A chemotherapy response prediction model derived from tumor-promoting B and Tregs and proinflammatory macrophages in HGSOC

Yue Xi, Yingchun Zhang, Kun Zheng, Jiawei Zou, Lv Gui, Xin Zou, Liang Chen, Jie Hao, Yiming ZhangFrontiers in Oncology 2023 · DOI

Abstract

The developed chemotherapy response prediction model can be used as a valuable clinical decision tool to guide chemotherapy in HGSOC patients.

Introduction

High-grade serous ovarian cancer (HGSOC) is the most common and lethal subtype of ovarian cancer. Although most patients initially respond to platinum-based chemotherapy, the majority eventually experience recurrence due to the development of cumulative chemoresistance. A major hurdle in clinical management is the lack of reliable biomarkers to predict treatment response, which often leads to a "one-size-fits-all" approach that may result in treatment failure.

Recent research suggests that the secret to predicting response lies within the Tumor Microenvironment (TME). Rather than focusing solely on malignant cells, investigating the complex interplay of immune cells—specifically B cells, T cells, and macrophages—may reveal the underlying mechanisms of resistance. This study introduces HGSOC.Sig, a 43-gene signature designed to predict chemotherapy response by leveraging high-resolution single-cell data.

Methodology

The researchers developed a sophisticated multi-step pipeline to identify and validate the predictive signature, moving from single-cell discovery to clinical validation.

Workflow for identifying chemotherapy response-related immune cell subtypes from single-cell RNA sequencing datasets.

Discovery and Subtype Identification

The study began with the re-analysis of two HGSOC single-cell RNA sequencing (scRNA-seq) datasets (GSE154600 and GSE165897). Using computational tools such as scCODE and Gene Set Enrichment Analysis (GSEA), the team identified specific functional subtypes of immune cells rather than broad categories. Four key subtypes were identified:

  • Resistance-associated: $XBP1^+$ B cells (linked to endoplasmic reticulum stress) and $ACTB^+$ Tregs (linked to invasive metastasis).
  • Response-associated: $FCN1^+$ and $CCL3^+$ pro-inflammatory macrophages.

Feature Selection and Validation

Using a recursive algorithm and functional enrichment analysis (GSVA), the team narrowed thousands of genes down to a final 43-gene signature. This model was then validated across multiple independent "bulk" RNA-seq datasets (including GSE156699, GSE30161, and GSE63885) and further verified using quantitative real-time PCR (qRT-PCR) on clinical samples from a local HGSOC patient cohort.

Key Insights: The Ribosomal Connection

A significant finding of the study was the identification of ribosomal biogenesis and protein translation as key drivers of the drug-response phenotype. The most diagnostic genes in the signature, such as EEF1A1, RPL31, and RPS12, are involved in these processes. This suggests that the metabolic state and "protein-factory" capacity of specific immune cell subtypes are primary indicators of how a tumor will respond to chemotherapy.

Results

The HGSOC.Sig model demonstrated unprecedented predictive accuracy compared to existing tools:

  • Unmatched Accuracy: The model achieved an Area Under the Curve (AUC) of 0.97 in the primary validation cohort, significantly outperforming previous models which typically reach an AUC of approximately 0.74.
  • Robustness: High predictive power (AUCs 0.87–0.93) was maintained across seven independent external datasets, proving the model is not overfit to a single population.
  • Survival Correlation: Patients classified as "high-score" (indicating resistance) had significantly poorer Progression-Free Survival (PFS) and Overall Survival (OS).
  • Clinical Hub Genes: Six specific genes (EEF1A1, RPL35A, RPL31, RPL11, RPS12, and RPS23) were confirmed via qRT-PCR to be significantly upregulated in non-responders, offering a potential path for a simplified clinical diagnostic test.

Conclusion and Future Directions

The HGSOC.Sig model represents a significant advancement in personalized medicine for ovarian cancer. By moving beyond broad cell categories to specific functional immune subtypes, the researchers have created a highly accurate tool for predicting chemotherapy response.

However, several areas for future work remain. The exact biological mechanisms by which ribosomal genes in immune cells influence the resistance of neighboring tumor cells require further investigation. Additionally, while retrospective validation was successful, prospective clinical trials are necessary to confirm the model's utility in a real-world clinical setting. Future studies incorporating spatial transcriptomics could also provide deeper insights into the physical interactions between these critical immune subtypes and the tumor.

Summary updated

Abstract

The developed chemotherapy response prediction model can be used as a valuable clinical decision tool to guide chemotherapy in HGSOC patients.
BibTeX
@inproceedings{xi2023chemotherapy,
  author = {Yue Xi and Yingchun Zhang and Kun Zheng and Jiawei Zou and Lv Gui and Xin Zou and Liang Chen and Jie Hao and Yiming Zhang and Yiming Zhang and Yiming Zhang},
  title = {{A chemotherapy response prediction model derived from tumor-promoting B and Tregs and proinflammatory macrophages in HGSOC}},
  booktitle = {Frontiers in Oncology},
  year = {2023},
  doi = {10.3389/fonc.2023.1171582},
}
Figure