Fig 1.
Experimental evidence for Ovol2-Zeb1 mutual repression.
A) Zeb1 binds to the Ovol2 promoter in mouse mammary epithelial cells. Top, diagram showing the presence of Zeb1 consensus motifs (triangles) in Ovol2 promoter. Bottom, ChIP-PCR using primers for the upstream (left) and downstream sites (right). B) RT-quantitative PCR analysis showing that OVOL2 and ZEB1 overexpression in MCF10A cells results in decreased level of Zeb1 and Ovol2 transcripts, respectively TSS, transcription start site.
Fig 2.
Incorporation of an Ovol2-Zeb1 mutual repression module results in the observation of four distinct states in EMT.
A) Influence diagram of the EMT/MET system. Blue icon: epithelial promoting factor. Yellow icon: mesenchymal promoting factor. Hexagon: extracellular input. B, C) One-parameter bifurcation diagrams of E-cadherin (Ecad) with respect to external TGF-β (B) and Ovol2 basal production rate (C). Solid curve: stable steady state. Dashed curve: unstable steady state. In B, only transition between I1 and I2 is reversible when TGF-β level is varied. In C, varying Ovol2 alone does not result in any reversible transition, but it can possibly reverse the following TGF-β-induced transitions: E-I1, E-I2 and E-M.
Fig 3.
Experimental evidence for bidirectional potential of MCF10A cells.
A, B, D, E) Flow cytometric analysis of epithelial marker (Ecad) and mesenchymal marker (Vim) profiles. A) Direct comparison of MCF10A with luminal (epithelial)-type cancer cell line MCF7 and basal (mesenchymal)-type cancer cell line MDA-MB231. MCF10A (green) falls in the intermediate state between MCF7 (blue) and MDA-MB231 (red). Analyses were performed at 90–100% confluency. B) Bidirectional potential of MCF10A cells. E(I)MT and M(I)ET was induced by forced expression of transcription factors Snail or Zeb1, and Ovol2, respectively. After 6 days of lentiviral infection, Snail (red) and Zeb1 (orange) induced EMT while Ovol2 (blue) induced MET as compared to the empty vector control (green). C) Stochastic simulations for a population of 2000 cells in three different conditions: basal parameter set (green), high basal production rate of Zeb1 ZEB1 (red, Zeb1 mRNA basal production rate was raised to 0.01 μM/hr) and high basal production rate of Ovol2 (blue, Ovol2 basal production rate was raised to 2 μM/hr). Initial conditions are all at I1 state. D) The histogram shows the expression status of CD44. “M” states (Snail;red and Zeb1;orange) correlate with high CD44 expression while cells in “E” state (Ovol2;blue) show low CD44 expression as compared to empty vector control (green). MCF7 is shown as a representative cell type in the “E” state with low CD44 expression. E) Ovol2 reprograms MDA-MB231 cells from M- to E- state. Cells were analyzed after 6 days of control (red) or Ovol2-expressing (blue) lentiviral infection. F) Stochastic simulations with a basal parameter set (red) and high basal production rate of Ovol2 (blue, Ovol2 basal production rate was raised to 2 μM/hr). Initial conditions are all at M state.
Fig 4.
Time series of Ecad/Vim profile change upon Ovol2 expression in MDA-MB231 cells.
Cells were infected with Ovol2-expressing lentivirus and Ecad/Vim profile was analyzed by flow cytometry at the indicated time points. Empty vector control at day 5 is shown at the top.
Fig 5.
Stepwise induction of EMT in MCF10A cells by different doses of TGF-β.
A) Cells were treated with various concentrations of TGF-β for 10 days and analyzed for Ecad/Vim expression by flow cytometry. Each panel shows a superimposed image of two treatment conditions. Note that the 25 μM TGF-β treatment gave rise to a heterogeneous population containing I2 cells and M cells (orange). B) The corresponding steady states of the cellular phenotypes (indicated as colored dots) observed under various TGF-β concentrations in A are mapped to the bifurcation diagram shown in Fig 2B. C) An illustrative summary of phenotypic transitions in the four-state system. Solid arrow represents transition with experimental evidence from this study. Dotted arrow represents hypothetical transition without experimental evidence.
Fig 6.
Roles of EMT-promoting and -inhibiting factors in the four-state EMT system.
A) Two-parameter bifurcation diagram with respect to external TGF-β and Ovol2 basal production rate. The red curves were computed by extending the saddle-node bifurcation points obtained in one parameter bifurcation analysis (S4 Fig), and they define different parameter regions that can be mono-stable, bi-stable, tri-stable or tetra-stable depending on the number of possible stable phenotypes (see labels), and each multi-stable region can be viewed as an area where multiple phenotypes co-exist (II-V). The size of each phenotype region is an indication of robustness of the phenotype when the two signals are varied. Green star: a basal parameter set and an intermediate TGF-β concentration that together give rise to four phenotypes. B, C) One-parameter bifurcation diagrams of Ecad with respect to external TGF-β (B) and Ovol2 basal production rate (C). Solid curve: stable steady state. Dashed curve: unstable steady state. A basal parameter set (blue) and a perturbed parameter set (orange) are compared in each plot. Triangles and diamonds denote the conditions under which both I1 and I2 are stable.
Table 1.
Influence of blocking mutual inhibition loops on the two intermediate states.
Fig 7.
Roles of the Ovol2-Zeb1 mutual inhibition loop in the four-state EMT system.
Comparison of the basal model (left column), reduced Ovol2-Zeb1 mutual inhibition (middle column), and blocked Ovol2-Zeb1 mutual inhibition (right column) on the four phenotypes. Each subplot is a two-parameter bifurcation diagram similar to Fig 6A. Subplots in each column highlight the various phenotypes in one condition. Shaded areas are highlighted phenotypes. Colors of the shading correspond to the colored labels on the right.
Fig 8.
Roles of the miR34a-Snail and miR200-Zeb1 mutual inhibition loops in the four-state EMT system.
Comparison of removing miR34a-Snail mutual inhibition (left column), miR200-Zeb1 mutual inhibition (middle column), or both (right column) on the four phenotypes. Each subplot is a two-parameter bifurcation diagram similar to Fig 6A. Subplots in each column highlight the various phenotypes in one condition. Shaded areas are highlighted phenotypes. Colors of the shading correspond to the colored labels on the right.
Fig 9.
Distinct differentiation propensities of the two intermediate states.
A, B, D, E) Stochastic simulations for a population of 5000 cells in four different conditions. Basal parameter set and intermediate external TGF-β concentration (0.5) were used (green star in Fig 6A). A) Initial condition: I1; small fluctuations. B) Initial condition: I1; large fluctuations. D) Initial condition: I2; small fluctuations. E) Initial condition: I2; large fluctuations. G, H) Stochastic simulations for a population of 5000 cells initially at I1 state. Ovol2 basal production level was reduced by 20% from basal parameter. J, K) Stochastic simulations for 5000 cells initially at I2 state. Ovol2 basal production level was increased by 100% from basal parameter. C, F, I, L) Metaphoric energy landscapes (green curves) for I1 (C), I2 (F) initial conditions, and reduced (I) or increased (L) Ovol2 basal expression rate. Orange circle represents the initial condition.