Université Libre de Bruxelles Modeling cell fate specification during early embryonic development in mouse Didier Gonze Unité de Chronobiologie Théorique Faculté des Sciences Université Libre de Bruxelles Lille, 15 November 2016 Workshop "Systems Biology: modeling and formal analysis in Biology"
Embryonic development Zygote Cell division Distinct cell types
Embryonic development
"Differentiation tree" Early embryonic development Zhou and Huang, 2011
Pre-implentation embryogenesis in the mouse E = Embryonic day Chazaud, 2008
Stem cells and Epiblast ES Organism itself Extra-embryonic tissues (yolk sac, placenta) Hemberger et al., 2009 Embryonic stem cells (ES) are pluripotent cells derived from the early embryo (epiblast). Meregalli et al., 2011
Transcription factors Cell fate choice is controled by genetic regulatory networks (involving key transcription factors) and by epigenetic regulation markers of pluripotent ICM cells markers of trophectoderm Burton, Torres-Padilla (2014) Nature Reviews Molecular Cell Biology 15:723 735
Binary cell fate and bistability Cross-inhibition between 2 transcription factors can account for binary cell fate choice Zhou & Huang, Trends Genet 2011
Binary cell fate and bistability Cross-inhibition between 2 transcription factors can account for binary cell fate choice Toggle switch (Gardner et al., 2000)
Nanog and Gata6 Nanog and Gata6 are the transcription factors that characterize the Epi and PrE fate, respectively Epiblastic cells : high level of the Nanog transcription factor Primitive endoderm cells: high level of the Gata6 transcription factor In Nanog -/- mutants, all cells adopt the PrE fate In Gata6 -/- mutants, all cells adopt the Epi fate Nanog Epi Gata6 PrE Guo et al., 2010
Nanog and Gata6 Guo et al., 2010 Gata6 and Nanog are co-expressed at increasing levels in the ICM before EPI/PrE differentiation à How? How does differentiation occur afterwards? mathematical modelling
Self-activations and tristability Nanog and Gata6 undergo self-activations. Cross-inhibition and auto-activation give rise to tristability. N (Nanog) G (Gata6) 2.5 Epi Nullclines Epi Trajectories 2 ICM Nanog 1.5 1 ICM 0.5 0 0 0.5 1 1.5 2 2.5 Gata6 PrE PrE See also Zhou and Huang, 2011
Role of FGF-ERK signalling ERK/Fgf4 signalling plays a key role in cell fate specification No ERK à no Gata6 à no PrE Over-activated ERK à all Gata6 à all PrE Fgf4 In presence of high level of Fgf4, no EPI cells are produced. (100% of the embryos do not contain Epi cells) Yamanaka et al., 2010 Frankenberg et al., 2013
Incorporation of FGF-ERK in the model ERK signalling interacts with the gene regulatory network G (Gata6) N (Nanog) FR (FgfR2 total) ERK (ERK actif/erk total) Fp (Fgf4)
Equations of the model 4 variables: G = concentration of Gata6 N = concentration of Nanog FR = concentration of Fgf4 receptor ERK = activity of the ERK pathway Key parameter: Fp = concentration of Fgf4 dg dt =! vsg1 ERK r Kag1 r + ERK + vsg2 G s $ Kig q # & " r Kag2 s + G s % Kig q + N kdg G q dn dt =! vsn1 Kin1 u Kin1 u + ERK + vsn2 N v $ # " u Kan v + N v Kin2 w & % Kin2 w + G kdn N w dfr dt = vsfr1 Kifr Kifr + N + vsfr2 G Kafr + G kdfr FR derk dt = va FR Fp Kd + Fp 1 ERK Ka +1 ERK vin ERK Ki + ERK
Tristability Fgf4 is the key parameter governing tristability in the core regulatory network
Dynamics and mechanism of specification Fgf4 secretion profile in vivo Fgf4 is high when Nanog and Gata6 start to be expressed Fgf4 decreases until the establishment of the ICM cell fate Fgf4 is not expressed by PrE cells Fgf4 is highly expressed by Epi cells Guo et al., 2010
Dynamics and mechanism of specification Dynamics of the system => cell-cell interactions Nanog activates the synthesis and secretion of Fgf4 in the extracellular medium FGFR2 In turn extracellular Fgf4 binds Fgf4 receptors (FgfR2) of neighbouring cells, which activate ERK signalling, leading to an increase of Gata6 and an inhibition of Nanog transcription Frankenberg et al., Dev Cell 2011
Dynamics and mechanism of specification G# (Gata6#)## N# (Nanog)## Incorporation of Fgf4 secretion in the model FR## (FgfR2)## Fp#(Fgf4)# FGF4 secretion ERK## (ac-ve#erk#/#total#erk)## dg dt =! vsg1 ERK r Kag1 r + ERK + vsg2 G s $ Kig q # & " r Kag2 s + G s % Kig q + N kdg G q dn dt =! vsn1 Kin1 u Kin1 u + ERK + vsn2 N v $ # " u Kan v + N v Kin2 w & % Kin2 w + G kdn N w dfr dt = vsfr1 Kifr Kifr + N + vsfr2 G kdfr FR Kafr + G derk dt = va FR Fp Kd + Fp 1 ERK Ka +1 ERK vin ERK Ki + ERK
Dynamics and mechanism of specification Average FGF4 concentration depends on the 2 cells Heterogeneous FGF4 distribution (=> parameter γ)
Dynamics and mechanism of specification Cell 1 Both cells initially co-express Nanog and Gata6 In a second phase, Cell 1 differentiates into Epi Cell 2 differentiates into PrE Cell 2
Mechanism of specification in a "toy" 2-cell model (1) Fgf4 concentration is initially in the domain of tristability (initial conditions). Gata6 and Nanog are very low in both cells. EPI PrE ICM ICM PrE EPI
Mechanism of specification in a "toy" 2-cell model (2) Fgf4 concentration decreases due to its degradation in the extracellular medium. Both cells tends to the intermediary ICM state EPI Fgf4 Fgf4 PrE ICM ICM PrE EPI
Mechanism of specification in a "toy" 2-cell model (3) The cell that perceives slightly less Fgf4 jumps to the Epi state and starts to secrete large amounts of Fgf4 (high Nanog). Fgf4 thus increases. EPI Fgf4 Fgf4 PrE ICM ICM PrE EPI
Mechanism of specification in a "toy" 2-cell model (4) At high Fgf4 concentration the second cell jumps to the PrE state. EPI Fgf4 Fgf4 PrE ICM ICM PrE EPI
Spatial organisation salt & pepper pattern After a transient increase of both Nanog & Gata6, they tend to be mutually exclusive. High-Nanog cells (Epi) and high- Gata6 cells (PrE) first appear to be well mixed (salt & pepper pattern, see stages E3.5-E4). In a sebsequent step (E4.5), a sorting process takes place: Epi cells move to the border, close to the cavity, whereas PrE cells rather concentrate in the inside of the embryo.
Model of a population of 25 cells - salt & pepper pattern
Model of a population of 25 cells - salt & pepper pattern
Model of a population of 25 cells - salt & pepper pattern t=0 t=26 t=35 t=50
Model of a population of 25 cells - prediction (1) Prediction: earlier specification of Epiblastic cells Model prediction: Average time for specification: Experiments: At the 32-cell stage Epi cells : 18 (arbitrary time units) PrE cells: 28 (arbitrary time units) Epi cells are specified a bit earlier than PrE cells. ICM EPI PrE
Model of a population of 25 cells - prediction (2) Prediction: Gata6 +/- heterozygous mutant Experiments: Model: EPI PrE Assumption : the level of Gata6 expression is reduced by 30% in the Gata6 +/- mutant
Model of a population of 25 cells - mutant Nanog -/- In Nanog -/- mutant, ERK inhibitor impacts the expression of Gata6 (D-D'') Cell composition of embryos after different timing of ERK inhibitor treatment. When ERK inhibitor is administrated early, Gata6 does not accumulate (D, D'). When ERK inhibitor is administrated late, Gata6 accumulates (D''). Frankenberg et al, Dev Cell 2011
Model of a population of 25 cells - mutant Nanog -/- Statistics on the 25-cell model Bifurcation analysis on the singlecell model
Effect of interfering with the Fgf/ERK signalling pathway Epi ICM PrE va=0 simulates the treatment with Fgf/Erk inhibitors vex=0.12 simulates the addition of exogenous Fgf4
Possible origin of randomness Key assumption in our model: Local extracellular concentration of Fgf4 is subject to noise (γ) The salt-and-pepper pattern can indeed not be rescued by the addition of exogenous (hence homogenous) Fgf4 in Fgf4 -/- embryos Kang et al., 2013 Other possible sources of heterogeneity: molecular noise (intrinsic variability) cell division (unequal partition of the molecules into the daughter cells, variability in the cell division time, spatial reorganization, etc)
Internal fluctuations as an alternative source of randomness? Simulation of the model in presence of molecular fluctuations (Gillespie)
Internal fluctuations as an alternative source of randomness? Heterogeneity in extracellular Fgf4 Internal fluctuations
Towards a 3D, multi-scale model...
Towards a 3D, multi-scale model...
Towards a 3D, multi-scale model... Effect of cell division Influence of period of division (τ) Too short period leads to an undeveloped embryo Influence of noise on individual cell division (δ) No influence Influence of noise on the extracellular Fgf4 (γ) A certain amount of noise (heterogeneity) is necessary for the appearance of the salt and pepper pattern Influence of noise on concentrations at division (η) Noise can substitute (or combine with) the Fgf4 heterogeneities
Conclusions New specification mechanism, based on self-modulated tristability. The model accounts for observations on the Wild-Type and on mutant embryos, in normal conditions and when the embryos are treated with substances interfering with the Fgf/Erk pathway. The model leads to testable predictions (earlier specification of Epiblastic Cells, Gata6 +/- mutants). The model suggests that the existence of heterogeneities in extracellular Fgf4 signalling is the most probable source of randomness responsible for the salt-and-pepper pattern. The model is currently extended to include cell division and the 3D dynamics. Bessonnard et al. (2014) Development 141: 3637. De Mot et al. (2016) Biophys. J 110: 710-22.
Acknowledgments Unité de Chronobiologie Théorique, ULB Laurane De Mot Alen Tosenberger Geneviève Dupont Albert Goldbeter Financial support Claire Chazaud GReD Clermont-Ferrand Sylvain Bessonnard Institut Pasteur Paris Michel Cohen-Tannoudji Institut Pasteur Paris
Appendix
Huang model Nanog (X) Gata6 (Y)
Network architecture