Arginase 2 Suppresses Renal Carcinoma Progression via Biosynthetic Cofactor Pyridoxal Phosphate Depletion and Increased Polyamine Toxicity

SUMMARY

Kidney cancer, one of the ten most prevalent malignancies in the world, has exhibited increased incidence over the last decade. The most common subtype is “clear cell” renal cell carcinoma (ccRCC), which features consistent metabolic abnormalities, such as highly elevated glycogen and lipid deposition. By integrating metabolomic, genomic, and transcriptomic data, we determined that enzymes in multiple metabolic pathways are universally depleted in human ccRCC tumors, which are otherwise genetically heterogeneous. Notably, the expression of key urea cycle enzymes, including arginase 2 (ARG2) and argininosuccinate synthase 1 (ASS1), is strongly repressed in ccRCC. Reduced ARG2 activity promotes ccRCC tumor growth through at least two distinct mechanisms: conserving the critical biosynthetic cofactor pyridoxal phosphate, and avoiding toxic polyamine accumulation. Pharmacological approaches to restore urea cycle enzyme expression would greatly expand treatment strategies for ccRCC patients, where current therapies only benefit a subset of those afflicted with renal cancer.

eTOC blurb

Ochocki et al. show that clear cell renal cell carcinoma (ccRCC) tumors have altered ammoniametabolism with multiple urea cycle enzymes being significantly underexpressed. Loss of the urea cycle enzyme arginase2 (ARG2) promotes ccRCC tumor progression by conserving essential biosynthetic cofactor pools and preventing toxic polyamine build up.

INTRODUCTION

Tumor progression is dependent on cancer cells simultaneously undergoing biomass expansion, while somehow remaining viable in a “harsh” nutrient/oxygen limited microenvironment. It is widely acknowledged that altered cellular metabolism contributes to the unregulated growth of cancer cells (Pavlova and Thompson, 2016). For example, enhanced glycolysis, first reported by Otto Warburg in the 1920s, provides a broad array of biosynthetic precursors required for the generation of amino acids, lipids, nucleotides, NADPH, and other biochemical intermediates (DeBerardinis and Chandel, 2016; Pavlova and Thompson, 2016). Changes in intracellular pools of glucose-derived acetyl-coA and TCA cycle intermediates have also been associated with altered histone acetylation and epigenetic gene regulation in multiple cancer cell types (Carrer and Wellen, 2015). Other metabolic adaptations promote cell survival under nutrient poor/hypoxic conditions, such as increased cellular uptake of exogenous unsaturated fatty acids to support lipid homeostasis (Kamphorst et al., 2013; Nakazawa et al., 2016; Young et al., 2013). Consequently, pharmacological inhibition of specific metabolic pathways may preferentially limit tumor cell growth and proliferation, and multiple metabolism-based therapies are currently in development (Pavlova and Thompson, 2016; Vander Heiden and DeBerardinis, 2017; Yen et al., 2017). The role of altered metabolism is of particular interest in cancers derived from the kidney, a tissue (along with the liver) responsible for gluconeogenic production of glucose from non-carbohydrate carbon substrates (Emery et al., 1988). In addition, the kidney removes metabolic waste products from the blood stream, most notably urea produced in the liver, while maintaining internal water/inorganic ion homeostasis and producing erythropoietin to regulate red blood cell mass.

Genetic and pathological analyses have revealed that altered metabolism is a primary defining feature of human ccRCC, which accounts for >75% of the 60,000 cases of kidney cancer diagnosed annually in the US (AC, 2014; TCGAR, 2013). ccRCC cells are characterized histologically by a clear cytoplasm that reflects elevated lipid and glycogen deposition (Rini et al., 2009; Varela et al., 2011). Although early-stage ccRCC patients can be treated effectively with surgery, approximately 30% are diagnosed with invasive or metastatic carcinoma, for which the five-year survival rate is less than 15% (Cohen and McGovern, 2005). ccRCC is highly resistant to chemo- and radio-therapies, and only a subset of patients (~20-30%) respond to immune checkpoint blockade (Cohen and McGovern, 2005; Hammers, 2016). A variety of other treatment options include multiple anti-angiogenics, mTOR inhibitors, and hypoxia inducible factor (HIF2α) antagonists (reviewed in (Ricketts et al., 2016); however, it remains unclear how to stratify patients for each of these more “targeted” therapies.

A hallmark genetic event in ccRCC is deletion or translocation of chromosome 3p (>90%), resulting in inactivation of the von Hippel-Lindau (VHL) gene (Gossage and Eisen, 2010; Keith et al., 2012; Nickerson et al., 2008). VHL encodes the substrate recognition component of an E3 ubiquitin ligase complex targeting the α subunits of HIFs for normoxic degradation (Ricketts et al., 2016). In VHL-deficient ccRCCs, constitutive HIF expression reprograms cellular metabolism and induces tumor angiogenesis, primarily through activation of downstream HIF target genes (Majmundar et al., 2010). Therefore, aberrant HIF signaling is believed to be a major driving force in ccRCC progression. Apart from VHL loss, ccRCC exhibits remarkable genetic heterogeneity (Gerlinger et al., 2012). Recent large-scale analyses identified frequent mutations in three genes, PBRM1 (~40%), SETD2 (~15%), and BAP1 (~15%), all of which encode epigenetic regulators and reside in a 43 Mb region on chromosome 3p that encompasses VHL (Dalgliesh et al., 2010; Pena-Llopis et al., 2012; Sato et al., 2013; TCGAR, 2013; Varela et al., 2011). In addition to these epigenetic factors, PI3K/mTOR signaling components are also mutated in a subset (<30%) of ccRCC tumors. These genetic alterations add substantial complexity to the genomic landscape of ccRCC and reflect considerable intratumoral heterogeneity. However, the extensive glycogen and lipid accumulation in ccRCC suggests that metabolic perturbations play a causative role in ccRCC tumor formation, as previously suggested (Hakimi et al., 2013; Linehan et al., 2010).

ccRCC tumors are increasingly characterized by dramatic changes in the metabolic pathways described above, as well as defects in one-carbon, nucleotide, and glycerophospholipid biochemistry (Hakimi et al., 2013; Hakimi et al., 2016). It appears that these common metabolic abnormalities, together with elevated mTORC1 signaling, endow ccRCC cells with significantly enhanced growth and survival. Previous expression profiling analyses demonstrated that ccRCC is distinguished by an overall decrease in mRNAs encoding multiple metabolic enzymes (relative to normal kidney) (Li et al., 2014), and gene set enrichment and metabolomics analyses revealed that gluconeogenesis/glycogen storage is the most significantly repressed metabolic pathway in ccRCC (Figure 1A). We subsequently determined that repression of the gluconeogenic enzyme fructose-1,6-bisphosphatase (FBP1) is critical for ccRCC cell growth, primarily through an unexpected mechanism in which FBP1 associates directly with chromatin to regulate gene expression (Li et al., 2014).

Figure 1. Urea Cycle mRNA, Protein, and Metabolite Alterations in ccRCC.

(A) Metabolic gene set analysis of RNAseq data provided by the TCGA, classified according to KEGG (Li et al., 2014). Generated metabolic gene sets were ranked based on their median fold expression changes in ccRCC tumor (n=480) vs. normal tissue (n=69), and plotted as median ± median absolute deviation. (B) The complete urea cycle as configured in the liver. Inset: TCGA-derived gene expression changes of urea cycle enzymes in ccRCC. (C) Copy number variation and mutational burden of urea cycle enzymes in 184 ccRCC tumors (data from TCGA). (D) ARG2 and ASS1 mRNA levels in ccRCC (TCGA). ***p <0.001, Welch’s t-test. (E) Copy number variation in ARG2 and ASS1 in ccRCC patients. Kaplan-Meier survival analysis of copy number loss in ARG2 and ASS1 (from TCGA data). Mantel-Cox log-rank test was performed. (F) Representative immunohistochemistry images of ARG2 or ASS1 protein in primary ccRCC, N = normal, T = tumor. Scale bars represent 100 μm (G) Violin plot of urea cycle metabolite abundance in primary ccRCC combining two independent datasets, n = 158 (Hakimi et al., 2016; Li et al., 2014). Data are displayed as the log₂ tumor/normal fold change and are pseudo-colored according to the intensity of the fold change in median. The internal bars represent the mean and SD of 158 tumor/normal pairs. Abbreviations: ARG2, arginase 2; ASS1, argininosuccinate synthase 1; ORNT1, ornithine translocase 1 (also called SLC25A15); CPS1, carbamoylphosphate synthase 1; OTC, ornithine transcarbamylase. See also Figure S1 and Table S1.

In this report, we combine analyses of DNA copy number variation, exome sequence changes, genome-wide RNA profiles, and metabolomics to demonstrate that urea cycle enzyme expression is also consistently repressed in ccRCC. Urea is generated from ammonia released by nucleotide and amino acid catabolism, and urea cycle activity in liver and kidney avoids toxic ammonia accumulation in the circulation (“hyperammonemia”). Urea cycle enzymes, including carbamoylphosphate synthase (CPS), convert free ammonia to carbamoylphosphate in hepatocyte mitochondria, which is then converted to cytosolic arginine by argininosuccinate synthase 1 (ASS1), argininosuccinate lyase (ASL), and arginase 2 (ARG2), and ultimately catabolized to urea and ornithine (Figure 1B). In normal human physiology, kidneys daily excrete grams of ammonia in the form of urea, and produce arginine that is largely exported to other organs. While the precise role of ammonia metabolism in ccRCC is unknown, the near-universal repression of genes encoding multiple urea cycle enzymes suggests that their normal physiological activities may be tumor suppressive. In this report, we identify two molecular mechanisms, namely enhanced pyridoxal phosphate (PLP) consumption and polyamine synthesis, by which urea cycle enzymes (especially ARG2) function to suppress ccRCC cell growth in vitro and in vivo. As such, urea cycle suppression promotes both ccRCC cell proliferation (where PLP conservation enhances anabolic metabolism), and viability (where polyamine uptake from the microenvironment results in toxic intracellular levels of putrescine and spermine). Our data suggest that these distinct metabolic adaptations, along with constitutive HIF activation, are defining hallmarks of ccRCC.

RESULTS

ccRCC Distinguished by Loss of Urea Cycle Enzymes

We employed metabolic gene set and direct metabolomic analyses of matched human ccRCC tumor and adjacent healthy kidney samples to identify the most significantly suppressed metabolic pathways in renal cancers (Li et al., 2014). These determined that genes encoding multiple urea cycle enzymes, including arginase 2 (ARG2), argininosuccinate synthase 1 (ASS1), and argininosuccinate lyase (ASL) (Figures 1A and 1B) were, along with those encoding carbohydrate storage enzymes (Li et al., 2014), the most significantly downregulated in ccRCC. Data from The Cancer Genome Atlas (TCGA) also revealed ARG2 and ASS1 copy number loss in 38% and 24% of ccRCC patients, respectively, and copy number gain of ASL in 40% (Figure 1C). Of note, ARG2 copy number loss is accompanied by ASS1 deletion in approximately 40% of ARG2-deficient ccRCCs (n=184), and ARG2 and ASS1 mRNA expression in multiple cohorts was lower in ccRCC tumors compared to normal kidney, irrespective of disease stage (Figures 1D, S1A, and S1B). Importantly, combined ARG2 and ASS1 loss in ccRCC tumors correlated with a nearly three-year reduction in patient survival, compared to patients whose tumors retained wild type ARG2 and ASS1 copy number (Figure 1E). These differences are reflected in markedly reduced ARG2 and ASS1 protein abundance in ccRCC tumors as assessed by immunohistochemistry (IHC) staining of primary ccRCC tissue (Figure 1F, and S1C) and primary sample cores (Figure S1D). Loss of ARG2 and ASS1 could reflect a less “differentiated” phenotype or reentry into the cell cycle by ccRCC cells. We assessed ARG2 and ASS1 expression in HK-2 cells arrested at different cell cycle stages using pharmacological approaches (see Experimental Procedures and Figure S1E-F). However, ARG2 and ASS1 mRNA levels were actually elevated in G1/S phases, relative to G2 and M. ARG2 and ASS1 protein levels were similarly increased during S phase (Figure S1F). These data indicate that ARG2 and ASS1 expression is not downregulated during proliferation, suggesting that that their underexpression in ccRCC is not a consequence of cellular dedifferentiation.

The abundance of multiple metabolites in the “urea cycle, arginine, and proline metabolism” KEGG classification was also significantly altered, as arginine, proline, and urea levels were all lower in ccRCC tumors compared to the normal kidney (Figure 1G, and Table S1). In particular, the nitric oxide synthase inhibitor dimethylarginine was reduced in ccRCC tumors, which may contribute to robust vascularization characteristic of ccRCC (Aziz et al., 1994; De Gennaro Colonna et al., 2009). Given the high frequency of gene copy number loss, reduced mRNA and protein expression, altered urea cycle metabolite levels, and correlative survival differences, we further explored the role of ARG2 and ASS1 as potential metabolic suppressors of ccRCC tumorigenesis.

ARG2 and ASS1 Re-expression suppress ccRCC cell growth

To assess the functional consequences of decreased ARG2 and ASS1 expression in ccRCC, we compared control HK-2 immortalized proximal tubule kidney epithelial cells, which retain ARG2 and ASS1, to several ccRCC cell lines that lack detectable ARG2 and/or ASS1 protein (Figure 2A). Whereas all ccRCC cell lines lack significant ARG2 expression (consistent with primary patient samples), only a subset (786-O, A498, RCC10, UOK101) exhibit decreased ASS1 abundance when cultured in vitro, and 786-O and A498 cells were consequently chosen for the majority of subsequent analyses. Stable knockdown of ARG2, ASS1, or both ARG2 and ASS1 (Double KD), mRNAs (Figure 2B) increased proliferation of control HK-2 cells in tissue culture (2D growth; Figure 2C) and anchorage-independent 3D soft agar growth assays (Figure 2D); moreover, the 3D growth effects were enhanced when both ARG2 and ASS1 were reduced simultaneously. HK-2 cells were grown in low serum/low glucose conditions (1% FBS, 1 mM glucose) to mimic nutrient availability typical of the tumor microenvironment. These results were subsequently confirmed using clonally selected HK-2 cells with CRISPR/Cas9 mediated deletion of ARG2 or ASS1 (Figure 2E). Remarkably, loss of a single urea cycle enzyme (i.e. ARG2 or ASS1) was sufficient to give normal kidney cells a significant growth advantage. Conversely, ectopic ARG2 and/or ASS1 re-expression in polyclonal 786-O and A498 ccRCC cells reduced 3D spheroid volume, the number of cells per spheroid, soft agar colony formation, and 2D growth (Figures 3A-B, and S2A). Interestingly, ARG2 re-expression only modestly inhibited the growth of UMRC2 ccRCC cells (Figure S2B), which may reflect the presence of endogenous ASS1 and slightly elevated baseline ARG2 in these cells, compared to other ccRCC cell lines (Figure 2A). Of note, ARG2 and ASS1 re-expression was carefully titrated in an attempt to match protein levels in HK-2 cells throughout these studies (Figures 3A and S2A-B).

Figure 2. Loss of ARG2 and ASS1 Enhances Proliferation of Normal Kidney Cells.

(A) ARG2 and ASS1 protein levels in HK-2 kidney epithelial and various ccRCC cell lines. Tubulin or actin served as loading controls. (B) Upper panel: short hairpin-mediated knockdown of ARG2, ASS1, or both simultaneously (Double KD) in HK-2 cells. Tubulin or actin served as loading controls. Lower panel: mRNA levels after knockdown with two independent hairpins. Error bars represent SEM of 3 independent replicates. (C-D) Knockdown of ARG2 or ASS1 augments HK-2 cell growth in (C) 2D growth and (D) 3D soft agar colony formation assays. Double KD experiments employed ARG2 SH-1 and ASS1 SH-4. In (C), cells were grown in 1% FBS and 1 mM glucose media (changed every 2 days). Error bars represent SEM from 3 replicate wells. **p < 0.01, ***p < 0.001, one-way ANOVA with Tukey’s multiple comparisons test. (E) Clonally selected CRISPR/Cas9-mediated ARG2^−/− or ASS1^−/− HK-2 cells demonstrate enhanced growth. Guide RNA against EGFP was used as a control, ns = non-specific band. Cells were grown in 1% FBS and 1 mM glucose media (changed every 2 days). Error bars represent SEM of 16 replicate wells. See also Figure S2.

Figure 3. ARG2 Enzymatic Activity is Required for ccRCC Growth Suppression.

(A) Left panel: Ectopic expression of ARG2 and/or ASS1 in 786-O ccRCC polyclonal populations. Numbers below ARG2 and ASS1 western images represent the quantified fold change in protein compared to normal HK-2 cells. Right panel: 3D spheroid growth assay of 786-O cells. After three weeks of growth, spheroid volume was measured via microscopy and the number of cells per spheroid was counted after accutase digestion of spheroids to a single cell suspension. Error bars represent SEM of three independent replicates containing 12 spheroids each. B) Soft agar growth assays of 786-O ccRCC cells. Error bars represent SEM of 5 replicate wells. (C) LC/MS quantitation of arginine, citrulline, and aspartate in 786-O cells stably expressing urea cycle enzymes. Error bars represent SEM of 4 replicates. (D) Expression and activity of wild type (WT) and H160F mutant ARG2 and (E) 786-O spheroid growth assay demonstrating ARG2 catalytic activity-dependent growth suppression. Error bars represent SEM of 3 independent replicates containing 12 spheroids each. (F) 786-O soft agar colony growth assay depicting ARG2 catalytic activity-dependent growth suppression. Error bars represent SEM of 3 replicate wells. *p < 0.05, **p < 0.01, ***p < 0.001, a one-way ANOVA with Tukey’s multiple comparisons test was conducted for each dataset unless otherwise specified.

ARG2 Suppresses ccRCC Growth via its Enzymatic Activity

To confirm that exogenously expressed ARG2 and ASS1 proteins are catalytically active, we performed liquid chromatography-mass spectrometry (LC/MS) to measure arginine, citrulline, and aspartate levels in transduced 786-O ccRCC cells. As expected, arginine levels were significantly reduced upon ARG2 expression, as were both citrulline and aspartate upon ASS1 expression (Figure 3C). Combined ARG2 and ASS1 expression decreased the abundance of all three urea cycle metabolites, most notably citrulline and aspartate. We reported previously that a key metabolic enzyme, FBP1, has catalytic activity- independent mechanisms of tumor suppression in ccRCC (Li et al., 2014). To test whether this was true for ARG2, we generated a missense mutation (ARG2 H160F) that abolished enzymatic activity (Figure 3D), consistent with previous reports (Xiong et al., 2014). In a 3D spheroid growth assay, expression of wild type ARG2 dramatically reduced ccRCC spheroid volume and the number of cells per spheroid, whereas ARG2 H160F failed to suppress spheroid growth (Figure 3E) and soft agar colony formation (Figure 3F). These data clearly demonstrate that ARG2 catalytic activity is required for ccRCC growth suppression, in contrast to FBP1 (Li et al., 2014). Loss of multiple urea cycle enzymes would be predicted to render ccRCC cells dependent on exogenous arginine for proliferation. Indeed, 786-O and A498 cells appear to be arginine auxotrophs, based on the absence of growth in arginine deficient media (Figure S2C). These findings are consistent with Yoon et al., who previously reported that kidney cancer cells exhibit low ASS1 levels and sensitivity to arginine deprivation (Yoon et al., 2007). However, unlike hepatocytes, normal kidney cells do not express all urea cycle enzymes (see Figure S3C). Therefore, they would not be expected to survive total deprivation of exogenous arginine either (Figure S2C). ccRCC cells typically grow in medium containing 5X lower arginine concentrations (84 mg/l) than HK-2 cells (421 mg/l); when HK-2 cells were grown in 50X less arginine, their growth was identical to that in replete conditions (Figure S2D, right panel), suggesting that they may be less sensitive to arginine depletion, as shown for 786-O cells grown in 50X less arginine concentrations (Figure S2D, left panel).

ARG2 and/or ASS1 suppress ccRCC growth in vivo, independent of mTORC1 activity and nucleotide synthesis

We evaluated the effects of ARG2 and/or ASS1 expression on ccRCC xenograft tumor growth in nude mice, as a fully accurate, autochthonous murine ccRCC model has only recently been described (Nargund et al., 2017). Ectopic ARG2 expression in 786-O ccRCC xenografts reduced tumor growth in vivo, which was further suppressed by co-expression of ARG2 and ASS1 (Figures 4A and S3A). Compared to ARG2, ASS1 expression produced a relatively modest effect on tumor volume at later time points (days 38-40) (Figure 4A), although the reduction in tumor mass at the experimental endpoint failed to achieve statistical significance (Figure S3A). It is noteworthy that ARG2 H160F failed to significantly suppress either tumor volume or mass (Figures 4A, S3A), consistent with in vitro assays described in Figure 3. Interestingly, the ASS1-, ARG2-, and ASS1/ARG2-expressing tumor cohorts exhibited decreased tumor cell proliferation and some evidence of reduced mTORC1 activity (Figures 4B-C), and ARG2- and combined ARG2/ASS1-expressing tumors displayed the expected patterns of altered urea concentration (Figure S3B). We then focused primarily on ARG2, as it exhibited the most consistent and robust suppression of growth and mTORC1 activity in multiple ccRCC cell lines both in vitro and in vivo (e.g. Figures 3A, 4A, 4C, and S3A). Moreover, the urea “cycle” is configured somewhat differently in the kidney than the liver, where ARG2 resides in the mitochondria rather than the cytoplasm in renal cells (compare Figures 1B and S3C), rendering ARG2 mediated reactions somewhat biochemically “independent”. An unbiased metabolomics analysis to assess the impact of expressing ARG2 in ccRCC xenograft tumors was performed (Figure 4D and Table S2). ARG2 expression caused striking widespread metabolic changes, most notably significant decreases in the concentration of multiple amino acid, lipid, nucleotide, and vitamin/cofactor metabolites. In particular, ARG2 expression decreased the abundance of 10 naturally occurring amino acids (Figure 4E).

Figure 4. ARG2 Suppresses in vivo Tumor Growth and Leads to Multiple Metabolic Changes.

(A) 786-O xenograft tumor growth in nude mice, where cells re-express ASS1, wild type ARG2, ARG2H160F mutant, or both ASS1 and ARG2 (n = 10 mice in each cohort, except for n=5 for the ARG2 H160F study. (B) Immunohistochemistry staining for Ki67 as a marker of proliferation in xenograft tumors expressing urea cycle enzymes, high power fields (HPF). (C) mTORC1 activity in ARG2- or ASS1- expressing xenograft tumors, as assessed by the mTORC1 targets phospho-S6K1, phospho-S6, and phospho-4E-BP1. (D) Unbiased metabolomic analysis of control or ARG2 expressing xenograft tumors, n = 7 for each cohort. ARG2 expression causes robust global reductions in several categories of metabolites, especially lipids, nucleotides, and biosynthetic co-factors. (E) Amino acid abundance in 786- O xenograft tumors (n = 7). Data are represented as the percent change in amino acid, after normalization to each corresponding control tumor. *p < 0.05, **p < 0.01, ***p < 0.001, Welch’s t-test was used for each dataset. See also Figures S3 and S4, and Table S2.

Given the requisite role of amino acids in controlling the activity of the mechanistic target of rapamycin (mTOR) nutrient sensing pathway (Gonzalez suppress ccRCC growth by decreasing amino acid abundance and thus, mTORC1 activation. Indeed, and Hall, 2017), we postulated that ARG2 might mTORC1 activity was reduced in xenograft tumors expressing either ARG2, ASS1, or both enzymes, as determined by phospho-S6K1, phospho-S6, phospho-4E-BP1 and phospho-ULK-1 levels (Figures 4C, S4A, and S4B). The phosphorylation status of these mTORC1 substrates was not accompanied by changes in the activity of upstream regulators Akt and TSC2 (Figure S4A), which respond to growth factors instead. However, several experimental results indicated that the ARG2-dependent phenotypes we observed are (at least in part) mTORC1-independent. For example, we initially hypothesized that elevated citrulline in ASS1-deficient ccRCC cells might activate mTORC1, as previously reported (Breuillard et al., 2015; Cynober et al., 2013; Le Plenier et al., 2012; Moinard and Cynober, 2007); however, exogenous citrulline (in contrast to leucine) had no consistent effect on S6 and 4E-BP1 phosphorylation in 786-O cells, even when present at supra-physiological levels (Figure S4C).

An alternative model posited that ARG2 and ASS1 expression could alter the subcellular distribution of mTORC1, which responds to amino acid stimulation by translocating to the lysosomal surface in a RAG GTPase-dependent manner (Bar-Peled and Sabatini, 2014). However, expression of these enzymes in ccRCC cells had no effect on subcellular mTORC1 lysosomal localization (Figure S4D). We also introduced two constitutively active mTORC1 mutants (R2505P and L1460P; (Grabiner et al., 2014)) into 786-O cells with and without ARG2 reconstitution. Figure S4D indicates that both constructs override ARG2-mediated mTORC1 suppression, based on phospho-S6 levels. Intriguingly, mTORC1 R2505P and L1460P, if anything, suppress growth of while failing to fully restore the growth of ARG2-expressing cells (Figure S4E). A direct comparison is control cells in soft agar colony formation assays, provided when one assesses the vector plus mTORC1 R2505P- transduced cells to ARG2 plus mTORC1 R2505P- expressing cells. These observations can be explained by suppression of basal autophagy (which may be active in 3D growth assays where colonies develop O₂/nutrient gradients) by constitutively active mTORC1. We also examined whether ARG2 instead modulates mTORC1 through global changes in amino acid availability, by expressing a constitutively active form of RagB, RagB Q99L (RagB^GTP), which confers high levels of mTORC1 activity (Sancak et al., 2008), even in the absence of amino acids (see Figure S4G). We reasoned that if ARG2 suppresses ccRCC tumor growth by depleting amino acids, thereby reducing mTORC1 activity, constitutively active RagB^GTP should rescue ARG2-dependent tumor growth suppression. Unexpectedly, the growth of 786-O ccRCC xenograft tumor cells co-expressing both ARG2 and RagB^GTP was equivalent to those expressing ARG2 alone, indicating that mTORC1 is not solely responsible for ARG2-dependent ccRCC growth suppression (Figure S4H).

Decreased expression of ASS1 and ASL has been reported in other cancers (Huang et al., 2013; Kobayashi et al., 2010; Rabinovich et al., 2015), although ccRCC appear to have suppressed additional urea cycle enzymes, especially ARG2. Of note, Rabinovich et al. demonstrated that ASS1 deficiency increases cytoplasmic aspartate levels (Rabinovich et al., 2015), which are critical for maintaining pyrimidine biosynthesis in multiple cancer cell lines (Birsoy et al., 2015; Sullivan et al., 2015). Specifically, ASS1-deficient sarcoma cells divert aspartate away from the urea cycle to support de novo nucleotide production and DNA replication (Rabinovich et al., 2015). Because ARG2-expressing xenograft tumors exhibited reduced aspartate levels (Figure 4E), we explored the possibility that supplying exogenous aspartate might reverse ARG2-dependent growth suppression, given that ARG2 re-expression also resulted in decreased nucleotide abundance (Figure 4D). We employed the 3D spheroid assay for this purpose, where ARG2 re-expression results in decreased spheroid volume (Figure S5A and B), mimicking reduced in vivo tumor growth (Figure 4A). However, supplementation with supra-physiological levels of aspartate (20 mM) was incapable of increasing cell proliferation in ARG2 reconstituted 786-O spheroids (Figure S5C). 786-O cells express relatively low levels of the glutamate-aspartate transporter SLC1A3, also known as the excitatory amino acid transporter, or EAAT1 (data not shown), which could result in poor aspartate transport (Birsoy et al., 2015). We therefore generated 786-O cells stably expressing human SLC1A3/hEAAT1, and cultured them in media containing 150 μM aspartate in a 2D growth assay, along with vector controls (Figure S5D). If anything, SLC1A3/hEAAT1 overexpression resulted in decreased rates of ARG2/786-O cell proliferation relative to control cells. To confirm that 786-O cells import aspartate from extracellular medium, cells expressing both exogenous ARG2 and the aspartate transporter hEAAT1 were cultured with ¹³C-labelled aspartic acid for 6 hours and intracellular ¹³C-aspartate measured using GC/MS. In the results shown in Figure S5E, we observed that 786-O cells take up aspartate irrespective of ARG2 expression, although exogenous hEAAT1 enhances aspartate transport by 30-40X in both settings. Therefore, the results presented in Figure S5C and S5D reflect cells importing exogenous aspartate, especially when expressing hEEAT1.

Kim et al. recently determined that increased expression of the urea cycle enzyme carbamoyl phosphate-1 in Kras/LKB1 mutant lung cancer cells is essential to maintain their intracellular pyrimidine pools, DNA synthesis, and genome integrity (Kim et al., 2017). However, culturing ARG2/786-O cells in various concentrations of cytidine, thymidine, or all four deoxyribonucleosides failed to change 3D spheroid growth rates over a two-week period (Figure S5E-F). Overall, these results strongly suggest that 1) decreased mTORC1 signaling, aspartate levels, and/or nucleotide availability are not sufficient to fully account for the growth suppressive effects of ARG2 in ccRCC, and 2) ARG2’s role as a metabolic tumor suppressor is dependent on its catalytic activity.

ARG2 Suppresses ccRCC Growth via Depletion of the Essential Cofactor Pyridoxal Phosphate

To define alternative mechanisms whereby urea cycle enzymes influence ccRCC tumor growth, we focused on the downstream metabolites altered by ARG2 activity, which hydrolyzes arginine to generate ornithine and urea. Ornithine has multiple biochemical fates, and can be converted by mitochondrial ornithine aminotransferase (OAT) to glutamate-gamma-semialdehyde, a substrate for the synthesis of both glutamate and proline (Figure 5A). ¹⁵N₄-arginine tracing in HK-2 cells revealed that M+1 enrichment of both glutamate and proline was abolished upon ARG2 knockout (Figure 5B, left panel). Conversely, ectopic ARG2 expression in 786-O ccRCC cells dramatically increased ¹⁵N₄-arginine incorporation into glutamate and proline (Figure 5B, right panel), suggesting that ARG2 directly controls the levels of downstream ornithine metabolites by increasing substrate availability for OAT. Of note, the absolute concentration of glutamate and ornithine did not change in these experiments (Figure S6A).

Figure 5. ARG2 Suppresses ccRCC Growth via Depletion of Pyridoxal Phosphate (PLP).

(A) Schematic representation of ¹⁵N₄-Arginine labeling into glutamate and proline synthesis. Ornithine decarboxylase (ODC) and ornithine aminotransferase (OAT) reactions and their requirement for pyridoxal phosphate (PLP) as a cofactor are highlighted. Closed red circles represent heavy nitrogen (¹⁵N) while open black circles represent carbon atoms (¹²C). (B) ARG2 controls the downstream synthesis of glutamate and proline. HK-2 (left panel) or 786-O cells (right panel) incubated with ¹⁵N₄-Arginine (same conditions as in Figure 6B). Synthesis of glutamate and proline from arginine results in one -¹⁵N incorporation (M1 enrichment), represented as atom percent excess (APE). Error bars represent SEM of 3 and 5 replicates for HK-2 and 786-O, respectively. (C) Schematic for the synthesis of PLP. Several precursor forms of pyridoxal exist and are interconverted between their phosphorylated versions; only PLP is the bioactive cofactor. (D) mRNA levels of pyridoxal kinase (PDXK) in ccRCC (data from TCGA). Box plots represent 69 normal and 480 tumor samples, one-way ANOVA with Tukey’s multiple comparisons test. (E) ARG2 directly modulates the levels of bioactive PLP. HK-2 (left panel) or 786-O cells (right panel) were grown in 1% FBS with 1 mM glucose for 24 hours (HK-2) or hypoxia (0.5% O₂) for 48 hours (786-O) prior to quantifying the total PL, PLP, PN, and PM levels via LC/MS. Error bars represent SEM of 3 and 5 replicates for HK-2 and 786-O, respectively. (F) Ectopic ARG2 and/or PDXKexpression in 786-O cells. (G) PDXK overexpression alters levels of pyridoxine pathway components. 786-O cells were cultured under hypoxia (0.5% O₂) for 48 hours prior to quantifying total PL, PLP, PN, and PM levels via LC/MS. Error bars represent SEM of 5 replicates. (H-I) PDXK expression reverses ARG2-dependent growth suppression. 3D soft agar colony formation assay of 786-O cells ectopically expressing either ARG2 and/or PDXK. (H) Quantification of colony number and (I) representative pictures of colony size. Scale bar represents 25 μm. Error bars represent SEM of 6 replicate wells for each condition, one-way ANOVA with Tukey’s multiple comparisons test. *p < 0.05, **p < 0.01, ***p < 0.001, a two-way ANOVA with Tukey’s multiple comparisons test was conducted for each dataset unless otherwise specified. See also Figures S5 and S6.

An alternative metabolic fate for ornithine is decarboxylation by ornithine decarboxylase (ODC) to generate the polyamine putrescine (Figure 5A). Interestingly, both OAT and ODC require the cofactor pyridoxal-5′-phosphate (PLP) for catalysis. PLP is a vitamin B₆ derivative involved in approximately 4% (~140) of all known cellular enzymatic reactions (Percudani and Peracchi, 2003). Vitamin B₆ encompasses six various inter-converted forms, including pyridoxine (PN), pyridoxal (PL), pyridoxamine (PM), and their phosphorylated versions (Figure 5C); however, only PLP is the bioactive cofactor used in enzymatic reactions (Galluzzi et al., 2013). Phosphorylation of the various B₆ precursors to generate active PLP requires pyridoxal-5’-phosphate kinase (PDXK), the expression of which is reduced across all stages of ccRCC (Figure 5D).

As PLP is a cofactor in numerous enzymatic reactions, we hypothesized that small perturbations in PLP levels could have significant cell growth effects. Furthermore, because ARG2 supplies the ornithine substrate for ODC and OAT (Figure 5A), enhanced reaction rates of these enzymes in ARG2- expressing cells might deplete PLP pools. Ectopic ARG2 expression significantly reduced the concentration of PLP, as expected, with no effect on the precursors PL, PN, etc. in 786-O ccRCC cells (Figure 5E, right panel). Conversely, deletion of ARG2 in HK-2 cells produced significantly elevated concentrations of PLP (Figure 5E, left panel), demonstrating that ARG2 activity modulates the pool of available PLP.

In complementary experiments, we ectopically expressed PDXK to increase PLP levels (Figure 5F and 5G), which rescued all ARG2-mediated growth defects, including slowed 2D proliferation (data not shown), decreased 3D soft agar colony numbers (Figure 5H), and reduced colony size (Figure 5I). This approach was adopted because PLP is too labile to add to semisolid culture media as a supplement (data not shown). To demonstrate that ectopic PDXK was having the predicted effects on intracellular levels of pyridoxine pathway components, we employed LC/MS for their quantification in ARG2- and PDXK-expressing 786-O cells (see Experimental Procedures). Certain pyridoxal derivatives (PM, PNP, PMP) were difficult to accurately measure, as they exist at the borderline of detection in our experimental system. Nevertheless, expressing PDXK in 786-O cells reduced intracellular levels of the PLP precursors PN, PL, and PM (Figures 5G and S6B). Interestingly, steady-state PLP levels were also decreased in PDXK-expressing vector control cells (Figure 5G), likely due to PLP consumption by multiple intracellular reactions. However, ectopic PDXK expression increases PLP levels in ARG2-expressing cells, albeit not to the levels observed in controls (Figure 5G). In aggregate, these data support the hypothesis that ARG2 inhibits ccRCC growth, at least in part, by depleting the intracellular PLP cofactor pool, resulting in reduced production of amino acids and numerous other metabolites.

Excess Polyamine Production Upon ARG2 Expression Leads to ccRCC Growth Suppression

We hypothesized that altered polyamine synthesis or metabolism might also contribute to the growth suppressive effects of ARG2 expression in ccRCC. Ornithine-derived putrescine is converted to higher order polyamines spermidine and spermine through the addition of decarboxy-SAM-derived aminopropyl groups (Michael, 2016) (Figure 6A). While polyamines promote tumor progression in some contexts (Soda, 2011; Tsujinaka et al., 2011), they cause cellular toxicity in other tumor cells (Pegg, 2013), although the precise mechanisms by which polyamines either promote or inhibit tumor growth are not well characterized. Treatment of 786-O ccRCC cells with exogenous putrescine resulted in a dose-dependent toxicity, whereas putrescine had no effect on the growth of control HK-2 cells, even at relatively high (10 mM) concentrations (Figure 6B). To examine the sensitivity of non-ccRCC cancer cells to exogenous polyamines, we studied prostate cancer cells (VCaP and LnCaP), reasoning that the prostate is an important physiological source of polyamines. Accordingly, prostate cells were cultured in the presence of 5 mM and 10 mM putrescine, and appeared to be more resistant to putrescine levels toxic for 786-O, even when growing rapidly (Figures 6B and S7B). Putrescine concentrations used for these experiments were chosen based on calculations described in the Experimental Methods. In aggregate, these data suggest that the metabolic wiring of ccRCC cells makes them particularly sensitive to changes in polyamine concentrations.

Figure 6. ccRCC Tumors Exhibit Increased Polyamine Levels and Correspondingly Alter Gene Expression to Lower de novo Synthesis.

(A) Schematic representation of polyamine biosynthesis. Each enzyme contains a corresponding box plot with TCGA mRNA data in normal (N) and ccRCC (T) samples. Each y-axis represents the normalized RNA-seq reads of the gene; n = 69 normal kidney and 480 tumor samples. *** p< 0.001 (B) 786-O cells grown under hypoxia (0.5% O₂) and supplemented with various concentrations of putrescine (Put), andHK-2 cells grown in 1 mM glucose, 1% FBS conditions, supplemented with Put. Cell growth was assessed by WST-1 assays, and error bars represent SEM from 10 replicate wells. Prostate cancer (VCaP) cells were supplemented with indicated concentrations of putrescine, and cell counts performed over the indicated times. Error bars represent SEM from 4 replicate wells. (C) Polyamine metabolite abundance in primary ccRCC combining two independent datasets, n = 158 (Hakimi et al., 2016; Li et al., 2014). Data are displayed as the fold change in metabolite abundance of each tumor, normalized to its control sample. (D) Quantification of immunohistochemistry staining of primary ccRCC tissue microarray,demonstrating reduced ODC protein in ccRCC tumors. *** p <0.001, Welch’s t-test. See image in Figure S6B. (E) Kaplan-Meier survival analysis of OAZ1 expression (from TCGA data) in ccRCC. ‘Low’ versus ‘High’ data are segregated based on the median expression of OAZ1. Mantel-Cox log-rank test was performed. Abbreviations: OAZ, ornithine decarboxylase antizyme; AZIN1, ornithine decarboxylase antizyme inhibitor; ODC1, ornithine decarboxylase1; AMD1, adenosylmethionine decarboxylase1; SRM, spermidine synthase, SMS, spermine synthase. See also Figures S6 and S7.

Hypoxia has been shown to promote the cellular uptake of polyamines in both normal and malignant cells, downstream of HIF1α stabilization (Aziz et al., 1994; Babal et al., 2002; Tantini et al., 2006; Tsujinaka et al., 2011). Of note, hypoxia can also stimulate ODC expression and activity (Babal et al., 2002). We speculated that naturally occurring tumor hypoxia and constitutive HIF activity might commit ccRCC cells to take up potentially toxic levels of polyamines from the microenvironment, thereby selecting for cells with lower endogenous polyamine synthesis (Svensson et al., 2008). This hypothesis is supported by unbiased metabolomics analyses of two independent ccRCC cohorts (n=158 samples in total; (Hakimi et al., 2016; Li et al., 2014) that demonstrated a marked overall elevation in putrescine and spermine concentrations, with minor elevations in the levels of spermidine (Figure 6C). mRNA expression data are also consistent with reduced de novo polyamine synthesis in ccRCC cells (Figure 6A). For example, transcripts encoding ARG2, ODC1, and spermine synthase (SMS) are expressed at lower levels in ccRCC than normal kidney tissue (Figure 6A). Additionally, upstream regulators of ODC, which catalzyes the committed step in polyamine synthesis, are modulated in a manner consistent with this model. Specifically, ornithine decarboxylase antizymes (OAZ) 1, 2, and 3 bind the ODC protein and target it for degradation via the 26S proteasome (Tajima et al., 2016). OAZ expression is elevated in ccRCC (Figure 6A), consistent with reduced ODC protein levels (Figures 6D and S6B). Elevated OAZ expression in ccRCC also correlates with worse patient survival (Figure 6E). Furthermore, expression of antizyme inhibitor 1 (AZIN1), which regulates polyamine synthesis by binding and inhibiting OAZ, thereby stabilizing ODC and promoting polyamine synthesis (Wu et al., 2015), was also reduced in ccRCC samples (Figure 6A), consistent with reduced polyamine synthesis. Taken together, these data support the hypothesis that altered expression of genes encoding polyamine pathway enzymes reduces endogenous polyamine synthesis in ccRCC as an adaptive response to the accumulation of toxic polyamines.

Metabolic tracing experiments were conducted to follow the fate of ¹⁵N₄-arginine in ccRCC cells with regard to downstream products such as ornithine and putrescine (Figure 7A). Ectopic expression of ARG2 in 786-O ccRCC cells led to robust incorporation of ¹⁵N₄-arginine into M+2 ornithine and urea, as well as the downstream metabolite M+2 putrescine (Figure 7B, bottom panel). Conversely, ¹⁵N₄-arginine incorporation into these metabolites was markedly reduced in ARG2-deficient HK-2 cells (Figure 7B, top panel). Ectopic ARG2 expression in 786-O cells elevated arginine-dependent putrescine synthesis nearly 3-fold, whereas ARG2 depletion in HK-2 cells reduced this by approximately 8-fold, independent of changes in ODC expression. Absolute levels of putrescine in 786-O and HK-2 cells were determined to be 2-4 mM in media lacking exogenous polyamines (Figure S6A), with ARG2 deletion reducing levels by 8X in HK-2 cells, and ARG2 re-expression increasing putrescine levels 4X in 786-O cells in these conditions. However, culturing the same renal cells in exogenous 2.5 mM putrescine dramatically increased its intracellular abundance >10-fold (Figures 7C and S7A). These findings indicate that kidney cells actively transport large quantities of polyamines from their microenvironment. As expected, both ARG2 expression and/or hypoxia increased intracellular putrescine levels (Figure 7C). ARG2 expression also increased the levels of spermidine and spermine, although the magnitude of this effect did not attain statistical significance. On the other hand, ARG2 knockout in HK-2 cells reduced the levels of putrescine (Figure S7A), while hypoxia had no effect (S7A), likely due to the lack of constitutive HIF activity in a VHL wild type setting.

Figure 7. ARG2 Expression Controls Polyamine Concentration and Subsequent Growth Modulation.

(A) Schematic representation of ¹⁵N₄-arginine contribution to polyamine biosynthesis. Closed red circles represent heavy nitrogen (¹⁵N) while open black circles represent carbon atoms (¹²C). (B) ARG2 increases arginine utilization for polyamines. Top panel, HK-2 cells grown in 1% FBS and 1 mM glucose for 24 hours and then supplemented with ¹⁵N₄-arginine for an additional 24 hours. M2 enrichment represents the proportion of each metabolite containing two - ¹⁵N atoms (atom percent excess [APE]). Bottom panel, 786-O cells grown under hypoxia (0.5% O₂) for 48 hours and then supplemented with ¹⁵N₄ Arginine for an additional 24 hours. Error bars represent SEM of 5 replicates. (C) 786-O cells grown under normoxia or hypoxia (0.5% O₂) for 48 hours followed by quantification of total polyamine pools with LC/MS. Error bars represent SEM of 8 replicates. (D-E) HK-2 cells grown in 1% FBS and 1 mM glucose and supplemented with various concentrations of putrescine or spermine. Error bars represent SEM of 8 replicates. (F) CRISPR/Cas9 mediated deletion of ornithine decarboxylase (ODC1) in ARG2 reconstituted 786-O cells. Numbers below ODC1 western blot images represent the quantified fold change in protein compared to 786-O lines with vector or ARG2, respectively. (G) ODC1 loss in 786-O cells reverses ARG2-mediated growth suppression in a soft agar colony formation assay. Error bars represent SEM of 3 replicate wells. *p < 0.05, **p < 0.01, ***p < 0.001, a two-way ANOVA with Tukey’s multiple comparisons test was conducted for each dataset unless otherwise specified. See also Figure S7.

As stated previously, HK-2 renal epithelial cells gain a proliferative advantage when ARG2 expression is reduced or eliminated (Figures 2C-E and Figure 7D). To test if reduced polyamine levels contribute to this growth advantage, we treated HK-2 cells with exogenous polyamines. CRISPR control HK-2 cells were unaffected by exposure to increasing levels of putrescine, spermine, or spermidine, (Figures 7D-E and S7B, left panels), while growth of ARG2-deficient HK-2 cells was inhibited by putrescine, spermine, and spermidine in a dose-dependent manner (Figures 7D-E and S7C, right panels). In contrast, exogenous ornithine had no effect on cell growth (Figure S6C). Stated another way, the growth advantage conferred by ARG2 loss in kidney cells is reversed specifically by supplying exogenous polyamines. To determine if ODC1 deficiency reverses polyamine overproduction in ARG2-reconstituted 786-O cells (see Figure 7A), CRISPR/Cas9 was used to achieve 70% deficiency in ODC1 protein in 786-O cell pools using ODC1 sgRNA-3 (Figure 7F). Soft agar colony assays revealed a significant rescue of colony formation when ARG2 expression and ODC1 deficiency were combined (Figure 7G). These data suggest that ARG2-dependent polyamine production, in addition to PLP consumption, directly modulates renal epithelial cell growth, where ccRCC cells experiencing tumor hypoxia avoid an overabundance of polyamines (due to uptake) by “rewiring” the biosynthetic machinery for these metabolites.

DISCUSSION

Nearly 40% of ccRCC patients exhibit ARG2 copy number loss, likely reflecting ARG2’s position at the nexus of dual mechanisms of tumor inhibition in ccRCC: 1) depleting available PLP pools, thereby removing a critical cofactor for multiple essential enzymatic reactions, and 2) increasing polyamines to toxic levels. It is now clear that altered metabolism plays an important role in cancers of the kidney, an organ responsible for excreting excess ammonia (as urea), maintaining inorganic ion homeostasis, and generating hormones (e.g. erythropoietin) and glucose. Of note, while ASS1 resides on chromosome 9 (9q34.11), ARG2 is located on 14q (14q24.1), a chromosome frequently deleted in ccRCC (Shen et al., 2011). Shen et al. previously proposed that multiple kidney cancer tumor suppressor genes exist on 14q (in addition to HIF1A), and ARG2 represents a very likely candidate. Only 3 out of 448 (0.7%) cases reported in the TCGA exhibit a deep deletion at the locus, whereas the remaining 38% cases with copy number variation exhibit mono-allelic copy number loss. As ARG2 mRNA levels are significantly decreased and ARG2 protein expression essentially undetectable, we conclude that any retained ARG2 alleles must be silenced in ccRCC tumors and cell lines. Interestingly, ASL exhibits copy number gain in ccRCC. We attribute this to its location on chromosome 7q (7q11.21), close to a pericentromeric region. Regions adjoining the centromere are often duplicated, which may explain the higher copy number of ASL detected in ~40% ccRCC.

Although gluconeogenesis and glycogen storage represent the most significantly down-regulated metabolic pathway in ccRCC, the urea cycle enzymes display a similar degree of repression in these tumors. We sought to determine if suppression of the urea cycle contributes directly to ccRCC disease progression. Whereas multiple cancers exhibit decreased ASS1 and/or ASL protein abundance (Huang et al., 2013; Kobayashi et al., 2010; Perroud et al., 2006; Rabinovich et al., 2015; Tang et al., 2006), ccRCC appear to have suppressed additional urea cycle enzymes, including ARG2. Re-expressing ASS1 (Figure 3) or ASL (unpublished data) in ccRCC cells inhibited proliferation, although our data indicate that ARG2 has more potent tumor suppressive effects in this setting. Nevertheless, which urea cycle intermediates are critical for these effects remains somewhat open-ended. ASS1-deficient sarcoma cells depend on aspartate diverted away from the urea cycle to support de novo nucleotide production and DNA replication (Rabinovich et al., 2015). In contrast, exogenous aspartate (or any deoxyribonucleoside) failed to reverse growth suppression in ARG2-reconstituted ccRCC cells (Figure S5), indicating that additional mechanisms must be at play. It is also important to contrast our results with those presented in the recent report by Kim et al., where Kras/LKB1mutant lung cancers overexpress the urea cycle enzyme carbamoyl phosphate-1 to maintain pyrimidine levels and DNA synthesis (Kim et al., 2017). ccRCC instead repress the expression of multiple urea cycle enzymes to also maintain nucleotide pools by different means, i.e. conservation of a critical biosynthetic cofactor (PLP) that ultimately supports their generation. ARG2- expressing ccRCC xenografts also exhibited decreased mTORC1 activity, coincident with decreased pools of arginine, aspartate, and glycine, suggesting that disrupted amino acid homeostasis might result in reduced mTORC1 signaling and cell growth. However, high levels of mTORC1 activity conferred by constitutive mTORC1 and RagB activity also failed to reverse ARG2-mediated growth suppression.

ARG2 expression had significant effects on pyridoxine metabolism, in that its deletion in renal epithelial cells, or reconstitution in ccRCC cells, changed overall intracellular PLP levels by a surprising 25-35%. Given the significant role of this essential vitamin B₆ component in numerous metabolic reactions critical for amino acid synthesis, nitrogen homeostasis, and glycogen production, ARG2 loss in ccRCC appears to maintain sufficient PLP to support more than 140 biosynthetic reactions that contribute to increased biomass and cell division. In fact, changes in PLP pools likely account for the substantial disruption of amino acid and subsequent nucleotide abundance (due to changes in aspartate levels) observed in ARG2 expressing xenografts. In aggregate, pyridoxine metabolism represents an understudied pathway for future therapeutic strategies for ccRCC treatment, with pyridoxal kinase a highly appealing potential drug target.

Metabolic tracing experiments with ¹⁵N₄-arginine revealed that ARG2 expression plays a critical role in arginine conversion to putrescine and urea. Our data indicate that human ccRCC tumors accumulate high levels of polyamines, despite having reduced ARG2 expression, suggesting that these cells import exogenous polyamines from the circulation or tumor microenvironment. The growth advantage conferred to normal renal epithelial cells by ARG2 deficiency is reversed by exogenous polyamines, suggesting that one benefit of ARG2 deficiency in ccRCC cells is avoiding toxic polyamine concentrations by reducing de novo synthesis of putrescine and spermidine. Relatively little is known about how polyamines regulate cancer cell behavior, particularly in ccRCC, although polyamine biosynthetic enzymes are controlled at the level of transcription, translation, and protein degradation by a variety of highly conserved feedback loops (see Figure 6A). While intracellular polyamines can support cell division and anabolic metabolism (Babal et al., 2002; Svensson et al., 2008; Tsujinaka et al., 2011), they can also be converted to toxic byproducts like acrolein (Moghe et al., 2015; Stevens and Maier, 2008). Preliminary experiments indicate that ARG2-reconstituted ccRCC cells accumulate higher levels of acrolein than control cells (data not shown), providing an explanation for why polyamine overabundance results in decreased cell vitality. Therefore, further investigation of polyamine influences on renal cancer progression is highly warranted. Taken together, polyamine toxicity and PLP depletion, rather than mTORC1 suppression, account for ARG2’s growth suppressive effects in ccRCC.

While metabolic adaptations supporting tumor cell growth and division have been extensively studied (Palm and Thompson, 2017; Vander Heiden and DeBerardinis, 2017), global metabolic shifts contributing to tumor cell survival remain less clear (Kamphorst et al., 2013; Young et al., 2013). Our findings demonstrate how a single metabolic pathway significantly contributes to both processes: PLP conservation promotes biomass expansion, while suppression of polyamine synthesis enhances cell survival when tumor hypoxia increases polyamine uptake. A third essential process underlying disease progression is evasion of the immune system. Urea cycle loss naturally makes ccRCC cells arginine auxotrophs (Figure S2). As tumor infiltrating T cells clearly rely on arginine for their cytotoxic activity (Geiger et al., 2016), ccRCC cell dependency on circulating arginine suggests that this metabolic adaptation contributes to immunosuppression in the tumor microenvironment. Competition for microenvironmental arginine (and serine, (Ma et al., 2017)) provides an explanation for the success of immune checkpoint blockade in only a subset of ccRCC patients (Hammers, 2016; Joseph et al., 2017). If ccRCC tumor cells deprive infiltrating lymphocytes (and other immune cells) of critical exogenous metabolites due to urea cycle deficiency, overcoming the effects of PD-1/PD-L1 activity would fail to fully restore anti-tumor immunity. Our findings also raise the intriguing possibility that immunotherapy approaches could be improved by dietary means (i.e. arginine supplementation) for individuals afflicted with renal cancer.

The overall goal of our studies is to identify consistent (if not universal) metabolic differences between healthy kidney tissue and ccRCC, a genetically diverse cancer exhibiting significant intratumoral genomic heterogeneity. Cancer metabolism research has focused primarily on central carbon, amino acid, and lipid pathways, and ccRCC has been previously characterized as a metabolic disease with consistent changes in glycolysis, gluconeogenesis, and lipid storage. We show here that ammonia metabolism can also be of central importance to ccRCC disease progression.

STUDY LIMITATIONS

Our report combines metabolomic, genomic, and transcriptomic analyses of a large number of primary kidney cancer and healthy renal tissue samples to reveal a critical role for urea cycle enzymes in ccRCC growth suppression. However, additional studies will be needed to assess Arg2 deficiency in genetically engineered mouse models of kidney cancer, such as those recently reported (Gu et al., 2017; Nargund et al., 2017). The xenograft tumors described here were generated in immunocompromised recipients, precluding analysis of ccRCC ARG2 loss on the immune system. Future work will tackle this important question, along with comprehensive flux analyses of ARG2 expressing ccRCC cells, rather than assays of steady state metabolite levels.

STAR METHODS

Contact for Reagent and Resource Sharing

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, M. Celeste Simon (celeste2@pennmedicine.upenn.edu).

Experimental Model and Subject Details

Mice

Xenograft tumor experiments were approved by the Animal Care and Use Committee at the University of Pennsylvania. Female NIH-III nude mice (Charles River, 4–6 weeks) were injected subcutaneously into each flank with 5 million ccRCC cells in a 1:1 mixture of PBS and Matrigel (Corning 356234) with an injection volume of 200 μL. Once palpable tumors were established, tumor volume was monitored by caliper measurements. Upon completion of the experiment, the animals were sacrificed by CO₂ inhalation and xenograft tumors were dissected for downstream analyses.

Cell lines and Cell Culture Conditions

Authenticated (short tandem repeat profiling) human cell lines HK-2, 786-O, 769-P, A498, RCC4, RCC10, UOK101, and UMRC2 were obtained from the American Type Culture Collection. RCC10 was a kind gift from W. G. Kailin (Dana Farber Cancer Institute, Boston, MA). Cells were cultured for a maximum of 4 weeks before thawing fresh, early passage cells. All cells are routinely confirmed to be Mycoplasma negative (MycoAlert; tested every 3 months). ccRCC cells were cultured in DMEM containing 10% FBS and Pen/Strep while HK-2 proximal tubular epithelial cells were cultured in Keratinocyte-SFM medium supplemented with human recombinant epidermal growth factor and bovine pituitary extract. To mimic tumor microenvironmental O₂ levels, 786-O cells were grown under hypoxia. Hypoxic conditions (0.5% O₂) were achieved in a Ruskinn in vivo₂ 400 workstation, by supplementing ambient air with balanced N₂ and CO₂. HK-2 cells don’t survive 0.5%-1% O₂ beyond 48 hours; therefore, 1% FBS and 1 mM glucose culture conditions were employed to mimic microenvironmental stress typical of renal tumors. For metabolic labeling assays, cells were maintained in arginine-free DMEM supplemented with 10% dialyzed FBS and 398.1 M ¹⁵N₄-arginine (standard DMEM arginine concentration). For low-glucose conditions, cells were incubated in glucose-free DMEM supplemented with 1 mM glucose.

Method Details

Constructs and Viral Transduction

Lentivirus was produced by transfecting HEK-293T cells with the indicated expression plasmid, pRSV- Rev, pMDL, and pCMV-VSV-G plasmids (4^th generation lentiviral system) using Fugene6 transfection reagent (Promega). The virus was harvested 48 hours after transfection. For viral transduction, cells were incubated with medium containing virus 24 hours. Cells were allowed to recover for 24 hours before antibiotic selection, and surviving pools were utilized for downstream analyses (with the exception of clonally selected HK-2 cells, detailed below).

The lentiviral vector pLKO.1 Scramble (plasmid no. 17920) was obtained from Addgene. pLKO.1 lentiviral vectors expressing hairpins against ARG2 SH-1 (TRCN0000051018), ARG2 SH-2 (TRCN0000051020), ASS1 SH-3 (TRCN0000045554), and ASS1 SH-4 (TRCN0000045553) were obtained from The RNAi Consortium (TRC) at the Broad Institute and GE Dharmacon. For genetic knockout using CRISPR/Cas9, the lentiviral vector lentiCRISPR v2 was obtained from Addgene (plasmid no. 52961). Guide RNA’s (sgRNA’s) were cloned into this plasmid and single knockout clones were screened and selected following previously established methods (Sanjana et al., 2014). The sgRNA sequences that were positive for the desired knockout were as follows: ARG2 (caccgGAAGAAATCCGTCCACTCCG and the complementary strand aaacCGGAGTGGACGGATTTCTTC), ASS1 (caccgCAGCCACACGAGGATGCACG and the complementary strand aaacCGTGCATCCTCGTGTGGCTG) and ODC1 (caccgGAAGGGGCTTTACATGTGCG and complementary aaacCGCACATGTAAAGCCCCTTCC).

Sequence verified cDNA constructs were obtained from the Mammalian Gene Collection (GE Dharmacon) and were subcloned into the pCDH-CMV-MCS-EF1-Puromycin mammalian expression vector (System Biosciences CD510B-1). Stable cell line generation with multiple cDNA’s was achieved using either of two additional mammalian expression vectors with different selectable markers: pCDH-CMV-MCS-Hygromycin or pCDH-CMV-MCS-Neomycin (System Biosciences). ARG2 cDNA construct: MHS6278-202800846 (Accession BC029050). ASS1 cDNA construct: MHS1010-202694229 (Accession BC013224). OAZ1 cDNA construct: MHS6278-202858287 (Accession BC112133).

The ARG2 (H160F) mutant was generated using Stratagene’s QuikChange II mutagenesis kit (Agilent). pLJM1-Flag-RagB^GTP-Q99L mutant plasmid was obtained from Addgene (plasmid no. 19315) and subcloned into pCDH-CMV-MCS-Neomycin (System Biosciences). The cDNA for human excitatory amino acid transporter 1 (hEAAT1) was obtained from Addgene (plasmid no. 32813) and was subcloned into pCDH-CMV-MCS-Neomycin. Constitutive mTORC1 constructs were a kind gift from David Sabatini (Addgene plasmid no. 69015 and 69006). These constructs were subcloned into pCDH-CMV-MCS-Puromycin.

Western Blot Analysis

Cells were harvested in lysis buffer (40 mM HEPES pH=7.4, 2 mM EDTA, 10 mM pyrophosphate, 10 mM glycerophosphate, 1% Triton X-100) containing Roche complete ultra protease/phosphatase inhibitor (cat. 05892791001). For western blots of xenograft tissue, approximately 5-10 mg of tissue was suspended in 500 L lysis buffer and homogenized on ice using a Tissue-Tearor (Biospec, 985370). Samples were centrifuged at 20,000 g for 15 min at 4°C. Protein lysates were resolved b y Tris-Glycine SDS-PAGE and were transferred to nitrocellulose membranes (Biorad #162-0115, 0.45 μm pore size for all experiments, except when probing for 4E-BP1 when 0.2 m pore size was used). All membranes were incubated with the indicated primary antibodies overnight at 4°C a nd were diluted in TBST (20 mM Tris, 135 mM NaCl, and 0.02% Tween 20) supplemented with 5% bovine serum albumin. Primary antibodies were detected with horseradish peroxidase-conjugated secondary antibodies followed by exposure to ECL reagents, except ARG2 and ASS1, which were incubated with fluorescently conjugated secondary antibodies and detected on the Li-COR Odyssey CLx imaging system.

TCGA RNA-seq Analysis

Level 3 RNA-seq data for 480 ccRCC and 69 normal kidney samples were downloaded and analyzed from the TCGA on April 2, 2013 as previously described (Li et al., 2014). Briefly: differential gene expression analysis of tumor and normal samples was performed using DeSeq (Bioconductor Version 2.12). Normalized counts, P values, false discovery rate, adjusted P values (p-adj), and fold expression change for each gene were exported. To perform the gene set analysis, a comprehensive list of 2,752 genes encoding all known human metabolic enzymes and transporters were clustered into functional sets according to the metabolic pathways annotated by the Kyoto Encyclopaedia of Genes and Genomes (KEGG) (http://www.genome.jp/kegg/). The cut-off value of p-adj was set to 0.1 to exclude genes not consistently detected by RNA-seq. Survival analysis of the data was performed in GraphPad Prism and statistical effects were analyzed using the log-rank (Mantel-Cox) test.

Immunohistochemistry

Normal and ccRCC kidney tissue sections (from Cooperative Human Tissue Network) and tissue microarray slides (KD802, KD804, KD481, KD482, KD244, KD901, KD601, KD701 from US Biomax) were deparaffinized by baking slides at 50°C for 20 min. The slides were rehydrated in series of ethanol solutions and endogenous peroxidase activities were quenched by 1% H₂O₂ in distilled water for 30 min. After three washes in TT buffer (500 mM NaCl, 10 mM Trizma, and 0.05% Tween-20), antigen retrieval was performed by boiling slides for 20 min in a citrate-based antigen unmasking solution (Vector Labs, H3300). After cooling down to room temperature, slides were blocked in 2% normal goat serum and 4% BSA in Tris buffer with Tween 20 and then blocked with avidin and biotin solutions (Vector Labs, SP- 2001). Next, tissue slides were incubated with various primary antibodies at 4°C overnight. After thre e washes in TT buffer, biotinylated secondary antibody was added onto these slides for 1 hour. Sections were then processed using the Vectastain Elite ABC Kit (Vector Labs, PK-6100) and DAB peroxidase substrate kit (Vector Labs, SK-4100), dehydrated in a standard ethanol/xylenes series, and mounted in 75% v/v Permount (Fischer, SP15-500) in xylenes.

Immunofluorescence

Immunofluorescence of mTOR in 786-O ccRCC cells was conducted as previously described (Clippinger and Alwine, 2012). Briefly: cells were plated in 35 mm dishes in which a 12 mm coverslip was placed on the bottom. After 24 hours, the cells were washed three times with PBS and fixed for 20 min in 4% paraformaldehyde at room temperature. Cells were permeabilized in PBS containing 0.5% Triton X-100 and blocked in PBS containing 5% bovine serum albumin (blocking buffer). Primary antibodies against mTOR (Santa Cruz, sc1549) and LAMP (BD Biosciences, 555803) and secondary antibodies (Alexa Fluor 594 and Alexa Fluor 647, Invitrogen) were diluted in blocking buffer. Coverslips were washed three times in PBS, rinsed in H₂O, and mounted on slides using VectaShield mounting medium containing 4′6′- diamidino-2-phenylindole (DAPI) (Vector Laboratories). Slides were examined using a Nikon Eclipse E600 (40× objective) microscope, and pictures were taken using a Hamamatsu camera.

Quantitative RT-PCR

Total RNA was isolated using the RNA easy purification kit (Qiagen, 74104). cDNA was synthesized using a High Capacity RNA-to-cDNA kit (Applied Biosystems, 4368814). qRT-PCR was performed on a ViiA7 Real-Time PCR system from Applied Biosystems. Predesigned Taqman primers were obtained from Life Technologies for the following genes: ACTB (HS01060665_G1), 18S (HS03928985_G1), ASS1 (HS01597989_G1), and ARG2 (HS00982833_M1).

Anchorage-independent growth assay

ccRCC cells stably expressing ARG2, ASS1, PDXK, or vector control were plated at a density of 6,000 cells per well (using 6-well plates) in complete DMEM containing 0.3% agarose (low-melt 2-hydroxyethylagarose, Sigma Aldrich A4018), onto underlays composed of DMEM containing 0.6% agarose. Additional media was added to the cultures once per week, and after three weeks of growth the colonies were quantified.

Liquid-overlay spheroid growth assay

Three-dimensional spheroid cultures were generated using a liquid overlay technique as previously described (Khaitan et al., 2006). Briefly: Twenty-four–well plates were coated with 1% agarose in DMEM before plating 100,000 cells per well in DMEM plus 10% FBS. To promote spheroid formation, plates were swirled before incubation. Media were changed every 2 days, and spheroids were harvested and measured after 9 days. Measurement of spheroid volume was conducted using ImageJ software by measuring two orthogonal diameters and calculating the geometric mean radius, given by $r=\frac{\sqrt{d1d2}}{2}$, where r is the radius and d1 and d2 represent the measured diameters. Spheroid volume was calculated as $V=\frac{4}{3}\pi r^3$. To calculate the number of cells per spheroid, 12 spheroids were pooled and dissociatedwith accutase for 10 min at room temperature. After centrifugation, the cells were counted with a hemocytometer; the count was divided by 12 to get the average number of cells per spheroid and this represented one replicate value.

Matrigel-based spheroid growth assay

Matrigel- based 3D spheroids were generated using a technique previously described (Lewis et al., 2015; Vinci et al., 2012). Briefly: 3000 cells per well were plated in a 96-well “low adherence” plate along with DMEM plus 10% FBS and 2.5% matrigel. Plates were centrifuged at 1700 rpm to promote spheroid formation, and then imaged every two days for two weeks using the EVOS FL Auto Imaging System. Spheroid volume was calculated using a previously published ImageJ macro (Ivanov et al., 2014).

Cell Growth Assays

Cell growth assays were performed with either standard counting or using WST-1 reagent. For standard counting: multiple cultures of HK-2 or ccRCC cells (ccRCC) or 100,000 cells/plate (HK-2) in complete DMEM. For growth assays performed in low-glucose were plated in 6 cm dishes at 50,000 cells/plate and low-serum conditions, cells were maintained in glucose-free DMEM supplemented with 1% dialyzed FBS and 1 mM glucose. Each day, one set of culture plates was collected and counted. For WST-1 growth assays: HK-2 or ccRCC cells were plated in 96-well plates at 700 cells/well (ccRCC) or 1,500 cells/well (HK-2) and allowed to attach overnight (1 - 96-well plate for each day of the assay). The following day, the media was changed with 200 μL of complete DMEM supplemented with assorted small molecules (e.g. putrescine, arginine, ornithine, etc.), depending on the assay (see figures for exact concentrations used in each experiment). The cells were subjected to WST-1 (Sigma Aldrich, 11644807001) assay following manufacturers protocols; this was considered Day 0. This was repeated for each day of the assay and the data in each experiment were normalized to the starting cell number at Day 0 of the assay.

For polyamine toxicity studies, we estimated the concentration of putrescine per cell using our data (Figure 7C) and published values for cell doubling time and the per cell volume (National Cancer Institute Developmental Therapeutics Program, NCI-60 Cell line Data: https://dtp.cancer.gov/discovery_development/nci-60/cell_list.htm) and protein content (Dolfi et al., 2013). With these estimations, we determined that normoxic 786-O vector control cells contain approximately 3.3 mM putrescine per cell, and that ARG2 re-expression increased this by approximately 25% to 4.5 mM/cell. Hypoxia increases the uptake of exogenous putrescine, and accordingly we found that hypoxic 786-O vector control cells contain approximately 6.7 mM putrescine/cell, twice the amount under normoxia. ARG2 expression raised the levels even further to 8.2 mM/cell in this setting. It is important to note that these calculations are estimates based on literature values of cell volume and protein content per cell. Nonetheless, we feel it provides an adequate estimation of the putrescine content per cell, justifies treating cells with 1-10 mM putrescine, spermine, and spermidine.

Arginase Activity Assay

Arginase activity was determined using the Abnova Arginase Activity Assay Kit (KA1609) following manufacturers protocols. Briefly: 786-O ccRCC cells expressing empty vector, ARG2 wild type, or ARG2 H160F were lysed in lysis buffer (see western blot analysis section for recipe), which was used directly in the assay. A urea standard curve was made and the samples were mixed with assay buffer and incubated for 1 hour at room temperature, after which the absorbance was read at 430 nm to quantify the urea produced. Data were normalized to protein concentration in each sample.

Urea concentration was measured in 786-O xenograft tissues using the BioVision Urea Colorimetric Assay Kit K375-100 following manufacturers protocols.

Metabolomics Analysis

Mass spectrometry–based metabolomics analysis of primary ccRCC was performed with Metabolon, as previously described (Li et al., 2014). Additionally, unbiased metabolomics analysis of 786-O ccRCC xenograft tissue expressing either empty vector or ARG2 was performed with Metabolon, in the same manner as (Li et al., 2014).

Arginine, Citrulline, and Aspartate Measurement via LC/MS

786-O ccRCC cells expressing empty vector, ARG2, and/or ASS1 were cultured for 24 hours, washed,and quenched by the direct addition of 1 mL −80°C 4 :1 methanol:water (v/v) to the cell culture dish. Plates were placed at −80°C for 20 min, scraped, an d transferred into Eppendorf tubes. Samples were pulse sonicated on ice for 30 seconds at a rate of 1 pulse/sec prior to centrifugation at 16,000 × g at 4°C for 10 minutes. The supernatants were then transferred to clean glass tubes and evaporated to dryness under nitrogen. Dried residues were resuspended in 100 μL of mobile phase A (5 mM ammonium acetate in water) for LC/MS analysis, which was conducted as previously described (Aird et al., 2015).

¹⁵N-Arginine Metabolite Tracing

Prior to metabolic labeling assays, cells were maintained in SILAC Flex arginine-free DMEM (A24939-01) supplemented with 10% dialyzed FBS and 398.1μM arginine (standard DMEM concentration). This media also lacks glucose, phenol red, glutamine, and lysine, which were added back to standard DMEM concentrations (25 mM, 15.0 mg/L, 4.0 mM, and 0.798 mM, respectively). For labeling studies, HK-2 cells were plated in 10 cm dishes in low glucose (1 mM) and low serum (1%) conditions (SG) and incubated for 48 hours prior to changing the media to SILAC FlexSG media supplemented with 398.1 μM ¹⁵N₄-arginine, which was then incubated for 24 hours. 786-O cells were plated in 10 cm dishes in complete SILAC Flex DMEM and placed under 0.5% O₂ hypoxic conditions for 48 hours prior to changing the media to SILAC Flex DMEM supplemented with 398.1μM ¹⁵N₄-Arginine, which was then incubated for an additional 24hours in hypoxia.

The cells were extracted and analyzed as previously described (Nissim et al., 1999; Nissim et al., 2011). Briefly, cells were washed twice with ice-cold PBS and then scraped in 4% perchloric acid (PCA) on ice, which was then freeze-thawed for three cycles. Cell extracts were collected and neutralized using 5 M KOH. The neutralized PCA extract was used for measurement of ¹⁵N enrichment in arginine, urea, ornithine and citrulline. The production of ¹⁵N-labeled urea (M+2, containing two ¹⁵N), or proline from [U-¹⁵N] arginine was measured using GC-MS, after separation from neutralized PCA extract with an AG-50 column and TBDMS derivatization, as in (Nissim et al., 1999; Nissim et al., 2011). Measurement of ¹⁵N- labeled arginine, citrulline, putrescine and ornithine was performed using LC/MS (Agilent LC 1260 Infinity and Agilent 6410 Triple-Quad MS). Samples prepared from neutralized PCA extracts and derivatized with ethylchloroformate. For measurement of ¹⁵N enrichment we used the MRM, ion-pair links 275-70 and 279-70 for arginine; 276-70 and 278-70 for citrulline; 233-187 and 235-189 for putrescine; and 305-70and 307-70 for ornithine.

Measurement of pyridoxal, pyridoxal phosphate, pyridoxine and pyridoxamine levels was done using LC/MS, using Poroshell 120EC-C18 column (Agilent). The MRM used were: 248-150 for pyridoxalphosphate; 168-150 for pyridoxal; 170-152 for pyridoxine; and 169-152 for pyridoxamine. All preparations and measurements were performed in the dark due to light sensitivity.

Cell Cycle Analysis

HK-2 cells were plated in 10 cm plates until ~60% confluency. To block cells in G1 phase, cells were treated with 20 μM lovastatin (Abcam) for 24 hours and harvested for analysis. To block cells in early S phase, a double thymidine block was performed by adding 200 μl of 100 mM thymidine (Sigma-Aldrich) to the media for overnight treatment. Cells were released from thymidine block by aspirating media, washing the cells and changing media intomedia containing 25 μM deoxycytidine (Sigma-Aldrich) for 8 hours. Media was changed back to thymidine containing media and incubated overnight, before cells were harvested for analysis. To synchronize cells in G2 phase and mitosis, following a double thymidine block, cells were first released as before for 2 hours. For G2-block, RO3306 (Sigma-Aldrich) was added to the cells (10 μM) for 10 hours. For mitosis block, nocodazole (0.1 μg/ml) (Sigma-Aldrich) was added for 10 hours. Mitotic cells were then collected by mechanical shake off.

The position of synchronized cells was confirmed by propidium iodide (PI) staining. Cells were collected in 15 ml conical tubes and centrifuged at 400 xg for 5 min. Supernatant was discarded, washed with 5 ml PBS, and centrifuged at 400 xg for 5 min. Supernatant was discarded and cells were re- suspended in 500 μl cold PBS. 100% ethanol (5 ml) was added drop wise while gently vortexing to fix the cells and kept overnight at −20 °C. Cells were spun down at 400 xg for 5 min and supernatant was discarded. Cells were washed with cold PBS and spun down at 400 xg for 5 min. Cells were re- suspended in 500 μl PI staining solution (PBS + 50 μg/ml PI (Sigma-Aldrich) + 100 μg/ml RNAse A (QIAGEN)), filtered using a 40 μM filter (BD Biosciences), and incubated on ice in the dark for 20 min. Samples were run on a BD FACSCalibur (BD Biosciences) and analyzed using FlowJo.

QUANTIFICATION AND STATISTICAL ANALYSIS

All statistical analyses were conducted using GraphPad Prism 7.0. All error bars represent mean ± SEM. One-way ANOVA with Tukey’s multiple comparisons test was used unless otherwise specified. Significance was defined as *p < 0.05, **p < 0.01, ***p < 0.001.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
ARG2	Santa Cruz	sc20151
β-actin	Santa Cruz	sc47778
mTOR	Santa Cruz	sc1549
ASS1	Abcam	ab124465
ODC1	Abcam	ab126590
OAZ1	Abcam	ab85221
Ki67	BD Biosciences	#550609
LAMP	BD Biosciences	#555803
phospho-p70 S6 Kinase (Thr389)	Cell Signaling	#9234
p70 S6 Kinase	Cell Signaling	#2708
β-Tubulin	Cell Signaling	#2146
phospho-4E-BP1 (Ser65)	Cell Signaling	#9451
4E-BP1	Cell Signaling	#9452
phospho-Akt (Thr308)	Cell Signaling	#13038
Akt	Cell Signaling	#9272
phospho-Tuberin/TSC2 (Ser939)	Cell Signaling	#3615
tuberin/TSC2	Cell Signaling	#3612
EAAT1	Cell Signaling	#5684
RagB	Cell Signaling	#8150
HSP90	Cell Signaling	#4874S
anti-rabbit IgG HRP-linked	Cell Signaling	#7074
anti-mouse IgG HRP-linked	Cell Signaling	#7076
IRDye 800CW conjugated anti-rabbit	Li-COR Biosciences	#926-32211
IRDye 800CW conjugated anti-mouse	Li-COR Biosciences	#926-32210
AlexaFluor 680 conjugated anti-rabbit	Life Technologies	#A21109
AlexaFluor 680 conjugated anti-mouse	Life Technologies	#A21058
Chemicals, Peptides, and Recombinant Proteins
DMEM	Life Technologies	11965-084
Pen/Strep	Life Technologies	15140-122
HK-2 Keratinocyte SFM Media	Life Technologies	10724-011
SILAC Flex DMEM	Life Technologies	A24939-01
DMEM for Silac	Life Technologies	88364
Glucose-free DMEM	Life Technologies	11966-025
Glutamine	Life Technologies	25030-081
Accutase	Life Technologies	A1110501
Standard FBS	Gemini	900-108
Dialyzed FBS	Gemini	100-108
15N4-Arginine	Cambridge Isotopes	NLM-396-PK
Glucose	Sigma Aldrich	G8270
Putrescine	Sigma Aldrich	P5780
Spermidine	Sigma Aldrich	S0266
Spermine	Sigma Aldrich	S4264
Ornithine	Sigma Aldrich	O6503
Aspartate	Sigma Aldrich	A7219
Lysine	Sigma Aldrich	L5501
Arginine	Sigma Aldrich	A8094
Phenol Red	Sigma Aldrich	P0290
WST-1	Sigma Aldrich	5015944001
Adenosine	Sigma Aldrich	A4036
Guanosine	Sigma Aldrich	G6264
Thymidine	Sigma Aldrich	T1895
Cytidine	Sigma Aldrich	C4654
Enhanced chemiluminescent substrate	Perkin Elmer	NEL105001EA
Matrigel matrix	Corning	354234
Critical Commercial Assays
RNeasy Mini kit	Qiagen	#74104
High Capacity RNA-to-cDNA kit	Applied Biosystems	#4368814
Arginase Activity Assay kit	Abnova	KA1609
Urea Colorimetric Assay kit	Biovision	K375-100
QuikChange II mutagenesis kit	Agilent	#200521
Experimental Models: Cell Lines
HK-2	ATCC	CRL-2190
786-0	ATCC	CRL-1932
A498	ATCC	HTB-44
Experimental Models: Organisms/Strains
Mouse : NIH III nude, female homozygous	Charles River	#201
Oligonucleotides
ARG2 sgRNA Forward: caccgGAAGAAATCCGTCCACTCCG Reverse: aaacCGGAGTGGACGGATTTCTTC	This paper	N/A
ASS1 sgRNA9br/)Forward: caccgCAGCCACACGAGGATGCACG Reverse: aaacCGTGCATCCTCGTGTGGCTG	This paper	N/A
ODC1 sgRNA Forward: caccgGAAGGGGCTTTACATGTGCG Reverse: aaacCGCACATGTAAAGCCCCTTCC	This paper	N/A
ACTB	Life Technologies	HS01060665_G1
18S	Life Technologies	HS03928985_G1
ASS1	Life Technologies	HS01597989_G1
ARG2	Life Technologies	HS00982833_M1
Recombinant DNA
pLKO.1 Scramble	Addgene	17920
Human excitatory amino acid transporter 1 cDNA	Addgene	32813
pLJM1-Flag-RagBGTP-Q99L	Addgene	19315
pCDNA3-FLAG-MTOR-R2505P	Addgene	69015
pCDNA3-FLAG-MTOR-L1460P	Addgene	69006
LentiCRISPR v2	Addgene	52961
ARG2 SH-1	GE Dharmacon	TRCN0000051018
ARG2 SH-2	GE Dharmacon	TRCN0000051020
ASS1 SH-3	GE Dharmacon	TRCN0000045554
ASS1 SH-4	GE Dharmacon	TRCN0000045553
pCDH-CMV-MCS-EF1-Puromycin	System Biosciences	CD510B-1
ARG2 cDNA	System Biosciences	MHS6278-202800846
ASS1 cDNA	System Biosciences	MHS1010- 202694229
OAZ1 cDNA	System Biosciences	MHS6278- 202858287
Software and Algorithms
GraphPad Prism 7.0	GraphPad Software	https://www.graphpad.com/scientific-software/prism/
Spheroid macro	ImageJ	Ivanov et al., 2014

Supplementary Material

2

Table S1. Significantly Altered Metabolites in ccRCC Tumors vs. Normal Tissue, related to Figure 1

3

Table S2. Significantly Altered Metabolites in ARG2 re-expressing xenograft tumors vs control, related to Figure 4

HIGHLIGHTS.

• Multiple urea cycle enzymes are significantly underexpressed in ccRCC

• ccRCC progression is dependent on alterations in ammonia metabolism

• Reduced urea cycle activity conserves the biosynthetic cofactor pyridoxal phosphate

• Polyamine accumulation represents a metabolic vulnerability for ccRCC