Tumour microenvironment characterisation to stratify patients for hyperthermic intraperitoneal chemotherapy in high-grade serous ovarian cancer (OVHIPEC-1)

Abstract

Background: Hyperthermic intraperitoneal chemotherapy (HIPEC) improves survival in patients with Stage III ovarian cancer following interval cytoreductive surgery (CRS). Optimising patient selection is essential to maximise treatment efficacy and avoid overtreatment. This study aimed to identify biomarkers that predict HIPEC benefit by analysing gene signatures and cellular composition of tumours from participants in the OVHIPEC-1 trial.

Methods: Whole-transcriptome RNA sequencing data were retrieved from high-grade serous ovarian cancer (HGSOC) samples from 147 patients obtained during interval CRS. We performed differential gene expression analysis and applied deconvolution methods to estimate cell-type proportions in bulk mRNA data, validated by histological assessment. We tested the interaction between treatment and potential predictors on progression-free survival using Cox proportional hazards models.

Results: While differential gene expression analysis did not yield any predictive biomarkers, the cellular composition, as characterised by deconvolution, indicated that the absence of macrophages and the presence of B cells in the tumour microenvironment are potential predictors of HIPEC benefit. The histological assessment confirmed the predictive value of macrophage absence.

Conclusion: Immune cell composition, in particular macrophages absence, may predict response to HIPEC in HGSOC and these hypothesis-generating findings warrant further investigation.

Clinical trial registration: NCT00426257.

Fig. 1. CONSORT diagram of OVHIPEC-1 participants and tissue availability.

HGSOC high-grade serous ovarian cancer, TCP tumour cell percentage, CRS cytoreductive surgery, HIPEC hyperthermic intraperitoneal chemotherapy.

Fig. 2. Summary of methods diagram.

a Samples from the OVHIPEC-1 study cohort were collected for HE-stained slides and whole-transcriptome sequencing at three time points: pre-treatment (TP1), surgical resection (TP2), and recurrence (TP3). Only patients with high-grade serous ovarian cancer were included in the analysis. b Cell-type-specific gene expression reference profiles were created using single-cell transcriptomic data from seven treatment-naïve high-grade serous ovarian cancer patients from a publicly available dataset (Olbrecht et al. [35]). Individual cells were isolated, RNA was extracted, and sequencing was performed. Clustering algorithms were then applied to group cells with similar gene expression profiles, and differentially expressed genes were calculated for each cluster. Clusters were annotated based on known marker genes or biological features, and reference profiles were constructed by averaging transcriptome counts within clusters. c A deconvolution algorithm, specifically Dampened Weighted Least Squares (DWLS), was employed to estimate the relative proportions of different cell types in samples obtained from the OVHIPEC-1 study cohort. This algorithm used the reference profiles derived from single-cell RNA sequencing to infer the cell-type fractions based on the gene expression signatures of the cells.

Fig. 3. Bulk differential gene expression analysis.

a Volcano plot showing the interaction coefficients between each gene’s expression level and HIPEC with survival in a Cox regression analysis, corrected for multiple hypothesis testing. No genes showed a significant association. b Volcano plot showing the difference of each gene’s expression level between metastatic and primary tumour samples, among the resection samples. c Volcano plot showing the difference of each gene’s expression level between pre-treatment samples and resection samples, with tissue location (primary vs metastatic) as covariate. d Gene set enrichment analysis for the comparison of resection samples to pre-treatment samples, given the MSigDB Hallmark gene sets. The graph shows the top 20 gene sets ordered by absolute enrichment score (bottom to top). The colour indicates the significance of the enrichment as a false discovery rate (FDR).

Fig. 4. Deconvolution of bulk RNA sequencing data to estimate cell-type abundance using single-cell expression reference profiles.

a UMAP projection of single cells based on their normalised expression counts. Cells with similar expression profiles are proximally located, allowing for clustering, and annotating single cells according to their respective cell type. In total, Seurat-based clustering revealed 25 clusters, representing the expected cell types. b Annotation of the UMAP projection based on sample origin shows no patient overlap in TC clusters, while immune cell clusters consist of cells from multiple patients. c Expression of cell-type-specific marker genes within each cluster. Cell-type-specific markers are used to provide cell-type labels to individual Seurat-based clusters. d Heatmap showing cell-type abundances as estimated by single-cell deconvolution for every patient. Hierarchical clustering of patients shows a heterogeneous landscape of cell-type composition, with varying degrees of immune infiltration. e Kaplan–Meier Curve using quartile cut-offs (first 25% vs 25–75% vs 75% and beyond), shows a significant association of deconvolution-based immune cell abundance with PFS (two-sided log-rank P value = <0.005). CAF cancer-associated fibroblasts, TC tumour cells.

Fig. 5. Predictive delta treatment score based on single-cell deconvolution.

Kaplan–Meier estimates for the (a) DTS^IC-low group and b DTS^IC-high group treated either with CRS or CRS-HIPEC. DTS^IC was dichotomised based on a median cut-off, where 50% of patients with the lowest DTS^IC were assigned to DTS^IC-low and 50% with the highest DTS^IC were assigned to DTS^IC-high. Patients in the DTS^IC-low group benefit significantly from HIPEC treatment (P < 0.005) over CRS, while in the DTS^IC-high group patients have significantly worse survival outcome when treated with HIPEC (P = 0.01). c Value of CPH interaction coefficients between treatment and cell-type features for DTS^IC averaged over all models in the leave-one-out validation. Error bars represent the 95% confidence intervals estimated by bootstrapping. A high positive interaction coefficient for a given cell type results in a lower DTS^IC and therefore better outcome for patients under HIPEC, while a high negative coefficient results in a higher DTS^IC and therefore better outcome under CRS, showing feature importance for DTS^IC. d Heatmap of cell-type abundances sorted according to DTS^IC. Given the prognostic significance of immune cells, both DTS^IC-low and DTS^IC-high patients show increased immune cell infiltration, while in DTS^IC-high patients, the TME contains a higher abundance of macrophages, contrasted by a higher B-cell abundance in DTS^IC-low patients. Cell-type abundances are normalised to the interval [0,1], with the interval of the unnormalized cell-type abundances with original feature value interval listed in brackets. e Absolute cell-type abundances for each immune cell cluster for DTS^IC-high and DTS^IC-low emphasising the absence of macrophages and the presence of B cells to be associated with differential HIPEC response. f Gene set enrichment analysis of the continuous deconvolution-derived DTS^IC shows upregulated immune-specific pathways in DTS^IC-low patients compared to DTS^IC-high patients. This finding provides further evidence supporting the validity of the deconvolution-derived DTS^IC, as the correlation between the patient-specific immune response and the differential HIPEC response remains even when reverting to the bulk mRNA-seq data. CRS cytoreductive surgery, DTS^IC immune cell delta treatment score, HIPEC hyperthermic intraperitoneal chemotherapy.

Fig. 6. Forest plot of exploratory subgroup analysis for overall survival based on histopathological assessment.

Histopathological assessment was performed on tissue resected during interval CRS. We tested whether the interaction between the subgroup variable and treatment was associated with OS in a CPH model. The size of each diamond is proportional to the number of patients. TIL tumour infiltrating lymphocytes.