Asparagine bioavailability governs metastasis in a model of breast cancer

Abstract

Using a functional model of breast cancer heterogeneity, we previously showed that clonal sub-populations proficient at generating circulating tumour cells were not all equally capable of forming metastases at secondary sites. A combination of differential expression and focused in vitro and in vivo RNA interference screens revealed candidate drivers of metastasis that discriminated metastatic clones. Among these, asparagine synthetase expression in a patient's primary tumour was most strongly correlated with later metastatic relapse. Here we show that asparagine bioavailability strongly influences metastatic potential. Limiting asparagine by knockdown of asparagine synthetase, treatment with l-asparaginase, or dietary asparagine restriction reduces metastasis without affecting growth of the primary tumour, whereas increased dietary asparagine or enforced asparagine synthetase expression promotes metastatic progression. Altering asparagine availability in vitro strongly influences invasive potential, which is correlated with an effect on proteins that promote the epithelial-to-mesenchymal transition. This provides at least one potential mechanism for how the bioavailability of a single amino acid could regulate metastatic progression.

Extended Data Figure 1. Analysis of ASNS expression levels in patient data

a, Expression level of genes identified as overexpressed in 4T1-T compared with 4T1-E in the primary tumours of patients with different disease subtypes (edges of the box are the 25th and 75th percentiles and error bars extend to the values q3 + w(q3 − q1) and q1 − w(q3 − q1), in which w is 1.5 and q1 and q3 are the 25th and 75th percentiles, which is also true for b, d, and e, ANOVA P < 0.0001). b, Expression level of the same genes in disease-free survivors and patients with relapse to the lung (rank-sum P < 0.01). c, For each gene that was identified in the screen, a prognostic value was calculated using three different datasets. One consisted of gene expression measurements in three patient-matched basal tumour and metastasis pairs (patients A1, A7, and A11). Here genes were classified as correlated with progression if expression was higher in each of the metastases and negatively correlated if expression was higher in each of the primaries. The other two datasets consisted of primary tumour gene-expression profiles with matched outcomes. For the UNC254 patient dataset, the site of relapse was not available and genes were deemed positively correlated with progression if they had significant (Cox P < 0.05) relapse-free survival hazard ratios greater than 1, and negatively correlated if these ratios were significant (Cox P < 0.05) and less than 1. As the UNC855 dataset also had site of relapse information, here both relapse-free and lung relapse-free survival (RFS and LRFS) hazard ratios were used to classify genes as positively or negatively correlated with progression based on the same criteria that were used for the UNC254 data. d, Expression level of ASNS in the primary tumours of patients with different disease subtypes (ANOVA P < 0.0001). e, Expression level of ASNS in the primary tumours of patients with non-specific relapse and relapse to the lymph node, bone, brain, liver, or lung compared with expression levels in patients without relapse to each corresponding site (rank-sum P < 0.005). f, Analysis of ASNS in three additional sets from patients with breast cancer (MDACC, METRABIC, and TCGA). Shown are survival plots and relevant statistics (Cox P < 0.01). g, Analysis of ASNS in the TCGA Pan-Cancer expression data. Shown are survival plots and relevant statistics for the ten non-breast solid tumours represented in the dataset (Cox P < 0.05 for colon, squamous head and neck, renal clear cell, and endometrial cancers). h, Analysis of ASNS across all tumours represented in the TCGA Pan-Cancer dataset (Cox P = 1.5 × 10−12).

Extended Data Figure 2. Primary validation of Asns as a driver of invasion and metastasis

a, Representative images of the lungs of mice that were intravenously injected with Asns-silenced or -expressing 4T1-T cells as described in Fig. 2a. b, Quantification of Matrigel invasion capacity for Asns-silenced and -expressing 4T1-T cells (n = 3 replicates per cell line). c, Quantification of mCherry-positive 4T1-T cells after roughly 50% of cells were infected with mCherry-expressing constructs harbouring shRNAs targeting Renilla luciferase and Asns. Cells were grown during the 24-h period that the Matrigel invasion assay described in Fig. 2b was being performed (n = 3 replicates per cell line). d, Violet cell-labelling intensity of Asns-silenced and -expressing 4T1-T cells, relative to the initial population. Cells were grown during the 24-h period that the Matrigel invasion assay described in Fig. 2b was being performed (n = 3 replicates per cell line). e, Free amino-acid quantification by HPLC for each amino acid in Asns-expressing and -silenced cells. Shown are the log-fold changes for each amino acid (n = 3 replicates per cell line). f, Quantification of mCherry-positive 4T1-T cells after roughly 50% of cells were infected with mCherry-expressing constructs harbouring shRNAs targeting Renilla luciferase and Asns. After infection, cells were grown in medium supplemented with l-asparagine or d-asparagine and mCherry percentages were measured at 48 and 96 h (n = 3 replicates per cell line). g, Quantification of Matrigel invasion for Asns-silenced and -expressing cells when assayed in medium supplemented with and without l-asparagine (n = 3 invasion chambers per cell line).

Extended Data Figure 3. Secondary validation of Asns as a driver of invasion and metastasis

a, Volume measurements of tumours resulting from orthotopic injection of Asns-silenced and -expressing parental 4T1 cells (n = 10 mice per cell line, edges of the box are the 25th and 75th percentiles and error bars extend to the values q3 + w(q3 − q1) and q1 − w(q3 − q1), in which w is 1.5 and q1 and q3 are the 25th and 75th percentiles, which is also the case for b–g). b, Quantification of lung metastases corresponding to the tumours described in a (rank-sum P < 0.002). c, Volume measurements of tumours resulting from orthotopic injection of parental 4T1 cells with basal (Empty) or enforced expression of Asns (n = 10 mice per cell line). d, Quantification of lung metastases corresponding to the tumours described in c (rank-sum P < 5.0 × 10−5). e, Average diameters of the metastases of each mouse described in d (rank-sum P < 0.001). f, Volume measurements for tumours resulting from orthotopic injection of MDA-MB-231 cells with basal (Empty) or enforced expression of ASNS (n = 10 mice per cell line). g, Quantification of lung metastases corresponding to the tumours described in f (rank-sum P < 0.005). h, Quantification of Matrigel invasion for the MDA-MB-231- derived cell lines described in f (n = 3 invasion chambers per cell line). i, Representative images of the collection wells for the invasion assays described in h. See Source Data.

Extended Data Figure 4. Primary validation that extracellular asparagine availability affects invasion and metastasis

a, HPLC quantification of cellular free amino-acid percentages for parental 4T1 cells when the medium is supplemented with each of the NEAAs lacking in the DMEM culture medium (n = 3 replicates per cell line). b, Quantification of MDA-MB-231 Matrigel invasion rates under the same conditions as described in Fig. 3a (n = 5 invasion chambers per condition, rank-sum P < 0.001). c, HPLC quantification of cellular free aminoacid percentages for MDA-MB-231 cells when cultured in the medium conditions described in a (n = 3 replicates per cell line). d, Violet celllabelling intensity of parental 4T1 cells when grown in asparagine-lacking or -supplemented medium for the same period that the Matrigel invasion assay described in Fig. 3a was being performed (n = 3 replicates per cell line). e, Violet cell-labelling intensity of MDA-MB-231 cells when grown in asparagine-lacking or -supplemented medium for the same period that the Matrigel invasion assay described in b was being performed (n = 3 replicates per cell line).

Extended Data Figure 5. Secondary validation that extracellular asparagine availability affects invasion and metastasis

a, Tumour volumes resulting from the orthotopic injection of parental 4T1 cells. Half of the mice received l-asparaginase and the other half received an equivalent volume of PBS at the same injection rate (n = 10 mice per condition, edges of the box are the 25th and 75th percentiles and error bars extend to the values q3 + w(q3 − q1) and q1 − w(q3 − q1), in which w is 1.5 and q1 and q3 are the 25th and 75th percentiles, which is also the case for b–g). b, Quantification of lung metastases detected in the animals described in a (rank-sum P < 0.001). c, Representative H&E-stained lung sections as described in b. d, Quantification of the lung metastases described in Fig. 3b, in which Asns-silenced and -expressing 4T1-T cells were injected into mice. Half of the mice received l-asparaginase and the other half received an equivalent volume of PBS at the same injection rate (n = 10 mice per condition, rank-sum P < 0.0005 for l-asparaginase versus control for each line and for Asns-silenced versus -unsilenced cells in each drug condition). e, Tumour volumes corresponding to the lung metastases described in d (rank-sum P < 0.005 for Asns-silenced versus -expressing cells in l-asparaginase-treated mice). f, Lung metastases resulting from the orthotopic injection of ASNS-silenced and -expressing MDA-MB-231 cells and subsequent treatment of the injected animals with l-asparaginase or PBS (n = 10 mice per cell line, rank-sum P < 0.05 for ASNS-silenced versus -expressing cells in both conditions and for silenced cells in treated versus untreated mice). g, Tumour volumes corresponding to the mice described in f (rank-sum P < 0.05 for Asns-silenced versus -expressing cells under both treatments and for PBS versus l-asparaginase-treated animals for each cell line). See Source Data.

Extended Data Figure 6. Tertiary validation that extracellular asparagine availability affects invasion and metastasis

a, Asparagine content in the serum free amino-acid pool, for mice fed 0%, 0.6%, or 4% asparagine diets (n = 5 mice per diet, edges of the box are the 25th and 75th percentiles and error bars extend to the values q3 + w(q3 − q1) and q1 − w(q3 − q1), in which w is 1.5 and q1 and q3 are the 25th and 75th percentiles, which is also the case for b, d, e and h, rank-sum P < 0.05 between each diet). b, Volumes of orthotropic tumours corresponding to the lung metastases described in Fig. 3c, in which Asns-silenced and -expressing 4T1-T cells were orthotopic injected into mice fed with 0%, 0.6%, and 4% asparagine diets (n = 10 mice per condition). c, Representative images of the lung metastases described for Fig. 3c, which also correspond to the mice described in b. d, Volumes of tumours resulting from the orthotopic injection of parental 4T1 cells into mice fed with 0%, 0.6%, or 4% asparagine diets (n = 10 mice per diet). e, Quantification of metastases in the lungs of the animals described in d (rank-sum P < 0.05 for mice receiving 0% versus 0.6% and 0% versus 4% diets). f, Representative images of H&E-stained sections of the lungs described in e. g, Relative expression of Asns in the mammary gland, serum, and lungs of mice treated with l-asparaginase or PBS, as measured by qPCR with two primer pairs P1 and P2 (n = 3 per condition). h, Transcripts per million (TPM) expression measurements for ASNS in human breast, lung, and whole-blood samples (n > 114 for each tissue, rank-sum P < 2.8 × 10−37 for blood versus breast and blood versus lung). See Source Data.

Extended Data Figure 7. Primary validation that asparagine availability regulates EMT

a, Protein-level changes between Asnssilenced and -expressing cells when genes are stratified by transcriptionlevel changes (top and bottom 10% of genes based on log-fold change in Asns-silenced cells, gene-up and -down, respectively) and asparagine content (top and bottom 10% of genes based on asparagine content, Asp-high and -low, respectively), edges of the box are the 25th and 75th percentiles and error bars extend to the values q3 + w(q3 − q1) and q1 − w(q3 − q1), in which w is 1.5 and q1 and q3 are the 25th and 75th percentiles, which is also the case for d and e, rank-sum P < 5.0 × 10−24 for both individual variables, and rank-sum P < 0.005 for interacting variables). b, Amino acid enrichment analysis of downregulated genes (bottom 25% based on log-fold change) on the basis of RNA and protein levels in Asns-expressing versus -silenced 4T1-T cells. Negative correlations indicate the amino acid is depleted in the downregulated genes, whereas positive correlations indicate the amino acid is enriched. For protein minus RNA level expression changes, amino acids with positive correlations are enriched in proteins in which depletion levels exceed what is predicted by corresponding RNA changes. Negative correlations indicate the amino acid is enriched in proteins in which depletion levels are less than what is predicted by corresponding RNA changes (rank-sum P < 1.0 × 10−5 for asparagine in protein and protein– RNA). c, Amino-acid enrichment in mouse and human EMT-up proteins (rank-sum P < 0.01 for both human and mouse). d, Position 15 asparagine codon enrichment in ribosome protected fragments from PC-3 cells grown with and without l-asparaginase, when all genes or only EMT-up genes are analysed (outliers were not plotted to improve interpretability, which is also the case for e, rank-sum P < 0.05 for EMT-up versus all genes in both untreated and l-asparaginase-treated cells). e, Increase in asparagine codon representation in ribosome protected fragments, when PC-3 cells are grown in l-asparaginase (relative to without), and all genes or EMT-up genes are analysed (rank-sum P < 0.05). See Source Data.

Extended Data Figure 8. Conservation of asparagine enrichment in EMT promoting proteins

Asparagine enrichment analysis of EMTpromoting protein orthologues in the 126 species listed in the Orthologous MAtrix database that harbour at least 10 orthologues (sign-rank P < 1.0 × 10−13 for all species and rank-sum P < 9.0 × 10−9 for mammals versus other species).

Extended Data Figure 9. Secondary validation that asparagine availability regulates EMT

a, Transcription-level changes in EMT-up and -down genes that occur in response to Asns silencing in 4T1-T cells (n = 2 replicates per condition, edges of the box are the 25th and 75th percentiles and error bars extend to the values q3 + w(q3 − q1) and q1 − w(q3 − q1), in which w is 1.5 and q1 and q3 are the 25th and 75th percentiles, which is also the case for b–e, h and i, rank-sum P < 0.001 for EMT-up genes, DESeq false discovery rate < 0.05 for Twist1 and Cdh1). b, Transcription-level changes in EMT-up and -down genes that occur in response to the medium of Asns-silenced 4T1-T cells being supplemented with l-asparagine (n = 2 replicates per condition, rank-sum P < 0.005 for EMT-up genes). c, Gene expression changes in EMT-up and -down genes that result from ATF4 knockout in near haploid KBM-7 chronic myelogenous leukaemia (HAP-1 cells, rank-sum P < 0.05 for EMT-up genes). d, Gene expression changes in EMT-up and -down genes, which result in the liver cells of homozygous ATF4-deleted mice when treated with l-asparaginase (rank-sum P < 0.05 for EMT-down genes in wild-type mice treated with l-asparaginase (WT+ L-asp) mice, and both EMT-up and -down genes in ATF4 mice treated with l-asparaginase (ATF4+ L-asp) mice). e, Volumes of tumours resulting from orthotopic injection of Tgf-β-silenced and -expressing 4T1-T cells (n = 10 mice per cell line). f, Percentage of Twist1-positive regions based on IHC staining of sections from tumours described in e (n = 5 tumour sections per cell line, rank-sum P < 0.01). g, Percentage of Cdh1-positive regions based on IHC staining of sections from tumours described in e (n = 5 tumour sections per cell line, rank-sum P < 0.01). h, Quantification of metastases resulting from the tumours described in e (rank-sum P < 0.05). i, Quantification of metastases resulting from intravenous injection of Tgf-β -silenced and -expressing cells (n = 10 mice per cell line, rank-sum P < 0.05). See Source Data.

Extended Data Figure 10. Tertiary validation that asparagine availability regulates EMT

a, Representative H&E-stained sections of the tumours described in Fig. 2c, in which Asns-silenced and -expressing 4T1-T cells were orthotopically injected into NSG mice. b, Images of cultured cells after they were isolated from the tumours and lungs of mice injected orthotopically with Asns-silenced and -expressing 4T1-T cells. c, Relative Twist1 expression, as measured by qPCR, which were sorted from the tumours and lungs of mice injected orthotopically with Asnssilenced and -expressing 4T1-T cells (n = 3 tumours and lungs per cell line). d, Relative Cdh1 expression, as measured by qPCR, in the tumours and lungs described in c (n = 3 tumours and lungs per cell line).

Extended Data Figure 11. Quaternary validation that asparagine availability regulates EMT

a, Representative images of IHC staining for Twist1 and Cdh1 on sections from lungs described in Fig. 4e, in which mice were injected orthotopically with Asns-silenced and -expressing 4T1-T cells. b, Quantification of all Twist1 stainings, described in Fig. 4e and a (n = 5 tumour sections and n > 5 lung metastases, edges of the box are the 25th and 75th percentiles and error bars extend to the values q3 + w(q3 − q1) and q1 − w(q3 − q1), in which w is 1.5 and q1 and q3 are the 25th and 75th percentiles, which is also the case for c–g, rank-sum P < 0.01 and P < 0.05 for Asns-silenced versus -expressing tumours and metastases, respectively). c, Quantification of all Cdh1 stainings, described in b (n = 5 tumour sections and n = 9 lung metastases, rank-sum P < 0.01 and P < 0.05 for Asns-silenced versus -expressing tumours and metastases, respectively). d, Quantification of Twist1-positive regions in the tumours resulting from orthotopic injection of Asns-expressing and -silenced 4T1-T cells into animals treated with PBS or l-asparaginase (n = 5 tumour sections per condition, rank-sum P < 0.01 for Asns-silenced versus -unsilenced cells and rank-sum P < 0.05 for each cell line in treated versus untreated mice). e, Quantification of Cdh1-positive regions in the tumours described in d (n = 5 tumour sections per condition, rank-sum P < 0.01 for Asns-silenced versus -unsilenced cells and rank-sum P < 0.05 for each cell line in treated versus untreated mice). f, Quantification of Twist1- positive regions in tumours resulting from orthotopic injection of Asnsexpressing and -silenced cells into mice fed a 0%, 0.6%, or 4% asparagine diet (n = 5 tumour sections per condition, rank-sum P < 0.01 between Asns-silenced and -expressing cells and between diets). g, Quantification of Cdh1-positive regions in the tumours described in f (n = 5 tumour sections per condition, rank-sum P < 0.01 between Asns-silenced and -expressing cells and between diets).

Figure 1. Identification of metastatic drivers

a, Relative proportions of 4T1-E and -T cells extracted from the lungs of NSG mice, into which mixtures of cells were introduced via tail vein at different concentrations. Each bar represents a sample or independent mouse. b, RNAi screening scheme to identify drivers of invasion in vitro and extravasation and colonization in vivo (n = 5 mice or n = 2 Matrigel six-well invasion chambers per approximately 50-construct shRNA pool, gene-level hit calls with empirical Bayes-moderated t-test false discovery rate < 0.05 and 0.1 for in vivo and in vitro screens, respectively). c, Overlap between genes identified in each arm of the RNAi screen depicted in b (hypergeometric test P < 0.01).

Figure 2. Validation of Asns as a driver of invasion and metastasis

a, Quantification of metastases in the lungs of mice intravenously injected with Asns-silenced or -expressing 4T1-T cells (n = 10 mice per cell line, edges of the box are the 25th and 75th percentiles, and error bars extend to the values q3 + w(q3 − q1) and q1 − w(q3 − q1), in which w is 1.5 and q1 and q3 are the 25th and 75th percentiles, which is also true for d–f, rank-sum test P < 0.001). b, Representative images of collection wells of the Matrigel assay after Asns-silenced and -expressing cells were applied 24 h previously (n = 3 invasion chambers per cell line). c, Tumour volumes resulting from the orthotopic injection of the cells described in a (n = 10 mice per cell line). d, Relative abundance of CTCs in animals corresponding to the tumours described in c (n = 6 mice per cell line, rank-sum P < 0.05 for shAsns-2). e, Quantification of metastases in H&E-stained lung sections, from mice described in d (rank-sum P < 0.0002). f, Average diameters of the metastases of each mouse described in e. See Source Data.

Figure 3. Extracellular asparagine availability drives invasion and metastasis

a, Quantification of parental 4T1 cell invasion rates, as measured by the Matrigel invasion assay, in culture medium supplemented with the indicated NEAAs (n = 5 invasion chambers, rank-sum P < 0.01). b, Representative H&E-stained lung sections from animals injected with Asns-silenced or -expressing 4T1-T cells. Animals were administered l-asparaginase or PBS (n = 10 mice per condition). c, Quantification of lung metastases in animals injected with Asns-silenced or -expressing 4T1-T cells. Animals were administered a diet with either 0%, 0.6%, or 4% asparagine content for the duration of the experiment (n = 10 mice per condition, ranksum P < 0.05 for Asns-silenced versus -expressing cells across all diets, for each cell line with 0% versus 4% diets, for shRenilla and shAsns-1 infected cells with 0% versus 0.6% diet, and for unsilenced cells with 0.6% versus 4% diet). d, Mass-spectrometric quantification of the asparagine levels in the mammary gland, blood serum, and lungs of animals administered l-asparaginase or PBS (relative abundance normalized by total metabolite peak area, n > 8 tissue sections per condition, rank-sum P < 0.005 for PBS versus l-asparaginase across all tissues, rank-sum P < 0.05 for mammary gland versus lung, and rank-sum P < 0.0005 for serum versus lung and serum versus mammary gland). See Source Data.

Figure 4. Asparagine availability regulates EMT

a, Relative protein abundances of EMT-up proteins in Asns-silenced and -expressing 4T1- T cells (n = 3 replicates per cell line, edges of the box are the 25th and 75th percentiles, and error bars extend to the values q3 + w(q3 − q1) and q1 − w(q3 − q1), in which w is 1.5 and q1 and q3 are the 25th and 75th percentiles, which is also true for b–d, rank-sum P < 0.01). b, Relative protein abundances of EMT-up proteins in 4T1-T cells when grown in normal and asparagine-supplemented medium (n = 3 replicates per cell line, rank-sum P < 0.05). c, Relative expression levels of EMT-up and -down genes in lung metastases versus primary tumours derived from orthotopic injection of 4T1-T cells (n = 4 mice, rank-sum P < 5.0 × 10−9 for EMT-up genes). d, Relative expression levels of Twist1, Cdh1, and EMT-up and -down genes in cells isolated from tumours and lungs derived from Asns-silenced versus -expressing 4T1-T cells (n = 4 replicates per condition, rank-sum P < 5.0 × 10−11 for EMT-down genes in the tumour and rank-sum P < 5.0 × 10−8 for EMT-up genes in the lung; Cdh1 was differentially expressed in the tumours and lungs, and Twist1 was differentially expressed in tumours; DESeq false discovery rate < 0.05). e, Representative images of immunohistochemistry (IHC) stainings for Twist1 and Cdh1 in orthotopic tumours derived from Asns-silenced and -expressing 4T1-T cells. See Source Data.