Altered Transcriptional Control Networks with Trans-Differentiation of Isogenic Mutant-KRas NSCLC Models

Abstract

Background: The capacity of cancer cells to undergo epithelial mesenchymal trans-differentiation has been implicated as a factor driving metastasis, through the acquisition of enhanced migratory/invasive cell programs and the engagement of anti-apoptotic mechanisms promoting drug and radiation resistance. Our aim was to define molecular signaling changes associated with mesenchymal trans-differentiation in two KRas mutant NSCLC models. We focused on central transcription and epigenetic regulators predicted to be important for mesenchymal cell survival.

Experimental design: We have modeled trans-differentiation and cancer stemness in inducible isogenic mutant-KRas H358 and A549 non-small cell lung cell backgrounds. As expected, our models show mesenchymal-like tumor cells acquire novel mechanisms of cellular signaling not apparent in their epithelial counterparts. We employed large-scale quantitative phosphoproteomic, proteomic, protein-protein interaction, RNA-Seq, and network function prediction approaches to dissect the molecular events associated with the establishment and maintenance of the mesenchymal state.

Results: Gene-set enrichment and pathway prediction indicated BMI1, KDM5B, RUNX2, MYC/MAX, NFκB, LEF1, and HIF1 target networks were significantly enriched in the trans-differentiation of H358 and A549 NSCLC models. Physical overlaps between multiple networks implicate NR4A1 as an overlapping control between TCF and NFκB pathways. Enrichment correlations also indicated marked decrease in cell cycling, which occurred early in the EMT process. RNA abundance time course studies also indicated early expression of epigenetic and chromatin regulators within 8-24 h, including CITED4, RUNX3, CMBX1, and SIRT4.

Conclusion: Multiple transcription and epigenetic pathways where altered between epithelial and mesenchymal tumor cell states, notably the polycomb repressive complex-1, HP1γ, and BAF/Swi-Snf. Network analysis suggests redundancy in the activation and inhibition of pathway regulators, notably factors controlling epithelial cell state. Through large-scale transcriptional and epigenetic cell reprograming, mesenchymal trans-differentiation can promote diversification of signaling networks potentially important in resistance to cancer therapies.

Keywords: EMT; epigenetic; systems biology; transcription; tumor heterogeneity.

Figure 1

Marker expression in H358 and A549 isogenic mesenchymal trans-differentiation models. (A) Immunoblot staining for E-cadherin, CD44, vimentin, and actin (control) in the H358-dox-TGFβ model in epithelial (−dox) or mesenchymal (+dox) states (5 days and ~180 days on doxycycline). (B) Deconvoluted fluorescence microscopy image for CD44 (green) and E-cadherin (red) in the H358-dox-TGFβ model in epithelial (−dox) or mesenchymal (+dox) states. (C) H358 cells induced to express TGFβ for 14 days (bottom panels) show increased aldefluor activity, a marker of aldehyde dehydrogenase activity and stemness, relative to control cells (top panels), as measured by FACS. (D) Immunoblot staining for fibronectin, E-cadherin, vimentin, and GAPDH (control) in the A549 cell in the presence or absence of exogenous TGFβ (10 ng/ml) for 7 or 14 days. (E) A workflow schema for RNA, protein, phosphopeptide abundance measurement, and cross-correlation.

Figure 2

(A) Regression analysis of duplicate mesenchymal RNA-Seq samples expressed at a ratio to the mean epithelial control for both H358 and A549 models (greater than twofold changes are indicated in black). (B) Concordant RNA and protein changes for both isogenic H358 and A549 models. (C) Transcription associated RNA transcripts with differential abundance between epithelial and mesenchymal cell states in both H358 and A549 models.

Figure 3

(A) Correlation of H358 and A549 EMT regulated genes with the LEF1 signature from LEF1 over-expressing epithelial DLD1 cells (from GSE3229). Normalized enrichment score (NES) was 2.63, nominal p-value, FDR q-value, and FWER p-value were <0.001. (B) The top 20 positively and negatively correlated genes were identified and heat mapped for H358 (B) control [0; A, B] and TGFβ [180; A, B] duplicate samples) and (C) for A549 (control [Cntrl; LB13, LB16] and TGFβ [TGFb; B11, B12] duplicate samples).

Figure 4

Nuclear translocation of β-catenin and TCF/LEF activation in steady-state mesenchymal H358 cells expressing Wnt5A. (A) Heat map of RNA abundance of Wnt signaling components and target genes in H358 and A549 isogenic models (mesenchymal/epithelial; log2) from duplicate samples. (B) Loss of membrane β-catenin localization and gain of punctate nuclear localization in mesenchymal H358/dox-TGFβ cells. Top panels: H358 cells in the absence of doxycycline. Bottom panels: H358/dox-TGFβ cells in a mesenchymal-like state. Cells were labeled with β-catenin antibody (red) and DAPI (blue) and imaged (60X). (C) Co-transfection of sTOP-TCF/LEF-luciferase (“TCF”) or control FOP-luciferase (“Cntrl”) was used to measure the activity of the TCF–LEF pathway. Renilla-luciferase was used to normalize transfection efficiency. H358/dox-TGFβ cells, in the presence or absence of doxycycline, were transfected and after 48 h luciferase measurements preformed under standard conditions. Both steady-state serum and 24 h serum starvation conditions (“-SFM”) were used with similar results. The y-axis units are relative light units (RLU). The means of two independent experiments are shown, each in triplicate, where the error bars reflect the standard error of the mean.

Figure 5

Activation of the NFκB pathway in mesenchymal H358. (A) Correlation of A549 EMT regulated genes with the NFκB signature from NFκB and RELA over-expression in keratinocytes (52). Normalized enrichment score (NES) was −1.85, nominal p-value <0.001, FDR q-value 0.01. (B) The top 20 positively and negatively correlated genes were identified and heat mapped for A549, as defined in Figure 4. (C) Heat map of RNA abundance of NFκB signaling components and target genes in H358 and A549 isogenic models (mesenchymal/epithelial; log2). (D) Protein–protein overlaps between TCF4 and NFκB protein interaction datasets from BioGrid, showing NR4A1 as a common physical node between the two pathways.

Figure 6

(A) Increased BMI target gene enrichment from RNA-Seq datasets comparing differential RNA expression between H358 and A549 cell states. Statistical significance (p < 0.0001 and FDR q-value <0.04 was observed. (B) The top 20 positively and negatively correlated genes were identified and heat mapped for H358 and (C) for A549, with labels as defined in Figure 3.

Figure 7

RNA expression ratios comparing H358 and A549 mesenchymal/epithelial cell states in duplicate. Complexes associated with RUNX2 (top) and RUNX3 (bottom) are shown.