Modelling hepatocellular carcinoma microenvironment phenotype to evaluate drug efficacy

Abstract

Hepatocellular carcinoma (HCC) is the third most common cause of cancer-related death worldwide. Treating HCC is challenging because of the poor drug effectiveness and the lack of tools to predict patient responses. To resolve these issues, we established a patient-centric spheroid model using HepG2, TWNT-1, and THP-1 co-culture, that mimics HCC phenotype. We developed a target-independent cell killing (TICK) exclusion strategy to monitor the therapeutic response. We demonstrated that our model reproduced the Barcelona Clinic Liver Cancer (BCLC) molecular classification, displayed known alterations of epigenetic players, and responded to tyrosine kinase inhibitors (TKIs) such as sorafenib, cabozantinib, and lenvatinib in a patient-dependent manner. Importantly, we reported for the first time that our model correctly predicted 34 clinical outcomes to TKIs out of 37 case studies on 32 HCC patients confirming that patient-centric spheroids, combined with our TICK exclusion strategy, are valuable models for drug discovery and opening a near perspective to personalized care.

Keywords: Hepatocellular carcinoma; Patient-centric spheroid model; Prediction of clinical outcomes; Targeted therapy; Tyrosine kinase inhibitors.

Fig. 1

Staging of HCC patient-derived serum educated spheroids based on the molecular signature described by Xu and colleagues. (a) Workflow showing the generation of PDSES using depleted serum from HCC patients with known BCLC staging. (b) Upper graph. Quantitative PCR analysis of the expression of the 13 hub genes in PDSES. Results are reported as Log expression relative to GusB gene. Depleted serum from 6 HCC patients with BCLC-A, BCLC-B, and BCLC-C were used to generate the spheroids. After 3 days of culture, total RNA was prepared for qPCR analysis of the 13 hub genes. Lower graph. Expression of 13 hub genes that are correlated with the overall survival in various stages. Results were extracted from the manuscript of Xu et al., 2018. Plos one (PMID: 30138346) and are shown as transcript per million. Thick lines show the median and thing lines depict the maximum and the minimum values.

Fig. 2

HCC patient-derived serum educated spheroids display an interindividual diversity of alteration of epigenetic players and sensitivity to the selective inhibitor of BRD4. (a) Blood sera from multiple centres were used to generate PDSES. The expression of epigenetic players including the writers, readers, and erasers was analyzed by quantitative PCR. Results are expressed as relative to GusB gene. (b) PDSES were generated with blood sera from 3 males and 3 females and then treated with JQ1. Cell viability was measured 3 days post-treatment. An inhibitory dose-response curve fit with constraints (top=100; bottom=0) was drawn for each patient.

Fig. 3

Determination of the target-independent cell killing (TICK) value for each TKI. (a) Individual-centric spheroids were generated using depleted blood sera from healthy individuals. Spheroids were then treated with sorafenib, cabozantinib, and lenvatinib. Cell viability was measured at day 3 post-treatment. For each TKI, an inhibitory dose-response curve fit with constraints (top = 100; bottom = 0) was drawn. Each condition was performed in triplicate for each individual. (b) Sorafenib and (c) cabozantinib curves show an average cell viability for each concentration of the cohort. TICK values were defined as the first percentage of cell death. (d) For lenvatinib, no cell death was observed up to 240 x C_max. TICK value was then set to the maximum concentration of 120 µM.

Fig. 4

HCC patient-derived serum educated spheroids show a diversity in responses to sorafenib and cabozantinib, and proportionally matched to results from clinical trials. (a) PDSES were generated with depleted blood sera from HCC patients and then treated with sorafenib or cabozantinib. Cell viability was measured 3 days post-treatment. An inhibitory dose-response curve fit with constraints (top = 100; bottom = 0) was drawn for each HCC patient and the true efficacy response was determined based on the TICK values. (b) True efficacy of sorafenib on a cohort of 37 HCC patients. Based on the TICK exclusion strategy, 17 HCC patients out of 37 were responding to sorafenib (shown with green curves). (c) 16 HCC patients out of 27 treated with cabozantinib were responding (shown with green curves). (d) Comparative analysis of the proportion of responders between in vitro data and results from 2 clinical trials.

Fig. 5

Evaluation of the performance of HCC patient-derived serum educated spheroids to reproduce the clinical outcomes to TKIs. (a) Blood sera from HCC patients were collected before the patient underwent a clinical treatment protocol. PDSES were generated and then treated with the same treatment as the one that he/she received in the clinic. A blind and unbiased in vitro analysis of the response to TKIs was performed. The true therapeutic response to TKIs was determined using the TICK exclusion strategy as described above. The full process from blood processing to final reports was 7 days. (b) Clinical outcomes and in vitro prediction responses to TKIs. PDSES from 29 HCC patients were generated and then treated with TKIs accordingly to the clinical treatment protocol. The true therapeutic efficacy was determined based on the TICK values for each TKI as described above.