NRF2-dependent Epigenetic Regulation can Promote the Hybrid Epithelial/Mesenchymal Phenotype

Abstract

The epithelial-mesenchymal transition (EMT) is a cellular process critical for wound healing, cancer metastasis and embryonic development. Recent efforts have identified the role of hybrid epithelial/mesenchymal states, having both epithelial and mesehncymal traits, in enabling cancer metastasis and resistance to various therapies. Also, previous work has suggested that NRF2 can act as phenotypic stability factor to help stablize such hybrid states. Here, we incorporate a phenomenological epigenetic feedback effect into our previous computational model for EMT signaling. We show that this type of feedback can stabilize the hybrid state as compared to the fully mesenchymal phenotype if NRF2 can influence SNAIL at an epigenetic level, as this link makes transitions out of hybrid state more difficult. However, epigenetic regulation on other NRF2-related links do not significantly change the EMT dynamics. Finally, we considered possible cell division effects in our epigenetic regulation model, and our results indicate that the degree of epigenetic inheritance does not appear to be a critical factor for the hybrid E/M state stabilizing behavior of NRF2.

Keywords: Nrf2; epigenetic regulation; epithelial mesenchymal plasticity; epithelial-mesenchymal transition; hybrid epithelial/mesenchymal state.

FIGURE 1

Epigenetic feedback-mediated dynamics of NRF2 coupled to core EMT circuitry. (A) A regulatory network for EMT that consists of two microRNA-TF mutually inhibitory circuits: miR-34/SNAIL and miR-200/ZEB. Signal I represents external EMT-inducing signals such as HGF, NF- $\kappa$ B, Wnt, TGF- $\beta$ and/or HIF1α. NRF2 module is added to the core networks by three distinct links. Blue links represent the inhibition from Keap1 and E-cad on NRF2 respectively, and the red link represents the inhibition on SNAIL from NRF2.(B–D) Bifurcation diagrams of miR-200 levels for the network shown in Panel 1A, with I as the bifurcation parameter. Solid lines represent stable states, dashed lines represent unstable states. Black lines correspond to the circuit without epigenetic regulation, and blue lines correspond to a circuit including epigenetic feedback. In (B), the epigenetic feedback is on the inhibition of SNAIL by NRF2. In (C), the epigenetic feedback is on the inhibitory link from E-cadherin to NRF2. In (D), the epigenetic feedback is on the inhibtion of NRF2 by Keap1.

FIGURE 2

Population dynamics of NRF2 mediated epigenetic feedback (A) A sample dynamic plot without epigenetic feedback. (B) A sample dynamic plot with feedback on the inhibition of SNAIL by NRF2. (C) Simulations showing the population change as a function of time. The percentage is calculated based on 1,000 independent simulations. There is no epigenetic regulation. (D) Same as (C) but now including epigenetic feedback on the inhibitory link from NRF2 to SNAIL. Epigenetic feedback by NRF2 stablizes hybrid state when competing with mesenchymal state.

FIGURE 3

(A) Bifurcation results (black lines represent no feedback, blue lines represent feedback on NRF2’s inhibition on SNAIL, and the mean of signal is fixed at 140 K (purple dotted line). (B) Population analysis without epigenetic feedback for 100 cells. (C) Population analysis with epigenetic feedback for 100 cells.

FIGURE 4

Population dynamics for the circuit including GRHL2 with/without epigenetic feedback. (A) New circuit with GRHL2. (B) Same as Figure 3A, purple dotted line represents low signal, and green dotted line represents high signal.(C) Starting from I = 170 K molecules, we show dynamic examples and population results for models with or without epigenetic feedback obtained by reducing the signal I to 102 K molecules after short high signal durations (high signal ending time = 4  $\varsigma$ ). (D) Same as Figure 3C with high signal ending time = 10  $\varsigma$ . (E) Same as Figure 3C with high signal ending time = 30  $\varsigma$ .

FIGURE 5

(A) Population changes as a function of time in a model including both cell division and epigenetic regulation, and where the noise of the threshold x0ms is increasing (top = 0 k, median = 100 K, bottom = 300 K). (B) Dynamic example for different initial x0ms values (top = 581.2 K, median = 562.5 K, bottom = 544.6 K; these are stable values of x0ms when I = 120 from Figure 4C). (C) Bifurcation diagram for x0ms based on inpit signal I, for the model including both epigenetic regulation on SNAIL’s inhibition from NRF2 and NRF2’s inhibition from Keap1.