Modeling the effects of EMT-immune dynamics on epithelial cancer progression

Daniel Bergman, Matthew Karikomi, Min Yu, Qing Nie, Adam L. MacLeanbioRxiv 2019 · DOI

TL;DR

This paper introduces a novel multiscale model to simulate the complex interplay between tumor cells undergoing epithelial-to-mesenchymal transition (EMT), the immune system, and cancer progression. By analyzing in silico patient data and real-world TCGA data, they identify key EMT-related properties that influence invasion-free survival and predict novel gene interactions that could be targeted for therapeutic intervention.

What Problem Does This Paper Address?

Cancer progression from an early, localized (in situ) stage to an invasive, metastatic disease is a complex process driven by multiple factors. Two critical players are the immune system, which attempts to clear cancerous cells, and epithelial-to-mesenchymal transition (EMT), a cellular plasticity mechanism that allows cells to become more mobile and resistant. While both are known to influence cancer, their intricate crosstalk and combined impact on progression, particularly the transition from in situ to invasive disease, remain poorly understood. This paper aims to bridge this knowledge gap by developing a computational model that integrates these three components to better understand and predict cancer progression.

Methods

The core of this research is a multiscale agent-based model (ABM). In this model:

  • Tumor cells are simulated as individual agents, each with its own state (epithelial or mesenchymal) and potential for mutation.
  • Immune cells (NK cells, CTLs, and Tregs) and cytokines (like TGF-β) are modeled as continuous variables, representing their overall populations and concentrations within the tumor microenvironment (TME).
  • Tumor evolution is driven by cell proliferation, apoptosis, and immune clearance, with mutations accumulating over time that can alter these processes and lead to invasiveness.
  • EMT is modeled as a dynamic process influenced by TGF-β, where cells can transition between epithelial and mesenchymal states. Mesenchymal cells are characterized by reduced proliferation (mesenchymal growth arrest, MGA) and increased immune evasion (MIE).
  • Immune dynamics involve cytotoxic cells (NKs, CTLs) clearing tumor cells, and regulatory cells (Tregs) suppressing this activity and producing TGF-β, creating a feedback loop.
  • Inflammation is incorporated as a cycling variable, alternating between high and low states, affecting immune cell activity.

To analyze the model's behavior and identify critical factors, the authors employed global sensitivity analysis using the Morris method. They also developed a novel TCGA data analysis framework to link gene expression data with patient survival outcomes, specifically focusing on EMT and inflammation-related genes.

Datasets

The primary dataset used for validation and prediction was the The Cancer Genome Atlas (TCGA). Specifically, they utilized:

  • Primary tumor bulk mRNA sequencing data and censored survival data for various cancer types.
  • Disease-free interval (DFI) data was used as a proxy for invasion-free survival.

Key Insights and Ideas

  • EMT is not just about metastasis: The study highlights that EMT plays a crucial role throughout the progression from in situ to invasive cancer, not just in the final metastatic steps.
  • A delicate balance is key: The model reveals that while increased mesenchymal immune evasion (MIE) generally worsens outcomes, mesenchymal growth arrest (MGA) can paradoxically improve invasion-free survival under specific conditions, particularly when periods of low inflammation are present. This suggests a "sweet spot" for MGA that can help contain the tumor.
  • Inflammation's dual nature: The inflammatory state of the TME has complex and sometimes counter-intuitive effects on cancer progression, influencing the impact of EMT properties.
  • TGF-β as a central hub: TGF-β emerges as a critical mediator, linking tumor-immune interactions with EMT and significantly shaping tumor outcomes.
  • Predicting therapeutic targets: The TCGA analysis, guided by the model's predictions, identified specific gene pairs (e.g., FGFR2 and FBXW4 in bladder cancer) whose co-expression is strongly linked to patient prognosis, offering potential new targets for therapy.

Key Results

  • The agent-based model successfully recapitulated invasion-free survival dynamics observed in silico.
  • Global sensitivity analysis identified Treg recruitment, CTL recruitment, and inflammation parameters as highly influential on time to invasion.
  • The model predicted that increasing mesenchymal immune evasion (MIE) consistently shortens invasion-free survival.
  • A non-monotonic relationship was found between mesenchymal growth arrest (MGA) and invasion-free survival, with an optimal MGA value maximizing survival, especially under low inflammation.
  • Analysis of TCGA data identified three cancer types (bladder, uterine, and liver) where EMT and inflammation-associated genes significantly predicted patient survival.
  • The study predicted novel gene interactions, such as the synergistic tumor-suppressive effect of FGFR2 and FBXW4 in bladder cancer, and opposing roles for FGFR2 in bladder versus uterine cancer.

Limitations and Future Work

The authors acknowledge several limitations and suggest avenues for future research:

  • Inflammation module simplification: The current model treats inflammation as a simple cycling scheme. Future work could link inflammation levels to tumor mutational burden or other biological factors.
  • TGF-β as the sole EMT driver: While TGF-β is a major player, EMT is influenced by multiple signaling pathways. Incorporating these additional factors would enhance model realism.
  • Tumor heterogeneity: While the model accounts for individual cell states, a deeper exploration of the consequences of increased heterogeneity during disease progression is warranted.
  • Experimental validation: The study calls for experimental work to validate the predicted effects, particularly regarding TGF-β's role in tumor suppression and the identified gene interactions.
  • Deeper analysis of transient dynamics: Further investigation into the short-term cell dynamics within tumors in situ and during progression could provide more granular insights into EMT's role.

Summary updated

Abstract

D uring progression from carcinoma in situ to an invasive tumor, the immune system is engaged in complex sets of interactions with various tumor cells. Tumor cell plasticity also alters disease trajectories via epithelial-to-mesenchymal transition (EMT). Several of the same pathways that regulate EMT are involved in tumor-immune interactions, yet little is known about the mechanisms and consequences of crosstalk between these regulatory processes. Here we introduce a multiscale evolutionary model to describe tumor-immune-EMT interactions and their impact on epithelial cancer progression from in situ to invasive disease. Through in silico analyses of large patient cohorts, we find controllable regions that maximize invasion-free survival. We identify that delaying tumor progression depends crucially on properties of the mesenchymal tumor cell phenotype: its growth rate and its immune-evasiveness. Through analysis of EMT-inflammation-associated data from The Cancer Genome Atlas, we find that association with EMT significantly worsens invasion-free survival probabilities in support of our model, and we predict new genes influencing outcomes in bladder and uterine cancer, including FGF pathway members. These results offer novel means to delay disease progression by regulating properties of EMT through specific gene interactions, and demonstrate the importance of studying cancer-immune interactions in light of EMT.
BibTeX
@misc{bergman2019modeling,
  author = {Daniel Bergman and Matthew Karikomi and Min Yu and Qing Nie and Adam L. MacLean},
  title = {{Modeling the effects of EMT-immune dynamics on epithelial cancer progression}},
  year = {2019},
  doi = {10.1101/615971},
  howpublished = {bioRxiv},
}
Figure