Tumor cell total mRNA expression shapes the molecular and clinical phenotype of cancer

Shaolong Cao, Jennifer Wang, Shuangxi Ji, Yang Peng, Matthew D. Montierth, Shuai Guo, John Paul Shen, Xiao Zhao, Jingxiao Chen, Jaewon J. Lee, Paola A. Guerrero, Nicholas Spetsieris, Nikolai Engedal, Sinja Taavitsainen, Kaixian Yu, Julie Livingstone, Vinayak Bhandari, Shawna M. Hubert, Najat C. Daw, P. Andrew Futreal, Eleni Efstathiou, Bora Lim, Andrea Viale, Jianjun Zhang, Matti Nykter, Bogdan Czerniak, Pavlos Msaouel, Anirban Maitra, Scott Kopetz, Peter J. Campbell, Terence P. Speed, Paul C. Boutros, Hongtu Zhu, Alfonso Urbanucci, Jonas Demeulemeester, Peter Van Loo, Wenyi Wang

Abstract

Cancers can vary greatly in their transcriptomes. In contrast to alterations in specific genes or pathways, differences in tumor cell total mRNA content have not been comprehensively assessed. Technical and analytical challenges have impeded examination of total mRNA expression at scale across cancers. To address this, we developed a model for quantifying tumor-specific total mRNA expression (TmS) from bulk sequencing data, which performs transcriptomic deconvolution while adjusting for mixed genomes. We used single-cell RNA sequencing data to demonstrate total mRNA expression as a feature of tumor phenotype. We estimated and validated TmS in 5,015 patients across 15 cancer types identifying significant inter-individual variability. At a pan-cancer level, high TmS is associated with increased risk of disease progression and death. Cancer type-specific patterns of genetic alterations, intra-tumor genetic heterogeneity, as well as pan-cancer trends in metabolic dysregulation and hypoxia contribute to TmS. Taken together, our results suggest that measuring cell-type specific total mRNA expression offers a broader perspective of tracking cancer transcriptomes, which has important biological and clinical implications.

Introduction

In the field of transcriptomics, research has traditionally focused on relative gene expression—identifying which genes are upregulated or downregulated compared to others. This focus is largely a byproduct of the technical requirements of RNA sequencing (RNA-seq). Standard normalization techniques flatten the total amount of RNA in every sample to a uniform level to allow for comparison. However, this process inadvertently masks a critical biological variable: the absolute volume of mRNA produced by a cell, or its "transcriptome ploidy."

Some cancer cells are biologically "louder" than others, producing a significantly higher volume of transcripts across the entire genome. This total mRNA content is often linked to increased cell size, rapid proliferation, and metabolic shifts. Until now, measuring this at scale was nearly impossible because bulk tumor samples are a mixture of cancerous and healthy cells, and technical noise often obscures the absolute signal. This study introduces TmS (Tumor-specific total mRNA expression), a novel metric designed to quantify this absolute transcriptional output and explore its clinical significance.

Methodology: Quantifying the Transcriptome Engine

The researchers developed a mathematical framework to estimate TmS from matched DNA and RNA sequencing data. The workflow addresses the inherent complexities of bulk sequencing through three primary steps:

  1. Transcriptomic Deconvolution: Using the DeMixT tool, the model separates the mRNA signal of the tumor from the surrounding healthy stroma and immune cells. This is achieved through a "profile-likelihood" approach that identifies stable "intrinsic tumor signature genes" to anchor the calculations.
  2. Purity and Ploidy Adjustment: Using matched DNA-seq data, the model estimates the tumor purity (the fraction of the sample that is malignant) and DNA ploidy (the number of chromosome sets).
  3. Dosage Correction: Crucially, the model adjusts mRNA levels by the DNA ploidy. This ensures that TmS reflects the mRNA output per haploid genome, isolating the biological "effort" or transcriptional activity of the cell from simple increases caused by having extra copies of DNA (aneuploidy).

To validate this model, the study utilized a massive dataset including single-cell RNA-seq (scRNA-seq) from over 46,000 cells across multiple tumor types, data from 5,015 patients in The Cancer Genome Atlas (TCGA), and specific cohorts from the ICGC and MD Anderson Cancer Center.

Biological Drivers of High TmS

The study revealed that high TmS is not merely a byproduct of genetic instability but represents a distinct biological state characterized by several key features:

  • Stem-like States: Cells with high TmS are frequently in a less-differentiated, "stem-like" state. This cellular plasticity often makes tumors more aggressive and resistant to conventional therapies.
  • Metabolic Reprogramming: A consistent pan-cancer link was found between high TmS and the Pentose Phosphate Pathway (PPP). This metabolic shift is logical: to sustain a massive volume of mRNA production, a cell must ramp up the pathways that provide the necessary nucleotide building blocks.
  • Hypoxia: Low oxygen levels (hypoxia) were consistently associated with elevated TmS. This suggests that environmental stress within the tumor microenvironment may push cells toward a high-transcriptome state as an adaptive survival mechanism.

Clinical Implications and Prognostic Value

The clinical impact of TmS is significant. In pan-cancer analyses, high TmS was a strong predictor of reduced overall survival (OS) and progression-free intervals (PFI). Interestingly, the study identified different patterns of association across cancer types:

  • Pattern I (e.g., Colorectal, Lung, Liver): High TmS consistently predicted poor outcomes regardless of the clinical stage.
  • Pattern II (e.g., Bladder Cancer): In some instances, high TmS predicted better survival in advanced stages. This is likely because these highly metabolically active cells are more sensitive to certain frontline chemotherapies that target rapidly dividing cells.

Furthermore, TmS showed promise as a dynamic biomarker for treatment response. In prostate cancer patients, treatment with the androgen receptor inhibitor Enzalutamide significantly reduced TmS levels, suggesting the metric could be used to monitor whether a drug is successfully "quieting" the tumor's transcriptional engine.

Conclusion

TmS offers a new dimension for understanding cancer transcriptomes. By moving beyond relative expression to measure the absolute "volume" of a tumor's transcriptional output, researchers can better identify aggressive disease and predict patient outcomes. While the current model requires a normal tissue reference and represents a bulk average of the tumor, it provides a foundation for future work into the "molecular volume" of cancer and its potential as a real-time monitor for treatment resistance.

Summary updated

Abstract

Abstract Cancers can vary greatly in their transcriptomes. In contrast to alterations in specific genes or pathways, differences in tumor cell total mRNA content have not been comprehensively assessed. Technical and analytical challenges have impeded examination of total mRNA expression at scale across cancers. To address this, we developed a model for quantifying tumor-specific total mRNA expression (TmS) from bulk sequencing data, which performs transcriptomic deconvolution while adjusting for mixed genomes. We used single-cell RNA sequencing data to demonstrate total mRNA expression as a feature of tumor phenotype. We estimated and validated TmS in 5,015 patients across 15 cancer types identifying significant inter-individual variability. At a pan-cancer level, high TmS is associated with increased risk of disease progression and death. Cancer type-specific patterns of genetic alterations, intra-tumor genetic heterogeneity, as well as pan-cancer trends in metabolic dysregulation and hypoxia contribute to TmS. Taken together, our results suggest that measuring cell-type specific total mRNA expression offers a broader perspective of tracking cancer transcriptomes, which has important biological and clinical implications.
BibTeX
@inproceedings{cao2021tumor,
  author = {Shaolong Cao and Jennifer Wang and Shuangxi Ji and Yang Peng and Matthew D. Montierth and Shuai Guo and John Paul Shen and Xiao Zhao and Jingxiao Chen and Jaewon J. Lee and Paola A. Guerrero and Nicholas Spetsieris and Nikolai Engedal and Sinja Taavitsainen and Kaixian Yu and Julie Livingstone and Vinayak Bhandari and Shawna M. Hubert and Najat C. Daw and P. Andrew Futreal and Eleni Efstathiou and Bora Lim and Andrea Viale and Jianjun Zhang and Matti Nykter and Bogdan Czerniak and Pavlos Msaouel and Anirban Maitra and Scott Kopetz and Peter J. Campbell and Terence P. Speed and Paul C. Boutros and Hongtu Zhu and Alfonso Urbanucci and Jonas Demeulemeester and Peter Van Loo and Wenyi Wang},
  title = {{Tumor cell total mRNA expression shapes the molecular and clinical phenotype of cancer}},
  booktitle = {Research Square},
  year = {2021},
  doi = {10.21203/rs.3.rs-600171/v1},
}
Figure