glycoproteins represent means ± SEM; n =4.(D) Isotopolog abundance of glutamate (M+1) in MCF7 and T47D (RhC and RhG) (28). Also, ammonia can diffuse 15 cells treated for 8 hours with 0.75 mM [ N]NH4Cl in control and GDH-depleted cells. Values are across the plasma membrane. scaled to account for total intracellular ammonia and represent means ± SEM; n =4.(E) Abundance We performed steady-state and tracing ex- 15 of N-isotopologs for metabolites downstream of glutamate treated for 8 hours with periments in the presence of 0.75 mM NH4Cl 15 0.75 mM [ N]NH4Cl in control and GDH-depleted cells. Values are scaled to account for total because it is the inflection point of ammonia intracellular ammonia and represent means ± SEM; n =4.*P < 0.05, ***P < 0.001, ****P < 0.0001 uptake and secretion and represents a low con- (two-tailed t test for all comparisons). centration of ammonia that is relevant to the

Spinelli et al., Science 358, 941–946 (2017) 17 November 2017 3of6 RESEARCH | REPORT

tumor microenvironment. We used Metabo- MCF7 T47D MCF7 T47D Analyst 3.0 to perform an unbiased pathway 2.0 2.0 **** analysis on the steady-state metabolites from *** cells cultured with or without ammonia (fig. 1.5 1.5

S8A). The most significantly altered pathway Control was glutamate, aspartate, and alanine metab- 1.0 1.0

olism. Exposure to NH Cl elicited a signature Cl 4 4 0.5 0.5 of increased transaminase activity, whereby the abundance of ketoacids decreased and that of Area Relative Sphere

(+) NH 0.0 0.0 amino acids derived from them increased Cl Cl 4 4 (Fig. 2B). Although amounts of nonessential NH amino acids increased, the abundance of other ControlNH Control amino acids remained unchanged by ammonia, indicating that ammonia did not affect shControl shGDH #1 shGDH #2 MCF7 T47D universal amino acid metabolism (fig. S8B). 2.5 *** 2.5 2.5 Nor did ammonia alter the abundance of me- 2.0 2.0 2.0 tabolites from the urea cycle and nucleotides 1.5 1.5 1.5 (fig. S8, C and D). ns ns Control MCF7 and T47D cells were treated with 0.75 mM 15 15 1.0 1.0 1.0 [ N]NH4Cl and scanned for N-isotopologs (fig. S9). Isotopolog abundances were scaled 0.5 0.5 0.5 Downloaded from

to represent total ammonia pools, because treat- Area Relative Sphere ment with 0.75 mM enriched ~35% of the in- 0.0 0.0 0.0 Conditioned tracellular ammonia pool (table S2). Consistent Cl Cl Cl 4 4 4 with tracing performed with glutamine-derived ControlNH Control NH Control NH 15 ammonia, we detected N labeling of gluta- Ammonia in mate and downstream metabolites, such as pro- shControl shGDH #1 shGDH #2 Conditioned Media line and aspartate (Fig. 2C). Upon tracing with 1500 3 *** 3 3 http://science.sciencemag.org/ ** low levels of ammonia, a striking 20% of the *** 200 glutamate pool was labeled, implying an important role for ammonia assimilation in gluta- M) 1000 2 2 2 150 * mate metabolism in cancer, as glutamate levels ns ns are high (millimolar) (11). Tracing with high 100 500 1 levels of ammonia that have been reported in 1 1 the tumor microenvironment (3 mM) also eli- Ammonia ( μ 50 cited the same signature of ammonia assimila- 0 Area Relative Sphere 0 0 0 nmoles ammonia secreted tion (fig. S10A). (Normalized to cell number) 0 Consistent with steady-state data, all of the Control Control Control Control amino acids labeled were generated through on June 17, 2019 glutamate-dependent transaminase reactions, ex- Conditioned Conditioned Conditioned Conditioned shControlshGDHshGDH #1 #2 cept proline and glutathione, which were made in direct synthetic pathways from glutamate (fig. Fig. 3. Ammonia stimulates breast cancer growth and proliferation. (A) Representative images S10B). Other nitrogen-containing metabolites of 3D culture models of MCF7 and T47D cells treated with 0.5 mM NH4Cl compared to control (particularly urea cycle intermediates) and es- conditions. (B) Quantification of average sphere area of 100 to 200 spheres per well in 3D culture sentialaminoacidswerenotlabeledbyammo- models of MCF7 and T47D cells treated with ammonia and control conditions for 7 days. Values n nia (fig. S9). Furthermore, even though ammonia are mean areas ± SEM from a representative experiment of five replicates; =4.(C) Quantification generated 15N-isotopologs of glutamine, we de- of average sphere area of 200 to 250 spheres per well in 3D culture models of MCF7 cells harboring tected no 15N-isotopologs of any nucleotides, stable shRNA-mediated knockdown of GDH or control hairpin. Cells were treated for 8 days. n which is distinct from ammonia metabolism in Values are mean areas ± SEM from a representative experiment of three replicates; =4. serine-threonine kinase II (STKII)–low tumors (D) Representative images of MCF7 and T47D cells in control conditions (medium changed (29). We speculate that because labeled gluta- daily) and conditioned media (medium changed every 72 hours). Cells were treated for 8 days. mine is generated in the mitochondria, this pool (E) Ammonia measurement in conditioned media compared to control after 8 days. (F) Quantification maynotaccessthecytoplasmwheredenovonu- of average sphere area of 200 to 250 spheres per well in 3D culture models of MCF7 control cleotide synthesis occurs, rendering nucleotides cells or cells depleted of GDH. Cells were treated in control or conditioned media for 8 days. n unlabeled. Values are mean areas ± SEM from a representative experiment of three replicates; =4. A time course of [15N]NH Cl tracing revealed (G) Nanomoles of ammonia secreted per cell after 72 hours in control cells or cells depleted of 4 n P P P P that ammonia was rapidly converted into gluta- GDH. Values are means ± SEM; =3.* < 0.05, ** < 0.01, *** < 0.001, **** < 0.0001 (two-tailed t mate, the first metabolite to reach steady state test for all comparisons); ns, not significant. (fig. S10, C to F). Thus, ammonia appears to be primarily assimilated to generate glutamate, and other labeled metabolites are produced in second- when shRNA-insensitive GDH1 was overexpressed cells depleted of GDH, indicating that adaptive ary reactions. We therefore investigated which (fig. S10, G and H). We did not observe adaptation reprogramming of ammonia assimilation into metabolic derivatives of ammonia required GDH through the ammonia-assimilating enzymes GS the urea cycle is not important in cultured breast 15 activity. In cells depleted of GDH, [ N]NH4Cl or CPS1 in cells lacking GDH. In both T47D and cancer cells (fig. S9). Our data reveal a general labeling of glutamate was diminished, as was MCF7 cells, glutamine and asparagine labeling mechanism by which free ammonia in the tumor labeling of metabolites downstream of glutamate did not change in cells depleted of GDH (fig. S11). microenvironment can be harnessed for bio- (Fig.2,DandE).Indeed,thislabelingwasrescued Metabolites of the urea cycle were unlabeled in synthetic pathways.

Spinelli et al., Science 358, 941–946 (2017) 17 November 2017 4of6 RESEARCH | REPORT

Ammonia assimilation in yeast has a funda- In vivo Ammonia Labeling mental role in supporting growth and proliferation 5 ** Argininosuccinate Glutamine Glutamate Proline Aspartate (30, 31). Because ammonia was not toxic to tu- (M+1) (M+1) (M+1) 4 80 (M+1) 30 6 (M+1) 3 6 mor cells (Fig. 2A), we tested whether ammo- 60 nia might facilitate growth and proliferation 3 20 4 2 4 Liver of breast cancer cells. As in yeast, addition of 2 40 10 2 1 2 NH Cl to cell culture media increased prolif- 20 4 1 eration in breast cancer cell lines (fig. S12, A Ammonia (mM) 0 0 0 0 0 and B). The culture medium was changed daily 0 150 40 6 3 6 to minimize ammonia accumulation, which Plasma Interstitial 30 Fluid 100 4 2 4 we measured to be approximately 0.3 mM per 20 Plasma day from glutamine degradation and cellular 50 2 1 2 10 metabolism (fig. S12, C and F). Moreover, in 3D Control shGDH culture, in which cells are suspended in matrigel 0 0 0 0 0 GDH 20 40 6 3 6 and form “spheres” for growth, addition of Isotopologue Abundance (%) 15 30 4 ammonia to media stimulated sphere growth α-tubulin 4 2 and cell proliferation (Fig. 3, A and B, and fig. 10 20 2 1 2 Tumor S12G). However, proliferation of primary hu- 5 10 man fibroblasts was not changed when am- In vivo Tumor Growth 0 0 0 0 0 124 124 124 124 124 monia was added to culture media (fig. S13A). Control 15 Time (Hours) Using [ N]NH4Cl tracing, we found that fibro- ) Downloaded from 3 GDH KD blasts centrally assimilated ammonia to gener- 150 ate glutamine (fig. S13B), in line with their high 32 Schematic of Systemic Ammonia Metabolism expression of glutamine synthetase ( ). More- 100 over, [15N]amide-glutamine tracing revealed that fibroblasts did not recycle glutamine-derived Ornithine Arginine ammonia to generate glutamate, aspartate, or 50 Liver NH Carbamoyl Urea Tumor Volume (mm Volume Tumor 3 Phosphate proline (fig. S13C). Thus, we hypothesized that Cycle http://science.sciencemag.org/ ammonia assimilation to generate glutamate 0102030 Citrulline Argininosuccinate TCA through GDH may be important for its role in Day NH NH Cycle 3 3 increased proliferation observed in breast can- α-KG Glu Gln cer cells. Indeed, depletion of GDH prevented In vivo Ammonia Labeling the accelerated growth of breast cancer cells NH3 treated with ammonia (Fig. 3C). Interestingly, Glutamate M+1 Proline M+1 the glutamate derivatives proline, aspartate, 4 * 1.5 ** Gln and glutathione are associated with prolifera- 3 tion and tumorigenesis (33–37). 1.0 Plasma Glu To assess the effect of tumor-generated am- 2

0.5

Gln on June 17, 2019 monia on growth and proliferation, we com- 1 pared the ability of cancer cells to grow in 3D cultures in which the medium was changed 0 0.0 Gln WT GDH WT GDH eitherdailyorevery3days,allowingammo- Isotopologue Abundance (%) KD KD Glu nia to accumulate. The latter procedure pro- NH3 vided a growth advantage for breast cancer Aspartate M+1 Glutamine M+1 NH cells, which correlated with ammonia accumu- 3 Glu Glu 2.5 ** 4 Gln lation in the media (Fig. 3, D and E). Therefore, ns Gln Gln 2.0 Glu Gln we tested whether ammonia recycling through 3 1.5 NH3 Tumor GDH was a critical aspect that influenced pro- 2 TCA 1.0 α-KG Glu liferation. Cells depleted of GDH had no growth 1 Cycle defect when the culture medium was changed 0.5 GDH daily, but the growth advantage when medium 0.0 0 Asp & Pro WTGDH WT GDH was changed after 3 days was abrogated (Fig. 3F Isotopologue Abundance (%) KD KD and fig. S14, A to C). Furthermore, cells depleted of GDH secreted more ammonia into the me- Fig. 4. Contributions of systemic and tumor-autonomous ammonia metabolism to amino acid dium, consistent with impairment of ammonia synthesis. (A) Measurement of ammonia in the interstitial fluids of the tumor microenvironment recycling (Fig. 3G). In addition, treatment of (TME) compared to plasma isolated from ER+ breast cancer xenograft models. Lines connect values MCF7 cells in 3D culture with high ammonia of ammonia in the plasma to that in the interstitial fluid of the TME. (B) Isotope abundance of 15 concentrations (3 mM NH4Cl) also stimulated N-isotopologs isolated from the liver, plasma, and tumor of mice intraperitoneally injected with a 15 proliferation (fig. S14D). bolus (9.0 mmol/kg) of [ N]NH4Cl. Tissues were harvested 1, 2, or 4 hours after injection. Values are To examine the physiological relevance of am- means ± SEM; n =4.15N-isotopologs were corrected for natural abundance of tissues harvested monia in the tumor microenvironment in vivo, from a control mouse treated with NH4Cl (9.0 mmol/kg) for 4 hours. (C) Western blot of GDH we measured concentrations of ammonia that knockdown in T47D xenograft tumors. (D) In vivo tumor growth of T47D control and GDH-depleted accumulated in the interstitial fluids of ER+ xeno- (GDH KD) xenograft models (n = 15 mice per group). Values are mean tumor volumes ± SEM. (E)In + 15 graft tumors. ER xenografts accumulated 0.8 to vivo tracing of [ N]NH4Cl in T47D control and GDH-depleted xenograft models. Values are mean 3 mM ammonia in the interstitial fluids of the isotopolog abundances ± SEM; n =4.(F) Schematic of systemic and tumor-autonomous ammonia tumor microenvironment, whereas plasma ammo- metabolism. TCA, tricarboxylic acid; a-KG, a-ketoglutarate. *P < 0.05, **P < 0.01 (two-tailed t test nia concentrations were approximately 300 mM for all comparisons).

Spinelli et al., Science 358, 941–946 (2017) 17 November 2017 5of6 RESEARCH | REPORT

(Fig. 4A). This range of concentrations did not We next generated a GDH-depleted xeno- REFERENCES AND NOTES induce autophagy and was shown to accelerate graft model to further investigate the mech- 1. N. N. Pavlova, C. B. Thompson, Cell Metab. 23,27–47 growth and proliferation in vitro. Plasma ammo- anism of ammonia assimilation in vivo (Fig. (2016). – nia concentrations in mice harboring tumors 4C). Tumor-specific depletion of GDH signif- 2. S. Vyas, E. Zaganjor, M. C. Haigis, Cell 166, 555 566 (2016). were not different from those in control mice icantly decreased tumor growth in vivo, con- 3. J. L. Coloff et al., Cell Metab. 23, 867–880 (2016). (fig. S15A). sistent with our findings that GDH-depleted 4. M. G. Vander Heiden, L. C. Cantley, C. B. Thompson, Science 324, We tested whether accumulated ammonia in cells grow slower in conditioned media and are 1029–1033 (2009). the tumor microenvironment was assimilated insensitive to ammonia-induced growth in vitro 5. C. H. Eng, K. Yu, J. Lucas, E. White, R. T. Abraham, Sci. Signal. 3, ra31 (2010). into metabolic pathways in vivo. Mice harbor- (Fig. 4D and fig. S18A). Because GDH-catalyzed 6. B. J. Altman, Z. E. Stine, C. V. Dang, Nat. Rev. Cancer 16, ing subcutaneous T47D breast tumors were ammonia assimilation mediates growth and 619–634 (2016). 15 7. M. Kappler et al., Clin. Oral Investig. 21, 211–224 (2017). intraperitoneally injected with [ N]NH4Cl, and proliferation in vitro, we used intraperitoneal 15 8. E. B. Tapper, Z. G. Jiang, V. R. Patwardhan, Mayo Clin. Proc. the tumor, liver, and plasma were assessed for injection of [ N]NH4Cl in control and GDH- – 15 90, 646 658 (2015). N-isotopologs over the next 1 to 4 hours (fig. depleted xenograft models to test whether GDH 9. R. J. DeBerardinis, T. Cheng, Oncogene 29,313–324 S15B). The liver and tumor used distinct meta- also assimilates ammonia in vivo. Glutamate, (2010). bolic pathways for ammonia assimilation (Fig. 4B aspartate, and proline labeling were significant- 10. S. Tardito et al., Nat. Cell Biol. 17, 1556–1568 (2015). and figs. S16 and S17A). In the liver, ammonia was ly decreased in GDH-depleted tumors relative 11. R. J. DeBerardinis et al., Proc. Natl. Acad. Sci. U.S.A. 104, 19345–19350 (2007). sequestered into the urea cycle, leading to labeling to control tumors (Fig. 4E). GDH depletion did 12. M. M. Adeva, G. Souto, N. Blanco, C. Donapetry, Metabolism 61, of ornithine, citrulline, argininosuccinate, and not abrogate glutamine labeling, underscoring 1495–1511 (2012). arginine (fig. S16). Although these labeled inter- the specificity of GDH for ammonia assimila- 13. S. M. Morris Jr., Annu. Rev. Nutr. 22,87–105 (2002). mediates of the urea cycle were also detected in tion to generate glutamate and the downstream 14. S. Tyanova et al., Nat. Commun. 7, 10259 (2016). 15. A. L. S. Lee, D. Roberts, R. E. Gerszten, C. B. Clish, Targeted the plasma, they were undetectable in the tumor, metabolites proline and aspartate. metabolomics. Curr. Protoc. Mol. Biol. 30.32.31 –30.32.24 Downloaded from indicating that these breast tumors did not en- To further validate that GDH-mediated (2012). gage the urea cycle for ammonia assimilation ammonia assimilation is tumor-autonomous, 16. M. M. Robinson et al., Biochem. J. 406, 407–414 (2007). in vivo (Fig. 4B and fig. S16). Proline and aspar- we traced ammonia ex vivo in control and 17. M. Meng, S. Chen, T. Lao, D. Liang, N. Sang, Cell Cycle 9, 3921–3932 (2010). tate, which were identified in vitro as metabolic GDH-depleted tumors. GDH depletion signif- 18. F. Frigerio, M. Casimir, S. Carobbio, P. Maechler, Biochim. derivatives of ammonia, were also labeled in the icantly decreased glutamate, aspartate, and Biophys. Acta 1777, 965–972 (2008). tumor in vivo. The metabolic pathways enabling proline labeling (fig. S18B). Taken together, 19. J. R. Treberg, M. E. Brosnan, M. Watford, J. T. Brosnan, Adv. Enzyme Regul. 50,34–43 (2010). proline and aspartate labeling were likely from these data show that tumor biomass is sup- http://science.sciencemag.org/ – tumor-autonomous metabolism, as labeled pro- ported by both tumor-autonomous metabo- 20. M. E. Brosnan, J. T. Brosnan, Am. J. Clin. Nutr. 90, 857S 861S (2009). line and aspartate were not detected in the plasma lism of ammonia and systemic assimilation, 21. J. Adam et al., Cell Rep. 3, 1440–1448 (2013). (Fig. 4B). especially for glutamate-derived amino acids 22. I. Zaganas et al., Metab. Brain Dis. 28, 127–131 (2013). We also observed labeling of glutamine and (Fig. 4F). 23. J. B. Spinelli, L. P. Kelley, M. C. Haigis, Sci. Rep. 7, 10304 glutamate in the tumor. Because labeled gluta- Ammonia accumulates in the tumor micro- (2017). 24. B. Masola, T. M. Devlin, Amino Acids 9, 363–374 (1995). mine and glutamate were also found in the liver environment because tumors are poorly vascu- 25. S. Dasarathy et al., Metab. Brain Dis. 32, 529–538 (2017). 15 and plasma, it is not clear whether these N- larized, making this a unique niche for ammonia 26. O. Braissant, V. A. McLin, C. Cudalbu, J. Inherit. Metab. Dis. 36, isotopologs were generated tumor-autonomously. metabolism in the human body. Because ammo- 595–612 (2013). Furthermore, the kinetics of glutamine labeling nia transport is mediated by diffusion, elevated 27. P. Hillmann, M. Köse, K. Söhl, C. E. Müller, Toxicol. Appl. Pharmacol. 227,36–47 (2008). in the tumor implied that a subset of the labeled ammonia in the microenvironment leads to its 28. F. Gruswitz et al., Proc. Natl. Acad. Sci. U.S.A. 107, 9638–9643

glutaminepoolinthetumormaybetakenupfrom accumulation inside of tumor cells (fig. S18C). (2010). on June 17, 2019 the plasma (Fig. 4B). Therefore, the ability to reassimilate this ammo- 29. J. Kim et al., Nature 546, 168–172 (2017). To distinguish systemic contributions of am- nia into metabolic pathways is critical in this 30. V. M. Boer, C. A. Crutchfield, P. H. Bradley, D. Botstein, J. D. Rabinowitz, Mol. Biol. Cell 21, 198–211 (2010). monia metabolism from tumor-autonomous me- context. In contrast, the liver reassimilates ammo- 31. S. Laxman, B. M. Sutter, L. Shi, B. P. Tu, Sci. Signal. 7, ra120 15 tabolic pathways, we traced [ N]NH4Cl and nia to generate urea, which is a sink for excess (2014). [15N]amide-glutamine in tumors ex vivo (fig. nitrogen and is excreted as metabolic waste to 32. T. Vermeulen et al., Arch. Biochem. Biophys. 478,96–102 S17, B to D). With [15N]NH Cl, labeling of glu- protect against toxicity associated with systemic (2008). 4 33. J. M. Phang, W. Liu, C. N. Hancock, J. W. Fischer, Curr. Opin. tamate, aspartate, proline, and glutamine was ammonia accumulation. Tumor cells strictly re- Clin. Nutr. Metab. Care 18,71–77 (2015). recapitulated. Consistent with in vivo studies, cycle this nitrogen to generate amino acids down- 34. N. Sahu et al., Cell Metab. 24, 753–761 (2016). the urea cycle intermediates, nucleotides, and stream of glutamate and do not engage the urea 35. L. Jin et al., Cancer Cell 27, 257–270 (2015). – other nitrogen-abundant metabolites were not cycle. 36. K. Birsoy et al., Cell 162, 540 551 (2015). 37. L. B. Sullivan et al., Cell 162, 552–563 (2015). labeled. These data underscore a fundamental Our findings show that ammonia is an im- role of ammonia for amino acid synthesis, par- portant nitrogen source for breast cancer me- ACKNOWLEDGMENTS ticularly glutamate, aspartate, and proline in vivo. tabolism. Ammonia is not simply a metabolic The authors acknowledge all members of the Haigis lab for Consistent with in vitro experiments, tumors waste product, and it can be recycled to sup- insightful discussions. Supported by the Ludwig Center at treated with [15N]amide-glutamine generated port the high demand for amino acid synthesis Harvard, Glenn Foundation for Medical Research, and NIH grant R01CA213062 (M.C.H.) and by NSF Graduate Research labeled glutamate as well as the downstream in rapidly proliferating cells. Although ammo- Fellowship Program grant DGE1144152 (J.B.S.). M.CH. and metabolites proline and aspartate, suggesting nia is sometimes considered a toxin, it stimu- J.B.S. are inventors on patent application HMV- 27560 submitted that glutamine-derived ammonia may be recycled lated growth and proliferation in breast cancer by Harvard Medical School that covers ammonia utilization in solid tumors (fig. S18A). In contrast to free cells. This stimulatory effect appears to be me- in cancer. 15 [ N]NH4Cl, which did not label metabolites of diated by GDH-catalyzed ammonia assimila- the urea cycle in cells in vivo or in solid tumors tion. Furthermore, ammonia accumulated in SUPPLEMENTARY MATERIALS ex vivo, [15N]amide-glutamine, when added to the tumor microenvironment and was used by www.sciencemag.org/content/358/6365/941/suppl/DC1 the tumors ex vivo, elicited labeling of the urea cancer cells for amino acid synthesis in vivo. Materials and Methods Figs. S1 to S18 cycle intermediate citrulline (fig. S17D). Thus, These biosynthetic pathways are supported in Tables S1 and S2 an alternative pathway that does not require both systemic and tumor-autonomous metab- References (38–42) ammonia may exist that connects the amide olism. Thus, metabolic recycling of ammonia 7 February 2017; accepted 29 September 2017 nitrogen on glutamine to citrulline production provides an important source of nitrogen for Published online 12 October 2017 ex vivo. breast cancer biomass. 10.1126/science.aam9305

Spinelli et al., Science 358, 941–946 (2017) 17 November 2017 6of6 Metabolic recycling of ammonia via glutamate dehydrogenase supports breast cancer biomass Jessica B. Spinelli, Haejin Yoon, Alison E. Ringel, Sarah Jeanfavre, Clary B. Clish and Marcia C. Haigis

Science 358 (6365), 941-946. DOI: 10.1126/science.aam9305originally published online October 12, 2017

Cancer cells put ammonia back to work Ammonia, often considered a metabolic waste product, can be recycled to build new amino acids. Rapidly proliferating cells produce extracellular nitrogen. Spinelli et al. used metabolic tracing of 15N to follow the fate of

extracellular ammonia and its incorporation into more than 200 components of the nitrogen metabolome (see the Downloaded from Perspective by Dang). Accumulation of ammonia enabled glutamate dehydrogenase to function in reductive amination, which allowed incorporation of nitrogen from ammonia back into amino acids. Experiments in mice also showed incorporation of ammonia into glutamate, aspartate, and proline. Science, this issue p. 941; see also p. 862 http://science.sciencemag.org/

ARTICLE TOOLS http://science.sciencemag.org/content/358/6365/941

SUPPLEMENTARY http://science.sciencemag.org/content/suppl/2017/10/11/science.aam9305.DC1 MATERIALS

RELATED http://science.sciencemag.org/content/sci/358/6365/862.full on June 17, 2019 CONTENT http://stm.sciencemag.org/content/scitransmed/7/274/274ra17.full http://stm.sciencemag.org/content/scitransmed/5/198/198ra108.full http://stm.sciencemag.org/content/scitransmed/2/31/31ra34.full http://stm.sciencemag.org/content/scitransmed/2/31/31ed1.full http://stke.sciencemag.org/content/sigtrans/10/510/eaao6604.full

REFERENCES This article cites 41 articles, 8 of which you can access for free http://science.sciencemag.org/content/358/6365/941#BIBL

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Use of this article is subject to the Terms of Service

Science (print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005. 2017 © The Authors, some rights reserved; exclusive licensee American Association for the Advancement of Science. No claim to original U.S. Government Works. The title Science is a registered trademark of AAAS.