Toward Modeling Context-Specific EMT Regulatory Networks Using Temporal Single Cell RNA-Seq Data

Abstract

Epithelial-mesenchymal transition (EMT) is well established as playing a crucial role in cancer progression and being a potential therapeutic target. To elucidate the gene regulation that drives the decision making of EMT, many previous studies have been conducted to model EMT gene regulatory circuits (GRCs) using interactions from the literature. While this approach can depict the generic regulatory interactions, it falls short of capturing context-specific features. Here, we explore the effectiveness of a combined bioinformatics and mathematical modeling approach to construct context-specific EMT GRCs directly from transcriptomics data. Using time-series single cell RNA-sequencing data from four different cancer cell lines treated with three EMT-inducing signals, we identify context-specific activity dynamics of common EMT transcription factors. In particular, we observe distinct paths during the forward and backward transitions, as is evident from the dynamics of major regulators such as NF-KB (e.g., NFKB2 and RELB) and AP-1 (e.g., FOSL1 and JUNB). For each experimental condition, we systematically sample a large set of network models and identify the optimal GRC capturing contextspecific EMT states using a mathematical modeling method named Random Circuit Perturbation (RACIPE). The results demonstrate that the approach can build high quality GRCs in certain cases, but not others and, meanwhile, elucidate the role of common bioinformatics parameters and properties of network structures in determining the quality of GRCs. We expect the integration of top-down bioinformatics and bottomup systems biology modeling to be a powerful and generally applicable approach to elucidate gene regulatory mechanisms of cellular state transitions.

Hypoxia, partial EMT and collective migration: Emerging culprits in metastasis

“…Mechanism-based mathematical modelling of these networks has revealed how hybrid E/M states can be attained and stably maintained; for instance, miR-200/ZEB feedback loop can allow three stable states: miR200 high /ZEB low (epithelial), miR200 low /ZEB high (mesenchymal), and miR200 med /ZEB med (hybrid E/M) [ 80 ]. Extensions of this model have revealed the existence and signatures of hybrid E/M phenotypes, and identified molecules that may be contributing to maintaining these phenotypes such as GRHL2 and OVOL1/2, called as ‘phenotypic stability factors’ [ [81] , [82] , [83] , [84] , [85] , [86] , [87] , [88] , [89] , [90] ].…”

Inference of Intercellular Communications and Multilayer Gene-Regulations of Epithelial–Mesenchymal Transition From Single-Cell Transcriptomic Data

“…The core gene circuits for EMT are known to involve multiple molecular components and interactions ( Jia et al, 2017 ; Tian et al, 2019 ; Yang et al, 2020 ), providing mechanisms of the EMT transition process ( Jolly and Levine, 2017 ). Recent time-series scRNA-seq data suggest that EMT is indeed highly context-specific ( Cook and Vanderhyden, 2020 ), calling for the need of inferring EMT regulation circuits from a data-driven approach ( Tanaka and Ogishima, 2015 ; Ramirez et al, 2020 ). Previous works have constructed the GRN of EMT based on the combination of prior knowledge, transcription factor predictions, and model validations from single-cell datasets ( Ramirez et al, 2020 ).…”

“…Recent time-series scRNA-seq data suggest that EMT is indeed highly context-specific ( Cook and Vanderhyden, 2020 ), calling for the need of inferring EMT regulation circuits from a data-driven approach ( Tanaka and Ogishima, 2015 ; Ramirez et al, 2020 ). Previous works have constructed the GRN of EMT based on the combination of prior knowledge, transcription factor predictions, and model validations from single-cell datasets ( Ramirez et al, 2020 ). Here we have incorporated the intercellular communications in the context of analyzing TCs and ICSs to inspect the dynamical change of regulation interactions along the EMT spectrum.…”

“…Computational methods have been developed to infer cell–cell communication networks based on ligand–receptor interactions ( Wang S. et al, 2019 ; Wang Y. et al, 2019 ; Cabello-Aguilar et al, 2020 ; Jin et al, 2020 ) and elucidate how cell–cell communications propagate to downstream target genes through transcription factors ( Browaeys et al, 2020 ). While methods have been developed to infer EMT gene regulatory network (GRN) from RNA-seq single-cell data ( Ramirez et al, 2020 ), the role of cell–cell communications on gene regulation dynamics along EMT trajectory is poorly understood.…”

Data_Sheet_1.PDF

“…Computational methods have been developed to infer cell-cell communication networks based on ligand-receptor interactions (Wang S. et al, 2019;Wang Y. et al, 2019;Cabello-Aguilar et al, 2020;Jin et al, 2020) and elucidate how cell-cell communications propagate to downstream target genes through transcription factors (Browaeys et al, 2020). While methods have been developed to infer EMT gene regulatory network (GRN) from RNAseq single-cell data (Ramirez et al, 2020), the role of cellcell communications on gene regulation dynamics along EMT trajectory is poorly understood.…”