Gene regulatory networking reveals the molecular cue to lysophosphatidic acid-induced metabolic adaptations in ovarian cancer cells

Abstract

Extravasation and metastatic progression are two main reasons for the high mortality rate associated with cancer. The metastatic potential of cancer cells depends on a plethora of metabolic challenges prevailing within the tumor microenvironment. To achieve higher rates of proliferation, cancer cells reprogram their metabolism, increasing glycolysis and biosynthetic activities. Just why this metabolic reprogramming predisposes cells towards increased oncogenesis remains elusive. The accumulation of myriad oncolipids in the tumor microenvironment has been shown to promote the invasiveness of cancer cells, with lysophosphatidic acid (LPA) being one such critical factor enriched in ovarian cancer patients. Cellular bioenergetic studies confirm that oxidative phosphorylation is suppressed and glycolysis is increased with long exposure to LPA in ovarian cancer cells compared with non-transformed epithelial cells. We sought to uncover the regulatory complexity underlying this oncolipid-induced metabolic perturbation. Gene regulatory networking using RNA-Seq analysis identified the oncogene ETS-1 as a critical mediator of LPA-induced metabolic alterations for the maintenance of invasive phenotype. Moreover, LPA receptor-2 specific PtdIns3K-AKT signaling induces ETS-1 and its target matrix metalloproteases. Abrogation of ETS-1 restores cellular bioenergetics towards increased oxidative phosphorylation and reduced glycolysis, and this effect was reversed by the presence of LPA. Furthermore, the bioenergetic status of LPA-treated ovarian cancer cells mimics hypoxia through induction of hypoxia-inducible factor-1α, which was found to transactivate ets-1. Studies in primary tumors generated in syngeneic mice corroborated the in vitro findings. Thus, our study highlights the phenotypic changes induced by the pro-metastatic factor ETS-1 in ovarian cancer cells. The relationship between enhanced invasiveness and metabolic plasticity further illustrates the critical role of metabolic adaptation of cancer cells as a driver of tumor progression. These findings reveal oncolipid-induced metabolic predisposition as a new mechanism of tumorigenesis and propose metabolic inhibitors as a potential approach for future management of aggressive ovarian cancer.

Keywords: ETS-1; HIF-1α; invasion; lysophosphatidic acid; metabolic adaptations; ovarian cancer.

Figure 1

LPA is associated with metabolic modulation of ovarian cancer (OC) cells. (A,B) OCRs (pmol·min⁻¹·μg⁻¹) were determined in both non‐cancer IOSE‐364 and PA‐1 OC cell lines using Seahorse XF‐24 flux analyzer upon exposure to LPA for 24 h. Oligomycin, FCCP, and antimycin A + rotenone were sequentially administered as indicated. (C) Basal OCR was calculated from the measurement prior to oligomycin addition minus non‐mitochondrial respiration, whereas maximal OCR was calculated from FCCP‐induced levels minus non‐mitochondrial respiration. Results are represented as mean ± SEM (*p < 0.05, **p < 0.01). (D) Reserve capacity was obtained as maximal minus basal OCR (**p < 0.01). (E) Glycolytic rate (mpH·min⁻¹·μg⁻¹) was determined and plotted from the ECAR measurements (F) in both IOSE‐364 and PA‐1 OC cell lines in response to LPA for 24 h. Glucose, oligomycin and 2‐deoxyglucose were injected sequentially to the cells in glucose‐free medium. Glycolytic rate was calculated by deducting the measurement after glucose injection (prior to oligomycin injection) from the measurement prior to glucose injection. Results are provided as mean ± SEM (*p < 0.05). (G) Quantitative ELISA was used to measure cellular lactate production in the conditioned medium from the treated cells. The result was given as fold change with respect to non‐treated cells in bar graph (**p < 0.01). (H) The effect of LPA on some key glycolytic genes was analyzed from the RNA‐Seq data. For the given gene sets of glycolysis, the analysis was achieved with the transcripts having greatest difference in control values, mentioned as form of fold change. This was analyzed by ranking genes by normalized expression or raw counts, and those taken with the greatest difference in observed values. (I) LDHA expression levels analyzed in LPA‐treated cells compared with non‐treated controls by quantitative PCR (i, *p < 0.05) and western detection (ii). Fold change for the protein expression was calculated using image j software (normalized against GAPDH) and is provided beneath the panel.

Figure 2

Gene‐regulatory networking unfolds modification of the LPA‐treated ovarian cancer (OC) cell transcriptome towards increased invasion. (A) Heat map showing standardized expression (Z‐scores) of genes differentially expressed upon long induction of LPA (20 μm) for 24 h in PA‐1 cells. (B) GSEA analysis revealing biological processes that promote cancer and gene families involved in signal transduction, transcriptional and translational regulation (p ≤ 0.05) on LPA treatment. (C) Venn diagram showing distribution of (i) upregulated and (ii) downregulated genes involved in multiple biological processes that promote tumor progression. (D) Global gene regulatory network encompassing differentially expressed genes and the key pathways involved. The genes are colored based on their fold change and the size is based on their connectivity score. (E) Core regulatory network encompassing key differentially expressed genes and the selective pathways involved, based on criteria as above. (F) IOSE cells were treated with LPA to show actin rearrangement resembling the mesenchymal phenotype through confocal microscopy after TRITC–phalloidin staining. Scale bar = 10 μm. (G) Confocal imaging for E‐cadherin expression was performed in (i) IOSE and (ii) PA‐1 cells after LPA treatment. In all cases, DAPI was used to stain the nucleus and merged images are shown. Scale bar = 10 μm. TGF‐β was used as a positive inducer of EMT in all cases. (H) Transwell invasion assay performed in PA‐1 and IOSE cells after treatment with LPA in the absence and presence of PTX for 22 h. Cells at three independent fields for each well were counted and are plotted with error bars (*p < 0.05, **p < 0.01). (I) MMP‐2, MMP‐9 and MMP‐13 mRNA expression was analyzed by quantitative PCR after treatment with LPA in the absence and presence of PTX (*p < 0.05, **p < 0.01 vs control; #p < 0.05, ##p < 0.01 vs LPA treatment) in (i) PA‐1 and (ii) SKOV‐3 cells. (J) Western blot analysis for the respective MMPs was performed with treatments similar to those mentioned for (i) PA‐1 and (ii) SKOV‐3 cells. (K) Gelatin zymography for MMP‐2/MMP‐9 activity assay in PA‐1 cells. (L) Matrigel invasion assay performed after cells were transfected with the indicated siRNAs followed by LPA induction for 22 h. Percent cell invaded was counted at three independent fields for each well (*p < 0.05 vs control, #p < 0.05 vs LPA treatment). (M) Wound healing assay performed with similar treatment to that described. Images at 0 and 24 h. Scale bar = 200 μm.

Figure 3

AKT pathway activation is critical for LPA‐induced ovarian cancer (OC) invasion. (A) Matrigel invasion assay performed with PA‐1 cells pretreated with the indicated pathway inhibitors for 1 h followed by LPA for 22 h. Scale bar = 100 μm. (B) Percent cell invasion plotted with cells counted in three independent fields for each well (*p < 0.05 vs control, #p < 0.05 vs LPA treatment). (C) Scratch wound healing assay performed with the mentioned treatments for 24 and 48 h. 0 h images are provided. Scale bar = 100 μm. Arrows indicate the width of the wound and the assay was repeated three times. (D) MMP‐9 activity was analyzed by gelatin zymography with the treatments as mentioned. (E) AKT pathway activation was performed with the indicated treatments by analyzing p‐AKT and total AKT levels in a western blot.

Figure 4

LPA induces ETS‐1 expression. (A) Heat map presenting clustering of all differentially expressed transcriptional regulators on LPA treatment. (B) Statistical ranking of the induced transcription factors were shown with fold change and p‐value. (C) Quantitative PCR (*p < 0.05, ***p < 0.001) and western blot were examined with 1, 5, 10 and 20 μm LPA on ETS‐1 expression. (D) Expression of ETS‐1 was analyzed by both quantitative PCR (*p < 0.05, ***p < 0.001 vs control; #p < 0.05 vs LPA treatment) and immunoblot upon treatment of PA‐1 cells with LPA in the absence and presence of PTX. (E) ETS‐1 expression was further analyzed in SKOV‐3 cells under similar treatments. In all cases, fold change for the protein expression was calculated using image j software (normalized against GAPDH) and is provided beneath the panel. (F) Quantitative expression of ETS‐1 was analyzed in PA‐1 cells by flow cytometry with the indicated treatments. Fold change with respect to control was plotted (**p < 0.01). (G) LPA‐induced nuclear localization of ETS‐1 in the absence and presence of PTX using confocal microscopy. DAPI is used to stain the nucleus and merged images are provided. Scale bar = 10 μm. (H) Nuclear and cytoplasmic distribution of ETS‐1 upon dose‐dependent exposure to LPA (0, 10 and 20 μm) for 24 h. GAPDH and histone H3 were used as the loading control for the cytoplasmic and nuclear fractions, respectively.

Figure 5

ETS‐1 knockdown abrogates LPA‐induced metabolic reorientation. (A) Gene matrix representing the common target transcripts expressed under the ETS subfamily members. All diagonal numbers represent specific genes expressed by the respective transcription factors. Representative (B) basal and (C) maximal OCR (pmol·min⁻¹·μg⁻¹) are calculated and plotted from the OCR analysis (D) in ETS‐1 knockdown ovarian cancer cells in the presence and absence of LPA using a Seahorse XF‐24 extracellular flux analyzer. Both graphs were plotted after deduction of the non‐mitochondrial respiration; error bars indicate SEM (*p < 0.05, ***p < 0.001 vs control; #p < 0.05 vs ETS‐1 knockdown). (E) ECAR analysis was obtained from similar treatment of the cells in glucose‐free medium and (F) the glycolytic rate (mpH·min⁻¹·μg⁻¹) was plotted. Glycolytic rate was analyzed by subtracting values obtained after glucose addition (prior to oligomycin) from those obtained prior to glucose injection; error bars indicate SEM (*p < 0.05 vs control; ##p < 0.01 vs LPA treatment). (G) PA‐1 cells were transfected with ETS‐1 siRNA and then treated with/without LPA followed by quantitative PCR analysis against MMP‐2, MMP‐9 and MMP‐13 expressions (**p < 0.01 vs control; #p < 0.05, ##p < 0.01 vs LPA treatment). Downregulation of ETS‐1 was also confirmed by quantitative PCR analysis in these cells. (H) MMP‐2/MMP‐9 activity was shown by gelatin zymography in the cells treated as mentioned. (I,J) Effect of ETS‐1 knockdown on invasion of PA‐1 cells in the presence/absence of LPA. Scrambled siRNA transfected cells were treated as a negative control. The invasion experiments were performed three times, and the mean ± SEM (error bars) plotted (**p < 0.01). (K) Similar transfection was performed to show the migration phenomenon in PA‐1 cells by wound closure assay. 0 and 24 h images are given. Scale bar = 200 μm.

Figure 6

LPAR2‐mediated induction of AKT‐signaling is involved in ETS‐1 expression. (A) Quantitative PCR was performed to show the ability of each of the three LPA receptor‐specific siRNAs (LPAR1/2/3) to significantly knockdown their own expression in PA‐1 cells. (B) ETS‐1 expression level was analyzed in these knockdown cells as indicated (**p < 0.01 vs control; #p < 0.05 and ##p < 0.01 vs LPA treatment). (C) PA‐1 cells were transfected with the indicated LPA receptor‐specific siRNAs, serum starved and stimulated with 20 μm LPA for 24 h, followed by western analysis of ETS‐1. GAPDH was used as a loading control. The densities of the respective bands were calculated using image j software (normalized against GAPDH) and are represented as ‘fold change’ beneath the panel. (D) LPAR2‐siRNA transfected PA‐1 cells treated with/without LPA followed by quantitative PCR analysis. Results show the mean ± SEM (**p < 0.01 vs control; ##p < 0.01 vs LPA treatment). (E) Western blot analysis of ETS‐1 expression was done following similar treatment. (F) Quantitative PCR analysis of MMPs was performed in the same transfected cells following the mentioned treatments (*p < 0.05 vs control, #p < 0.05 vs LPA treatment). (G) MMP‐2/MMP‐9 activity was assessed by gelatin zymography in the conditioned media of the transfected cells. (H) Quantitative PCR (*p < 0.05 vs control, #p < 0.05 vs LPA treatment) and (I) immunoblot analysis of ETS‐1 expression was performed in PA‐1 cells pretreated with PtdIns3K/AKT inhibitors for 1 h followed by LPA induction.

Figure 7

LPA mimics hypoxia through activation of HIF‐1α in ovarian cancer (OC) cells. (A) Western blot analysis of HIF‐1α expression upon treatment with 20 μm LPA in a time‐dependent manner in PA‐1 cells. GAPDH was used as a loading control. HIF‐1α expression was analyzed by (B) quantitative PCR (*p < 0.05) and (C) Western blot in PA‐1 cells treated with LPA in the presence/absence of PTX. In all cases, fold change for the protein expression was calculated using image j software (normalized against GAPDH) and is provided beneath the panel. (D) Nuclear localization of HIF‐1α upon treatment by LPA with or without PTX was observed by confocal imaging. DAPI is used to stain the nucleus (scale bar = 10 μm). (E) OCR in control, LPA‐ and CoCl₂‐treated PA‐1 cells under basal condition was analyzed using Seahorse extracellular flux analyzer. Representative basal OCR (pmol·min⁻¹·μg⁻¹) from three independent experiments are plotted; error bars indicate mean ± SEM (*p < 0.05, **p < 0.01). (F) GSEA analysis represented the differentially expressed target genes that gets activated under the control of HIF‐1α transcription factor. (G) Quantitative PCR (**p < 0.01 vs control, #p < 0.05 vs LPA treatment) and (H) immunoblot analysis for HIF‐1α expression in cells pretreated with AKT inhibitor for 1 h followed by LPA‐induction. VEGF expression was also observed as a known transcriptional target of HIF‐1α under this condition.

Figure 8

LPA‐induced HIF‐1α transcriptionally upregulates ETS‐1 in ovarian cancer (OC) cells. (A) Quantitative PCR and (B) immunoblot analysis were used to analyze the HIF‐1α and ETS‐1 expression in PA‐1 cells transfected with HIF‐1α‐specific siRNA for 24 and 48 h, followed by LPA treatment (*p < 0.05, **p < 0.01 vs control; #p < 0.05 vs LPA treatment). (C) Expression of ETS‐1 was also determined in HIF‐1α‐knockdown SKOV‐3 cells for 24 and 48 h, followed by treatment with LPA. Fold change for the protein expression was calculated using image j software (normalized against GAPDH) and is provided beneath the panel. (D) ETS‐1 expression was further checked in the presence of HIF‐1α inhibitor. (E) ChIP analysis of HIF‐1α towards binding to ets‐1 promoter region having hypoxia response elements sequences was performed in HIF‐1α‐overexpressed PA‐1 cells and (F) LPA‐treated PA‐1 and OAW‐42 cells. No antibody control and IgG sets were used as negative control. (G) Similar treatments were performed to check the MMP‐9 activity in a gelatin zymogram from the conditioned media. (H) Matrigel Invasion Assay of the HIF‐1α knockdown cells was performed in the absence/presence of LPA and percent cell invasion was plotted (*p < 0.05 vs control; #p < 0.05 vs LPA treatment).

Figure 9

LPA is functionally involved in the upregulation of ETS‐1 in vivo. (A) Representative images of tumors formed in BALB/c mice injected with either PBS alone or ID8 cells with PBS or LPA. (B) Effect of LPA treatment on tumor growth in BALB/c mice. (C) Tumor volume was measured and represented as fold change (*p < 0.05). (D) Western blotting was performed with three sets of tumor samples as indicated, for ETS‐1 and HIF‐1α expression analysis. Proliferating cell nuclear antigen expression was observed in these samples as a marker of proliferation and GAPDH was used as a loading control. (E) ETS‐1 and (F) HIF‐1α expression levels were analyzed in these tissue sections by fluorescence‐immunohistochemistry using confocal microscopy (scale bar = 10 and 100 μm). (G) ETS‐1 expression was analyzed in ID8 cells upon treatment with LPA in the presence and absence of its receptor inhibitor to validate the in vivo data. (H) Schematic representation of LPA‐induced ovarian cancer aggressiveness. LPA, through LPAR2‐specific AKT signaling, activates HIF‐1α and subsequently ETS‐1 leading to reduced mitochondrial respiration and increased glycolysis, as well as cell‐type‐specific proteolytic enzyme expression, thereby specifying a sustained invasive potential for the cancer cells.