The current standard of care for patients with microsatellite-stable metastatic colorectal cancer (mCRC) consists of the combination of a cytotoxic chemotherapy backbone and a monoclonal antibody based on RAS/BRAF testing.1 2 In clinical practice, there are two choices for cytotoxic agent, namely oxaliplatin or irinotecan in combination with 5-fluorouracil, and two for monoclonal antibodies targeting vascular endothelial growth factor (VEGF) or epidermal growth factor receptor (EGFR). Most research has focused on the efficacy of monoclonal antibodies, leading to the successful identification of predictive and prognostic biomarkers for EGFR inhibitors such as RAS and BRAF mutation status and primary tumor sidedness.3–6 However, there are no predictive biomarkers for the selection of oxaliplatin or irinotecan which have distinct mechanisms of action: oxaliplatin is a platinum compound forming intra-strand and inter-strand cross-links; irinotecan is a topoisomerase I (TOP1) inhibitor trapping TOP1 cleavage complex with DNA replication and transcription. The choice of these two drugs remains an important clinical question for optimizing treatment in individual patients.

Immunogenic cell death (ICD) is a novel concept of tumor cell death that engages both innate and adaptive immune responses, thereby conferring an additional immunogenic antitumor effect to cytotoxic agents.7–9 However, most chemotherapeutics kill tumor cells in a non-immunogenic manner, and only a few cytotoxic agents can prime antitumor immunity by inducing ICD.10 11 The mechanisms of ICD induction are characterized by the ability to induce endoplasmic reticulum stress, which in turn leads to the release of damage-associated molecular patterns (DAMPs) from dying tumor cells and the subsequent activation of pattern recognition receptors (PRRs) of host innate immune cells such as dendritic cells.12 The engagement of innate immune responses cross-primes adaptive responses controlled by cytotoxic T lymphocytes through proinflammatory cytokine release, potentially leading to the continuous elimination of tumor cells.12 13 Oxaliplatin has been recently well-recognized as an ICD inducer, similarly as other drugs such as anthracyclines and bortezomib.7 In a preclinical study, CRC cells treated with oxaliplatin exhibited increased pre-apoptotic exposure of calreticulin (CALR) and post-apoptotic release of high-mobility group box 1 (HMGB1), which are the signals of DAMPs required for ICD induction.14 By contrast, irinotecan lacks the mechanism of action that drives ICD, and thus, its cytotoxic effects depend on non-immunogenic mechanisms.10 15 These findings suggested that genetic variants in the ICD pathway are related to the efficacy of oxaliplatin but not irinotecan, prompting us to test whether these novel biomarkers will be useful for patient selection. Therefore, we investigated the predictive utility of single nucleotide polymorphisms (SNPs) in ICD-related genes encoding essential DAMPs and PRRs using genetic and clinical data from two first-line randomized clinical trials comparing oxaliplatin- (or oxaliplatin-containing) and irinotecan-based treatments in patients with mCRC.

This is the first study to identify genetic variants in ICD-related pathways that may predict the efficacy of oxaliplatin-based first-line chemotherapy in patients with mCRC. No predictive biomarkers for the personalized use of cytotoxic agents in mCRC have been identified so far. DNA damage response genes and the transcriptome-based consensus molecular subtype have been examined in biomarker research, but their predictive value has not been validated.16 20 Gray et al reported that genetic variants of three PRR genes (FPR1, TLR3, and TLR4) were not related to the efficacy of oxaliplatin in patients with CRC in the adjuvant setting (using data from the SCOT trials) and the first-line setting for patients with metastasis (using data from the COIN/COIN-B trials).19 However, DAMPs are also essential tumor-sided factors for ICD induction, whereas PRRs are host-side factors. Therefore, in our study, we extensively assessed the functional variants of both DAMPs and PRRs. According to our hypothesis, ICD-related SNPs were significantly associated with the efficacy of oxaliplatin-based combination. These findings imply the linkage between ICD pathway and oxaliplatin efficacy, which can be coupled with the potential synergy of oxaliplatin and immunotherapy.21 22

Our findings identified ANXA1 rs1050305 and LRP1 rs1799986 as SNPs exhibiting significant associations with the efficacy of oxaliplatin but not irinotecan as confirmed in the discovery and validation cohorts. ANXA1 is one of the DAMPs released from the cytoplasm of dying cells, and it interacts with FPR1, which is expressed on dendritic cells.23 The extracellular accumulation of ANXA1 constitutes a hallmark of ICD.12 ANXA1 rs1050305 is a synonymous SNP located at an exonic splicing enhancer/silencer site that disrupts splicing activity and causes alternative splicing (https://snpinfo.niehs.nih.gov/snpinfo/snpfunc.html). LRP1 is a PRR activated via docking of cell surface-exposed CALR, which is another hallmark of chemotherapy-induced ICD.12 24 LRP1 rs1799986 is a missense SNP that may participate in splicing regulation. This SNP is reported to be associated with susceptibility to Alzheimer’s disease, supporting its functional significance.25 Of note, while the predictive value of ANXA1 rs1050305 was consistently validated on OS, that of LRP1 rs1799986 was observed on inconsistent clinical outcomes between the discovery and validation cohorts: the discovery cohort showed a significant association with OS, and the validation cohort showed that with PFS. This discrepancy might be due to some unidentified factors such as different contribution of second-line treatment between two cohorts, which might affect the associations between biomarker and clinical outcome.

Our study had great strengths in its design, featuring discovery, validation, and control cohorts obtained from two randomized clinical trials. The validated effects of the SNPs on clinical outcomes in the discovery and validation cohorts, lacking of that in the two control cohorts, strongly supports the predictive value of ANXA1 rs1050305 and LRP1 rs1799986 specifically for oxaliplatin efficacy. The multiple significant associations observed in the discovery and validation cohorts but not in the control cohorts further indicate that the ICD pathway plays an important role in the efficacy of oxaliplatin but not irinotecan. Another strength was that the treatment-by-SNP interaction test confirmed the predictive value of ANXA1 rs1050305 and LRP1 rs1799986 and highlighted the predictive utility of 6 out of 10 tested SNPs.26

This study had two key limitations. First, this was a prospective–retrospective study; thus, the results need to be validated in prospective clinical trials. According to published guidelines, one or more validation studies are required to reach a sufficient level of evidence for medical utility.27 28 Second, the validation cohort consisted of patients treated with FOLFOXIRI plus bevacizumab, which is both an oxaliplatin-containing and irinotecan-containing regimen. To validate the different effects of SNPs on oxaliplatin and irinotecan efficacy, it will be better to use a treatment arm containing only oxaliplatin-based regimens as the validation cohort. However, our study provides reliable evidence describing the specific connection of ICD-related SNPs with the efficacy of “oxaliplatin-containing” treatment based on the use of “oxaliplatin-non-containing” cohorts as controls.

In conclusion, our study provides the first evidence that genetic variants in the ICD pathway are significantly associated with the efficacy of oxaliplatin-based, but not irinotecan-based, first-line chemotherapy in patients with mCRC. Further translational studies are warranted to validate our novel findings and realize the personalized use of conventional cytotoxic chemotherapeutic agents.