A Transcriptional Regulatory Loop of Master Regulator Transcription Factors, PPARG, and Fatty Acid Synthesis Promotes Esophageal Adenocarcinoma

Abstract

Although obesity is one of the strongest risk factors for esophageal adenocarcinoma, the molecular mechanisms underlying this association remain unclear. We recently identified four esophageal adenocarcinoma-specific master regulator transcription factors (MRTF) ELF3, KLF5, GATA6, and EHF. In this study, gene-set enrichment analysis of both esophageal adenocarcinoma patient samples and cell line models unbiasedly underscores fatty acid synthesis as the central pathway downstream of three MRTFs (ELF3, KLF5, GATA6). Further characterizations unexpectedly identified a transcriptional feedback loop between MRTF and fatty acid synthesis, which mutually activated each other through the nuclear receptor, PPARG. MRTFs cooperatively promoted PPARG transcription by directly regulating its promoter and a distal esophageal adenocarcinoma-specific enhancer, leading to PPARG overexpression in esophageal adenocarcinoma. PPARG was also elevated in Barrett's esophagus, a recognized precursor to esophageal adenocarcinoma, implying that PPARG might play a role in the intestinal metaplasia of esophageal squamous epithelium. Upregulation of PPARG increased de novo synthesis of fatty acids, phospholipids, and sphingolipids as revealed by mass spectrometry-based lipidomics. Moreover, ChIP-seq, 4C-seq, and a high-fat diet murine model together characterized a novel, noncanonical, and cancer-specific function of PPARG in esophageal adenocarcinoma. PPARG directly regulated the ELF3 super-enhancer, subsequently activating the transcription of other MRTFs through an interconnected regulatory circuitry. Together, elucidation of this novel transcriptional feedback loop of MRTF/PPARG/fatty acid synthesis advances our understanding of the mechanistic foundation for epigenomic dysregulation and metabolic alterations in esophageal adenocarcinoma. More importantly, this work identifies a potential avenue for prevention and early intervention of esophageal adenocarcinoma by blocking this feedback loop. SIGNIFICANCE: These findings elucidate a transcriptional feedback loop linking epigenomic dysregulation and metabolic alterations in esophageal adenocarcinoma, indicating that blocking this feedback loop could be a potential therapeutic strategy in high-risk individuals.

Figure 1.. EAC-specific MRTFs promote fatty-acid synthesis pathway

(A) Heatmap of GSEA results of top 15 enriched hallmark pathways in MRTF-high expression (top 40%) vs MRTF-low expression (bottom 40%) EAC samples from TCGA. The average level of the 4 MRTFs was used as the “Mean” group. (B) Individual GSEA plots of fatty-acid metabolism pathway as in panel (A). (C) GSEA plots of RNA-seq upon silencing of either ELF3 or KLF5 in ESO26 cells, and silencing of GATA6 in OE19 cells. The RNA-Seq data of knockdown of ELF3 or KLF5 were generated by our group, while the data of GATA6 knockdown was conducted by Rogerson et al (33). (D) Heatmap of fold changes of mRNA levels of key enzymes for de novo fatty-acids synthesis following siRNA knockdown of either ELF3, KLF5 or GATA6 in ESO26 and OE33 cells. (E) Confocal images (left panel) and flow cytometry analyses (right panel) of lipid droplet in the presence and absence of either ELF3, KLF5 or GATA6 knockdown in ESO26 and OE33 cells.

Figure 2.. MRTFs directly activate both the promoter and enhancer of PPARG in an EAC-specific manner

(A) Changes in mRNA expression of eight regulators of fatty-acid metabolism upon knockdown of MRTFs. (B) Western blot validating the change of PPARG. (C) Flow cytometry analyses of lipid droplet and (D) mRNA levels of central enzymes for de novo fatty-acid synthesis after either KLF5 knockdown alone or combined with PPARG over-expression. (E) IGV line plots of the H3K27Ac ChIP-Seq, MRTF ChIP-Seq and ATAC-Seq in indicated samples. Signal values of normalized peak intensity are shown on the upper left corner. Number of samples for each track is shown on the upper right corner. Scatter plots at the bottom show the correlation between ATAC-Seq peaks and PPARG mRNA expression. Each dot is a TCGA EAC sample. All ChIP-seq data were generated internally; ATAC-Seq were from either TCGA or Rogerson et al (bottom 2 tracks)(33). (F) Luciferase reporter assays in ESO26 and OE33 cells. PPARG enhancer and a negative control (Ctrl) region were separately cloned into luciferase reporter vector. Mean±SD are shown, n=3. *, p<0.05; **, p<0.01. (G) mRNA expression of PPARG in esophageal cancer and nonmalignant distal esophageal tissues from TCGA.

Figure 3.. PPARG regulates fatty-acid synthesis in EAC cells

(A) Flow cytometry analyses of lipid droplet in the presence and absence of PPARG knockdown. (B) Top 10 enriched pathways of down-regulated genes upon PPARG knockdown in ESO26 cells. (C) Volcano plot of LC-MS/MS-based lipidomics after PPARG knockdown. Each dot is one lipid species. (D) Top ranked TF-binding motifs in PPARG ChIP-Seq in both ESO26 and OE33 cells. (E) IGV plots of H3K27Ac and PPARG ChIP-Seq at the FASN loci in ESO26 and OE33 cells. (F) Schematic diagram showing the regulation of lipid synthesis pathways by PPARG via integration of RNA-Seq, ChIP-Seq and lipidomics data. (G) Western blotting of the key enzymes for de novo synthesis of fatty-acids following PPARG-knockdown.

Figure 4.. Inhibition of PPARG/fatty-acid synthesis pathway suppresses EAC cell viability

Knockdown of PPARG by individual siRNAs inhibited (A) cell proliferation and (B) colony growth, while (C) increased cell apoptosis in different EAC cell lines. (D) PPARG was overexpressed and verified by Western blotting in different EAC cell lines. (E) Overexpressed (OE) PPARG promoted cell proliferation and (F) colony growth compared with empty vector control (EV). (G) Knockdown of either FASN or SCD decreased lipid droplet (H) and cell proliferation (I). Mean±SD are shown, n=3. *, p<0.05; **, p<0.01;***, p<0.001.

Figure 5.. T0070907, a PPARG-specific inhibitor, shows potent anti-EAC function

(A) Short-term cell viability assay measuring the IC50s of T0070907 in different cell lines. (B) Relative number of colonies formed at different concentrations of T0070907. (C) Treatment of T0070907 increased cell apoptosis. (D) Short-term cell viability assays measuring T0070907 IC50s in either (D) PPARG knockdown or (E) over-expression cells. (F) ChIP-PCR using PPARG antibody either with or without T0070907 treatment. (G) mRNA levels of PPARG-target genes after T0070907 treatment. (H) Xenograft weight, (I) mouse weight, (J) xenograft photos, and (K) gene expression from indicated groups. Mean±SD are shown, n=3. *, p<0.05; **, p<0.01; ***, p<0.001; N.S., not significant.

Figure 6.. PPARG cooperates with ELF3 and directly co-activates ELF3 super-enhancer

(A) Top enriched motifs of PPARG ChIP-Seq shared in ESO26 and OE33 cells. Note that EHF and CDX2 are not overlapped. (B) Heatmap showing ChIP-Seq signals at PPARG/ELF3 co-binding regions, ordered by the intensity of PPARG peaks. Lines, peaks. (C) Line plots showing the distribution of PPARG/ELF3 ChIP-Seq signals in their co-binding regions. (D) Inflection plot ranking all enhancers co-occupied by PPARG/ELF3. Pie charts showing the percentage of super-enhancers and typical-enhancers. (E) IGV plots of H3K27ac and TF ChIP-Seq signals in ESO26 and OE33 cells. Connecting lines show enhancer-promoter interactions detected by 4C as we published recently(5). (F) Enhancer activity measured by luciferase reporter assays. (G) mRNA levels of MRTFs after siRNA knockdown of PPARG. (H) Western blot of indicated proteins upon PPARG knockdown. (I) Schematic graph of the regulatory relationship between PPARG and MRTFs summarized from Figure 6. Mean±SD are shown; n=3. *, p<0.05; N.S., not significant.

Figure 7.. A transcriptional feedback loop of Fatty-acid/PPARG/MRTF in EAC

(A) Heatmap of fold changes of MRTF mRNA levels (B) and enhancer activity measured by luciferase assays following treatment with different fatty-acids (10 μM) for 48h. (C) Schematic diagram of a transcriptional feedback loop involving fatty-acid/PPARG/MRTF in EAC. (D) Xenograft images, (E) growth curves and (F) tumor weights of xenograft expressing either scramble shRNA or PPARG-shRNA which were grown in mice fed with either high-fat diet (HFD) or low-fat diet (LFD). (G) mRNA levels of indicated genes were measured by qRT-PCR. Mean±SD are shown, n=3 *, p<0.05; **, p<0.01; ***, p<0.001; N.S., not significant.