Arginine starvation kills tumor cells through aspartate exhaustion and mitochondrial dysfunction

Abstract

Defective arginine synthesis, due to the silencing of argininosuccinate synthase 1 (ASS1), is a common metabolic vulnerability in cancer, known as arginine auxotrophy. Understanding how arginine depletion kills arginine-auxotrophic cancer cells will facilitate the development of anti-cancer therapeutic strategies. Here we show that depletion of extracellular arginine in arginine-auxotrophic cancer cells causes mitochondrial distress and transcriptional reprogramming. Mechanistically, arginine starvation induces asparagine synthetase (ASNS), depleting these cancer cells of aspartate, and disrupting their malate-aspartate shuttle. Supplementation of aspartate, depletion of mitochondria, and knockdown of ASNS all protect the arginine-starved cells, establishing the causal effects of aspartate depletion and mitochondrial dysfunction on the arginine starvation-induced cell death. Furthermore, dietary arginine restriction reduced tumor growth in a xenograft model of ASS1-deficient breast cancer. Our data challenge the view that ASNS promotes homeostasis, arguing instead that ASNS-induced aspartate depletion promotes cytotoxicity, which can be exploited for anti-cancer therapies.

Fig. 1

Metabolomics analysis of MDA-MB-231 cells upon arginine starvation. a Selected glycolysis (blue text) and TCA cycle (green text) intermediates, as well as amino acids, were quantified using mass spectrometry in MDA-MB-231 cells. Equal numbers of cells (about 70% confluency) were harvested after culture in arginine-replenished (Ctrl) or arginine-starved (-Arg) medium for 24 and 48 h. Data are shown as mean ± S.D.; n = 3; *p < 0.05; **p < 0.01; ***p < 0.001. b, c Arginine starvation (48 h) suppresses flux from [U-¹³C]glucose to glycolysis (b) and TCA cycle (c) in MDA-MB-231. Equal numbers of cells (about 70% confluency) were harvested after culture in arginine-replenished (Ctrl) or arginine-starved (-Arg) medium for 24 and 48 h. Data are shown as mean ± S.D.; n ≥ 3; *p < 0.05; **p < 0.01; ***p < 0.001. Experiments were repeated at least three times. d Overview of pathway analysis. The scatterplot represents the pathway impact value and p-value from pathway topology analysis of the differentially expressed metabolites from MDA-MB-231 cells cultured in full and arginine-starved (48 h) medium. The size and color of each node is based on its pathway impact value and p-value, respectively. Pathways with statistical significance (p < 0.05) are shown in red.

Fig. 2

Mitochondria are important targets of arginine starvation. a Heatmap of altered expression of mitochondrial complex and glycolysis genes. Gene expression was assessed using RNA-seq in MDA-MB-231 cells maintained in full (Ctrl) and arginine-starved (-R, 48 h) media (n = 2 per group). Data are shown in relative reads for each gene. b A representative Western blot shows mitochondrial complex protein levels in mitochondria-depleted ρ⁰ and TFAM-KO MDA-MB-231 cells; n = 3. The uncropped blot can be found in Supplementary Fig. 10A. c, d Acid phosphatase (ACP) assays of viability of mitochondria-depleted ρ⁰ (c) and TFAM-KO (d) cells incubated with decreasing concentrations of arginine for 72 h; n = 3. e Oxidative stress in mitochondria-depleted ρ⁰ cells after arginine starvation (-Arg). Relative oxidized DCF levels were calculated by designating the value in MDA-MB-231 cells cultured in full medium as 1; n = 3. f, g DNA damage in mitochondria-depleted ρ⁰ cells after arginine starvation. The alkaline comet assay was used for measuring DNA fragmentation. f Representative images. Scale bar: 100 µm. g The comet tail moment is shown as mean ± S.D.; n ≥ 20. For bar graphs, data are shown as mean ± S.D.; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001

Fig. 3

Arginine starvation epigenetically impairs mitochondrial bioenergetics. a The relative acetyl-CoA level was assessed in MDA-MB-231 cells after arginine starvation (-Arg) for 48 h by liquid chromatography-mass spectrometry. The relative level was calculated by designating control (Ctrl) cell extracts as 1; n = 3. b Overall histone H3K9 acetylation (H3K9Ac) after arginine starvation. The cells were cultured in control or arginine-starved medium, as indicated, for 24 h. Acetate (10 mM) or arginine (Arg rescue, 84 mg/L) was added to arginine-starved medium for an additional 24 h. Relative change (over with arginine control) of each sample is calculated by designating the densitometry tracing value in the control as 1 after normalization with actin. The original blot of this panel is included in Supplementary Fig. 10B. c qRT-PCR analyses of mitochondrial complex NDUFB6, NDUFB7, NDUFB10, and NDUFV1 genes upon arginine starvation; n = 3. d ChIP assay for H3K9Ac occupancy in NDUFB6, NDUFB7, and NDUFB10 promoter regions after arginine starvation; n = 3. e qRT-PCR analyses of mitochondrial Complex I NDUFB7 and NDUFB10 expression after arginine starvation and supplementation of arginine (84 mg/L) or acetate (10 mM); n = 3. f The oxygen consumption rate (OCR) was measured in MDA-MB-231 cells maintained in control or arginine-starved medium in the presence or absence of dimethyl-ketoglutarate (DKG; 10 mM) or succinate (10 mM) using a Seahorse bioanalyzer; n = 5. Relative OCR was calculated by designating the basal OCR of cells in full medium as 100%. O: oligomycin; F: FCCP; R: rotenone. g Effect of arginine starvation and supplementation with acetate (10 mM) on basal respiration, maximal respiration, and ATP production from Supplementary Fig. 3D; n = 5. For bar graphs, data are shown as mean ± S.D.; *p < 0.05; **p < 0.01; ***p < 0.001

Fig. 4

ATF4-dependent ASNS expression links arginine starvation and cell death. a Heatmap of the UPR pathway. Gene expression was assessed using RNA-seq in control MDA-MB-231 (231) and mitochondria-depleted MDA-MB-231 ρ⁰ cells treated with full medium (Ctrl) or arginine starvation (-Arg, 48 h). The gene list was established using the UPR RT² profiler PCR array (Qiagen). Data are shown in relative reads for each gene. b Representative Western blots of p-eIF2α, ATF4 and ASNS abundance in parental and ρ⁰ cells with or without arginine (-Arg). Treatment with the ROS scavenger, N-Acetyl-cysteine (NAC), dampens phosphorylation of eIF2α. MDA-MB-231 cells under arginine starvation were treated with the indicated concentrations of NAC and subjected to Western blot analyses for p-eIF2α signal. A representative blot is shown here, n ≥ 3. The original blots of this panel can be found in Supplementary Fig. 11. c Representative Western blots of p-p70S6K, p-eIF2α, ATF4 and ASNS after arginine starvation in both MDA-MB-231 and MCF7 cells; n = 3. The unprocessed images are included in Supplementary Fig. 12. d Representative Western blots of TSC1, TSC2, ATF3, ATF4, or GCN2 knockdown on ATF4, ASNS, and p-eIF2α induction after arginine starvation (24 h); n = 3. The uncropped blots of the panel are included in Supplementary Fig. 13. e, f MDA-MB-231 cell viability in response to arginine starvation with or without ATF3/4-knockdown (e; n = 4) and ASNS-knockdown (f; n = 3). g MDA-MB-231 cell viability under treatment of ASNase, with or without ASNS knockdown. For the bar graphs, data are shown as mean ± S.D.; n = 3; *p < 0.05; **p < 0.01; ***p < 0.001

Fig. 5

Aspartate rescues the viability of arginine-starved cells. a Cell viability of MDA-MB-231 cells after arginine starvation with or without supplementation of asparagine, aspartate glutamine or glycine (10 mM each) for 72 h; n = 3. b qRT-PCR analyses of ASNS, ATF4 and the mRNA abundance of the key components of the malate-aspartate shuttle after arginine starvation (-R; 24 h) in siCtrl- and siATF4-cells; n = 3. c Effect of ASNS-knockdown on basal respiration, maximal respiration, and ATP production measured in Supplementary Fig. 5C; n = 9. d OCR was measured in SLC1A3- or SLC25A10-knockdown cells; n = 4. O: oligomycin; F: FCCP; R: rotenone. e Effect of arginine starvation and replenishment with aspartate (10 mM) on the basal respiration, maximal respiration, and ATP production measured in Supplementary Fig. 5D; n = 5. f MDA-MB-231 cells were cultured with [U-¹³C]aspartate for 6 h after 48 h of incubation in arginine-depleted (-Arg) or full (Ctrl) media. The relative aspartate-derived m + 4 fractions of intracellular asparagine (Asn), aspartate (Asp), fumarate (Fum), malate (Mal), and citrate (Cit) were measured with gas chromatography mass spectrometry. The relative m + 4 isotopologue was calculated by designating the respective mean value in MDA-MB-231 ctrl cells as 1; n = 3. g NAD⁺/NADH ratio and NADH abundance in arginine-starved MDA-MB-231 cells with or without aspartate (10 mM) supplementation; n = 3. h qRT-PCR analyses of NDUFB6, NDUFB7, and NDUFB10 expression in arginine-starved shCtrl and shASNS cells (48 h). The values are normalized to the 18 S rRNA levels and the mean expression level in the control cells; n = 3. For bar graphs, data are shown as mean ± S.D.; n.s., not significant; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001

Fig. 6

Arginine starvation reduces tumor size in a xenograft model. a, b The effect of an arginine-free diet, compared to the matched control diet (Ctrl), on orthotopically xenografted luciferase-tagged breast cancer BT-549 cells was measured by tumor volume (a) and bioluminescence imaging (b). c Effect of an arginine free-diet on mouse body weight. d, e The weights of tumors dissected from BT549 cell xenografted mice (d) and MDA-MB-231 cell xenografted mice (e) fed with either control or arginine-free diet (Mice failed to develop tumor after arginine starvation was assigned value “0” to for analysis; n = 6). f–i Representative histological analysis (hematoxylin and eosin) of tumors harvested from mice maintained on a control diet (f, g) and arginine-free diet (h, i) at day 39 post-tumor cell orthotopic implantation. Arrowheads indicate mitotic cells. Expanded view of the mitotic cells (black arrowheads) and multi-polar anaphase cell (red arrowhead) are shown; scale bar: 50 µm (f, h); scale bar: 500 µm (g, i). j Mitotic cells were quantified from five high-power fields (HPF) from each group. Data are shown as mean ± S.D.; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001