Abstract
Chemotherapy resistance is a critical contributor to cancer mortality and thus an urgent unmet challenge in oncology. To characterize chemotherapy resistance processes in high-grade serous ovarian cancer, we prospectively collected tissue samples before and after chemotherapy and analyzed their transcriptomic profiles at a single-cell resolution. After removing patient-specific signals by a novel analysis approach, PRIMUS, we found a consistent increase in stress-associated cell state during chemotherapy, which was validated by RNA in situ hybridization and bulk RNA sequencing. The stress-associated state exists before chemotherapy, is subclonally enriched during the treatment, and associates with poor progression-free survival. Co-occurrence with an inflammatory cancer–associated fibroblast subtype in tumors implies that chemotherapy is associated with stress response in both cancer cells and stroma, driving a paracrine feed-forward loop. In summary, we have found a resistant state that integrates stromal signaling and subclonal evolution and offers targets to overcome chemotherapy resistance.
INTRODUCTION
Platinum-based chemotherapy is the most widely prescribed drug in metastatic cancer treatment (1). It is curative in testicular cancers and effective in other cancers, such as in high-grade serous ovarian cancer (HGSOC) where the introduction of platinum-based combination therapy improved the 10-year survival rate by more than 10% and doubled the number of complete responses (1, 2). However, most patients with HGSOC develop platinum resistance leading to almost invariably fatal refractory disease and only 43% 5-year survival (3). HGSOC is a copy number–driven cancer that has exceptionally high intratumor heterogeneity and almost 100% prevalence of TP53 mutations (4, 5), which impedes overcoming platinum resistance.Patients with platinum-sensitive HGSOC with homologous recombination–deficient (HRD) tumors benefit from poly(adenosine diphosphate–ribose) polymerase (PARP) inhibitors (6). However, approximately half of the patients with HGSOC do not have HRD tumors and face very limited treatment options at the chemotherapy-resistant stage. On cellular level, clinically observed chemotherapy resistance is a continuum from a Darwinian selection process of intrinsically resistant cell populations to an adaptive induction of a fitness phenotype (7, 8). Most studies of drug resistance in the clinical setting have so far focused on genetic changes, such as MET amplification with kinase inhibitors (9), BRCA reversal mutations with chemotherapy (10), or genomic signatures in a heterogeneously treated patient cohort (11). The number and complexity of resistance mechanisms to chemotherapy surpass those of targeted therapies (12), which warrant homogeneously treated patient cohorts that allow high-resolution analysis of cancer cells before and after chemotherapy.Chemotherapy affects transcriptional programs of cancer cells, which provides an opportunity to comprehensively decipher the most relevant chemotherapy-induced processes using single-cell RNA sequencing (scRNA-seq) data. Data from scRNA-seq also enable addressing the interplay between cancer cells and tumor microenvironment (TME). scRNA-seq and genomic analysis performed before and after treatment in paired samples from four patients with metastatic breast cancer revealed that while chemotherapy selected preexisting genetic abnormalities, it also induced adaptive transcriptional changes related to epithelial-to-mesenchymal transition (EMT), AKT1 signaling, and hypoxia (13). In paired samples from four patients with non–small cell lung cancer, the surviving cells underwent a primitive state change to alveolar cells in residual disease (14). While these studies demonstrate the importance of paired samples, they each had cancer cells containing pairwise specimens from only four patients and, more importantly, limited clinical data from the patients, such as the patient outcome after therapy or survival times, which hinders making clinically relevant conclusions from the data.Here, we characterized transcriptional patterns of chemotherapy resistance in HGSOC using patient-derived prospective tissue sample pairs before and after treatment at single-cell resolution. Our cohort consists of scRNA-seq data from treatment-naïve and post–neoadjuvant chemotherapy (post-NACT) pairs from 11 homogeneously treated patients with HGSOC with full clinical information. To validate our findings, we used RNA in situ hybridization (RNA-ISH) data of 10 treatment-naïve versus post-NACT sample pairs, 49 bulk RNA-seq samples including 18 treatment-naïve versus post-NACT pairs, and 8 treatment-naïve versus relapse pairs in the HERCULES cohort (http://project-hercules.eu/) and bulk RNA-seq data of 271 treatment-naïve samples in The Cancer Genome Atlas (TCGA) cohort (5). Our unbiased analysis reveals how chemotherapy modulates cancer cell states by both subclonal selection and microenvironment-boosted transcriptional induction across the homogeneously treated sample cohort. Our results define a cell state that allows biomarker-based prediction and targeting of chemoresistance.
RESULTS
Obtaining scRNA-seq data from HGSOC patient samples before and after chemotherapy
We collected prospective tissue samples from 11 patients with HGSOC before and after chemotherapy and measured transcriptomes of 93,650 cells using scRNA-seq (Fig. 1A and see Materials and Methods). All patients in the study were treated with NACT, i.e., diagnostic laparoscopy followed by three cycles of platinum-taxane, interval debulking surgery (IDS), and adjuvant chemotherapy, and four patients further received bevacizumab maintenance therapy. NACT is typically recommended for patients who are inoperable at diagnosis and often have poor prognosis. Accordingly, in our cohort, the median platinum-free interval (PFI; Fig. 1A), which measures the time from treatment end to relapse, is only 4.2 months. Our sample cohort with metastatic tumors from poorly responsive patients represents many understudied aspects of HGSOC as described in Materials and Methods. Further clinical information of the cohort is given in Table 1.
