Highly proliferative tumors consume glucose and amino acids, including glutamine, to sustain the metabolic demands that support biomass production [1
]. Glutamine is a conditionally essential amino acid that can be obtained from the microenvironment or synthesized de novo. Glutaminolysis is the process whereby glutamine is converted to glutamate and then to α-ketoglutarate, which enters the tricarboxylic acid (TCA) cycle. The TCA cycle supports oxidative phosphorylation and energy generation and provides a carbon source for fatty acid synthesis, a nitrogen source for synthesis of amino acids and nucleotides, and intermediates necessary for the synthesis of reduced glutathione (GSH), which neutralizes reactive oxygen species [2
In mammals, including humans, two enzymes, GLS and GLS2, catalyze the conversion of glutamine to glutamate. GLS and GLS2 have distinct patterns of expression and regulation in different organs and tumor types [3
]. GLS is ubiquitously expressed and is also highly expressed in several types of tumors as a result of direct regulation by oncogenes such as KRAS and MYC [4
]. GLS2 is expressed mostly in the liver, brain, and pancreas and is directly regulated by p53, p63, and p73 [6
]. In glioblastoma and cancers of liver and colon, GLS2 expression is lost due to DNA methylation of the GLS2
Many tumors depend on glutamine for growth, and glutamine addiction is associated with GLS [4
]. Pharmacological agents that target GLS directly, such as CB-839, are currently in clinical trials [12
]. Although glutamine addiction has been observed in many cancers, recent studies employing three-dimensional organoid cultures and in vivo models using fluorinated glutamine have demonstrated that not all tumor types metabolize glutamine [13
]. The observed glutamine independence of some tumors could confer resistance to glutaminase inhibitors [14
]. The contribution of GLS2 to glutamine dependence in these tumors has not been examined.
Considerable evidence suggests that the epithelial to mesenchymal transition (EMT) program contributes to the development of therapy resistance and metastasis [15
]. We have previously demonstrated that EMT promotes acquisition of stem-cell properties by cancer cells [25
]. In this study, we found that the induction of EMT results in the suppression of GLS2
expression and the promotion of glutamine independence even in low-glucose conditions and in the presence of GLS. In addition, we observed that GLS2 re-expression enhanced glutamine consumption and reduced sphere formation. The transcription factor FOXC2 is critical to maintaining mesenchymal and stem-cell properties [27
] and has been shown to direct metabolic activities in adipocytes [29
]. We found that inhibition of FOXC2 expression (and thus inhibition of EMT) also restored GLS2 expression and glutamine dependency in cells that had undergone EMT. We evaluated GLS2
expression in breast cancer patients and found that, in line with our data, high GLS2
expression is inversely correlated with the EMT gene signature. Further, we found that GLS2
copy number deletions were over-represented in the basal breast cancer subtype; a subtype with poor clinical outcomes and high metastatic potential [39
]. In support of the idea that tumor cells with high GLS2 expression have less aggressive characteristics, we found that high GLS2
expression correlates with improved overall survival in breast cancer patients.
We previously identified changes in metabolite abundances, particularly in glutamine and glutamate, that occur with the onset of EMT [43
]. Here we demonstrated a striking difference in the expression of two glutaminases, GLS and GLS2, in cells that have and have not undergone EMT. Furthermore, in breast tumors the presence of an EMT signature positively correlates with GLS
expression and negatively correlates with GLS2
expression. The highly metastatic basal subtype is enriched with GLS
amplifications and GLS2
deletions, but in less aggressive subtypes such as luminal A and luminal B, there are more frequent GLS2
amplifications and GLS
deletions. Previous studies have described the opposing roles that GLS and GLS2 have in cancer [44
]. For example, the inverse correlation of GLS
has been described in the glioblastoma cell line T98G [46
]. Although both GLS and GLS2 catalyze the conversion of glutamine to glutamate, their localization, expression, and regulation are distinct [3
transcription is directly induced by oncogenes KRAS and MYC, and up-regulation of GLS contributes to glutamine addiction in primary tumors [5
We sought to understand how the distinct expression patterns of these two glutaminases might metabolically benefit the EMT. We metabolically characterized cells that were induced to undergo EMT and assessed their abilities to utilize glucose and glutamine compared to their epithelial counterparts. We observed that cells that have undergone EMT were viable in low glucose conditions, even in the absence of glutamine, unlike epithelial control cells. We reason that this metabolic flexibility enables cancer cells to survive during the metastatic cascade where they encounter conditions of fluctuating nutrient availability. It has been previously demonstrated that, depending on the substrate used for energy synthesis, some cancer cells acquire metabolic flexibility upon transformation [49
]. However, metabolic rigidity can also arise, as in glutamine-dependent cancer cells, and can enhance tumorigenic potential [4
]. In addition to glutamine independence, we observed that the cells that had undergone EMT had lower oxygen consumption, lower levels of ATP production, and reduced levels of GSH indicative of reduced mitochondrial activity. Previous work demonstrated that GLS2 mediates mitochondrial activity and GSH, both of which impact tumorigenic potential [6
We hypothesized that the loss of GLS2 expression upon the induction of EMT is responsible for the glutamine-independent phenotype and reduced mitochondrial activity. To test this, we over-expressed GLS2 in the GLS2-negative and EMT-enriched breast cancer cell line SUM159. We found that GLS2 over-expression led to an enhancement of mitochondrial activity, glutamine consumption, and GSH and ATP production even in the presence of GLS. Furthermore, this enhanced mitochondrial activity mediated by GLS2 led to a decrease in secreted lactate and reduced the frequency of mammosphere formation. Notably, these effects occurred without altering the expression of EMT markers implying that the metabolic program alone can significantly impact EMT-induced stem-cell properties.
We have previously demonstrated strong evidence that the EMT program imparts stem-cell properties to cancer cells [25
]. EMT is reversible: upon reaching a distant organ, cancer cells turn off the EMT program to proliferate and establish a secondary tumor [53
]. Our findings suggest that epithelial-like cancer cells can depend on both glucose and glutamine at the primary site, but upon the activation of the EMT, cancer cells not only acquire migratory and invasive properties but also become glutamine independent. It is reasonable to speculate that the inverse regulation of the two glutaminases provides the metabolic support necessary for the evolving energetic demands during the EMT. Although we observed a reduction in mammosphere formation upon over-expression of GLS2, we did not observe a reduction in expression of EMT transcription factors FOXC2
, or Snail
. In future studies it would be valuable to test if inhibiting GLS while over-expressing GLS2 confers a more robust attenuation of EMT properties or tumorigenic potential.
We, and others, have established the functional link between EMT and FOXC2 [15
]. Based on these studies, we tested if inhibiting FOXC2 (and thus EMT) would impact GLS2 expression and have metabolic consequences. Indeed, we found that inhibition of FOXC2 expression of activity restored GLS2 expression, mitochondrial activity, and glutamine utilization. Interestingly, GLS expression was not altered by FOXC2 inhibition. These data suggest that the role of GLS and GLS2 is not redundant and that their orchestrated expression provides support for the EMT. In our models, GLS2 expression and EMT features are inversely correlated; therefore, we tested whether GLS2 status was predictive of outcome in breast cancer patients. We found that high levels of GLS2
are associated with improved survival over 5 years (p
= 0.019); however, GLS
levels were not correlated with survival (p
= 0.045). These findings indicate that the inverse relationship between GLS2
may have a stronger predictive value for outcomes in breast cancer than the correlation between GLS2
and warrants further validation.
Considerable evidence links the EMT program with the development of resistance to chemotherapy and tumor relapse [54
]. For example, triple negative breast cancer tumors with mesenchymal properties are highly resistant to chemotherapies [55
]. Our data reveal, for the first time, a GLS2-dependent glutamine metabolic pathway that is influenced by the EMT. Tumors enriched with aggressive and mesenchymal features also exhibit glutamine-independent metabolism suggests that targeting pathways that inhibit the EMT may impact susceptibility to glutaminase inhibitors and is a hypothesis that warrants investigation. Overall, our findings demonstrate that GLS2 has an important role in glutamine metabolism and that re-expression of GLS2 and its downstream metabolism is critical to metastasis.
4. Materials and Methods
4.1. Cell Culture
All cell lines were cultured at 37 °C in 5% CO2. All cell lines derived from HMLE cells were cultured in MEGM (Lonza CC-3051)/DMEM F12 (Mediatech, Wembley, WA, Australia; 10090CV) (1:1) supplemented with penicillin and streptomycin (Gibco/Life Technologies), insulin (Sigma, St. Louis, MO I9278, USA), hEGF (Sigma 9644), hydrocortisone (Sigma H0888), and bovine pituitary extract. MCF7 and MDA-MB-231 cells were cultured in DMEM/F12 (Fisher, Waltham, MA, USA; 10-090-cv) supplemented with 10% fetal bovine serum (FBS) (Sigma), and penicillin and streptomycin. SUM159 cells were cultured in Ham’s F12 (Corning 10-090-cv) medium containing 10% FBS and penicillin and streptomycin. HEK293T cell lines were cultured in DMEM (Corning, Corning, NY, USA; 10-013-CV) supplemented with 10% FBS and penicillin and streptomycin. MCF10A cells were cultured in DMEM/F12 (Corning 10-090-CV) supplemented with 5% horse serum (Sigma H1138), 20 ng/mL human epidermal growth factor (Sigma E9644), 0.5 µg/ml hydrocortisone, 100 ng/mL cholera toxin (Sigma C-8052), 5 µg/mL insulin, and penicillin and streptomycin.
4.2. Biolog Assay
Cells were plated at 1 × 104 cells per well into 96-well PM-M2 plates (Biolog, Hayward, CA, USA 13102). Cells were cultured in 50 µL of Biolog M1 medium (Biolog 72301) supplemented with penicillin and streptomycin, 200 mM glutamine, and 5% FBS. After 24 h, 10 µL of Biolog MA redox dye (Biolog 74351) was added to each well, and the plate was placed in the Biolog OmniLog2.3 plate reader to obtain kinetic data at 15 min intervals for 24 h. Data were analyzed with Biolog software packages PMM_Kinetic and PMM_Parametric to obtain the average of the initial rate values at 8 h. Each cell line was tested in triplicate.
4.3. Glucose Sensitivity Assay
Cells were plated at 3000 cells per well in 96-well plates in growth medium. After 24 h, medium was removed by aspiration, and cells were washed with PBS. Medium was replenished with 100 µL of glucose and glutamine-free DMEM containing 12 mM, 3 mM, 1.5 mM, 0.8 mM, 0.4 mM, or 0 mM of glucose (Sigma). Each concentration was evaluated in four replicates. After 24 h, 20 µL of CellTiter 96 Aqueous One Solution (Promega, Madison, WI, USA) was added to each well. After incubation at 37 °C for 60 min, absorbance at 490 nm was determined. Raw data were converted to relative percent viabilities, transformed to log, and the IC50 was calculated using GraphPad Prism software.
4.4. Seahorse Assays
For the MitoStress test, cells were seeded at 15,000 cells per well in an XF96 cell culture plate with the appropriate medium. The XF96 probes were calibrated with 200 µL of calibrant solution and incubated at 37 °C in the absence of CO2. After 12 h, the medium was changed to Seahorse base medium (Seahorse Biosciences, Santa Clara, CA, USA) supplemented with 10 mM glucose, 2 mM glutamine, and 1 mM sodium pyruvate and adjusted to pH 7.4. The probe cartridge was prepared to have 0.5 µM oligomycin in port “A”, 0.25 µM FCCP in port “B”, and 0.25 µM Rotenone with 0.25 µM Antimycin in port “C”. For the glycolysis stress test, the probe cartridge was prepared to have 20 µL of 50 mM glucose in port “A”, 11 mM oligomycin in port “B”, and 650 mM 2-deoxyglucose in port “C”. ECAR was normalized to mpH/min. Oxygen consumption rate and extracellular acidification rate were measured by the Seahorse XFe96 Analyzer (Seahorse Biosciences).
4.5. ATP Determination
Cells were plated in triplicate at 1 × 104 per well in a 96-well plate. After 24 h, an ATP standard curve was prepared and 100 µL of CellTiter-Glo reagent was added to each test sample and standard. After incubating at room temperature for 10 min, the luminescent signal was recorded with a luminometer.
4.6. Glutamate Determination
Samples were analyzed with the Glutamine/Glutamate Determination Kit (Sigma GLN-1) following the manufacturer’s recommendations. Briefly, 5 × 105 cells were plated in triplicate in 6-well plates. After 24 h, growth medium was removed by aspiration, and cells were washed with PBS. Cells were lysed with 250 µL of RIPA buffer. A glutamate standard curve was prepared in a 96-well plate. Test wells contained 25 µL of sample, 175 µL of reaction buffer, and 2 µL of L-GLDH enzyme. The absorbance at 340 nm was measured at time 0 and in 40 min increments until the signal plateaued.
4.7. GSH Determination
Cells were plated at 10,000 cells per well in 6 replicates in a 96-well plate. After 24 h, reduced glutathione was assayed with GSH-Glo Glutathione Assay (Promega V6911). Briefly, the growth medium was removed from each well, and cells were washed with PBS. A GSH standard curve was prepared, and cells were incubated with GSH-Glo reagent. After 30 min, Luciferin detection reagent was added to each well and incubated for 15 min. The luminescence was measured using a luminometer.
4.8. MitoTracker Red
Cells were plated on glass coverslips at 50% confluency. After 24 h cells were stained with 20 nM MitoTracker Red CMXRos (Life Technologies) under the same culture conditions. After 30 min, cells were fixed with 4% paraformaldehyde. After washing with PBS, coverslips were treated with 5% glycine in PBS for 15 min. Actin was stained with 488 Phalloidin (Alexa Fluor, Waltham, MA, USA), and the nuclei were stained with DAPI. Fluorescent images were obtained with a Zeiss fluorescent microscope. The fluorescence intensities of 100 cells were analyzed and quantified by ImageJ software to obtain the relative fluorescence intensity (RFI).
4.9. 13C-Glutamine Flux
U-13C-Glutamine (17 mg) was added to 50 mL of HMLE medium and filtered. HMLER cells expressing FOXC2 or HMLER control cells were seeded into wells of 10 cm plates at 70% confluency; each cell type was analyzed in four replicates. After 24 h, medium was aspirated, and 5 mL of labeled glutamine solution was added to the cells. A sample was collected immediately after addition. After 6 h, medium was removed by aspiration. Cells were washed with ice-cold PBS and 1 mL of methanol/water (1:1) was added to the cells. Cells were scraped into 15 mL Falcon tubes and immediately flash frozen by submerging tubes into liquid nitrogen. Samples were stored at −80 °C until metabolite extraction. Cell pellets were thawed at 4 °C and subjected to freeze–thaw cycles in liquid nitrogen and ice three times to rupture the cell membrane. Following this, 750 µL of ice-cold methanol/water (4:1) containing 20 µL of internal standards were added to each cell extract. Next, ice-cold chloroform/water (3:1) was added. Organic and aqueous layers were separated. The aqueous portion was deproteinized using a 3-KDa filter (Amicon Ultracel-3K membrane, Millipore Corporation, Burlington, MA, USA), and the filtrate containing metabolites was dried under vacuum (Genevac EZ-2plus, SP Scientific, Warminster, PA, USA). Prior to mass spectrometry, the dried extracts were re-suspended in 100 µL methanol/water (1:1) containing 0.1% formic acid and analyzed by liquid chromatography-mass spectrometry (LC/MS).
Liquid chromatography was performed using an Agilent 1290 Series HPLC system equipped with a degasser, binary pump, thermostatted auto sampler, and column oven (all from Agilent Technologies, Sana Clara, CA, USA). All samples were kept at 4 °C. A 5 µL aliquot of sample was delivered to a 4.6 mm i.d. × 10 cm Amide XBridge HILIC column (Waters) at 300 µL/min. Chromatography was performed using 0.1% formic acid in water (solvent A) and 0.1% formic acid in acetonitrile (solvent B). Gradients were run starting from 85% solvent B to 30% B from 0–3 min, 30% B to 2% B from 3–12 min, 2% B was held from 12–15 min, 2% B to 85% B from 15–23 min, and 85% B was held for 7 min to re-equilibrate the column.
Samples were analyzed by a 6495 QQQ triple quadrupole mass spectrometer (Agilent Technologies). Metabolites were measured using positive ionization mode with an electrospray ionization (ESI) voltage of 3000 eV. Approximately 9–12 data points were acquired per detected metabolite. Selected reaction monitoring (SRM) was used to determine 13C incorporation into glutamine by measuring the expected precursor/product ion pairs: M + 0: 147.08/130.1, M + 1: 148.08/131.1, M + 2: 149.08/132.1, M + 3: 150.08/133.1, M + 4: 151.08/134.1, and M + 5 152.08/135.1. SRM was used to determine 13C incorporation into GSH by measuring the expected precursor/product ion pairs: M + 0: 308.0911/76, M + 1: 309.0911/76, M + 2: 310.0911/76, M + 3: 311.0911/76, M + 4: 312.0911/76, and M + 5: 313.0911/76. Mass isotopomer distribution was calculated and corrected for natural abundance.
4.10. Steady-State Mass Spectrometry
For profiling of metabolites associated with glycolysis and TCA pathways, 5 × 106 cells were evaluated in triplicate. Cells were collected using trypsin. All cell pellets were stored at −80 °C until analysis. Cell pellets were thawed at 4 °C and subjected to freeze–thaw cycles in liquid nitrogen and ice three times to rupture the cell membrane. Following this, 750 µL of ice-cold methanol/water (4:1) containing 20 µL of internal standards were added to each cell extract. Next, ice-cold chloroform/water (3:1) was added. Organic and aqueous layers were separated. The aqueous portion was deproteinized using a 3-kDa filter (Amicon Ultracel-3K membrane, Millipore Corporation), and the filtrate containing metabolites was dried under vacuum (Genevac EZ-2plus). Prior to mass spectrometry, the dried extracts were re-suspended in 100 µL methanol/water (1:1) containing 0.1% formic acid and analyzed by LC/MS.
Liquid chromatography was performed using an Agilent 1290 Series HPLC system equipped with a degasser, binary pump, thermostatted auto sampler, and column oven (all from Agilent Technologies). All samples were kept at 4 °C. An aliquot of 5 µL of sample was analyzed. Samples were delivered into the spectrometer via normal phase chromatographic separation using a Luna Amino (NH2) 4 µm, 100 Å, 2.1 × 150 mm column (Phenominex, Torrance, CA, USA) at 200 µL/min. Chromatography was performed using 5 mM ammonium acetate in water at pH 9.9 (solvent A) and 100% acetonitrile (solvent B). Gradients were run starting from 80% solvent B to 2% B over a 20 min period followed by 2% B to 80% B for a 5 min period, further followed by 80% B for a 13 min period. The flow rate was gradually increased during the separation from 200 µL/min (0–20 min) to 300 µL/min (20.1–25 min) to 350 µL/min (25–30 min) to 400 µL/min (30–37.99 min) and was then returned to 200 µL/min (5 min). Samples were analyzed using a 6495 QQQ triple quadrupole mass spectrometer using SRM. Metabolites associated with glycolysis and TCA pathways were measured using negative ionization mode with ESI voltage of −3500 eV. Approximately 9–12 data points were acquired per detected metabolite.
4.11. RNA Extraction and qRT-PCR
Total RNA was isolated using the RNeasy Plus kit (Qiagen, Germantown, MD, USA) according to the manufacturer’s instructions. RNA was quantified with a Thermo Scientific Nanodrop 2000c instrument. RNA sample purity was verified based on the ratio of absorbance at 260 nm to that at 280 nm. RNA samples were normalized to 1 µg. qRT-PCR experiments were run in triplicate, and the mean was used for the determination of mRNA levels. Relative quantification of the mRNA levels was performed using the comparative Ct method with human RPLPO as the reference gene and with the formula 2−ΔΔCt. The following primer pairs were used: forward RPLPO 5′-GCGACCTGGAAGTCCAACTA-3′ and reverse RPLPO 5′-ATCTGCTTGGAGCCCACAT-3′, forward GLS2 5′-ACACCCTCAGCCTCATGCAT-3′ and reverse GLS2 5′-ATGGCTCCTGATACAGCTGACTT-3′, forward GLS 5′-CACTGCCCTCCCATTACCTAG-3′ and reverse GLS 5′-GAAGCTCAAGCATGGGAACAG-3′, forward FOXC2 5′-GCCTAAGGACCTGGTGAAGC-3′ and reverse FOXC2 5′-TTGACGAAGCACTCGTTGAG-3′, and forward ZEB1 5′-GCACAACCAAGTGCAGAAGA-3′ and reverse ZEB1 5′-CATTTGCAGATTGAGGCTGA-3′.
Cells were plated at 50% confluency on glass coverslips in 6-well plates. After 24 h, the medium was removed aspiration, and cells were washed with PBS. Cells were fixed in 4% paraformaldehyde at room temperature for 20 min and then washed with PBS. Cells were quenched with 4% glycine for 15 min at room temperature and washed with PBS. Cells were blocked with 5% BSA, 0.1% Triton-X100 in PBS for 1 h at room temperature. After three washes with PBS, cells were incubated with (1:200) primary GLS2 (LGA) antibody in 1% BSA for 2 h at room temperature. Cells were washed three times with PBS and then incubated with a secondary antibody at 1:1000 in 1% BSA for 1 h at room temperature in the dark. Cells were washed with PBS and nuclear stain DAPI was added to the cells at 1:1000. After 5 min in the dark. Cells were washed three times with PBS, and coverslips were mounted on glass slides with DAKO mounting solution and incubated at 4 °C overnight to dry. Fluorescent images of the slides were taken with the same exposure time for each slide. The anti-GLS2 was generously provided by Dr. José M. Matés and Dr. Javier Márquez, University of Málaga, and was generated as previously described [18
4.13. Plasmids and Viral Transduction
HMLER-derived cell lines that express Snail and FOXC2, were generated as described previously [42
]. The S367E and S367A FOXC2
mutants were generated as described previously [28
]. The HMLE-ZEB1, HMLER-ZEB1, HMLE-Snail/shFOXC2/pZEB1, HMLER-Snail/shFOXC2/pZEB1, and SUM159-shFOXC2/pZEB1 cell lines were made with a lentiviral FUW-2A-mStrawberry ZEB1 expression construct (a gift from Li Ma, MD Anderson, 1515 Holcombe Blvd, Houston, TX 77030, UA) and selected with 4 µg/mL puromycin. For stable GLS2 expression, a GLS2 cDNA expression vector (purchased from MD Anderson core) was subcloned into the pHAGE-EF1alpha-PURO vector (a gift from Kenneth Scott, Baylor College of Medicine, Houston, TX, USA) using the Gateway system (Invitrogen) and selected with 4 µg/mL puromycin. pGIPZ-based shRNAs designed to deplete cells of GLS2 and ZEB1 were purchased from MD Anderson shRNA core.
4.14. Breast Cancer Data Analysis
We had previously assembled the EMT-74 data set, a panel of 74 samples across six data sets representing the gene expression profiles of cells before and after an EMT [56
]. We examined the gene expression of GLS
in the EMT-74 panel samples and compared the expression in the epithelial or mesenchymal state using a Student’s t
-test. As controls, we evaluated expression of an epithelial marker (CDH1
), mesenchymal markers (CDH2
), and a housekeeping gene (GAPDH
4.15. Copy Number Analysis
The copy numbers and RNA-seq data for 1075 patients in TCGA was downloaded from the Firehose of the Broad Institute (http://gdac.broadinstitute.org/
, January 2016 version). For copy number and RNA expression association analysis, the LinkedOmics portal (http.linkedomics.org
) was used. A threshold for copy number data was set by the GISTIC2 algorithm. Primary solid tumors were used for analysis. Statistical analysis was performed using the computing environment R (version 3.4). The one-way analysis of variance (ANOVA) was performed using aov function and p
-value was obtained. Loss of one or two copies of a gene was considered as a deletion whereas gain in one or two copies of gene was considered as amplification.
4.16. L-Lactate Measurement
In a 96 well plate, 1 × 104 cells were plated per well in triplicate. After 24 h, the spent medium was diluted 1:50, and lactate was measured with EnzyChrom L-Lactate Assay Kit (BioAssay Systems, Hayward, CA, USA).
4.17. Proliferation Assay
Cells were plated in triplicate at 1 × 104 cells per well in 6 well-plates. Aliquots stained with Trypan Blue were counted manually every 24 h for 72 h.
4.18. Mammosphere Assay
Mammosphere culture medium was made with 1% methylcellulose and MEGM without bovine pituitary extract. In a 96-well low attachment plate, 500 cells per well in six replicates were cultured for 14 days. Every 3 days, 50 µL of fresh media was added. Mammospheres were photographed, and spheres with a diameter greater than 80 µm were counted.
4.19. Pan-Cancer Correlation and Survival Analysis
The correlation between expression of GLS2
and EMT score was performed across pan-cancer using the Spearman correlation. The normalized RNA expression data for TCGA samples was downloaded from Broad Firehose release 28 January 2016 run (Firehose, 28 January, 2016 #11383) ([57
). Primary tumor samples but not normal tissue or metastases samples were selected for analysis. The correlation of the RSEM-normalized RNA expression values between genes in 33 TCGA cancer types was determined using the correlation function cor.test in R (Bioconductor, version 3.4). The EMT score was calculated by single sample gene set enrichment analysis (ssGSEA) using normalized RNA expression of each cancer type and Hallmark gene set (MSigDB database version 6.2) by GSVA package (version 1.3).
Survival analysis was carried out using vital status and overall survival (days from diagnosis to last follow-up) for the TCGA pancancer patients. Only patients with 5 years of follow-up were selected and used for the analysis. Cox proportional-hazards model implemented in survival (v2.44) package in R was used for the survival analysis and Cox coefficient was calculated for each cancer type. Visualization was performed using ggplot2 (v3.2.1) and survminer (v0.4.6) packages in R.
4.20. Gene Expression Analysis
Gene expression analysis was performed on a previously published dataset as previously described in [43
]. Analysis was performed using the ‘limma’ and ‘affy’ packages in R. All p
-values were adjusted for multiple testing using the Benjamini Hochberg method, significance was set at an adjusted p
-value < 0.001. Metabolic enzymes and transporters were identified using a previously published comprehensive gene set [58
4.21. Statistical Analysis
Unless otherwise stated, all samples were assayed in triplicate. All in vitro experiments were repeated at least three independent times, and the animal study included 10 mice in each group. Unless otherwise indicated, data are presented as means ± standard deviation (SD), and significance was calculated using the Student’s unpaired two-tailed t-test.
4.22. Western Blot Analysis and Antibodies
Cell pellets were obtained by trypsinization. Cell pellets were lyzed with RIPA buffer (Sigma R0278) containing phosphatase inhibitor (Roche, Indianapolis, IN, USA; 4906837001) and protease inhibitor (Roche 11697498001) and vortexed vigorously then incubated on ice for 30 minutes, spun down, and cell debris removed. After quantification by Bradford assay, 50 µg of protein was denatured in β-mercaptoethanol at 100 °C for 10 minutes. Samples were loaded onto a 10% SDS-agarose gel and run at 110 V. For blots, antibodies were used at 1:1000 in 5% milk in TBST. Antibodies used were anti-Snail (Cell Signalling, Danvers, MA, USA; 3879S; 1:1000 in BSA), anti-FOXC2 (Bethyl Laboratories, Montgomery, TX, USA; A302–383A; 1:1000 in 5% milk), anti-ZEB1 (Santa Cruz sc-25388; 1:1000 in 5% milk), anti-actin (Santa Cruz, Dallas, TX, USA; sc-1616-R; 1:1000 in 5% milk), and anti-tubulin (Cell Signaling 2144s; 1:1000 in BSA). To quantify the protein band intensity, densitometry analyses were performed by evaluating band intensity of mean gray value using ImageJ software [59
]. In brief, the mean gray rectangular areas of the same size were measured in each band. The background was measured above each band. The pixel intensity was inverted for all values by calculating 255 minus the band mean gray value. The background was subtracted and then the protein:loading control ratio was calculated for all samples.