Oxidative stress inhibits distant metastasis by human melanoma cells

Abstract

Solid cancer cells commonly enter the blood and disseminate systemically, but are highly inefficient at forming distant metastases for poorly understood reasons. Here we studied human melanomas that differed in their metastasis histories in patients and in their capacity to metastasize in NOD-SCID-Il2rg(-/-) (NSG) mice. We show that melanomas had high frequencies of cells that formed subcutaneous tumours, but much lower percentages of cells that formed tumours after intravenous or intrasplenic transplantation, particularly among inefficiently metastasizing melanomas. Melanoma cells in the blood and visceral organs experienced oxidative stress not observed in established subcutaneous tumours. Successfully metastasizing melanomas underwent reversible metabolic changes during metastasis that increased their capacity to withstand oxidative stress, including increased dependence on NADPH-generating enzymes in the folate pathway. Antioxidants promoted distant metastasis in NSG mice. Folate pathway inhibition using low-dose methotrexate, ALDH1L2 knockdown, or MTHFD1 knockdown inhibited distant metastasis without significantly affecting the growth of subcutaneous tumours in the same mice. Oxidative stress thus limits distant metastasis by melanoma cells in vivo.

Extended data figure 1. Expression of melanoma markers by xenografted melanomas

a) M405, M481, M514, M528, M498, M597, M610, and UT10 tumours were consistently positive for S100, a marker used clinically to diagnose melanoma. b) Flow cytometric analysis of xenografted tumour cells that were HLA-ABC+ and negative for mouse CD31/CD45/Ter119 showed that these cells were usually positive for Melanoma Cell Adhesion Molecule (MCAM) and Melanoma-associated Chondroitin Sulfate Proteoglycan (MCSP). Both of the tumors that lacked MCSP staining (M514 and M597) were heavily pigmented and expressed other melanoma markers (such as S100 and MCAM).

Extended data figure 2. Clinical data on the melanomas used in this study and summary of their metastatic behavior in NSG mice

a) Summary of the clinical characteristics of the melanomas used in this study at the time of banking, as well as patient outcome after banking, and metastasis patterns upon transplantation of banked tumours into NSG mice. Melanomas were stratified into efficient and inefficient metastasizers. Efficient metastasizers formed distant metastases in patients and in NSG mice. Inefficient metastasizers did not form distant metastases in patients or distant macrometastases in NSG mice. They did form micrometastases in the lung, but not outside of the lung in the period of time it took for subcutaneous tumours to grow to 2 cm in diameter (when the mice had to be euthanized in these experiments). Nonetheless, most of the inefficient metastasizers have the ability to form macrometastases if given enough time (data not shown). b) Growth rates of subcutaneous tumours in NSG mice after subcutaneous transplantation of 100 cells. Statistical significance was assessed using two-tailed Student’s t-test. c) Clinical characteristics of the patients from whom melanomas were obtained at the time of banking and upon subsequent clinical follow up. The tumours were confirmed to be melanomas by clinical dermatopathology. The tumours were independently confirmed to be melanomas after xenografting in mice by histological and flow cytometric analysis of melanoma markers (Extended data figure 1) as well as examination by a clinical dermatopathologist.

Extended data figure 3. Barriers to distant metastasis in vivo

a) Live human melanoma cells were identified by flow cytometry based on the expression of DsRed (all melanomas in this study stably expressed DsRed) and human HLA and the lack of expression of mouse CD45, CD31 and Ter119 (to exclude mouse hematopoietic and endothelial cells). Human melanoma cells were observed in the blood of NSG mice bearing efficiently metastasizing melanomas. b) Mice xenografted with efficiently metastasizing melanomas (n=43 mice with tumours derived from 4 patients) had significantly higher frequencies of circulating melanoma cells (CMCs) in their blood than mice xenografted with inefficiently metastasizing melanomas (n=13 mice with tumours derived from 4 patients) or control mice that had not been xenografted (n=18 mice). Blood was collected by cardiac puncture. Statistical significance was assessed using one-way analysis of variance (ANOVA) followed by Tukey’s test for multiple comparisons (**, p< 0.005). c–f) Bioluminescence analysis of total photon flux (photons/second) from mouse organs after intravenous injection (c, d) or intrasplenic injection (e, f) of luciferase-tagged melanoma cells derived from efficiently metastasizing (c, e) or inefficiently metastasizing (d, f) melanomas. Each melanoma was derived from a different patient and was studied in an independent experiment. g) Schematic of the experiment shown in Table 2a. h) Schematic of the experiment shown in Table 2b. i) Summary of mean limiting dilution frequencies of tumour-forming cells after subcutaneous, intravenous, or intrasplenic transplantation into NSG mice.

Extended data figure 4. Unsupervised clustering suggests that melanoma cells undergo reversible metabolic changes during metastasis

Hierarchical clustering from two independent experiments reflecting (a) subcutaneous (SQ) tumours and metastatic nodules from the liver, pancreas, lung, and kidneys of mice transplanted with melanomas M405, M481 and M514 (see Extended data table 1 for data on individual metabolites) and (b) subcutaneous (SQ) tumours and metastatic nodules from the liver, pancreas, and kidneys of mice transplanted with melanomas M405, M481 and UT10 (n=2–3 mice/melanoma in each experiment; see Extended data table 2 for data on individual metabolites). c) Hierarchical clustering of metabolites extracted from flow cytometrically sorted human melanoma cells isolated from subcutaneous tumours or metastatic nodules (UT10, M481, n=3 mice/melanoma in two independent experiments). d and e) Hierarchical clustering of metabolites extracted from subcutaneous tumours and metastatic nodules from mice transplanted subcutaneously with either subcutaneous, circulating or metastatic melanoma cells (n=4 mice for each melanoma in two independent experiments). f) GSH/GSSG ratios from each of the experiments that compared subcutaneous tumours and metastasizing cells. i) Metabolites were extracted in the presence of 0.1% formic acid to inhibit spontaneous oxidation in two independent experiments comparing subcutaneous and metastatic tumours from mice with three different melanomas in each experiment (M405, M481, and UT10). ii and iii) GSH/GSSG ratios from the experiments shown in panels a and b, respectively. iv) GSH/GSSG ratios in melanoma cells isolated by flow cytometry from subcutaneous tumours and the blood (CMCs) of mice bearing M405 and M481. v) Metabolites were extracted in the presence of 0.1% formic acid in two independent experiments in which melanoma cells were isolated by flow cytometry from subcutaneous and metastatic tumours (M405 and M481). While the GSH/GSSG ratio was always significantly higher in melanoma cells from subcutaneous tumours as compared to circulating cells or metastatic nodules the ratio varied among experiments as a result of technical differences in cell isolation and metabolite extraction as well as differences in MS sensitivity to GSH and GSSG. g) GSH/GSSG ratios in subcutaneous tumours that arose from the transplantation of melanoma cells obtained from subcutaneous tumours or metastatic nodules, as well as the metastatic nodules from the same mice (M405; n=2 to 3 mice per treatment in one experiment). These data suggest that the decline in GSH/GSSG ratio in metastasizing melanoma cells is reversible upon subcutaneous transplantation. h) Histogram showing mitochondrial mass in subcutaneous tumour cells that arose from the transplantation of subcutaneous cells (SQ from SQ), subcutaneous tumour cells that arose from the transplantation of metastatic cells (SQ from Mets), and metastatic cells (Metastases). These histograms reflect the data shown in Figure 1g. All error bars represent standard deviation. Statistical significance was assessed using, two-tailed Student’s t-tests (f and g) (*, p<0.05)

Extended data figure 5. Metastatic nodules exhibited increased enrichment of labeled serine and glycine as compared to subcutaneous tumours

In vivo isotope tracing of uniformly ¹³C-labelled (a) lactate (M+3), (b) 3-phosphoglycerate (M+3), (c) serine (M+3), and (d) glycine (M+2) in subcutaneous tumours versus metastatic nodules from the same mice (UT10, n=3–4 mice per time point in 2 independent experiments). The fractional enrichment of labeled lactate, and 3-PG did not significantly differ among plasma, subcutaneous tumours, or metastatic tumours at any time point. In contrast, the fractional enrichment of labeled serine and glycine were significantly higher in metastatic as compared to subcutaneous tumors. This is consistent with increased de novo serine synthesis in metastatic tumors but could also reflect altered serine/glycine exchange with circulating serine/glycine pools in metastatic as compared to subcutaneous tumors. e) NADPH/NADP ratios in subcutaneous tumours and metastatic nodules from the same mice shown in Figure 2d and 2e. f and g) Western blot of ALDH1L2 (f) and MTHFD1 (g) protein after shRNA knockdown in melanoma cells. Uncropped western blots are shown in Supplementary Figure 1. h and i) Amount of GSH (h) and GSSG (i) per mg of subcutaneous or metastatic tumour as measured by LC-MS (M405, M481 and UT10, n=2–3 mice/melanoma in 2 independent experiments). All data represent mean±sd. Statistical significance was assessed using two-tailed Student’s t-tests (e, h, and i) and one-way analyses of variance (ANOVAs) followed by Dunnett’s tests for multiple comparisons (a – d; *, p<0.05; ***, p< 0.0005). (j) Schematic of the folate pathway including NADPH generating (green box) and NADPH consuming (red box) reactions.

Extended data figure 6. Immunofluorescence analysis of ALDH1L2 in subcutaneous melanoma as well as metastatic nodules in the liver, pancreas and lung

Melanoma cells can be distinguished from host stromal cells based on staining for the melanoma marker, S100.

Figure 1. Metastasizing melanoma cells experience high levels of oxidative stress

a) GSH/GSSG ratio in subcutaneous tumours as compared to metastatic nodules (n=15 mice from 2 independent experiment with 3 melanomas, M481, M405, UT10; note that extractions were performed with 0.1% formic acid to prevent spontaneous oxidation). Total amounts of GSH and GSSG are shown in Extended data Figure 5h and 5i. b) GSH/GSSG ratio in subcutaneous tumours as compared to circulating melanoma cells (n=7 mice from 3 independent experiments with 2 melanomas, M405 and UT10; these were different experiments than those in panel a, performed under different technical conditions). c, d) cytoplasmic (c) and mitochondrial (d) ROS levels in dissociated melanoma cells from subcutaneous tumours, the blood, and metastatic nodules obtained from the same mice (n=9 mice from 3 independent experiments using 3 different melanomas). e, f) Mitochondrial mass (e) and mitochondrial membrane potential (f) in dissociated melanoma cells from subcutaneous tumours, the blood, and metastatic nodules obtained from the same mice (n=6 mice from 2 independent experiments using 3 different melanomas). g) Melanoma cells underwent reversible changes in mitochondrial mass during metastasis: mitochondrial mass in dissociated melanoma cells from subcutaneous tumours versus metastatic nodules obtained from the same mice transplanted with subcutaneous, circulating, or metastatic melanoma cells. All data represent mean±sd. Statistical significance was assessed using two-tailed Student’s t-tests (a and b) and one-way analyses of variance (ANOVAs) followed by Dunnett’s tests for multiple comparisons (c–g; *, p<0.05; ***, p< 0.0005; ****, p<0.00005).

Figure 2. Melanoma cell metastasis, but not subcutaneous tumour growth, is promoted by anti-oxidants in vivo

a–c) Growth of established subcutaneous tumours in NSG mice treated with either PBS (Control) or N-acetyl-cysteine (NAC) by daily subcutaneous injection. Tumor diameter source data are shown in Supplementary Figure 1. Frequency of circulating melanoma cells in the blood (b) and metastatic disease burden (c) assessed based on total bioluminescence signal from the visceral organs of the same mice. Data in panels a–c represent 8 independent experiments with total replicates/treatment shown in the bars of panel b. Panel a shows a single representative experiment per melanoma due to the difficulty of reflecting tumour growth measurements from independent experiments in the same graph. No statistically significant differences among treatments were observed in subcutaneous tumor growth in any experiment. d and e) Levels of NADPH (d) and NADP (e) in subcutaneous tumours versus metastatic nodules (n=4 mice from 2 independent experiments with M481 and UT10). f and g) In vivo isotope tracing of uniformly ¹³C-labelled glucose into serine (f), and glycine (g) in subcutaneous tumours versus metastatic nodules from the same mice (n=6 mice in 2 independent experiments for M405; n=3 mice in one experiment for each of M481 and UT10). The fragments for uniformly labeled serine (M+3) and glycine (M+2), which come from labeled glucose via de novo serine synthesis, are shown. All data represent mean±sd. Statistical significance was assessed using two-tailed Student’s t-tests (d–f, h and i), or the Mann-Whitney test (g, due to unequal variance), and repeated-measures two-way analyses of variance (ANOVAs) (a–c; *. p<0.05; **, p<0.005; ***, p< 0.0005, ****, p<0.00005).

Figure 3. During metastasis, some melanoma cells reversibly increase their expression of folate pathway enzymes that generate NADPH and folate pathway inhibition selectively impairs metastasis

a) Western blot analysis of folate pathway enzymes in subcutaneous tumours versus metastatic liver and kidney nodules from NSG mice transplanted with three different melanomas. b–d) Growth of subcutaneous tumours in mice bearing three different melanomas (M405, M481, UT10) treated with DMSO (control) or methotrexate (n=5 mice/treatment). The frequency of circulating melanoma cells in the blood (c) and metastatic disease burden (d) in the same mice (n=10 mice/treatment for each melanoma except n=8 for M405). Data in panels b–d reflect 6 independent experiments but only one representative experiment per melanoma is shown in panel b. e–g) Growth of subcutaneous tumours in mice transplanted with two different melanomas expressing scrambled control shRNA versus two shRNAs against ALDH1L2. The frequency of circulating melanoma cells (f) and metastatic disease burden in visceral organs based on total bioluminescence signal (g). The data in panels e–g reflect 6 independent experiments (n=10 mice per shRNA for M405 and n=19 mice per shRNA for M481) but only one representative experiment per melanoma is shown in panel e. h) Western blot analysis of ALDH1L2 expression in subcutaneous tumours versus metastatic liver nodules from a donor mouse or from recipient mice subcutaneously transplanted with subcutaneous, circulating, or metastatic melanoma cells from the donor mouse. The increase in ALDH1L2 expression in metastatic liver nodules was reversible upon subcutaneous transplantation. Data in panels a and h are from two independent experiments. i–k) Growth of subcutaneous tumours in mice transplanted with cells from two melanomas expressing either scrambled control shRNAs or two shRNAs against MTFHD1. Frequency of circulating melanoma cells (j) and metastatic disease burden in visceral organs (k) from the same mice. Data in panels i–k reflect four independent experiments with a total of 9 mice per control shRNA and 10 mice per shRNA against MTFHD1 for each melanoma. Panel i shows one representative experiment per melanoma. All error bars represent standard deviation. Statistical significance was assessed using one-way analyses of variance (ANOVAs) followed by Dunnett’s tests for multiple comparisons (f, g, j and k), two-tailed Student’s t-tests (c and d) and repeated measures two-way ANOVAs (b, e and i) (*, p<0.05; **, p,0.005, ***, p< 0.0005; ****, p<0.00005). Tumor diameter and western blot source data are in Supplementary Figure 1.