Gene networks reveal stem-cell state convergence during preneoplasia and progression to malignancy in multistage skin carcinogenesis

Mark A. Taylor, Eve Kandyba, Kyle Halliwill, Reyno Delrosario, Matvei Koroshkin, Hani Goodarzi, David A. Quigley, Leslie A. Lange, Di Wu, Saumya Bollam, Olga K. Mirzoeva, Rosemary J. Akhurst, Allan BalmainbioRxiv 2023 · DOI

Abstract

Adult mammalian stem cells play critical roles in normal tissue homeostasis, as well as in tumor development, by contributing to cell heterogeneity, plasticity, and development of drug resistance. The relationship between different types of normal and cancer stem cells is highly controversial and poorly understood. Here, we carried out gene expression network analysis of normal and tumor samples from genetically heterogeneous mice to create network metagenes for visualization of stem-cell networks, rather than individual stem-cell markers, at the single-cell level during multistage carcinogenesis. We combined this approach with lineage tracing and single-cell RNASeq of stem cells and their progeny, identifying a previously unrecognized hierarchy in which Lgr6+ stem cells from tumors generate progeny that express a range of other stem-cell markers including Sox2, Pitx1, Foxa1, Klf5, and Cd44. Our data identify a convergence of multiple stem-cell and tumor-suppressor pathways in benign tumor cells expressing markers of lineage plasticity and oxidative stress. This same single-cell population expresses network metagenes corresponding to markers of cancer drug resistance in human tumors of the skin, lung and prostate. Treatment of mouse squamous carcinomas in vivo with the chemotherapeutic cis-platin resulted in elevated expression of the genes that mark this cell population. Our data have allowed us to create a simplified model of multistage carcinogenesis that identifies distinct stem-cell states at different stages of tumor progression, thereby identifying networks involved in lineage plasticity, drug resistance, and immune surveillance, providing a rich source of potential targets for cancer therapy.

Introduction

One of the central debates in oncology concerns the relationship between normal tissue stem cells and cancer stem cells (CSCs). While phenotypic plasticity—the ability of a cell to switch its identity—is a recognized hallmark of cancer, the rules governing these transitions are poorly understood. Technical limitations often force a choice between the statistical breadth of bulk-tissue data and the high resolution of single-cell RNA sequencing (scRNA-seq). This study addresses this gap by using gene networks to identify functional cell states that individual markers might overlook.

Mapping Stem-Cell States via Network Metagenes

To track skin carcinogenesis, the researchers employed a multi-layered approach. They used Weighted Gene Co-expression Network Analysis (WGCNA) on bulk transcriptomics from hundreds of genetically heterogeneous mice to identify modules of genes that consistently function together. From these, they developed "metagenes"—the aggregate expression of the top 100 genes most highly correlated with a specific "seed" gene (e.g., Lgr6). This provides a robust signal for functional networks at the single-cell level.

Overview of the experimental pipeline, network metagene approach, and the single-cell landscape of skin carcinogenesis.

This computational framework was integrated with lineage tracing in specialized mouse models and scRNA-seq of over 57,000 cells, allowing the team to physically track Lgr6+ stem cells and their descendants as they progressed from normal skin to benign papillomas and malignant carcinomas.

Convergence of Stem-Cell Networks

A defining discovery of this work is the "convergence" of stem-cell pathways during tumor progression. In normal skin, various stem-cell markers and their associated networks are distinct and poorly correlated. However, as tissue progresses toward malignancy, these networks converge.

Spearman correlation heatmaps demonstrating the convergence of stem-cell marker networks in carcinoma compared to normal skin.

This convergence reveals a hierarchy where Lgr6+ cells act as the progenitors for a population expressing a wide array of markers, including Sox2, Pitx1, Foxa1, Klf5, and Cd44. In benign papillomas, this convergence identifies a unique population termed the "lower spike" (LS).

The "Lower Spike": A Hub for Plasticity and Resistance

The LS population is characterized by markers of oxidative stress, wound healing, and "lineage infidelity," where cells lose their specific tissue identity. While the "upper spike" (US) population is focused on rapid division and self-renewal, LS cells are relatively quiescent but highly plastic and stress-resistant.

Significantly, the LS metagene network in untreated benign tumors already expresses the same gene signatures found in human lung, skin, and prostate cancers that have survived chemotherapy. This suggests that the machinery for drug resistance is built into the tumor's architecture long before treatment begins. Indeed, treating mouse carcinomas with cisplatin led to a significant upregulation of the LS metagene network in the surviving, resistant cell population.

Conclusion

By mapping the evolution of stem-cell networks, this study provides a simplified model of multistage carcinogenesis. The identification of the LS population as a precursor to malignancy and a reservoir for drug resistance identifies it as a critical target for future therapies. Future work will investigate whether inhibiting the stress and plasticity pathways within the LS population can prevent the emergence of chemotherapy resistance.

Summary updated

Abstract

Adult mammalian stem cells play critical roles in normal tissue homeostasis, as well as in tumor development, by contributing to cell heterogeneity, plasticity, and development of drug resistance. The relationship between different types of normal and cancer stem cells is highly controversial and poorly understood. Here, we carried out gene expression network analysis of normal and tumor samples from genetically heterogeneous mice to create network metagenes for visualization of stem-cell networks, rather than individual stem-cell markers, at the single-cell level during multistage carcinogenesis. We combined this approach with lineage tracing and single-cell RNASeq of stem cells and their progeny, identifying a previously unrecognized hierarchy in which <i>Lgr6</i>+ stem cells from tumors generate progeny that express a range of other stem-cell markers including <i>Sox2, Pitx1, Foxa1, Klf5</i>, and <i>Cd44.</i> Our data identify a convergence of multiple stem-cell and tumor-suppressor pathways in benign tumor cells expressing markers of lineage plasticity and oxidative stress. This same single-cell population expresses network metagenes corresponding to markers of cancer drug resistance in human tumors of the skin, lung and prostate. Treatment of mouse squamous carcinomas <i>in vivo</i> with the chemotherapeutic <i>cis</i>-platin resulted in elevated expression of the genes that mark this cell population. Our data have allowed us to create a simplified model of multistage carcinogenesis that identifies distinct stem-cell states at different stages of tumor progression, thereby identifying networks involved in lineage plasticity, drug resistance, and immune surveillance, providing a rich source of potential targets for cancer therapy.
BibTeX
@misc{taylor2023gene,
  author = {Mark A. Taylor and Eve Kandyba and Kyle Halliwill and Reyno Delrosario and Matvei Koroshkin and Hani Goodarzi and David A. Quigley and Leslie A. Lange and Di Wu and Saumya Bollam and Olga K. Mirzoeva and Rosemary J. Akhurst and Allan Balmain},
  title = {{Gene networks reveal stem-cell state convergence during preneoplasia and progression to malignancy in multistage skin carcinogenesis}},
  year = {2023},
  doi = {10.1101/2023.05.08.539863},
  howpublished = {bioRxiv},
}
Figure