The Application Of Core Endogenous Network Hypothesis And The Dynamic Methods In Deciphering Pancreatic Cell Fate Decisions And Regulatory Networks

The maturation of different pancreatic cells is coordinated by sequential cell fate decisions during development.The classical differentiation hierarchy of pancreatic cells describes that multipotent progenitors(MPs)first bifurcate into tip cells and trunk cells,and then these cells give rise to acinar cells and endocrine/ductal cells separately.However,lineage tracings reveal that pancreatic progenitors are highly heterogeneous in tip domain and tip-trunk adjacent region in embryonic pancreas.The pancreatic progenitor fate decisions from multipotency to unipotency are insufficiently characterized.Moreover,in pancreas development,lineage-tracing experiments show that a proportion of PTF1A+ progenitors in the acinar lineage and SOX9+ progenitors in the ductal lineage can experience lineage conversions and mature into cells in other lineages.These lineage conversion paths cannot be explained by the already established classical hierarchy.Compared to cell lineage tracing method and single-cell sequencing data analysis,dynamics modeling of gene regulatory network can well reflect the logic and complexity of cell fate decisions.However,due to the complexity of the genetic regulation,it is difficult to accurately obtain the complete regulatory network from known knowledge.Moreover,gene regulatory networks(GRNs)inferred from sequencing data based on bioinformatic or statistical methods cannot give reliable predictions of cell fate decisions from dynamic modeling sine the dynamic modeling heavily relies on the accuracy of the network structure.This requires us to propose a "simplified" regulatory network model for development that can be used to simulate and predict cell fate decisions.The molecular-cellular core endogenous network hypothesis was originally proposed to explain the mechanism of carcinogenesis.In this paper,we extend the core endogenous network hypothesis to the development process and propose the core endogenous network hypothesis for development.The core endogenous network hypothesis for development and the quantification methods make it possible to decipher the complex cell fate decision at the core regulatory network level.The core endogenous network hypothesis for development is further applied to decipher the complex pancreatic cell fate decisions.We construct the core endogenous network for pancreatic cell fate decisions based on genetic regulations and quantified its intrinsic dynamic properties using dynamic modeling.A developmental landscape that has not been clarified previously is revealed from the dynamics.Not only well-characterized pancreatic cells are reproduced,but also previously unrecognized progenitors—tip progenitor(TiP),trunk progenitor(TrP),late endocrine progenitor(LEP),and acinar progenitors(AciP/AciP2)are predicted in the development landscape.Further analysis shows that TrP and LEP mediate endocrine lineage maturation,while TiP,AciP,AciP2,and TrP mediate acinar and ductal lineage maturation.The predicted cell types are validated by analyzing single-cell RNA-Sequencing(scRNA-Seq)data.Significantly,to our known knowledge,this is the first time that a redefined hierarchy with detailed early pancreatic progenitor fate commitments corresponding to the classical development hierarchy is obtained.Moreover,we find a set of hyper-transition states mediate the lineage conversions of progenitors TiP and TrP,which naturally explained the "multipotency" of lineage-biased progenitors.Two parallel transition paths from acinar to ductal cells which are mediated by a set of transition and hyper-transition states separately are predicted.We then verified the existence of these transition/hyper-transition states in the mouse pancreatic scRNA-Seq data.The lineage conversion paths that do not lie in the classical hierarchy are also fully revealed from the dynamics of the core endogenous network.Elucidating the structure of GRNs is of significant importance for the mechanistic understanding of the development process and disease progression.With the advance of single-cell techniques,different network inference methods that deal with single-cell gene expression data emerge.However,since the biochemical reactions in the gene transcription process are generally nonlinear,inference methods using linear models cannot capture the nonlinear regulatory effects.Several existing nonlinear models based on the regulatory dynamics are limited to small size networks.Here we designed a nonlinear method by modeling the pairwise gene regulatory dynamics using a sigmoidal function to infer the regulatory relationships from single-cell gene expression data.This method is applicable to network inference including hundreds of genes.We applied this method to infer GRNs regulating the pancreas development and other cell differentiation processes.Results show that this method can give the right inference from nonlinear regulatory effects even when the linear correlation relationships are very weak.This method gives the highest AUC(area under the curve)values in several datasets when compared to other GRN inference methods,which indicates this method has a strong prediction power.In conclusion,the core endogenous network hypothesis and the dynamic methods have important application values in deciphering complex cell fate decisions and gene regulatory networks.