|Home | About | Journals | Submit | Contact Us | Français|
Melanoma is the deadliest form of commonly encountered skin cancers because of its rapid progression towards metastasis1,2. Although metabolic reprogramming is tightly associated with tumor progression, whether metabolic regulatory circuits affect metastatic processes is poorly understood. PGC1α is a transcriptional coactivator that promotes mitochondrial biogenesis, protects against oxidative stress3, and reprograms melanoma metabolism to influence drug sensitivity and survival4,5. Here, we provide data to indicate that PGC1α suppresses melanoma metastasis, acting through a pathway distinct from its bioenergetic functions. Elevated PGC1α expression inversely correlates with vertical growth in human melanomas. PGC1α silencing makes nonmetastatic melanoma cells highly invasive and conversely, PGC1α reconstitution suppresses metastasis. Within populations of melanoma cells, there is a marked heterogeneity in PGC1α levels, which predicts their inherent high or low metastatic capacity. Mechanistically, PGC1α directly increases transcription of ID2, which in turn binds to and inactivates the transcription factor TCF4. Inactive TCF4 causes downregulation of metastasis-related genes including integrins that are known to influence invasion and metastasis6–8. Inhibition of BRAFV600E using vemurafenib9, independently of its cytostatic effects, suppresses metastasis by acting on the PGC1α-ID2-TCF4-integrin axis. Taken together, our findings reveal that PGC1α maintains mitochondrial energetic metabolism and suppresses metastasis through direct regulation of parallel acting transcriptional programs. Consequently, components of these circuits define new therapeutic opportunities that may help curb melanoma metastasis.
Whereas the landscape of genetic alterations and multiple driver mutations have been discovered in melanoma10,11, less is understood about the genes driving metastasis2. Nevertheless, it is thought that efficient metastasis requires the malignant cell to balance proliferation with invasion and migration12,13. We recently found that elevated expression of the metabolic integrator and transcriptional coactivator peroxisome proliferator-activated receptor-gamma coactivator-1α (PGC1α or PPARGC1A) defines a subset of melanomas wherein it promotes mitochondrial metabolism, protects against oxidative stress, and enhances survival4,5. Although high PGC1α expression is associated with worse prognosis in metastatic melanomas4,5, reduced levels coincide with invasive/vertical growth in primary specimens (Fig. 1a). Therefore, we investigated the effects of PGC1α in invasion and metastasis. Gene set enrichment analysis (GSEA) of expression data (GSE36879)4 upon PGC1α knockdown in the non-metastatic melanoma cell line A375P revealed a coordinated upregulation of genes implicated in metastasis, including genes that control focal adhesion or extracellular matrix (ECM) interactions, integrins and components of the transforming growth factor-β (TGFβ) and Wnt signaling (Extended Data Fig. 1a, b)14–17. In addition, PGC1α expression showed an inverse correlation with gene sets involved in melanoma metastasis (Extended Data Fig. 1c, d). Upregulation of pro-metastatic genes following PGC1α suppression was confirmed by qPCR in PGC1α-positive melanoma cell lines (Fig. 1b, Extended Data Fig. 2a–d). Conversely, the increase in integrin transcripts was reverted upon ectopic PGC1α expression (Extended Data Fig. 2e, f). Targeting PGC1α using the CRISPR/Cas9 system led to similar gene expression changes (Extended Data Fig. 2g).
Consistent with lower PGC1α expression during vertical growth and acquisition of the metastatic phenotype, the PGC1α-suppressed invasive and metastatic gene signature was associated with worse primary melanoma survival (Fig. 1c). Changes in integrin expression upon PGC1α depletion were accompanied by activation of the downstream kinase focal adhesion kinase (FAK)18 (Fig. 1d), and increases in migration and invasion (Fig. 1e). FAK inhibition blocked the enhanced migration induced by PGC1α depletion (Extended Data Fig. 2h–j). Remarkably, PGC1α silencing by either shRNA (Fig. 1f, Extended Data Fig. 3a, b) or CRISPR/Cas9 (Fig. 1g) converted these low-invasive, PGC1α-positive cells into highly metastatic entities as assessed by tail-vein injection experiments. To fully recapitulate the metastatic process in vivo, we used MeWo cells in an orthotropic metastasis model19 and found that, again, PGC1α suppression caused the subcutaneously implanted tumors to generate systemic disease (Fig. 1h). Conversely, reconstitution of PGC1α in the PGC1α-negative cells A375 and A2058 decreased integrin expression (Fig. 1i, Extended Data Fig. 3c, d) and compromised their invasiveness in vitro and in vivo (Fig. 1j, k). Together, these results indicate that PGC1α inhibits a pro-metastatic program in melanoma cells resulting in the suppression of invasion and metastasis.
Melanomas are highly heterogeneous and might switch between a proliferative and an invasive/metastatic phenotype13. Using MitoTracker to label mitochondria and FACS analysis, we found that melanoma cell lines with heightened PGC1α expression displayed heterogeneous mitochondrial mass (Fig. 2a), which was dynamically regulated (Extended Data Fig. 4a), and hence could implicate alternate mitochondrial biogenesis and PGC1α function during phenotype switching. The sorted mitochondria-high (mito/PGC1α-high) population expressed significantly higher PGC1α and mitochondrial components, as well as lower integrins compared to the mitochondria-low (mito/PGC1α-low) population (Fig. 2a, b). The PGC1α-low population showed enhanced migration in vitro and metastasis in vivo (Fig. 2c, d). Compared to the non-migrating population, melanoma cells that had migrated through the trans-well membrane expressed lower amounts of PGC1α and higher pro-metastatic transcripts (Fig. 2e, Extended Data Fig. 4b). To strengthen the link between PGC1α and metabolic heterogeneity with metastatic spread, we isolated circulating tumor cells (CTCs) from the blood of mice bearing subcutaneous PGC1α-positive MeWo tumors (Extended Data Fig. 4c). Notably, these CTCs exhibited lower levels of PGC1α, but elevated integrins compared to the primary tumors (Fig 2f). However, in the corresponding lung metastases, which had formed from CTCs, PGC1α transcripts increased to similar levels as in the primary tumors (Fig. 2g). Demonstrating that increases in PGC1α in established metastases confer growth advantages similar to primary melanomas4,20, restoration of PGC1α in lung metastases derived from PGC1α knockdown cells enhanced tumor progression (Fig. 2h, Extended Data Fig. 4d). In aggregate, these results indicate that melanoma cells display heterogeneous levels of PGC1α and mitochondria. The mito/PGC1α-low population expresses a pro-metastatic gene program, while the mito/PGC1α-high population drives a proliferation phenotype.
To assess the mechanisms by which PGC1α suppresses this pro-metastatic program, we surveyed genes upregulated upon PGC1α suppression for potential negative transcriptional regulators. We identified two Inhibitor of DNA binding (ID) proteins—ID2 and ID3, among the top differentially expressed genes. Levels of ID2 and ID3, but not ID1 or ID4, were reduced in melanoma cells upon PGC1α knockdown and increased by PGC1α (Extended Data Fig. 5a–c). Chromatin immunoprecipitation (ChIP) revealed that PGC1α was bound at the ID2 promoter, suggesting direct transcriptional regulation (Extended Data Fig. 5d). Next, we depleted ID2 and ID3 in PGC1α-positive melanoma cells and found that suppression of ID2, but not ID3, increased integrin expression and downstream signaling (Fig. 3a, Extended Data Fig. 5e–i). Similar to PGC1α knockdown, ID2 depletion also strongly promoted migration, invasion and lung metastasis (Fig. 3b, c, Extended Data Fig. 5j). To test whether ID2 mediates the repressive effect of PGC1α, we ectopically expressed ID2 in PGC1α-depleted cells (Extended Data Fig. 6a). ID2 expression suppressed the induction of the prometastatic programs, invasion and metastasis enforced by PGC1α depletion (Fig. 3d–f, Extended Data Fig. 6b, c). Similar results were observed when ID2 was ectopically expressed in PGC1α-negative melanoma cells (Extended Data Fig. 6d–f). However, depletion of ID2, in contrast to PGC1α, did not alter glucose metabolism (Extended Data Fig. 6g). Taken together, these data indicate that the ID2 inhibitor is a downstream target of PGC1α suppressing pro-metastatic transcriptional programs without affecting PGC1α metabolic function.
ID2 functions as a transcriptional inhibitor through direct heterodimerization with basic helix-loop-helix (bHLH) factors, blocking binding to promoters21–23. To find bHLH factor(s) that could regulate integrin expression and metastasis driven by PGC1α suppression, we surveyed two different protein-protein interaction databases. BioGRID24 displayed 34 unique ID2 interactors including the bHLH transcription factors TCF3, TCF4, MyoD and TCF12 (Extended Data Fig. 7a). STRING25 revealed three bHLH transcription factors (MYC, TCF3, TCF4) in the top 10 predicted partners (Extended Data Fig. 7b). Among these factors, only suppression of TCF4 (Transcription factor 4) was able to consistently decrease integrin expression in both A375P-shPGC1α and PGC1α-negative cells (Extended Data Fig. 7c–e). Knockdown of TCF4 prevented the induction of integrins and FAK phosphorylation upon PGC1α or ID2 suppression (Fig. 3g, Extended Data Fig. 8a, b). Co-immunoprecipitation showed that TCF4 binds to ID2 in A375P cells (Fig. 3h). Consistent with this interaction, while the recruitment of TCF4 to promoters of integrins was increased upon PGC1α or ID2 knockdown, ectopic expression of ID2 blunted TCF4 recruitment (Extended Data Fig. 8c). Ectopic expression of TCF4 was sufficient to induce integrin expression and signaling (Fig. 3i, Extended Data Fig. 8d), concordant with TCF4 recruitment to their promoters (Fig. 3j). TCF4 knockdown abrogated the enhanced migration and metastasis of cells with PGC1α or ID2 suppression (Fig. 3k, l), and in PGC1α-negative cells (Extended Data Fig. 8e). Notably, expression of PGC1α and TCF4 in cell lines and tumors was mutually exclusive (Extended Data Fig. 8f, g), further supporting the opposite link between PGC1α and TCF4. Similar to ID2, manipulation of TCF4 levels did not alter cellular metabolism (Extended Data Fig. 8h), indicating that PGC1α’s effects on metastasis is separable from its metabolic functions. Collectively, TCF4 is required for the pro-metastatic transcriptional program upon PGC1α suppression leading to increased invasion and metastasis.
PLX4032, a BRAFV600E inhibitor, has been shown to increase PGC1α expression in melanoma cells harboring this mutation5,9,26. Based on our results described here, PLX4032 could inhibit metastasis by acting on the PGC1α transcriptional axis. Treatment of BRAFV600E melanoma cells with PLX4032 or MEK inhibitors highly induced PGC1α and ID2 expression (Fig. 4a), and reduced levels of most integrins tested (Fig. 4b, c, Extended Data Fig. 9a). Consistently, PLX4032 increased the recruitment of PGC1α to the ID2 promoter (Extended Data Fig. 9b) and strongly induced the interaction between ID2 and TCF4 (Fig. 4d), decreasing TCF4 promoter occupancy at four integrin genes (Fig. 4e). Despite FAK activation, measured as phospho-Y397-FAK levels, was slightly inhibited upon MAPK blockage (Fig. 4c), PLX4032 was able to repress melanoma invasion in vitro and metastasis in vivo, which was largely reversed by PGC1α or ID2 depletion (Fig. 4f, g, Extended Data Fig. 9c). Within the time frame of the in vitro assay (24 hours), PLX4032 did not decrease cell growth (Extended Data Fig. 9d), and the dose of PLX4032 used in mice (1 mg/kg) was lower than the dose used to induce tumor regression27. Taken together, these data indicate that PLX4032 can suppress invasion and metastasis, independently of its cytotoxic effects. PLX4032-induced metastasis inhibition is mediated, at least in part, through transcriptional activation of PGC1α.
Our results overall indicate that the metabolic transcriptional coactivator PGC1α is an apical regulator of melanoma progression through protection against oxidative stress which confers survival and proliferative advantages4,5,20, and suppression of cell motility, cell-cell interaction, adhesion and invasion that promotes metastatic drive. Strikingly, PGC1α expression inversely correlates with invasive growth in local disease; while in metastatic melanomas, it associates with worse outcomes. Although PGC1α defines a subset of melanomas with specific characteristics, its heterogeneous expression within tumors allows different proliferative or invasive abilities (Fig. 4h). We argue that the heterogeneity of PGC1α levels within melanomas reflects a dualistic nature of PGC1α function—promoting growth and survival of tumors, while suppressing metastatic spread. This heterogeneity might be important during melanoma progression through PGC1α changes in response to different signals including nutrients, and switching between survival/proliferation and invasion/metastasis. From a therapeutic standpoint, independent of the cytotoxic/cytostatic effects, our results extend the clinical benefits of the BRAFV600E-targeted drugs to impact metastasis. For melanoma treatment, BRAFV600E-inhibitors may have heightened therapeutic benefits if applied at an earlier stage by inducing PGC1α and reducing metastatic propensity. Moreover, selection of cells with lower PGC1α may promote metastasis, such as within BRAFV600E inhibitor-treated RAS-mutant melanomas or BRAFV600E inhibitor-resistant melanomas28, as these samples reactivate ERK/MEK signaling and reduce PGC1α expression. Finally, targeting the components downstream of PGC1α that drive metastasis could provide new therapeutic opportunities for melanoma and other malignancies such as prostate cancer29.
PLX4032, PD98059, AZD6244 and PF-573228 were purchased from Selleck Chemicals. The siRNAs against TCF4 (sc-61657), c-Myc (sc-29226), TCF3 (sc-36618) or TCF12 (sc-35552) were purchased from Santa Cruz Biotechnology. Antibodies against ITGA4, ITGA5, ITGB1, ITGB3, ITGB4, ITGB5, FAK, c-Myc, TCF12, and Porin were purchased from Cell Signaling Technology; p-FAK (Y397) antibodies were from Cell Signaling and Thermo Fisher Scientific, ID2 antibodies were from Cell Signaling, Santa Cruz Biotechnology and Thermo Fisher; V5, HMB45, COX5A, NDUFS4 and NDUFA9 antibodies were from Abcam; TCF4 antibodies were from Abnova and Santa Cruz; and PGC1α antibodies were from Santa Cruz and Millipore.
The GSEA software v2.0 (http://www.broadinstitute.org/gsea)30 was used to perform the GSEA analysis. In all the analysis, the KEGG gene sets were used. The values of the 219195_at probe (corresponding to PPARGC1A) were used as phenotype. For the analysis of the CCLE dataset, the gene expression data was downloaded from the CCLE portal (wwww.broadinstitute.org/CCLE) and the data from 61 melanoma cell lines was used for the analysis. The GSEA default parameters were used with the exception that Pearson correlation was computed to rank the genes for the analysis of the CCLE data and permutation was changed to gene set for the analysis of the GSE36879 dataset.
Published datasets GSE318931 and GSE1239132 with associated pathological stages for each sample as Invasive/Vertical or Superficial/Radial were analyzed for relative deviation from median-normalized PGC1α intensities (linear) within each dataset (significance based on 2-sample, 2-sided t-test statistics). To examine the enrichment of the PGC1α-regulated metastasis/invasion signature genes (ITGA3, ITGA4, ITGA10, ITGB3, ITGB5, CAV1, CAV2, ACTN2, LAMA4A, COL4A1, INHBA, TGFBI, TGFBR3, TGFBR2, SMAD3, SMAD7, IL8, IL11, LEF1, TCF7L2, DKK3, PPP3CA and SFRP1), we performed ssGSEA projections33 to yield a percentile-based Normalized Enrichment Score within each of GSE3189 and GSE12391, which were used to combine the datasets (2-sample, 2-sided t-test statistics). The association between primary melanoma survival and PGC1α-regulated metastasis/invasion signature was based on ssGSEA for signature closeness within GSE57715 and calculation of log-rank survival.
For the analysis of PGC1α and TCF4 gene expression, data obtained from the TCGA skin cutaneous melanoma dataset34 consisting of 471 samples with RNAseq data was downloaded from the cBioPortal35,36 (www.cbioportal.org). Data were represented as Z-scores of RNAseq V2 RSM. The dotted lines denote Z-scores of 0. Samples were classified as expressing if Z-score >0 and a mutually exclusivity report from the cBioPortal was generated.
pDONR223-LacZ entry control vector was purchased from Addgene (25893) and pLX304-LacZ control vector was generated using LR clonase II (Invitrogen). V5 tagged pLX304-ID2 and -TCF4 vectors were kindly provided from Marc Vidal Lab at Dana-Farber Cancer Institute. Luciferase-expressing FUW-Luc was kindly provided by Andrew Kung (Dana-Farber Cancer Institute) and pMSCV-Luciferase-hygro plasmid was purchased from Addgene. Full-length PGC1α was amplified by KOD polymerase (F: GCTTGGGACATGTGCAGCGAA and R: TTACCTGCGCAAGCTTCTCTGAGC), and then PCR product was ligated into pDONR223 by BP reaction. PGC1α expressing destination vectors (pLX304 for constitutive expression and pInducer2037 for doxycycline-inducible expression) were generated by LR reaction with entry vector (pDONR223-PGC1α).
Melanoma cells were obtained from ATCC and their authentication was confirmed by either DNA fingerprinting with small tandem repeat (STR) profiling or in-house PCR testing of melanoma marker genes and BRAF mutation status. Mycoplasma contamination was tested negative in house with the PCR Mycoplasma Detection Kit (Lonza). Melanoma cells were cultured in high glucose DMEM containing 10% FBS. For detachment culture conditions, cells were plated on plates coated with poly-2-hydroxy methacrylate (poly-HEMA). Lentiviruses encoding shRNAs or cDNAs were produced in HEK293T cells with packaging vectors (pMD2G and psPAX2) using Polyfect (Qiagen). Lentiviruses particles were collected 48 h after post-transfection and used to infect melanoma cells in the presence of 8 μg/ml polybrene, and infected cells were selected with 2 μg/ml of puromycin or 7 μg/ml blasticidin for 4 days prior to experiments. siRNA transfection was performed using Lipofectamine 2000 (Invitrogen) according to the manufacturer’s instructions. Guide-RNAs were cloned into pLX-sgRNA (Addgene #50662 for PGC1α) or lentiCRISPR (Addgene #49535 for ID2). Cells were subsequently infected with lentiviruses encoding Cas9 (pCW-Cas9, Addgene #50661) and sgRNAs followed by selection with blasticidin and puromycin as described above, and 1 μg/mL of doxycycline for 7 days.
Cells were lysed in a buffer containing 1% IGEPAL, 150 mM NaCl, 20 mM HEPES (pH7.9), 10 mM NaF, 0.1 mM EDTA, 1 mM sodium orthovanadate and 1X protease inhibitor cocktail. Protein concentration was quantified using BCA protein concentration assay kit (Pierce). Cell lysates were electrophoresed on SDS-polyacrylamide gels and transferred to Immobilon-P membrane (Millipore). Membranes were incubated with primary antibodies in 5% bovine serum albumin containing 0.05% Tween-20 overnight at 4°C. The membrane was then incubated with HRP-conjugated secondary antibody for 1 h at room temperature, and visualized using an ECL Prime (GE Healthcare).
Total RNA was isolated with Trizol (Invitrogen) by Direct-zol RNA MiniPrep kit (Zymo Research), and 2 μg of total RNA was used for cDNA synthesis using high capacity cDNA reverse transcription kit (Applied Biosystems). qPCR was carried out using SYBR Green PCR Master Mix (Applied Biosystems). Experimental Ct values were normalized to 36B4 where not otherwise indicated, and relative mRNA expression was calculated. Sequences for all the primers are provided in the Supplementary Information.
For PGC1α overexpression by adenovirus, A375P-shPGC1α and A375 cells were infected with adenoviruses expressing GFP or Flag-PGC1α for 36 h, followed by qPCR. PGC1α targets such as GSTM4 and COX5A were used as positive controls. For inhibitor treatment, cells were incubated with indicated concentrations of inhibitors for 6 h and mRNAs were analyzed by qPCR. For the RNA from migratory and non-migratory cells, migration of the A375P and G361 cells was initiated as described below. The non-migratory cells in suspension in the upper chambers were collected by centrifugation and resuspension in lysis buffer from the Cells-to-cDNA II kit (Invitrogen). The migratory cells were collected by directly applying the lysis buffer to the membrane, following the wash and clearing of the non-migratory cells in the upper chambers. 18S rRNA was used as internal control. For cells from paraffin-embedded tissue sections, Pinpoint™ Slide RNA Isolation System II (Zymo Research) was used to extract RNAs.
Cells with different mitochondrial contents were sorted based on the labeling of MitoTracker Green (Invitrogen). Briefly, MitoTracker Green was spiked in the medium of 100% confluent melanoma cells at the final concentration of 75 μM, and incubated with the cells for 20 min, followed by FACS sorting at DFCI Flow Cytometry Core. The top 10% cell population with the highest mitochondrial contents (mito/PGC1α-high) and the bottom 10% cell population with the lowest mitochondrial contents (mito/PGC1α-low) were harvested for qPCR, Western blot, migration assay (1×105/well for overnight) and metastasis assay.
For the circulating tumor cells, whole blood of the tumor-bearing mice was collected by cardiac perfusion with PBS containing 0.5 mM EDTA. After red blood cell lysis, the pelleted cells were stained with anti-mouse CD31 and CD45, along with anti-human HLA (eBioscience, as depicted in Extended Data Fig. 4c). The CD31−CD45−hHLA+ cells were directly sorted into RNAprotect Cell Reagent (Qiagen), and then converted into cDNA using the Cells-to-cDNA II kit. The primary tumors were subjected to enzymatic digestion for single cell suspension and FACS sorting to make them equivalent controls. qPCR was performed with SYBR Green, following the unbiased, target-specific preamplification of cDNA using SsoAdvanced PreAmp Supermix (BioRad). Experimental Ct values were normalized to 18s rRNA, and relative mRNA expression was calculated.
Lactate and glucose assay kits (BioVision Research Products) were used to measure extracellular lactate and glucose, following manufacturer’s instructions. Briefly, equal number of cells were seeded in 6-well plates and cultured in Phenol Red-free DMEM for 24 h or 36 h. Cultured medium was then mixed with the reaction solution. Lactate and glucose levels were measured at 450 nm and 570 nm, respectively, using a FLUO star Omega plate reader. Values were normalized to cell number.
Melanoma cell lines were lifted by 0.5 mM EDTA in PBS and washed with 1X PBS. For the intravenous injection, a total of 3 × 105 (A375) or 1 × 106 (G361 and MeWo) or 2 × 106 (A375P and FACS-sorted MeWo) cells in 0.2 mL of DMEM were injected into the tail vein of 6-week-old male nude mice. No randomization or blind techniques were applied in this study. To assess the degree of tumor formation in the lung, bioluminescence imaging of living mice were performed on a Xenogen IVIS-50TM imaging systems equipped with an isoflurane (1~3%) anesthesia system and a temperature controlled platform19, three weeks (G361 and MeWo) or four weeks (A375 and A375P) post injection. For the doxycycline induction experiment, upon detection of lung metastasis following tail vein implantation, PGC1α expression was induced by feeding mice with chow or doxycycline-containing diet (200 g/kg, Harlan Laboratories) for one week. For the orthotopic metastasis model, 1 × 106 cells were injected subcutaneously into one side of the 6-week-old male NOD/SCID mouse, with two injections per animal, followed by surgical tumor removal when the subcutaneous tumors reached the size of 2 mm in diameter. Metastasis was monitored by in vivo imaging at 8–10 weeks post surgery. After the measurement of bioluminescence, animals were sacrificed and the lungs were harvested. Collected lung tissues were fixed in 10% buffered formalin solution (Sigma-Aldrich) overnight. Fixed tissues were stained with hematoxylin and eosin (H&E) or antibodies against p-FAK-Y397 (Invitrogen) or HMB45 (Abcam), and one image per sample was shown as representative of one or three pictures captured, as indicated specifically in corresponding figure legend. Scale bar represents 200 microns unless otherwise indicated. All procedures were conducted in accordance with the guidelines of the Beth Israel Deaconess Medical Center Institutional Animal Care and Use Committee, and none of the tumors exceeded the size limit dictated by the IACUC guidelines.
For cell migration assays, transwell chambers were purchased from Corning Life Science. Generally, A375P (5×103) or G361 (4×103) cells in 0.1 mL of FBS-free medium were seeded into the upper chamber and incubated for 6 h if not otherwise indicated. For invasion assays, A375P (5×103), A375 (3×103), G361 (4×103), A2058 (3×103), RPMI7951 (5×103) or WM115 (4×103) cells in 0.1 mL of FBS-free medium were seeded into the upper chamber of an 8 μM matrigel coated chamber (BD Bioscience) and incubated for 16 h if not otherwise indicated. Specifically, for the migration and invasion assays on sorted cells (Fig. 2c) or A375P cells with ID2 knockdown (Fig. 3b), 1×105 cells were seeded and incubated for 24 h. Cells that had migrated and invaded through the matrigel were then fixed and stained with H&E if not otherwise indicated. The membrane attached with migrated and invaded cells was placed on a glass slide and total cell numbers from three or four random fields under 20–40X magnifications were quantified with an Olympus IX51 or a Nikon 80i Upright microscope, by counting cells on 20–50% of one field area and extrapolated to 100% of the field17.
Specifically, for the experiments with FAK inhibitor, shScr or shPGC1α stably expressing cells (A375P 1×105/well, G361 2.5×104/well) were cultured in transwell chambers with either DMSO or indicated concentration of PF-573228, followed by staining with Crystal Violet and quantification after migration for 24 h. For the experiments with PLX4032, cells were incubated with DMSO or 1 μM PLX4032 for 10 h in matrigel-coated transwell chambers, followed by quantification.
Nuclear lysates were incubated with specific antibodies overnight at 4°C, followed by precipitation with protein G Dynabeads (Invitrogen) at 4°C for 2 h. For Figure 3h, nuclear lysates from V5-ID2 stably-expressing A375P cells were subjected to co-IP with 1 μg ID2 antibody (C-20, Santa Cruz Biotechnology), followed by Western blot with TCF4 (M03, Abnova) and ID2 (4E12G5, Thermo Scientific); for Figure 4d, 10 mg of nuclear lysates from A375P cells treated with DMSO or 1 μM PLX4032 for 16 h were subjected to co-IP with 4 μg ID2 antibody (C-20). ChIP was performed with the MilliPore ChIP Kit with slight modification. Following sonication, nuclear lysates were precleared with protein A/GDynabeads (Invirogen) for 1 h. Equal amounts of precleared lysates were incubated with IgG or gene specific antibodies (PGC1α 4C1.C from Millipore, or PGC1α H-300, and TCF4 K-15 from Santa Cruz Biotechnology) overnight, followed by precipitation with protein A/G-Dynabeads for 2 h. qPCR with SYBR Green was performed to quantify the promoter occupancy. For Figure 4e, A375P cells stably expressing V5-TCF4 were cultured with PLX4032 at 5 μM for 16 hours and followed by ChIP and qPCR.
ToxiLight™ Non-destructive Cytotoxicity BioAssay Kit (Lonza) was used to quantify the cytotoxic effects of indicated compounds according to the manufacturer’s instruction. The measurement of dead cells in DMSO group was set as 1, and was used to normalize other treatment groups (Extended Data Figure 2j). For the cell growth assay with PLX4032 (Extended Data Figure 9d), cells were cultured with DMSO or PLX4032 for indicated time under either attachment or detachment conditions, followed by cell counting with hemocytometer. For detachment culture conditions, cells were plated on tissue culture plates coated with poly-2-hydroxy methacrylate (poly-HEMA).
All statistics are described in figure legends. In general, for two experimental comparisons, a two-tailed unpaired Student’s t-test was used unless otherwise indicated. For multiple comparisons, one-way ANOVAs were applied. When cells were used for experiments, three replicates per treatment were chosen as an initial sample size. All n values defined in the legends refer to biological replicates. If technical failures such as tail-vein injection failure or inadequate intraperitoneal injection occurred before collection, those samples were excluded from the final analysis. Statistical significance is represented by asterisks corresponding to *p < 0.05, **p < 0.01 and ***p < 0.005.
We thank Dr. Rod Bronson (Rodent Histopathology Core, Dana-Farber/Harvard Cancer Center) for his critical analysis of the mouse histology. We thank the Nikon Imaging Center at Harvard Medical School for help with light microscopy. We appreciate the important discussions on this project from members of the Puigserver lab. J-HL was supported in part by a postdoctoral fellowship from the American Heart Association (13POST14750008) and the National Research Foundation from Korean government. These studies were funded in part by the Claudia Adams Barr Program in Cancer Research (to PP), Dana-Farber Cancer Institute internal funds (to PP) and NIH R01CA181217 (to PP), as well as the Friends of Dana-Farber Award (to CL).
The authors disclose no potential conflicts of interest.
Author ContributionsP.P. and F.V. conceived the project. C.L. and J-H.L. designed and performed all the experiments with direction and discussions from P.P.; Y.L. contributed to animal experiments and edited the manuscript. A.T. contributed to immunoblotting experiments. F.V. and H.R.W. designed and performed the bioinformatic analyses. S.R.G. performed immunohistochemistry experiments. P.P., F.V., J-H.L., H.R.W. and C.L. prepared the manuscript.
The authors declare no competing financial interests. Readers are welcome to comment on the online version of the paper.