KLF4 Induces Mesenchymal-Epithelial Transition (MET) by Suppressing Multiple EMT-Inducing Transcription Factors

Abstract

Epithelial-Mesenchymal Plasticity (EMP) refers to reversible dynamic processes where cells can transition from epithelial to mesenchymal (EMT) or from mesenchymal to epithelial (MET) phenotypes. Both these processes are modulated by multiple transcription factors acting in concert. While EMT-inducing transcription factors (TFs)-TWIST1/2, ZEB1/2, SNAIL1/2/3, GSC, and FOXC2-are well-characterized, the MET-inducing TFs are relatively poorly understood (OVOL1/2 and GRHL1/2). Here, using mechanism-based mathematical modeling, we show that transcription factor KLF4 can delay the onset of EMT by suppressing multiple EMT-TFs. Our simulations suggest that KLF4 overexpression can promote a phenotypic shift toward a more epithelial state, an observation suggested by the negative correlation of KLF4 with EMT-TFs and with transcriptomic-based EMT scoring metrics in cancer cell lines. We also show that the influence of KLF4 in modulating the EMT dynamics can be strengthened by its ability to inhibit cell-state transitions at the epigenetic level. Thus, KLF4 can inhibit EMT through multiple parallel paths and can act as a putative MET-TF. KLF4 associates with the patient survival metrics across multiple cancers in a context-specific manner, highlighting the complex association of EMP with patient survival.

Keywords: Epithelial–Mesenchymal Plasticity (EMP); KLF4; Mesenchymal–Epithelial Transition (MET); epigenetics; mathematical modeling.

Figure 1

KLF4 inhibits EMT. (A) Schematic representation of KLF4 coupled to an EMT regulatory network consisting of miR-200, ZEB1, SNAIL, and SLUG. Green arrows denote activation, and red bars indicate inhibition. Solid arrows represent transcriptional regulation; a dotted line represents microRNA-mediated regulation. The circuit shown within the dotted rectangle is the control case (i.e., core EMT network without KLF4). (B) Bifurcation diagrams indicating ZEB1/2 mRNA levels for increasing the external signal (I) levels for the coupled EMT–KLF4 circuit (solid blue and dotted red curve) and the core EMT circuit (solid green and dotted black curve). (C) Temporal dynamics of the ZEB1/2 mRNA levels in a cell starting in an epithelial phenotype when exposed to a high level of an external EMT signal (I_ext = 100,000 molecules) (pink-shaded region) for the circuits shown in (A)). (D–F) Phase diagrams for the KLF4–EMT network driven by an external signal (I_ext) for (D) varying strengths of repression on SNAIL by KLF4, (E) for varying strengths of repression on KLF4 by SNAIL, and (F) for varying strengths of repression on SLUG by KLF4.

Figure 2

KLF4 promotes an epithelial phenotype. (A) (Left): Heatmap showing the steady-state values of all the components of the KLF4–EMT-coupled circuit, obtained across the parameter sets simulated by RACIPE. (Right) Same as the left panel but for KLF4 overexpression. (B) Change in the fraction of the epithelial phenotype due to the simulated overexpression or downregulation of either KLF4 or ZEB1 or SLUG. Error bars for n = 3 independent simulations. (C) Frequency density of the epithelial, hybrid, and mesenchymal phenotypes obtained using the EMT scores. (Top) KLF4 circuit, (middle) KLF4 circuit with KLF4 down-expression (DE), and (bottom) KLF4 circuit with KLF4 overexpression (OE). (D) Pairwise Pearson’s correlation matrix. (i) Correlation of the epithelial and mesenchymal players and EMT scoring metrics (76GS, KS, and MLR) in the CCLE cell lines. (ii) Correlation of the RACIPE-simulated expression values of the nodes in the KLF4 network. Squares indicate a p-value > 0.01. (E) Change in the size of the epithelial and mesenchymal clusters upon OE and DE of GRHL2 and KLF4. (F) Schematic showing the variations of the KLF4, GRHL2, ZEB1, and SNAIL levels across the EMT spectrum. **: p < 0.01.

Figure 3

KLF4 is inhibited during EMT. KLF4 expression across comparison groups in GEO datasets. (A) GSE59922, (B) GSE40690, (C) GSE58252, (D) GSE118407, (E) GSE85857, and (F) GSE110677. #: p < 0.1, *: p < 0.05, and **: p < 0.01 for a Student’s two-tailed t-test with unequal variances. (G,H) KLF4 and ZEB1 expression in TCGA types, arranged by mean KS scores (color scheme given on the right).

Figure 4

Epigenetic modulations involving KLF4 can alter the population dynamics of EMT states. (A) Scatter plot for KLF4 expression and its methylation status in TCGA types. (B) Bifurcation diagrams indicating the ZEB mRNA levels for increasing the EMT-inducing external signal (I_ext) levels for the coupled EMT–KLF4 circuit (solid blue and dotted red curve), for the circuit with higher α₁ and lower α₂ values (solid yellow and dotted brown curve), and for the circuit with lower α₁ and higher α₂ values (solid green and dotted black curve). (C) Stochastic simulation of the KLF4–EMT network for varied values of α₁ and α₂. (Top) α₁ = α₂ = 0, (middle) α₁ = 0.75 and α₂ = 0.1, and (bottom) α₁ = 0.25 and α₂ = 0.75. (D) Population distribution of E (top), hybrid E/M (middle), and M (bottom) cells for varying values of α₁ and α₂. In panel A; 1.5374e-05 means 1.5374 × 10⁻⁵.

Figure 5

KLF4 correlates with patient survival in a cancer-specific manner. (A,B) Relapse-free survival trends in GSE42568 and GSE3494 (breast cancer), respectively. (C,D) Same as (A,B) but for the overall survival. (E,F) Overall survival trends in GSE26712 and GSE30161 (ovarian cancer), respectively. (G,H) Overall survival trends in GSE30219 and CaArray (lung cancer), respectively. HR denotes the hazard ratio, and logrank P denotes the p-value. The mean value and 95% confidence interval are shown. In panel B; 2.2e-05 means 2.2 × 10⁻⁵.