Despite being the leading cause of cancer deaths, metastasis remains a poorly understood process. To identify novel regulators of metastasis in melanoma, we performed a large-scale RNA sequencing screen of 48 samples from patient-derived xenograft (PDX) subcutaneous melanomas and their associated metastases. In comparison with primary tumors, expression of glycolytic genes was frequently decreased in metastases, whereas expression of some tricarboxylic acid (TCA) cycle genes was increased in metastases. Consistent with these transcriptional changes, melanoma metastases underwent a metabolic switch characterized by decreased levels of glycolytic metabolites and increased abundance of TCA cycle metabolites. A short isoform of glyceraldehyde-3-phosphate dehydrogenase, spermatogenic (GAPDHS) lacking the N-terminal domain suppressed metastasis and regulated this metabolic switch. GAPDHS was downregulated in metastatic nodules from PDX models as well as in human patients. Overexpression of GAPDHS was sufficient to block melanoma metastasis, whereas its inhibition promoted metastasis, decreased glycolysis, and increased levels of certain TCA cycle metabolites and their derivatives including citrate, fumarate, malate, and aspartate. Isotope tracing studies indicated that GAPDHS mediates this shift through changes in pyruvate carboxylase activity and aspartate synthesis, both metabolic pathways critical for cancer survival and metastasis. Together, these data identify a short isoform of GAPDHS that limits melanoma metastasis and regulates central carbon metabolism.

Significance:

This study characterizes metabolic changes during cancer metastasis and identifies GAPDHS as a novel regulator of these processes in melanoma cells.

Cancer metastasis is a complex and inefficient process whereby most cancer cells are unable to survive (1). Successfully metastasizing cancer cells must have metabolic plasticity to manage the heterogeneous environments and stresses during the metastatic journey (2). Although metabolism has been of interest in oncogenesis for quite some time, very little is understood about the critical metabolic regulators involved in cancer metastasis.

Glycolytic enzymes are frequently upregulated in developing cancers (3) and enhanced glycolysis is an important requirement for meeting the metabolic demands of cancer cell proliferation (4). Additionally, alterations in glycolysis and its associated enzymes are known to occur during the process of metastasis, beginning as early as tumor disassociation (5). A recent analysis of single-cell RNA sequencing (RNA-seq) comparing primary and metastatic breast cancer cells found glycolysis to be one of the most significantly altered transcriptional pathways in metastasis (6). The functional effect of glycolysis on metastasis appears to be context dependent as some reports have shown it enhances metastasis (7), whereas others show it inhibits it (8).

Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) is a ubiquitously expressed enzyme that lies at the center of the glycolytic pathway. Although traditionally thought of as a housekeeping gene, it functions in other aspects of metabolism as well as nonmetabolic moonlighting functions (9). Like glycolysis itself, GAPDH expression and function have been described in multiple tumor types with seemingly contradictory phenotypes, underscoring its pleiotropic and likely contextual role in tumor progression. In human melanoma samples, GAPDH expression is altered during tumor progression, with increased expression in thick primary melanomas and decreased expression in lymph node metastases (10). Although the functional role of GAPDH has not been well studied in melanoma, it has been shown to play various roles in other cancers, with functions ranging from promoting tumor induction (11) to impacting therapy response (12).

GAPDH, spermatogenic (GAPDHS) is a variant of GAPDH encoded on a separate chromosome, with an expression pattern limited only to developing spermatocytes (13, 14), where it plays a critical role in glycolysis, ATP generation, and sperm motility (15). GAPDHS shares 70% amino acid sequence homology with its somatic counterpart GAPDH (14), and biochemical assays suggest these differences endow GAPDHS with enhanced protein stability (16) and increased catalytic efficiency (17). Although GAPDHS has been noted to be expressed in a small subset of melanoma cell lines (18), its metabolic functions have not been investigated. Herein, we identify a short GAPDHS isoform that is uniquely expressed in melanoma, alters metastatic efficiency, and affects the central carbon metabolism of melanoma cells.

Melanoma specimens

Melanoma specimens were obtained with informed consent from all patients according to Institutional Review Board (IRB)-approved protocols at the University of Michigan (IRBMED approvals HUM00050754 and HUM00050085; ref. 19) and the University of Texas Southwestern Medical Center (IRB approval 102010-051). Materials used in this article are available either commercially or from the authors, although there are restrictions imposed by the IRB and institutional policy on the sharing of patient materials. Igr1 (ABC-TC537S, Accegen Biotechnology) and Mewo cell lines (ATCC, HTB-65) were obtained from commercial vendors. Mycoplasma testing and cell line authentication were not performed except as documented by the initial vendor. All experiments were performed using early passages (<p10) after receipt of the cell lines.

Melanoma tumor disaggregation

Single-cell suspensions of melanoma tumors were obtaining by dissociating tumors in Kontes tubes with disposable pestles (VWR) followed by enzymatic digestion in 200 U/mL collagenase IV (Worthington), DNase (50 U/mL), and 5 mmol/L CaCl2 for 20 minutes at 37°C. Cells were filtered through a 40-μm cell strainer to remove clumps and obtain a single-cell suspension.

Mouse studies and xenograft assays

All mouse experiments complied with the relevant ethical regulations and were performed according to protocols approved by the Institutional Animal Care and Use Committee at the University of Texas Southwestern Medical Center (protocol 2016-101360). No formal randomization techniques were used, although animals were allocated randomly to groups and specimens were processed in an arbitrary order. For all experiments, the maximum tumor diameter permitted was 2.5 cm and this limit was not exceeded in any experiment. For all experiments, mice were kept on normal chow and fed ad libitum.

Patient-derived melanomas were transplanted into 4-to-8-week-old male and female NOD.CB17-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice. Subcutaneous injections were performed in the right flank of mice in a final volume of 50 μL using 100 cells per injection. Subcutaneous tumor diameters were measured weekly with calipers until tumors reached 2.5 cm in its largest diameter. At that point, mice in the cohort or subcohort were killed, per approved protocol, for analysis of subcutaneous tumor diameter, circulating melanoma cell frequency and metastatic disease burden. Metastatic disease burden was evaluated by macroscopic examination (for nodule counts) and/or bioluminescence imaging (see details below).

Intravenous injections were performed by injecting 1,000 human melanoma cells into the tail vein of mice in 100 μL of phosphate-buffered saline (PBS).

Lentiviral/shRNA transduction of human melanoma cells

All PDX melanomas expressed dsRed and luciferase as previously described (1). All shRNAs were expressed from a pGFP-C-shLenti vector (Origene). For knockdown of GAPDHS, the following Origene shRNA clones were used: shA (TL312840A), shB (TL312840B), and shC (TL312840C). For each PDX, the two shRNA constructs providing the best knockdown were used for subsequent studies. For overexpression of GAPDHS, the human open reading frame was obtained from the Precision LentiORF collection (Dharmacon) in a bicistronic lentiviral construct coexpressing turbo green fluorescent protein (pLOC-GAPDHS-IRES-tGFP). This construct was modified to remove the N-terminal domain of GAPDHS and begin with the alternative start site known to be utilized in melanoma cells as described herein. A turbo red fluorescent protein (tRFP) was expressed in place of GAPDHS in the same construct for use as a control.

For production of virus, 0.9 μg of the appropriate plasmid along with 1 μg of helper plasmids (0.4 μg pMD2G and 0.6 μg of pSPAX2) was transfected into 293T cells using PolyJet (SignaGen) per the manufacturer's instructions. The viral supernatants were then collected at 48 hours after transfection and filtered through a 45-μm filter. Approximately 300,000 freshly dissociated melanoma cells were infected with viral supernatants supplemented with 10 μg/mL polybrene (Sigma) for 4 hours. Cells were next washed with staining medium (L15 medium, BSA 1 mg/mL, 1% penicillin/streptomycin, and 10 mmol/L HEPES pH7.4) and pelleted. Cell pellets were resuspended in staining medium with 25% high-protein Matrigel (354248; BD Biosciences) and injected subcutaneously into NSG mice. After growing into tumors approximately 1–2 cm in diameter, the tumors were removed, dissociated into single-cell suspensions (see above), and sorted for dsRed and GFP double-positive cells. These double-positive cells were transplanted into NSG mice for propagation and in vivo studies examining tumor growth, metastasis, and metabolism. For melanoma cell lines, the same procedure was performed above except posttransfected cells were expanded in vitro (in DMEM, 10% FBS, and penicillin/streptomycin) rather than in mice.

Bioluminescence imaging

Metastatic burden in live animals was monitored with bioluminescence imaging (melanomas were tagged with stable expression of luciferase). Five minutes prior to imaging, mice were injected intraperitoneally with 100 μL of PBS with D-luciferin monopotassium salt (40 mg/mL, Biosynth). Mice were anesthetized with isoflurane just prior to imaging. All mice were imaged with an IVIS Imaging System 200 series (Caliper Life Sciences) with Living Image software. The bioluminescence signal (total photon flux) was quantified with the “region of interest” tool in Living Image (PerkinElmer) software.

Flow cytometry/sorting

To isolate melanoma cells from xenograft tumors and metastases, tumors were digested as described above. Melanoma cells were identified and sorted by flow cytometry as previously described (1). Sorting was performed with a FACSAria II SORP (Becton Dickinson) or FACSAria Fusion (Becton Dickinson).

To quantify circulating melanoma cells in the blood, blood was collected by cardiac puncture with a syringe pretreated with citrate-dextrose solution (sc-2141744, Santa Cruz). Specimens were first sedimented using Ficoll following the manufacturer's instructions (17144002, Ficoll Paque Plus, GE Healthcare), to eliminate red blood cells. The remaining cells in suspension were washed with Hank's balanced buffered solution (14025076, HBSS, Gibco) and stained with antibodies in preparation for flow-cytometric analysis. Antibody labeling was performed for 20 minutes on ice, then washed with HBSS and centrifuged at 200 × g for 5 minutes. Cells were stained with directly conjugated antibodies against mouse CD45 (violetFluor 450, eBiosciences), mouse CD31 (390-eFluor450, BioLegend), mouse Ter119 *eFluor450, BD Biosciences) and human HLA-A, HLA-B, and HLA-C (G46-2.6-FITC, BD Biosciences). Human melanoma cells were isolated as cells that were negative for mouse endothelial and hematopoietic markers and positive for human HLA markers and dsRed (stably transfected in melanoma cells). Cells were washed with staining medium and resuspended in 4′,6-diamidino-2-phenylindole (DAPI; 1 μg/ml; Sigma) to eliminate dead cells from analysis. Cells were analyzed using an LSRFortessa cell analyzer (Becton Dickinson), FACSAria II SORP (Becton Dickinson), or FACSAria Fusion (Becton Dickinson).

RNA isolation and RNA-seq

For RNA isolation of melanoma samples, up to 100,000 melanoma cells were sorted directly into TRIzol LS (Thermo Fisher Scientific). RNA was isolated using the Direct-zol RNA Microprep column purification kit (Zymo Research). RNA quality was assessed on an Agilent 2100 Bioanalyzer, and only samples with high RNA quality were used. RNA was quantitated using a Qubit RNA Assay Kit (Thermo Fisher Scientific). The RNA library was prepared using the TruSeq Stranded Total RNA kit (Illumina) following the manufacturer's Low Sample (LS) Protocol. Libraries were quantitated and assessed using the Agilent 2200 Tapestation system. Samples were sequenced using paired-end 75-bp reads on an Illumina NextSeq 500 system.

RNA-seq analysis

PDX subcutaneous tumors and metastases

RNA-seq libraries were sequenced as 75-bp paired-end reads by an Illumina NextSeq 500 at 46,677,291 ± 10,592,103 read pairs per sample. Quality of raw reads was checked using FastQC 0.11. Raw reads were trimmed using TrimGalore 0.6 and mapped to the Ensembl GRCh37 human reference genome using TopHat2 with Bowtie2. Mapped reads were quality-filtered using SAMtools 1.9 to keep uniquely mapped reads only. 81.6% ± 4.6% of the raw read fragments were uniquely mapped, and they were quantified using HTSeq 0.9. For sample clustering, heatmap, and gene differential expression analyses, quantified reads were first normalized using the relative log expression method by DESeq2, and graphs were generated using R 4.0.2. For gene-expression measurements, quantified reads were normalized to fragments per 1,000 exonic bases per million mapped reads (FPKM), and gene-expression levels were measured as FPKMs using DESeq2. To discover novel gene isoforms, uniquely mapped reads were quantified using DEXSeq to obtain counts per sample that fall into genes' exonic regions. Pathway analysis was performed using iPathwayGuide (Advaita). Raw RNA-seq data cannot be made available for patient privacy reasons, but see Supplementary Table S1 for a complete list of genes and FPKMs obtained from all samples.

The Cancer Genome Atlas analysis

RSEM and batch-normalized Illumina HiSeq RNA-seq V2 data of GAPDHS were sourced from The Cancer Genome Atlas (TCGA) PanCancer Atlas and downloaded from the cBioPortal website (https://www.cbioportal.org/). RPKM-normalized exonic counts, RSEM-normalized gene expressions, sample types (primary vs. metastatic), and the BRAF mutation status were sourced from the Skin Cutaneous Melanoma (SKCM) cohort of TCGA and downloaded from the UCSC Xena cancer browser (https://xenabrowser.net/).

GAPDHS in human tissues

TPM-normalized RNA-seq expression data of GAPDHS in normal human tissues were sourced from the GTEx project (https://gtexportal.org/home/) and downloaded from the Human Protein Atlas website (https://www.proteinatlas.org/).

GAPDHS in human sperm samples

Exon-level expression data of GAPDHS in human sperm samples were downloaded from the NCBI's GEO study GSE65683 (20).

Protein extraction and Western blotting

Tumor tissue was dissected and snap-frozen onto dry ice or in liquid nitrogen. Protein was extracted using RIPA buffer with the addition of protease inhibitors and mechanical homogenization with pestles. Protein was quantified using a BCA kit and equal amounts (5–15 μg) loaded across wells. Testis lysate was obtained from Novus Biologicals (NB820-60623). Protein samples were resolved on SDS-PAGE gels and transferred to polyvinylidene difluoride membranes (product 1620177, Bio-Rad). After blocking in 5% milk diluted in TBS with 0.1% Tween20 (TBS-T), membranes were probed with primary antibody diluted in the blocking buffer overnight at 4°C. Membranes were washed with TBS-T and then probed with horseradish peroxidase–conjugated secondary antibodies (Cell Signaling Technology) diluted in the blocking buffer at room temperature for 1 to 2 hours. Signals were most often developed with SuperSignal West (34580, Thermo Fisher Scientific). The following primary antibodies were used for Western blot analysis: GAPDHS (sc-293335, Santa Cruz Biotechnology, 1:500), GAPDH (14C10, Cell Signaling Technology, 1:10,000), ACTIN (D6A8, Cell Signaling Technologies, 1:1,000), pyruvate carboxylase (16588-1-AP, Proteintech, 1:1,000).

IHC

Mouse organs were fixed in 4% paraformaldehyde and then submitted to the UTSW Histo Pathology Core for processing, paraffin embedding, sectioning, and hematoxylin and eosin staining. IHC was performed on human primary and metastatic melanoma sections using EDTA steamer retrieval and primary antibody GAPDHS (PA5-60070, Thermo Fisher Scientific 1:100). The latter antibody was tested on a large-scale tissue array to optimize conditions and ensure expected specificity for testis with limited signal in other tissues. Mitotic index was obtained by counting the number of mitotic figures within the equivalent of a 1-mm2 area of the indicated samples.

Reactive oxygen species measurement

Intracellular reactive oxygen species (ROS) was measured as previously described (21). In brief, indicated subcutaneous tumors were removed and mechanically dissociated in staining medium. Single-cell suspensions were obtained by passing cells through a 40-μm strainer. Equal cell numbers of cells from each group were stained for 30 minutes at 37°C with 5 μmol/L CellROX DeepRed (Invitrogen) or 100 nmol/L Mitotracker DeepRed (Invitrogen) in HBSS (Ca and Mg-free) and DAPI (to select live cells). The cells were then washed and analyzed by flow cytometry using a FACS Fusion or Fortessa (BD Biosciences) to determine ROS and/or mitochondrial potential levels in live human melanoma cells (HLA+, dsRed+, DAPI, mouse CD45/CD31/Ter119). Control cells were normalized to 1 for presentation clarity.

Mito-TEMPO treatment

For Mito-TEMPO (CAS 1569257-94-8, Santa Cruz) treatments, mice were implanted with 100 cells of the indicated PDX melanoma as described above. When tumors became palpable, mice were treated daily with intraperitoneal injections of Mito-TEMPO (0.7 mg/kg) or vehicle control (PBS) as described (22).

Sample preparation for metabolic studies

For all metabolic studies, melanoma tissue was quickly dissected at the time of euthanasia and immediately snap-frozen in liquid nitrogen. Pieces were obtained from three separate regions of the subcutaneous tumor in every mouse to ensure adequate representation of different areas, which we have previously found to exhibit some level of heterogeneity. At the time of extraction, frozen tissues were homogenized manually with a pestle in ice-cold 80:20 methanol:water (v/v). After homogenization, samples were spun for 13,000 × g for 15 minutes at 4°C. Supernatants were transferred to a new tube, and equivalent amounts (based on BCA quantification) were dried down and resuspended in 80% acetonitrile for analysis.

Metabolite detection and analysis

Metabolite detection was performed as previously described in detail (21). In brief, HILIC chromatographic separation of metabolites was achieved through use of a Millipore ZIC-pHILIC column (5 μm, 2.1 × 150 mm) with a binary solvent system of 10 mmol/L ammonium acetate in water, pH 9.8 (solvent A), and acetonitrile (solvent B) with a constant flow rate of 0.25 mL/min. Metabolites were measured with a Thermo Scientific QExactive HF-X hybrid quadrupole orbitrap high-resolution mass spectrometer (HRMS) coupled to a Vanquish UHPLC.

Metabolite identities were confirmed in three ways: (i) precursor ion m/z was matched within 5 ppm of the theoretical mass as predicted by the chemical formula; (ii) fragment ion spectra were matched within 5 ppm to known metabolite fragments; and (iii) the retention time of metabolites was within 5% of the retention time of a purified standard run with the same chromatographic method. Metabolites were relatively quantitated by integrating the chromatographic peak area of the precursor ion searched within a 5 ppm tolerance. For calculating relative abundance, the peak area was divided by the overall total ion chromatogram (TIC) for that sample.

13C isotope tracing studies and analysis

In vivo isotope tracing studies were performed when PDX subcutaneous tumors reached 2 cm in diameter. Prior to infusing, mice were fasted for 16 hours and a 27-gauge catheter was placed in the lateral tail vein under anesthesia. [1,2-13C]glucose (CLM-504, Cambridge Isotope Laboratories) was infused intravenously with a bolus of 0.4125 mg/g of body mass per minute for 3 hours (in a volume of 150 μL/h; ref. 23). At the end of the infusion, mice were euthanized, and tumors were harvested and immediately snap-frozen in liquid nitrogen. To assess fractional enrichment in plasma, 20 μL of blood was obtained retro-orbitally after 30, 60, 120, and 180 minutes of infusion.

In vitro isotope tracing studies were performed with specified PDX tumor cells. Equivalent numbers of cells were placed in 12-well tissue-culture–treated plates in DMEM plus 10% FBS for 24 to 48 hours and cultured at 37°C. At t = 0, wells were washed with 1 mL of 1× PBS two times, and the medium was replaced with a new medium containing the isotope indicated. For [U-13C]-glucose tracing, the medium was replaced with DMEM (-glucose; A11966-025, Gibco), 10% dialyzed FBS, and [U-13C]-glucose (CLM-1396, Cambridge Isotope Laboratories) at a final concentration of 4.5 mg/mL and cells were harvested after 1 hour. For [U-13C]-lactate tracing, the medium was replaced with DMEM, 10% dialyzed FBS, and [U-13]C-lactate 20% w/w (CLM-1579, Cambridge Isotope Laboratories) at a final concentration of 3.6 mmol/L and cells were harvested after 4 hours. For [U-13C]-glutamine tracing, the medium was replaced with DMEM (-glucose/-glutamine; A14430-01, Gibco), 4.5 mg/mL glucose, 10% dialyzed FBS, and [U-13C]-glutamine (CLM-1822, Cambridge Isotope Laboratories) at a final concentration of 3.6 mmol/L and cells were harvested after 4 hours. For [U-13C]-palmitic acid tracing, the medium was replaced with DMEM, 10% charcoal-stripped FBS, and [U-13C]-palmitic acid (CLM-409, Cambridge Isotope Laboratories) at a final concentration of 100 μmol/L, and cells were harvested after 24 hours. For etomoxir experiments, cells were treated with 0, 3, or 30 μmol/L etomoxir (CAS 828934-41-4, Cayman Chemical Company) in DMEM with 10% FBS or with [U-13C]-palmitic acid-containing medium as described above. At indicated timepoints, wells were gently washed with ice-cold PBS and then scraped in 80:20 methanol:water (v/v). Samples were transferred into Eppendorf tubes and spun for 13,000 × g for 30 minutes at 4°C. Supernatants were taken, dried down, and resuspended in 80% acetonitrile for analysis.

For 13C isotope tracing analysis, the theoretical masses of 13C isotopes of glycolytic and tricarboxylic acid (TCA) metabolites were calculated and added to a library of predicted isotopes. These masses were next searched with a 5 ppm tolerance and integrated only if the peak apex showed less than 1% difference in retention time from the [U-12C] monoisotopic mass in the same chromatogram. After analysis of the raw data, theoretical natural abundance was calculated. Natural isotope abundances were corrected using a customized R script, found at the GitHub repository (https://github.com/wencgu/nac). The script was written by adapting the AccuCor algorithm (24). For in vivo tracing studies, all tissue fractional enrichments were normalized to the m+2 glucose enrichment in the peripheral circulation of the host mouse at the infusion endpoint (t = 3 hours). Pieces were obtained from three separate regions of the subcutaneous tumor in every mouse to ensure adequate representation of different areas. Any tumor samples with <1% labeled glucose (likely due to necrosis) were removed from the analysis.

GAPDH activity

GAPDH activity was performed using a commercially available assay (MAK277, Sigma) and following the manufacturer's instructions. The optical density (OD) for each sample was measured in kinetic mode, and two time points in the linear range were used to calculate the GAPDH activity of the samples.

Statistical methods

In each type of experiment, multiple mice were tested in multiple independent experiments performed on different days. Mice were allocated to experiments randomly and samples processed in an arbitrary order, but formal randomization techniques were not used. No formal blinding was applied when performing the experiments or analyzing the data. Sample sizes were not predetermined based on statistical power calculations but were based on our experience with these assays. No data were excluded.

For comparison of data between two groups or more in Figs. 1 and 6 and Supplementary Figs. S3A, S4C-D, S6, S7, S8, and S9, Student t tests (unpaired) and one-way ANOVA with Dunnett correction for multiple comparisons were performed as indicated in figure legends. For all other data, prior to analyzing the statistical significance of differences among treatments, we tested whether data were normally distributed and whether variance was similar among groups. To test for normality, we performed the Shapiro–Wilk tests when 3≤ n < 20 or D'Agostino Omnibus tests when n ≥ 20. To test whether variability significantly differed among groups, we performed F tests (for experiments with two groups) or Levene's median tests (for experiments with more than two groups). When the data significantly deviated from normality or variability significantly differed among groups, we log2 transformed the data and tested again for normality and variability. If the transformed data no longer significantly deviated from normality and equal variability, we performed parametric tests on the transformed data. If log2 transformation was not possible or the transformed data still significantly deviated from normality or equal variability, we performed nonparametric tests on the nontransformed data.

Figure 1.

Melanoma metastases show decreased glycolysis and elevated levels of TCA cycle intermediates fumarate and malate. A, Pathway enrichment analysis was performed with the set of genes whose expression level was significantly up or down in FACS-sorted melanoma cells isolated from metastatic nodules as compared with subcutaneous tumors. Gene sets from three patient melanomas, M481, M405, and UT10, were combined for the analysis. B and C, Heat map representation of the RNA-seq data demonstrating fold change in expression of glycolytic (B) and TCA (C) enzymes. All samples were normalized to the average FPKM of their respective PDX group. Green, fold changes that are lower than the average FPKM within individual melanomas; red, higher than average FPKM fold changes. D and E, Relative abundance of the metabolites from glycolysis (D) or the TCA (E) cycle in primary subcutaneous tumors and spontaneous metastatic nodules dissected from multiple organs in mice xenografted with melanoma M481 (n = 8 mice). F and G, Isotope tracing in primary subcutaneous tumors and metastatic nodules from mice xenografted with M481 after infusion with [1,2-13C]-glucose. The enrichment of metabolites generated through glycolysis (F) or the TCA cycle (G) is shown as the enrichment of the m+2 fraction normalized to fractional enrichment of m+2 glucose in the host mouse at the end of the infusion experiment (t = 3 hours; n = 8 mice). Metabolites were detected and measured with a ThermoScientific QExactive HF-X hybrid quadrupole orbitrap HRMS coupled to a Vanquish UHPLC as described in the Materials and Methods. For each metabolite in D–G, data are the mean ± SEM. Unpaired t tests were performed between the two groups. *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; ****, P ≤ 0.001; ns, nonsignificant.

Figure 1.

Melanoma metastases show decreased glycolysis and elevated levels of TCA cycle intermediates fumarate and malate. A, Pathway enrichment analysis was performed with the set of genes whose expression level was significantly up or down in FACS-sorted melanoma cells isolated from metastatic nodules as compared with subcutaneous tumors. Gene sets from three patient melanomas, M481, M405, and UT10, were combined for the analysis. B and C, Heat map representation of the RNA-seq data demonstrating fold change in expression of glycolytic (B) and TCA (C) enzymes. All samples were normalized to the average FPKM of their respective PDX group. Green, fold changes that are lower than the average FPKM within individual melanomas; red, higher than average FPKM fold changes. D and E, Relative abundance of the metabolites from glycolysis (D) or the TCA (E) cycle in primary subcutaneous tumors and spontaneous metastatic nodules dissected from multiple organs in mice xenografted with melanoma M481 (n = 8 mice). F and G, Isotope tracing in primary subcutaneous tumors and metastatic nodules from mice xenografted with M481 after infusion with [1,2-13C]-glucose. The enrichment of metabolites generated through glycolysis (F) or the TCA cycle (G) is shown as the enrichment of the m+2 fraction normalized to fractional enrichment of m+2 glucose in the host mouse at the end of the infusion experiment (t = 3 hours; n = 8 mice). Metabolites were detected and measured with a ThermoScientific QExactive HF-X hybrid quadrupole orbitrap HRMS coupled to a Vanquish UHPLC as described in the Materials and Methods. For each metabolite in D–G, data are the mean ± SEM. Unpaired t tests were performed between the two groups. *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; ****, P ≤ 0.001; ns, nonsignificant.

Close modal

When data or log2 transformed data were normal and equally variable, statistical analyses were performed using Student t tests (when there were two groups) or one-way ANOVAs (when there were more than two groups). When the data and log2 transformed data were abnormal or unequally variable, statistical analysis was performed using Mann–Whitney tests (when there were two groups), Kruskal–Wallis tests (when there were more than two groups), or nparLD tests (when there were two or more groups measured at multiple time points). P values from multiple comparisons were adjusted using the Dunnett method after ANOVAs, Dunn method after Kruskal–Wallis tests, or Benjamini–Hochberg (FDR) method after nparLD tests. Comparisons between metastatic efficiency percentages of the shRNAs were performed using Fisher exact tests with the Holm–Sidak method for multiple comparisons correction. All statistical tests were two-sided where applicable. Statistical tests were performed using GraphPad Prism V9.1.0 or R 4.0.2.

Data availability statement

The data generated in this study are available within the article and its supplementary data files. The human sequence data generated in this study are not publicly available due to patient privacy requirements.

Identification of transcriptional alterations during melanoma metastasis

Patient-derived xenografts (PDX) have previously been shown to recapitulate the metastatic properties seen in the melanoma patients from which they were derived (19). When transplanted subcutaneously into NOD-SCID Il2rg−/− (NSG) mice, efficient metastasizers spontaneously give rise to circulating melanoma cells and distant macrometastases. In order to discover novel regulators of melanoma metastasis, we performed large-scale RNA-seq of sorted melanoma cells from primary subcutaneous tumors and their respective metastases in mice transplanted subcutaneously with three efficiently metastasizing PDXs (M481, M405, and UT10). A total of 11 mice were used (4 × M481, 4 × M405, and 3 × UT10) to generate 48 unique melanoma samples from primary tumors and metastatic nodules from a range of metastatic organ sites (Supplementary Fig. S1A and Supplementary Table S1). Melanoma cells were identified and sorted based on FSC/SSC, DAPI, mouse CD31/CD45/Ter119, human HLA+, and dsRed+ as previously described (1).

Unsupervised clustering analysis revealed that all samples clustered by patient, rather than by tumor location (primary subcutaneous tumor vs. metastatic nodule), suggesting that patient tumors have a transcriptional “fingerprint” that is maintained across passages and sites of metastasis (Supplementary Fig. S1B). Consistent with this finding, the biggest transcriptional alterations among samples were seen between the three different patient melanomas (9%–12%). In contrast, we found that <1%–4% of the annotated detectable transcriptome was significantly altered when comparing metastatic samples to their primary subcutaneous site (Supplementary Fig. S1A). The genes that were significantly altered during metastasis were associated with pathways such as HIF signaling, central carbon metabolism, glycolysis, and general metabolism (Fig. 1A), although each tumor also had its own unique pathways altered (Supplementary Fig. S1C). Transcripts coding for a number of glycolytic enzymes were significantly decreased in metastatic nodules as compared with subcutaneous tumors (Fig. 1B; Supplementary Table S2) whereas some TCA cycle enzymes were increased in metastases (Fig. 1C). However, there were no individual genes that were consistently up- or downregulated in metastatic nodules from different organ sites and across all three patient melanomas, suggesting melanoma cells use a variety of strategies to overcome barriers to metastasis.

Melanoma cells that have metastasized decrease glycolysis and maintain elevated levels of fumarate and malate

To test whether there were changes in glycolysis or the TCA cycle, we compared the relative abundance of metabolites in melanoma samples from subcutaneous tumors and metastatic nodules. In order to account for metabolic influences that different microenvironments would potentially impose, we isolated metastatic nodules that ranged in size (from approximately 2 to 8 mm) and were from a variety of sites (including liver, lung, kidney, and pancreas). Consistent with the decreased expression of glycolytic enzymes that we observed in metastatic nodules (Fig. 1B; Supplementary Table S2), glycolytic intermediates were also depleted in these samples as compared with subcutaneous samples from the same mice (Fig. 1D). In turn, the TCA metabolites fumarate and malate were significantly increased in metastatic nodules (Fig. 1E), suggesting that melanoma cells may alter their central carbon metabolism during metastasis. To further understand the relative activity of these metabolic pathways, we performed isotope infusions of [1,2-13C]-glucose in eight mice bearing M481 melanoma PDX tumors. The fraction of m+2 labeling of the metabolites 2/3-PG and lactate was significantly decreased in metastatic nodules as compared with subcutaneous tumors from the same mice (Fig. 1F), suggesting decreased flow of glucose into lower glycolysis (i.e., after the GAPDH enzymatic step) and downstream lactate production. The m+2 fractional enrichment in TCA metabolites citrate, α-ketoglutarate, and succinate was lower in metastases, suggesting this portion of the TCA cycle was not as glucose dependent in metastases. In contrast, metastatic nodules had significantly higher overall levels of fumarate and malate and no change in the fractional enrichment of these metabolites from glucose as compared with subcutaneous tumors (Fig. 1G). Thus, there was a correlation between decreases in glycolysis and an elevation of fumarate and malate in melanoma cells that had metastasized.

GAPDHS expression is decreased in melanoma metastases

The glycolytic enzymes glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and glyceraldehyde-3-phosphate dehydrogenase, spermatogenic (GAPDHS) were downregulated in several metastatic sites in the RNA-seq experiment (Fig. 2A). In addition, the most notable decline in the fractional enrichment of [1,2-13C]-glucose tracing in metastases was between glyceraldehyde 3-phosphate (the substrate of GAPDH/GAPDHS) and its next detectable downstream product, 2/3-phosphoglycerate (Fig. 1F). Thus, we sought to explore whether these enzymes may have a role in the metabolic rewiring of metastasizing melanoma cells.

Figure 2.

GAPDHS is decreased in melanomas that have metastasized. A, FPKM of GAPDHS and GAPDH transcripts in primary subcutaneous tumor (SQ) samples (black) and selected sites of metastasis (red) in M405, UT10, and M481 PDXs. The number of samples per group is indicated in Supplementary Fig. S1A. B, Western blot analysis of GAPDHS and GAPDH in primary subcutaneous tumors (SQ) and metastatic tumors in mice bearing the indicated PDX melanoma tumor. C, Normalized counts of the GAPDHS transcript in TCGA melanoma samples (SKCM) derived from primary or metastatic melanoma tissue. Samples are grouped by BRAF mutation status. See Materials and Methods for details. D, Hematoxylin and eosin (H&E; top) and GAPDHS IHC (bottom) on sections from a matched primary and metastatic melanoma from a patient. Scale bar, 100 μm. All graphed data represent mean ± SD. Statistical significance was assessed using one-way ANOVA (A; with or without log2 transformation) and Kruskal–Wallis tests (A and C). *, P ≤ 0.05; **, P ≤ 0.01; ****, P ≤ 0.001.

Figure 2.

GAPDHS is decreased in melanomas that have metastasized. A, FPKM of GAPDHS and GAPDH transcripts in primary subcutaneous tumor (SQ) samples (black) and selected sites of metastasis (red) in M405, UT10, and M481 PDXs. The number of samples per group is indicated in Supplementary Fig. S1A. B, Western blot analysis of GAPDHS and GAPDH in primary subcutaneous tumors (SQ) and metastatic tumors in mice bearing the indicated PDX melanoma tumor. C, Normalized counts of the GAPDHS transcript in TCGA melanoma samples (SKCM) derived from primary or metastatic melanoma tissue. Samples are grouped by BRAF mutation status. See Materials and Methods for details. D, Hematoxylin and eosin (H&E; top) and GAPDHS IHC (bottom) on sections from a matched primary and metastatic melanoma from a patient. Scale bar, 100 μm. All graphed data represent mean ± SD. Statistical significance was assessed using one-way ANOVA (A; with or without log2 transformation) and Kruskal–Wallis tests (A and C). *, P ≤ 0.05; **, P ≤ 0.01; ****, P ≤ 0.001.

Close modal

We next performed Western blot analysis comparing subcutaneous tumors and metastatic tumors (<4 mm) isolated from the same mice. We analyzed metastatic nodules from a variety of organs in five different PDX melanomas. In all cases, GAPDHS protein levels were dramatically decreased in metastatic nodules as compared with subcutaneous tumors from the same mice, with a dramatically larger fold-change effect than what was seen at the transcript level (Fig. 2B). In contrast, somatic GAPDH protein levels were similar between subcutaneous and metastatic samples. To determine whether this trend holds true among samples directly isolated from patients, we compared RNA-seq data of TCGA primary and metastatic melanomas and also found a significant decline in GAPDHS expression in melanoma metastases (Fig. 2C). This effect was independent of BRAF mutation status, which is known to influence glycolysis (25). We next performed GAPDHS IHC on biopsies obtained from melanoma patients, which showed prominent staining in primary melanomas but significantly less expression in metastases (Fig. 2D). Overall, GAPDHS protein levels were robustly decreased in melanoma metastases, both in patients and in PDX models.

A novel isoform of GAPDHS is expressed in the majority of melanomas

Among normal tissues, GAPDHS is exclusively expressed in developing spermatocytes within the testis (Supplementary Fig. S2A; ref. 15). Although GAPDHS expression has been noted in cell lines (18), its expression pattern in human cancers has not been described. When we looked at its RNA expression across all cancers in the TCGA database, GAPDHS transcripts were abundant in the majority of melanoma tumors (including the 473 cutaneous and 80 uveal melanomas contained in TCGA) but rare in other cancers (Fig. 3A). Consistent with these findings, GAPDHS transcripts were also frequently found in RNA-seq from 22 melanoma PDXs in our laboratory (Fig. 3B).

Figure 3.

A novel isoform of GAPDHS is uniquely expressed in melanoma. A, Normalized RNA levels of the GAPDHS transcript among cancers in TCGA database. UVM, uveal melanoma. B, FPKM values of the GAPDHS transcript from the RNA-seq data of 22 melanoma PDXs. C, Exon annotation of the human GAPDHS gene located on chromosome 19. Shown are the known ATG-starting codon in exon 1 as well as an in-frame ATG in the latter part of exon 2. The proline-rich domain and NAD(P) binding domain span exons 1–4, whereas the catalytic domain is in exons 7–11. D, Exon counts within GAPDHS transcripts detected in the SKCM cohort of the TCGA database. E, Exon counts within GAPDHS transcripts detected in RNA-seq data of 22 PDXs. F, Western blot of GAPDHS protein in lysates from testis, PDXs, and melanoma cell lines. The dashed line separates two different exposures from the same blot. Arrows, bands corresponding to the full-length isoform and short isoform that is missing the N-terminal domain. Data in A and B, mean ± SD. Statistical significance in A was determined using Kruskal–Wallis tests comparing either SKCM or UVM with all other tumor types.

Figure 3.

A novel isoform of GAPDHS is uniquely expressed in melanoma. A, Normalized RNA levels of the GAPDHS transcript among cancers in TCGA database. UVM, uveal melanoma. B, FPKM values of the GAPDHS transcript from the RNA-seq data of 22 melanoma PDXs. C, Exon annotation of the human GAPDHS gene located on chromosome 19. Shown are the known ATG-starting codon in exon 1 as well as an in-frame ATG in the latter part of exon 2. The proline-rich domain and NAD(P) binding domain span exons 1–4, whereas the catalytic domain is in exons 7–11. D, Exon counts within GAPDHS transcripts detected in the SKCM cohort of the TCGA database. E, Exon counts within GAPDHS transcripts detected in RNA-seq data of 22 PDXs. F, Western blot of GAPDHS protein in lysates from testis, PDXs, and melanoma cell lines. The dashed line separates two different exposures from the same blot. Arrows, bands corresponding to the full-length isoform and short isoform that is missing the N-terminal domain. Data in A and B, mean ± SD. Statistical significance in A was determined using Kruskal–Wallis tests comparing either SKCM or UVM with all other tumor types.

Close modal

The first and second exons of the GAPDHS gene are known to encode a proline-rich N-terminal domain (Fig. 3C) thought to tether the enzyme to the fibrous sheath in spermatocytes (14). Interestingly, the RNA transcript sequences of GAPDHS present in TCGA and in our PDX melanomas (Fig. 3D and E) lacked reads corresponding to exon 1 and the first part of exon 2 of the gene. Instead, melanoma transcripts of GAPDHS began near an in-frame alternative start site within the second exon (Fig. 3C). No overlapping transcripts could be detected upstream to suggest alternative splicing (data not shown) and only proteins corresponding to the shorter, delta-N-terminal domain isoform of GAPDHS were detected in human melanomas by Western blot (Fig. 3F). Consistent with these findings, a similar-sized GAPDHS protein was previously described in melanoma cell lines, although the etiology of this size difference was not explored (18). In contrast to melanoma cells, we detected proteins corresponding to both the full-length and delta-N-terminal domain isoforms of GAPDHS in the testis (Fig. 3F). The protein corresponding to the shorter isoform detected in human testis appears to be regulated posttranscriptionally as there were no transcripts lacking exon 1 or 2 detected by RNA-seq analysis of 11 human spermatocyte samples (Supplementary Fig. S2B). Thus, melanoma cells exclusively express both a shorter GAPDHS transcript and corresponding GAPDHS protein.

GAPDHS knockdown enhances melanoma metastasis

Given that melanoma metastases had decreased glycolytic activity and GAPDHS levels compared with subcutaneous tumors, we hypothesized that downregulation of GAPDHS provided a functional advantage to metastatic efficiency. To test this, we generated three GAPDHS knockdown PDX melanomas (M481, M405, and UT10) using stable transduction with lentiviral shRNAs (Fig. 4AC). All of the melanomas were tagged with constitutive dsRed and luciferase allowing us to distinguish these cells from recipient mouse cells by flow cytometry and to track metastatic efficiency using bioluminescence imaging, respectively. Knockdown of GAPDHS did not affect the ability of melanomas to form subcutaneous tumors (data not shown) and did not generally change the growth or mitotic index of subcutaneous tumors (Fig. 4DF; Supplementary Fig. S3A) though knockdown of GAPDHS led to slightly earlier tumor engraftment in one PDX (Fig. 4F). Melanomas were allowed to spontaneously metastasize, and mice were analyzed when subcutaneous tumors reached 2.5 cm. Mice bearing melanomas with GAPDHS knockdown were more likely to have macrometastases and on average had significantly higher numbers of macrometastases compared with mice with control melanomas (Fig. 4GL). This was true for two melanomas that predominantly metastasize to the liver (M481 and M405) as well as a melanoma that predominantly metastasizes to the kidney (UT10).

Figure 4.

Low GAPDHS expression enhances metastatic efficiency in melanoma PDXs. M481, M405, and UT10 melanomas that express control hairpins (shSCR) or hairpins against GAPDHS (shA, shB, or shC) were injected subcutaneously into NSG mice. A–C, Western blots for GAPDHS and ACTIN from subcutaneous tumors with the indicated hairpins. D–F, Growth of subcutaneous tumors was measured over time. G–I, The percentage of mice with macrometastases at endpoint (left) and number of indicated organ macrometastases per mouse from mice bearing control or knockdown tumors (right). Mouse numbers are indicated on the corresponding axis labels. J–L, Representative photographs (top) and hematoxylin and eosin images (bottom) illustrating metastatic burden in livers (J and K) and kidneys (L) in mice bearing control or GAPDHS knockdown M481 (J), M405 (K), or UT10 (L) melanomas. Dashed lines outline individual metastases in hematoxylin and eosin image. Scale bar, 1 mm. All data represent mean ± SD. Statistical significance was determined using the nparLD (nonparametric analysis of longitudinal data) test (D–F), Fisher exact tests (G–I, left), and Kruskal–Wallis tests (G–I, right). *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; ****, P ≤ 0.001; ns, nonsignificant.

Figure 4.

Low GAPDHS expression enhances metastatic efficiency in melanoma PDXs. M481, M405, and UT10 melanomas that express control hairpins (shSCR) or hairpins against GAPDHS (shA, shB, or shC) were injected subcutaneously into NSG mice. A–C, Western blots for GAPDHS and ACTIN from subcutaneous tumors with the indicated hairpins. D–F, Growth of subcutaneous tumors was measured over time. G–I, The percentage of mice with macrometastases at endpoint (left) and number of indicated organ macrometastases per mouse from mice bearing control or knockdown tumors (right). Mouse numbers are indicated on the corresponding axis labels. J–L, Representative photographs (top) and hematoxylin and eosin images (bottom) illustrating metastatic burden in livers (J and K) and kidneys (L) in mice bearing control or GAPDHS knockdown M481 (J), M405 (K), or UT10 (L) melanomas. Dashed lines outline individual metastases in hematoxylin and eosin image. Scale bar, 1 mm. All data represent mean ± SD. Statistical significance was determined using the nparLD (nonparametric analysis of longitudinal data) test (D–F), Fisher exact tests (G–I, left), and Kruskal–Wallis tests (G–I, right). *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; ****, P ≤ 0.001; ns, nonsignificant.

Close modal

GAPDHS knockdown does not increase circulating tumor cells

Some metastatic suppressors function by limiting tumor cell intravasation or tumor cell survival in circulation. As downregulation of somatic GAPDH has been associated with enhanced buffering of ROS (26), we wanted to know whether inhibiting GAPDHS would decrease ROS and therefore increase the survival of circulating tumor cells. Interestingly, there was not a significant difference in the frequency of circulating tumor cells in the blood from mice bearing GAPDHS knockdown melanomas (Supplementary Fig. S3B–S3D), suggesting that their increased metastatic efficiency was not due to more tumor cells intravasating or surviving in the blood. To further support this, we injected an equal number of GAPDHS knockdown cells directly into the bloodstream via tail vein and measured metastasis over time using bioluminescence imaging. Mice injected with GAPDHS knockdown melanoma cells had detectable metastasis at earlier timepoints as compared with mice injected with control melanoma cells. Furthermore, the overall magnitude of the metastatic burden was increased in mice injected with GAPDHS knockdown cells (Supplementary Fig. S3E–S3G).

We measured ROS levels and found similar or slightly higher levels of ROS in GAPDHS knockdown melanoma cells as compared with controls (Supplementary Fig. S4A). This increase in ROS was not a reflection of higher mitochondrial potential, which was equal in both control and GAPDHS knockdown tumors (Supplementary Fig. S4B). To test whether the metastatic capacity of GAPDHS knockdown cells was secondary to increased mitochondrial ROS activity as previously described (22), we also assessed metastasis in the context of the mitochondrial antioxidant Mito-TEMPO. Mito-TEMPO did not impede the metastatic capacity of GAPDHS knockdown cells (Supplementary Fig. S4C–S4D) but rather enhanced it in a manner similar to control cells. Together, these data suggest that GAPDHS expression does not limit circulating melanoma cells or ROS buffering.

Ectopic expression of GAPDHS limits melanoma metastasis

Given that metastatic nodules have significant reductions in GAPDHS and that decreasing GAPDHS enhanced melanoma metastasis, we next tested whether increasing GAPDHS expression could limit metastasis. We transduced three different melanomas (M481, M405, and UT10) with lentiviral expression vectors encoding the melanoma-specific isoform of GAPDHS (Fig. 5AC) and transplanted them subcutaneously in mice. GAPDHS overexpression had little or no effect on the subcutaneous growth of most melanomas (Fig. 5DF) but significantly reduced the metastatic burden compared with the same melanomas transduced with control vectors in two of the three human melanomas (Fig. 5G and H). Mice transplanted with UT10 cells that overexpressed GAPDHS also had decreases in metastatic burden; however, the difference did not reach statistical significance (Fig. 5I). Circulating tumor cells were similar across all samples, once again suggesting that GAPDHS does not function by limiting the ability of tumor cells to enter the circulation or to survive there (Supplementary Fig. S5).

Figure 5.

High GAPDHS expression limits metastatic efficiency in melanoma PDXs. M481, M405, and UT10 melanomas that express control (pLOC) or GAPDHS overexpression vectors (GAPDHS) were injected subcutaneously into NSG mice. A–C, Western blots showing GAPDHS and ACTIN expression from subcutaneous tumors with the indicated vector. D–F, Growth of subcutaneous tumors over time. Number of mice per group is indicated on each graph. G–I, Percentage of mice with macrometastases (left) and number of organ macrometastases per mouse (right) at endpoint from mice bearing control or GAPDHS-overexpressing melanomas. The number of mice per group is indicated on the axes' labels. All data represent mean ± SD. Statistical significance was determined using nparLD (nonparametric analysis of longitudinal data). D–F, Fisher exact tests (G–I, left), and Mann–Whitney tests (G–I, right). *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; ns, nonsignificant.

Figure 5.

High GAPDHS expression limits metastatic efficiency in melanoma PDXs. M481, M405, and UT10 melanomas that express control (pLOC) or GAPDHS overexpression vectors (GAPDHS) were injected subcutaneously into NSG mice. A–C, Western blots showing GAPDHS and ACTIN expression from subcutaneous tumors with the indicated vector. D–F, Growth of subcutaneous tumors over time. Number of mice per group is indicated on each graph. G–I, Percentage of mice with macrometastases (left) and number of organ macrometastases per mouse (right) at endpoint from mice bearing control or GAPDHS-overexpressing melanomas. The number of mice per group is indicated on the axes' labels. All data represent mean ± SD. Statistical significance was determined using nparLD (nonparametric analysis of longitudinal data). D–F, Fisher exact tests (G–I, left), and Mann–Whitney tests (G–I, right). *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; ns, nonsignificant.

Close modal

The short form of GAPDHS is catalytically active in melanoma cells

GAPDHS catalyzes the conversion of glyceraldehyde 3-phosphate to 1,3-bisphosphate (16, 17, 27). To confirm that GAPDHS possesses enzymatic activity in melanoma cells, we overexpressed the short form of GAPDHS in two melanoma cell lines Igr1 and Mewo (Supplementary Fig. S6A) and performed a colorimetric assay to assess for GAPDH activity. In both cell lines, overexpression of GAPDHS led to a significant increase in total GAPDH activity (Supplementary Fig. S6B). To assess the enzymatic activity of GAPDHS in vivo, we performed metabolomics to examine levels of the substrate glyceraldehyde 3-phosphate in control or GAPDHS knockdown melanomas growing as subcutaneous tumors. Knockdown of GAPDHS led to a significant accumulation of glyceraldehyde 3-phosphate (consistent with decreased enzymatic activity) in all three melanomas we tested, although the difference did not reach statistical significance in one of the lines (Supplementary Fig. S6C). GAPDHS utilizes NAD+ as a cofactor to produce NADH. Therefore, we examined the impact of GAPDHS knockdown on the NAD+/NADH ratio. The NAD+/NADH ratio was significantly increased in two of the melanomas when blocked, also consistent with a decreased level of GAPDHS enzymatic activity (Supplementary Fig. S6D).

GAPDHS knockdown limits glycolysis and increases fumarate and malate levels through increased pyruvate carboxylase activity

We next examined whether the metabolic changes in glycolysis and the TCA cycle that we observed in metastatic nodules (Fig. 1BG) were dependent on the downregulation of GAPDHS. To test this, we first examined the relative abundance of glycolytic intermediates in control or GAPDHS knockdown melanoma cells cultured ex vivo. In GAPDHS knockdown melanoma cells, the relative abundance of several glycolytic intermediates was significantly decreased (Fig. 6A). Interestingly, the TCA cycle intermediates fumarate and malate were increased in GAPDHS knockdown cells (Fig. 6B), similar to our observations in metastatic nodules (Fig. 1E). To compensate for decreases in glycolysis, drug-resistant melanoma cells activate pyruvate carboxylase to replenish TCA cycle intermediates (28). We therefore tested whether GAPDHS was impacting pyruvate carboxylase activity in melanoma cells. We cultured UT10 cells where GAPDHS expression was depleted (shRNAs) or increased (GAPDHS) with [U-13C]-labeled glucose for 1 hour and examined 13C-labeling in the TCA cycle to determine pyruvate carboxylase activity (Supplementary Fig. S7A). GAPDHS knockdown cells (Fig. 6C) showed increased fractional enrichment of m+3 metabolites in the TCA cycle, including fumarate and malate, whereas GAPDHS-overexpressing cells (Fig. 6D) showed decreased fractional enrichment of m+3 fumarate and malate. Consistent with these findings, we found GAPDHS knockdown cells had significantly elevated ratios of citrate m+3 to pyruvate m+2 (Fig. 6E, left), whereas GAPDHS-overexpressing cells had significantly lower ratios (Fig. 6E, right), suggesting that GAPDHS influences anaplerosis and pyruvate carboxylase activity. Protein levels of pyruvate carboxylase enzyme were similar between control and GAPDHS knockdown melanoma cells, implying pyruvate carboxylase activity is regulated posttranslationally (Supplementary Fig. S7B). For example, acetyl-CoA, a known positive allosteric regulator of pyruvate carboxylase activity (29), was elevated in GAPDHS knockdown melanomas (Supplementary Fig. S7C).

Figure 6.

GAPDHS knockdown leads to decreased glycolysis and an increase in pyruvate carboxylase activity, with accumulation of downstream malate, fumarate, and aspartate. A and B, UT10 melanoma cells with GAPDHS knockdown (shA, shB), overexpression (GAPDHS), and corresponding control vectors (shSCR or pLOC) were cultured ex vivo, and the relative abundance of metabolites in glycolysis (A) and the TCA cycle (B) was normalized to TIC. Each sample was then normalized to relative abundance of its corresponding control vector. Samples were run in triplicate. C and D, UT10 melanoma cells were cultured for 1 hour with [U-13C]glucose as described in Materials and Methods. M+3 fractional enrichment was plotted for each metabolite of interest for GAPDHS knockdown (shA and shB; C) and GAPDHS overexpression (GAPDHS) cells (D). E, Pyruvate carboxylase activity is shown as measured by the ratio of m+3 citrate to m+3 pyruvate in the indicated samples. F, Relative abundance of aspartate in the indicated samples. G, Fractional enrichment of aspartate labeled with m+2 or m+3 in samples from C and D. H–J, Relative abundance of malate (H), fumarate (I), and aspartate (J) in control and GAPDHS knockdown M481, M405, and UT10 PDXs. All data are represented as mean ± SD. For each metabolite with two groups, unpaired t tests were performed. For metabolites with three or more groups, one-way ANOVA was performed with Dunnett correction for multiple comparisons. P values are indicated. *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; ****, P ≤ 0.001; ns, nonsignificant.

Figure 6.

GAPDHS knockdown leads to decreased glycolysis and an increase in pyruvate carboxylase activity, with accumulation of downstream malate, fumarate, and aspartate. A and B, UT10 melanoma cells with GAPDHS knockdown (shA, shB), overexpression (GAPDHS), and corresponding control vectors (shSCR or pLOC) were cultured ex vivo, and the relative abundance of metabolites in glycolysis (A) and the TCA cycle (B) was normalized to TIC. Each sample was then normalized to relative abundance of its corresponding control vector. Samples were run in triplicate. C and D, UT10 melanoma cells were cultured for 1 hour with [U-13C]glucose as described in Materials and Methods. M+3 fractional enrichment was plotted for each metabolite of interest for GAPDHS knockdown (shA and shB; C) and GAPDHS overexpression (GAPDHS) cells (D). E, Pyruvate carboxylase activity is shown as measured by the ratio of m+3 citrate to m+3 pyruvate in the indicated samples. F, Relative abundance of aspartate in the indicated samples. G, Fractional enrichment of aspartate labeled with m+2 or m+3 in samples from C and D. H–J, Relative abundance of malate (H), fumarate (I), and aspartate (J) in control and GAPDHS knockdown M481, M405, and UT10 PDXs. All data are represented as mean ± SD. For each metabolite with two groups, unpaired t tests were performed. For metabolites with three or more groups, one-way ANOVA was performed with Dunnett correction for multiple comparisons. P values are indicated. *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; ****, P ≤ 0.001; ns, nonsignificant.

Close modal

In some cancers where aspartate is limited (30, 31), tumor cells rely on pyruvate carboxylase activity to support aspartate biosynthesis (32). Interestingly, GAPDHS knockdown cells had higher levels of aspartate, whereas cells overexpressing GAPDHS had significantly lower levels of aspartate as compared with controls (Fig. 6F). This change appeared to be due to an alteration in aspartate synthesis rather than uptake because [U-13C]-glucose labeling into aspartate was increased in GAPDHS knockdown cells and diminished in GAPDHS overexpression cells (Fig. 6G). These results are consistent with the hypothesis that GAPDHS regulates pyruvate carboxylase activity, perhaps to support aspartate biosynthesis.

We also performed metabolomics on the same GAPDHS knockdown and overexpression lines growing as subcutaneous tumors in mice. Interestingly, in vivo, GAPDHS knockdown and overexpression PDXs had significantly higher and lower levels of citrate, respectively (Supplementary Fig. S8A–S8C). Consistent with our in vitro findings, GAPDHS expression led to diminished fumarate and malate (Supplementary Fig. S8D–S8I), whereas GAPDHS knockdown led to fumarate and malate accumulation in all three PDX tumors (Fig. 6H and I). Two of the GAPDHS knockdown PDXs also had a statistically significant increase in aspartate (Fig. 6J).

GAPDHS knockdown cells use glucose to fuel the TCA cycle

To confirm that the increases we saw in malate and fumarate in GAPDHS knockdown cells were indeed from glucose and not derived from an increase in other carbon sources, we also performed isotope tracing using lactate, glutamine, and fatty acids. As lactate has been shown to be a carbon source for the TCA cycle in efficiently metastasizing melanomas, we first examined [U-13C]-lactate isotope tracing in UT10 control (shSCR) and GAPDHS knockdown (shGAPDHS) melanoma cells ex vivo (Supplementary Fig. S9A). Both groups had similar labeling of lactate (m+3; Supplementary Fig. S9B), suggesting GAPDHS does not regulate lactate uptake. However, GAPDHS knockdown cells had diminished pyruvate (m+3) labeling and decreased m+2 labeling in the TCA cycle, suggesting decreased utilization of lactate (Supplementary Fig. S9B). Despite this lower labeling of pyruvate, GAPDHS knockdown melanomas had higher (m+3) labeling in citrate and TCA cycle intermediates (Supplementary Fig. S9C), consistent with increased pyruvate carboxylase activity (Supplementary Fig. S9D) as we previously saw with glucose isotope tracing.

We next performed isotope tracing with [U-13C]-labeled glutamine to examine whether GAPDHS knockdown might lead to increased glutamine incorporation into the TCA cycle (Supplementary Fig. S9E). Both control and GAPDHS knockdown melanomas had similar labeling of glutamine (m+5) and glutamate (m+5), suggesting similar uptake and utilization (Supplementary Fig. S9F). GAPDHS knockdown cells had a slight but significant increase in labeled alpha-ketoglutarate, consistent with a more anaplerotic state (Supplementary Fig. S9F). However, GAPDHS knockdown cells did not have increased labeling in the forward or reverse direction of the TCA cycle (Supplementary Fig. S9F and S9G), suggesting that these cells are not using more glutamine to fuel the TCA cycle. In fact, GAPDHS knockdown cells had less (m+3) incorporation (Supplementary Fig. S9G), consistent with a larger proportion of this coming from other carbon sources, like glucose.

Lastly, we performed [U-13C]-palmitic acid tracing to test whether GAPDHS regulates lipid uptake and incorporation into the TCA cycle (Supplementary Fig. S9H). GAPDHS knockdown cells had strikingly less incorporation of 13C-label in downstream acetyl-CoA and TCA cycle pools (Supplementary Fig. S9I), suggesting that GAPDHS knockdown cells do not rely on lipid uptake as a source for their increased fumarate and malate. To confirm this more broadly, we also treated control and GAPDHS knockdown cells with etomoxir, an inhibitor of CPT1 and fatty acid oxidation. We chose two doses of etomoxir (3 and 30 μmol/L), which are both well below concentrations reported to have off-target effects on complex I (33). We confirmed that these doses did indeed inhibit fatty acid entry into mitochondria through isotope tracing of [U-13C]-palmitic acid, where both doses of etomoxir completely eliminated 13C-labeling of acetyl-CoA (Supplementary Fig. S9J). Importantly, GAPDHS knockdown melanoma cells continued to have increased malate and fumarate levels in the presence of both doses of etomoxir (Supplementary Fig. S9K and S9L), confirming that accumulation is independent of fatty acid uptake and incorporation into the TCA cycle.

Thus, we have shown a novel isoform of GAPDHS is uniquely expressed in melanoma cells, which serves to maintain glycolysis and limit metastasis. Loss of GAPDHS leads to metabolic rewiring of central carbon metabolism, with increased pyruvate carboxylase activity and accumulation of citrate, fumarate, malate, and aspartate, which increased melanoma metastasis (Fig. 7).

Figure 7.

GAPDHS limits melanoma metastasis by fueling glycolysis at the expense of carbon flux to the TCA cycle and aspartate synthesis. In primary melanomas or with overexpression of GAPDHS (left), GAPDHS activity maintains high rates of glycolysis and lactate production, whereas pyruvate carboxylase activity and TCA cycle activity remain low. In melanoma metastases or with knockdown of GAPDHS (right), glycolytic activity is reduced, favoring elevated pyruvate carboxylase activity and increases in the TCA cycle intermediates citrate, fumarate, malate, and downstream aspartate.

Figure 7.

GAPDHS limits melanoma metastasis by fueling glycolysis at the expense of carbon flux to the TCA cycle and aspartate synthesis. In primary melanomas or with overexpression of GAPDHS (left), GAPDHS activity maintains high rates of glycolysis and lactate production, whereas pyruvate carboxylase activity and TCA cycle activity remain low. In melanoma metastases or with knockdown of GAPDHS (right), glycolytic activity is reduced, favoring elevated pyruvate carboxylase activity and increases in the TCA cycle intermediates citrate, fumarate, malate, and downstream aspartate.

Close modal

Our data suggest that melanoma cells from individual patients retain a transcriptional signature whereby <1% to 4% of their transcriptome is significantly altered during the course of metastasis (Supplementary Fig. S1A). Alterations were enriched in transcriptional pathways associated with glycolysis, central carbon metabolism, and other metabolic pathways (Fig. 1A). The transcriptional changes we observed were accompanied by metabolic changes including a decrease in glycolysis and an increase in the TCA cycle within metastatic nodules. One potential confounding variable in the PDX model system is that primary subcutaneous tumors are inherently larger (2–2.5 cm) than their corresponding metastases. To minimize the impact of this, we obtained metastatic nodules of a variety of sizes and from a range of distant organs, with the goal of identifying metabolic features that might be generalizable across metastases. Although metastatic tumors are undoubtedly influenced by oxygenation status, vascularity, and their local microenvironment, we think our data suggest there may be at least some generalizable metabolic features that are critical in the establishment of metastases.

Our findings suggest that the metabolic switch seen in metastatic nodules is at least in part mediated through the actions of GAPDHS and pyruvate carboxylase activity. The presence of GAPDHS correlated with high glycolysis, low levels of the TCA metabolites malate and fumarate, and low metastatic potential, whereas suppression of GAPDHS resulted in decreases in glycolysis, increases in malate and fumarate, and increases in metastasis. These changes phenocopy the differences observed in subcutaneous tumors and metastatic nodules (Fig. 1DG).

GAPDHS knockdown or overexpression did not affect circulating tumor cell frequency in the blood (Supplementary Figs. S3B–S3D and S5A–S5C) or ROS levels (Supplementary Fig. S4A). As circulating tumor cells are of extremely low frequency, it is possible that a larger sample size would be required to detect significant alterations between groups. However, our data set thus far suggest that GAPDHS is not critical for intravasation or survival in the blood. The metabolism of metastasizing cells is dynamic, and successfully metastasizing cells must be able to adapt to the changing microenvironments they encounter at each stage of the process (34). We hypothesize that the transcriptional and metabolic shift from glycolysis to the TCA cycle seen in PDX melanoma metastases is likely a reflection of these traits having a selective advantage in early metastatic tumor establishment. It is worth noting that although we frequently confirmed GAPDHS overexpression and knockdown in primary tumors of our cohorts, we did not routinely examine GAPDHS expression in the metastatic nodules where tissue is more limited. It would be interesting to see if metastatic nodules in GAPDHS-overexpressing tumors represent “escape clones” that managed to silence their GAPDHS overexpression or, alternatively, were able to compensate for its presence with further metabolic or molecular adaptations. Future work will be needed to determine the exact step in the late metastatic cascade (e.g., extravasation vs. establishment) in which GAPDHS loss is most critical.

Our work supports the conclusion that like somatic GAPDH, the short isoform of GAPDHS catalyzes the conversion of glyceraldehyde 3-phosphate to 1,3-bisphosphate in melanoma cells. Serendipitously for our studies, GAPDHS enzymology studies have largely focused on the short isoform (rather than the full-length spermatocyte isoform) because the presence of the N-terminal domain prevented protein purification due to solubility issues. In the work by three independent groups (16, 17, 35), all found the short isoform of GAPDHS to be catalytically active. It is worth noting that unlike spermatocytes, melanoma express somatic GAPDH in addition to GAPDHS. The presence of somatic GAPDH, in addition to the substrate accumulation of GAP and the cofactor NAD+ (both known to increase enzymatic activity) that we see in GAPDHS knockdown cells, is likely what allow for continued flow of glucose through glycolysis albeit at lower levels.

Although some distinguishing features of GAPDHS have been identified such as its increased catalytic activity (17) and NAD binding cooperativity (36) compared with somatic GAPDH, a thorough metabolic and biochemical study needs to be performed to understand the unique influences of GAPDH and GAPDHS on melanoma cells. Given that GAPDHS and GAPDH share only 70% sequence homology, there are many potential sites that could allow for alternative regulation of both protein stability and protein function. These differences may be important in understanding the unique contributions of these two similar enzymes within melanoma cells.

Interestingly, GAPDHS had a striking impact on select TCA metabolite levels and pyruvate carboxylase activity (Fig. 6; Supplementary Figs. S8–S9). Although increased pyruvate carboxylase activity itself has been associated with increased metastasis (37), additional downstream TCA metabolites and derivatives have also been implicated in increasing metastatic capacity. Citrate, fumarate, and malate, the metabolites with the most consistent increases in GAPDHS knockdown cells, have been documented in a number of prometastatic contexts (38–40), including EMT induction (41). Importantly, isotope tracing studies reveal that GAPDHS knockdown cells have increased synthesis of aspartate, a metabolite known to be critical for cancer cell survival, especially in nutrient- and oxygen-deprived environments (30) such as those seen when a metastasizing tumor establishes itself outside the circulation in a distant metastatic site. This ability may be what allows GAPDHS knockdown melanoma cells to efficiently establish metastatic nodules in a range of distant organs. Future studies are needed to determine whether the impact of GAPDHS on metastasis is due entirely to its impact on metabolism or whether it may also be impacting other cellular functions in “moonlighting capacities” as seen in somatic GAPDH.

Lastly, it is intriguing to think about what might be regulating GAPDHS expression and why melanoma cells (and not any other type of cancers in TCGA) uniquely express this otherwise spermatogenic-specific enzyme. The fact that we see discrepancies in the transcript and protein levels suggests that GAPDHS has significant posttranscriptional regulation, which has also been noted in spermatocytes (16). Future work in our lab is aimed at understanding the fundamentals of this question and determining whether this enzyme may offer melanoma cells a selective advantage in certain contexts. Despite being a metastatic suppressor, a review of large data sets containing melanoma sequences shows that genomic deletion or loss-of-function mutations in GAPDHS are virtually absent. Although this enzyme offers a selective disadvantage during early metastatic dissemination, it may be important in earlier or later stages of proliferation and anabolic growth.

J.G. Gill reports grants from Cancer Prevention and Research Institute of Texas, Dermatology Foundation, and NIH during the conduct of the study. S. Muh reports grants from Cancer Prevention and Research Institute of Texas during the conduct of the study. T. Mathews reports grants from Cancer Prevention and Research Institute of Texas during the conduct of the study. A.B. Aurora reports grants from Cancer Prevention and Research Institute of Texas during the conduct of the study. No disclosures were reported by the other authors.

J.G. Gill: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. S.N. Leef: Validation, investigation. V. Ramesh: Validation, investigation. M.S. Martin-Sandoval: Formal analysis, methodology. A.D. Rao: Conceptualization, writing–review and editing. L. West: Investigation. S. Muh: Formal analysis. W. Gu: Formal analysis. Z. Zhao: Data curation, formal analysis, visualization. G.A. Hosler: Formal analysis, investigation. T.W. Vandergriff: Formal analysis, investigation. A.B. Durham: Resources. T.P. Mathews: Formal analysis, methodology. A.B. Aurora: Resources, funding acquisition, investigation, writing–original draft, writing–review and editing.

The research was supported by the Cancer Prevention and Research Institute of Texas (MIRA RP180778 to J.G. Gill, S.N. Leef, V. Ramesh, M.S. Martin-Sandoval, Z. Zhao, and T.P. Mathews, and A.B. Aurora) and the NIH (UL1TR002240 to A.B. Durham). J.G. Gill was supported by the Dermatology Foundation and the NIH (T32 AR065969). A.D. Rao was supported by the Victorian Cancer Agency. A.B. Durham was supported by the NIH Clinical and Translational Science Awards Program (UL1TR002240). The authors thank S.J. Morrison for the generous use of his laboratory's resources as well as helpful conversations and guidance. They thank M. Nitcher for mouse colony management as well as N. Loof and the Moody Foundation Flow Cytometry Facility. The authors thank A. Tasdogan and B. Faubert for assistance in metabolic protocols. They thank V. Khivansara for help in RNA-seq library preparation and E. Piskounova for help in sorting cells for RNA-seq preparation. They thank Nisha Meireles and Sherry Fu for database and data management of the University of Michigan melanoma tumors. The authors would also like to thank J. Xu and the CRI Sequencing Facility; R.J. DeBerardinis, P. Mishra, A. DeVilbiss, A. Solmonson, Lauren Zacharias, and the CRI Metabolomics Facility.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Piskounova
E
,
Agathocleous
M
,
Murphy
MM
,
Hu
Z
,
Huddlestun
SE
,
Zhao
Z
, et al
.
Oxidative stress inhibits distant metastasis by human melanoma cells
.
Nature
2015
;
527
:
186
91
.
2.
Lehuede
C
,
Dupuy
F
,
Rabinovitch
R
,
Jones
RG
,
Siegel
PM
.
Metabolic plasticity as a determinant of tumor growth and metastasis
.
Cancer Res
2016
;
76
:
5201
8
.
3.
Altenberg
B
,
Greulich
KO
.
Genes of glycolysis are ubiquitously overexpressed in 24 cancer classes
.
Genomics
2004
;
84
:
1014
20
.
4.
Lunt
SY
,
Vander Heiden
MG
.
Aerobic glycolysis: meeting the metabolic requirements of cell proliferation
.
Annu Rev Cell Dev Biol
2011
;
27
:
441
64
.
5.
Park
JS
,
Burckhardt
CJ
,
Lazcano
R
,
Solis
LM
,
Isogai
T
,
Li
L
, et al
.
Mechanical regulation of glycolysis via cytoskeleton architecture
.
Nature
2020
;
578
:
621
6
.
6.
Davis
RT
,
Blake
K
,
Ma
D
,
Gabra
MBI
,
Hernandez
GA
,
Phung
AT
, et al
.
Transcriptional diversity and bioenergetic shift in human breast cancer metastasis revealed by single-cell RNA sequencing
.
Nat Cell Biol
2020
;
22
:
310
20
.
7.
Wiel
C
,
Le Gal
K
,
Ibrahim
MX
,
Jahangir
CA
,
Kashif
M
,
Yao
H
, et al
.
BACH1 stabilization by antioxidants stimulates lung cancer metastasis
.
Cell
2019
;
178
:
330
45
.
8.
van Weverwijk
A
,
Koundouros
N
,
Iravani
M
,
Ashenden
M
,
Gao
Q
,
Poulogiannis
G
, et al
.
Metabolic adaptability in metastatic breast cancer by AKR1B10-dependent balancing of glycolysis and fatty acid oxidation
.
Nat Commun
2019
;
10
:
2698
.
9.
Sirover
MA
.
On the functional diversity of glyceraldehyde-3-phosphate dehydrogenase: biochemical mechanisms and regulatory control
.
Biochim Biophys Acta
2011
;
1810
:
741
51
.
10.
Ramos
D
,
Pellin-Carcelen
A
,
Agusti
J
,
Murgui
A
,
Jorda
E
,
Pellin
A
, et al
.
Deregulation of glyceraldehyde-3-phosphate dehydrogenase expression during tumor progression of human cutaneous melanoma
.
Anticancer Res
2015
;
35
:
439
44
.
11.
Mondragon
L
,
Mhaidly
R
,
De Donatis
GM
,
Tosolini
M
,
Dao
P
,
Martin
AR
, et al
.
GAPDH overexpression in the T cell lineage promotes angioimmunoblastic T cell lymphoma through an NF-kappaB-dependent mechanism
.
Cancer Cell
2019
;
36
:
268
87
.
12.
Chiche
J
,
Reverso-Meinietti
J
,
Mouchotte
A
,
Rubio-Patino
C
,
Mhaidly
R
,
Villa
E
, et al
.
GAPDH expression predicts the response to R-CHOP, the tumor metabolic status, and the response of DLBCL patients to metabolic inhibitors
.
Cell Metab
2019
;
29
:
1243
57
.
13.
Welch
JE
,
Schatte
EC
,
O'Brien
DA
,
Eddy
EM
.
Expression of a glyceraldehyde 3-phosphate dehydrogenase gene specific to mouse spermatogenic cells
.
Biol Reprod
1992
;
46
:
869
78
.
14.
Welch
JE
,
Brown
PL
,
O'Brien
DA
,
Magyar
PL
,
Bunch
DO
,
Mori
C
, et al
.
Human glyceraldehyde 3-phosphate dehydrogenase-2 gene is expressed specifically in spermatogenic cells
.
J Androl
2000
;
21
:
328
38
.
15.
Miki
K
,
Qu
W
,
Goulding
EH
,
Willis
WD
,
Bunch
DO
,
Strader
LF
, et al
.
Glyceraldehyde 3-phosphate dehydrogenase-S, a sperm-specific glycolytic enzyme, is required for sperm motility and male fertility
.
Proc Natl Acad Sci U S A
2004
;
101
:
16501
6
.
16.
Elkina
YL
,
Kuravsky
ML
,
El'darov
MA
,
Stogov
SV
,
Muronetz
VI
,
Schmalhausen
EV
.
Recombinant human sperm-specific glyceraldehyde-3-phosphate dehydrogenase: structural basis for enhanced stability
.
Biochim Biophys Acta
2010
;
1804
:
2207
12
.
17.
Chaikuad
A
,
Shafqat
N
,
Al-Mokhtar
R
,
Cameron
G
,
Clarke
AR
,
Brady
RL
, et al
.
Structure and kinetic characterization of human sperm-specific glyceraldehyde-3-phosphate dehydrogenase, GAPDS
.
Biochem J
2011
;
435
:
401
9
.
18.
Sevostyanova
IA
,
Kulikova
KV
,
Kuravsky
ML
,
Schmalhausen
EV
,
Muronetz
VI
.
Sperm-specific glyceraldehyde-3-phosphate dehydrogenase is expressed in melanoma cells
.
Biochem Biophys Res Commun
2012
;
427
:
649
53
.
19.
Quintana
E
,
Piskounova
E
,
Shackleton
M
,
Weinberg
D
,
Eskiocak
U
,
Fullen
DR
, et al
.
Human melanoma metastasis in NSG mice correlates with clinical outcome in patients
.
Sci Transl Med
2012
;
4
:
159ra49
.
20.
Jodar
M
,
Sendler
E
,
Moskovtsev
SI
,
Librach
CL
,
Goodrich
R
,
Swanson
S
, et al
.
Absence of sperm RNA elements correlates with idiopathic male infertility
.
Sci Transl Med
2015
;
7
:
295re6
.
21.
Tasdogan
A
,
Faubert
B
,
Ramesh
V
,
Ubellacker
JM
,
Shen
B
,
Solmonson
A
, et al
.
Metabolic heterogeneity confers differences in melanoma metastatic potential
.
Nature
2020
;
577
:
115
20
.
22.
Porporato
PE
,
Payen
VL
,
Perez-Escuredo
J
,
De Saedeleer
CJ
,
Danhier
P
,
Copetti
T
, et al
.
A mitochondrial switch promotes tumor metastasis
.
Cell Rep
2014
;
8
:
754
66
.
23.
Faubert
B
,
Li
KY
,
Cai
L
,
Hensley
CT
,
Kim
J
,
Zacharias
LG
, et al
.
Lactate metabolism in human lung tumors
.
Cell
2017
;
171
:
358
71
.
24.
Su
X
,
Lu
W
,
Rabinowitz
JD
.
Metabolite spectral accuracy on orbitraps
.
Anal Chem
2017
;
89
:
5940
8
.
25.
Haq
R
,
Shoag
J
,
Andreu-Perez
P
,
Yokoyama
S
,
Edelman
H
,
Rowe
GC
, et al
.
Oncogenic BRAF regulates oxidative metabolism via PGC1alpha and MITF
.
Cancer Cell
2013
;
23
:
302
15
.
26.
Kuehne
A
,
Emmert
H
,
Soehle
J
,
Winnefeld
M
,
Fischer
F
,
Wenck
H
, et al
.
Acute activation of oxidative pentose phosphate pathway as first-line response to oxidative stress in human skin cells
.
Mol Cell
2015
;
59
:
359
71
.
27.
Lamson
DR
,
House
AJ
,
Danshina
PV
,
Sexton
JZ
,
Sanyang
K
,
O'Brien
DA
, et al
.
Recombinant human sperm-specific glyceraldehyde-3-phosphate dehydrogenase (GAPDHS) is expressed at high yield as an active homotetramer in baculovirus-infected insect cells
.
Protein Expr Purif
2011
;
75
:
104
13
.
28.
Delgado-Goni
T
,
Galobart
TC
,
Wantuch
S
,
Normantaite
D
,
Leach
MO
,
Whittaker
SR
, et al
.
Increased inflammatory lipid metabolism and anaplerotic mitochondrial activation follow acquired resistance to vemurafenib in BRAF-mutant melanoma cells
.
Br J Cancer
2020
;
122
:
72
81
.
29.
Jitrapakdee
S
,
St Maurice
M
,
Rayment
I
,
Cleland
WW
,
Wallace
JC
,
Attwood
PV
.
Structure, mechanism and regulation of pyruvate carboxylase
.
Biochem J
2008
;
413
:
369
87
.
30.
Garcia-Bermudez
J
,
Baudrier
L
,
La
K
,
Zhu
XG
,
Fidelin
J
,
Sviderskiy
VO
, et al
.
Aspartate is a limiting metabolite for cancer cell proliferation under hypoxia and in tumours
.
Nat Cell Biol
2018
;
20
:
775
81
.
31.
Sullivan
LB
,
Gui
DY
,
Hosios
AM
,
Bush
LN
,
Freinkman
E
,
Vander Heiden
MG
.
Supporting aspartate biosynthesis is an essential function of respiration in proliferating cells
.
Cell
2015
;
162
:
552
63
.
32.
Cardaci
S
,
Zheng
L
,
MacKay
G
,
van den Broek
NJ
,
MacKenzie
ED
,
Nixon
C
, et al
.
Pyruvate carboxylation enables growth of SDH-deficient cells by supporting aspartate biosynthesis
.
Nat Cell Biol
2015
;
17
:
1317
26
.
33.
Raud
B
,
Roy
DG
,
Divakaruni
AS
,
Tarasenko
TN
,
Franke
R
,
Ma
EH
, et al
.
Etomoxir actions on regulatory and memory T cells are independent of Cpt1a-mediated fatty acid oxidation
.
Cell Metab
2018
;
28
:
504
15
.
34.
Bergers
G
,
Fendt
SM
.
The metabolism of cancer cells during metastasis
.
Nat Rev Cancer
2021
;
21
:
162
80
.
35.
Sexton
JZ
,
Danshina
PV
,
Lamson
DR
,
Hughes
M
,
House
AJ
,
Yeh
LA
, et al
.
Development and implementation of a high throughput screen for the human sperm-specific isoform of glyceraldehyde 3-phosphate dehydrogenase (GAPDHS)
.
Curr Chem Genomics
2011
;
5
:
30
41
.
36.
Kuravsky
ML
,
Barinova
KV
,
Asryants
RA
,
Schmalhausen
EV
,
Muronetz
VI
.
Structural basis for the NAD binding cooperativity and catalytic characteristics of sperm-specific glyceraldehyde-3-phosphate dehydrogenase
.
Biochimie
2015
;
115
:
28
34
.
37.
Christen
S
,
Lorendeau
D
,
Schmieder
R
,
Broekaert
D
,
Metzger
K
,
Veys
K
, et al
.
Breast cancer-derived lung metastases show increased pyruvate carboxylase-dependent anaplerosis
.
Cell Rep
2016
;
17
:
837
48
.
38.
Lu
X
,
Bennet
B
,
Mu
E
,
Rabinowitz
J
,
Kang
Y
.
Metabolomic changes accompanying transformation and acquisition of metastatic potential in a syngeneic mouse mammary tumor model
.
J Biol Chem
2010
;
285
:
9317
21
.
39.
Peng
M
,
Yang
D
,
Hou
Y
,
Liu
S
,
Zhao
M
,
Qin
Y
, et al
.
Intracellular citrate accumulation by oxidized ATM-mediated metabolism reprogramming via PFKP and CS enhances hypoxic breast cancer cell invasion and metastasis
.
Cell Death Dis
2019
;
10
:
228
.
40.
Wei
Q
,
Qian
Y
,
Yu
J
,
Wong
CC
.
Metabolic rewiring in the promotion of cancer metastasis: mechanisms and therapeutic implications
.
Oncogene
2020
;
39
:
6139
56
.
41.
Sciacovelli
M
,
Goncalves
E
,
Johnson
TI
,
Zecchini
VR
,
da Costa
AS
,
Gaude
E
, et al
.
Fumarate is an epigenetic modifier that elicits epithelial-to-mesenchymal transition
.
Nature
2016
;
537
:
544
7
.