The Systems-based Identification of Biomarkers for Regorafenib in Metastatic Colorectal Cancer

Regorafenib (REG) is a tyrosine kinase inhibitor approved for treatment refractory metastatic colorectal cancer (mCRC) and has been shown to improve median overall (OS) and progression-free survival (PFS) by 1.4 and 0.2 months respectively. As >50% of patients experienced adverse REG-induced toxicities, biomarkers to predict responders are urgently required. To address this, we established a pre-clinical platform of molecularly subtyped cell line xenografts (CDXs) and patient derived xenografts (PDXs), and employed a multi-‘omic strategy to interrogate biomarkers of response.

CDXs representing CRC consensus molecular subtypes (CMS) were established and REG (10mg/kg) was administered orally for a maximum of 4 weeks. Contrast-enhanced ultrasound (CEUS) was used to monitor early effects of REG on tumour vasculature in CDX models. A CRC PDX study was performed models were assigned to CMS and novel copy number clusters. To better understand subtype specific responses to REG unsupervised differential gene expression analysis was used to identify mechanisms of resistance. Finally, a validated systems model of BCL2 proteins (‘DR_MOMP’) and a gene-based proliferation signature were employed in an attempt to predict REG response withing molecular subtypes.

CDXs and PDXs replicated the subtype specific drug response observed in the clinic. CEUS analysis indicated no treatment effects of REG on vessel function at early time points. CMS4 subtyped CDXs and PDXs showed the best response to REG. We further observed a significant (P = 0.04) predictive correlation between proliferation index and tumour response in CMS4 derived PDXs. Differential gene expression analysis revealed a potential role for EPHA2, UTG1A1, MAPK11, KIT and CYP2B6 in mediating REG resistance and revealed that EPHA2 expression was predictive of REG response in PDXs. There was also a linear correlation between DR_MOMP stress dose (P = 0.03), and tumour growth in CMS4 models. Tumours with high copy number exhibited favourable response to REG vs copy number stable tumours. This was verified in patient samples from the REARRANGE trial (NCT02835924). PDXs maintained consistent copy number profiles across consecutive passages.

Overall, we have provided evidence for a differential effect of REG across CRC subtypes. We further demonstrated the utility of employing multi-omic analysis in clinically relevant models to identify putative biomarkers for multi-targeted tyrosine kinase inhibitors.