Bistable switches control memory and plasticity in cellular differentiation

Abstract

Development of stem and progenitor cells into specialized tissues in multicellular organisms involves a series of cell fate decisions. Cellular differentiation in higher organisms is generally considered irreversible, and the idea of developmental plasticity in postnatal tissues is controversial. Here, we show that inhibition of mitogen-activated protein kinase (MAPK) in a human bone marrow stromal cell-derived myogenic subclone suppresses their myogenic ability and converts them into satellite cell-like precursors that respond to osteogenic stimulation. Clonal analysis of the induced osteogenic response reveals ultrasensitivity and an "all-or-none" behavior, hallmarks of a bistable switch mechanism with stochastic noise. The response demonstrates cellular memory, which is contingent on the accumulation of an intracellular factor and can be erased by factor dilution through cell divisions or inhibition of protein synthesis. The effect of MAPK inhibition also exhibits memory and appears to be controlled by another bistable switch further upstream that determines cell fate. Once the memory associated with osteogenic differentiation is erased, the cells regain their myogenic ability. These results support a model of cell fate decision in which a network of bistable switches controls inducible production of lineage-specific differentiation factors. A competitive balance between these factors determines cell fate. Our work underscores the dynamic nature of cellular differentiation and explains mechanistically the dual properties of stability and plasticity associated with the process.

Fig. 1.

Human myogenic progenitors derived from bone marrow stromal cells. (A) Marrow stromal cells (MSC) expressed the CD105, CD90, and CD44 antigens. WB15-M cells lost expression of these antigens and gained expression of α7-integrin. (B) WB15-M cells expressed MyoD and Myf5. (C) Expression of myogenin was enhanced when cells were cultured in low serum. (D) Low-serum culture induced formation of multinucleated myofibers that expressed α7-integrin, sarcomeric α-actinin, myosin heavy chain (MyHC), dystroglycan, and dystrophin.

Fig. 2.

MAPK-inhibited WB15-M cells expressed satellite cell markers and exhibited osteogenic potential. (A) PD and SB treatment of WB15-M cells suppressed MyoD expression and enhanced Pax7 and Msx1 expression (red nuclear signals). DAPI staining was in blue. (B) Effect of MAPK inhibition could be reversed after removal of MAPK inhibitors. (C) MAPK-inhibited cells expressed ALP when stimulated with BMP2 (red reaction), and untreated cells did not. (D) BMP-stimulated ALP expression was dose dependent. (E) PD-inhibited and BMP-stimulated cells produced von Kossa-positive calcium deposits (arrows).

Fig. 3.

MAPK-inhibited WB15-M cells demonstrated an all-or-none and ultrasensitive response. (A) Colony-forming assays of WB15-M cells treated with MAPK inhibitors and stimulated with BMP2. (B) Each PD-treated colony showed an all-or-none response, and ALP expression exhibited a bimodal population distribution (histograms). (C) Experimental dose–response curve showing a sharp rise in ALP-positive colonies in PD-treated cells (squares) with increasing BMP2 doses. Untreated cells (triangles) had no response.

Fig. 4.

Bistable switch model of cellular differentiation. (A) Schematic representation of a bistable switch model. (B) Phase diagram showing x against β at different values of α. (C) The boundary between bistable and monostable domains at different values of n. The dashed line indicates α = 1.1. (D) A curve of x as a function of BMP2, showing a sharp jump in x when the BMP2 dose crosses the upper boundary of the bistable domain (arrow). Hill parameter n = 8. (E) Simulated behavior of cells at different noise levels, showing probabilistic transitions between states. (F) Effect of noise on the shape of the dose–response curves. (G) Simulated dose–response curve for PD-treated cells (squares) and untreated cells (triangles) at a 5% noise level.

Fig. 5.

MAPK-inhibited WB15-M cells exhibited BMP-induced memory. (A) MAPK-inhibited WB15-M cells were untreated (Upper) or pretreated with 300 ng/mL of BMP2 (Lower) before being challenged with BMP2. (B) The dose–response curves for the BMP2-pretreated cells (squares) and untreated cells (circles) formed a hysteresis loop. (C) The schematic path by which a BMP2-pretreated cell switched from an “on” state to an “off” state. (D) Boundaries of the bistable domain in relationship to the value of L. The dashed line marked L = 100 ng/mL. (E) Interruption of BMP2 treatment in PD-treated WB15-M cells for 0, 2, or 4 days caused loss of memory. (F) Exposure of the cells to cycloheximide (10 or 100 μg/mL) for 1 day also caused loss of memory. (G) Decline of x with time as a function of α. (H) PD-treated WB15-M cells were stimulated with varying doses of BMP2, and ALP expression was measured after various days of culture. (I) Expected changes in ALP-positivity with days, as predicated by the model, in colonies stimulated with BMP2 at 150, 200, and 250 ng/mL (dotted, dashed, and solid lines), contrasting with the changes actually observed (filled circles, open circles, and filled triangles). (J) Expanded views of individual colonies analyzed on day 8 and day 10 showing an all-or-none response.

Fig. 6.

Bistable switches control developmental plasticity. (A) MAPK-inhibitor pretreated cells (Upper Left and Center) and untreated cells (Lower Left and Center) were stimulated with BMP2. MAPK inhibitor pretreatment enhanced the response to BMP2 stimulation, even when the inhibitors were interchanged (Right). (B) WB15-M cells pretreated with PD for various numbers of days were stimulated with BMP2 (Left). The number of ALP-positive colonies was proportional to the days of PD pretreatment (Right). (C) Interruption of PD treatment for 2 or 4 days before BMP2 challenge erased the memory induced by PD. (D) In a 2-step differentiation assay, PD- and BMP2-pretreated cells formed either ALP+ or MyHC+ colonies under the conditions for osteogenesis (PD+BMP2; Upper Left) or myogenesis (Low Serum; Upper Center). Cells without pretreatment only formed myogenic colonies (Lower Center and Right). (E) Model of interacting bistable switches that control cell fate and differentiation. (F) Energy landscape of the model showing trajectories (lines) and final states (circles) of lineage and cell fate factors under untreated (−PD) or osteogenic (+PD+BMP2) conditions. (G) Simulated changes in factor levels with time in cells with or without PD pretreatment, placed under osteogenic (+PD+BMP2) or myogenic (−PD+Low Serum) conditions.