Modeling hepatocellular carcinoma and its microenvironment on a chip

Abstract

Hepatocellular carcinoma (HCC) is the most common type of liver cancer. Its incidence is increasing and is closely related to advanced liver disease. Interactions in the HCC microenvironment between tumor cells and the associated stroma actively regulate tumor initiation, progression, metastasis, and therapy response. Effective drug development increasingly requires advanced models that can be utilized in the earliest stages of compound and target discovery. Here we report a phenotypic screen on an advanced HCC patient-derived chip (PDChip) model. The vascularized HCC PDChip models include relevant cellular players of the HCC microenvironment. We assessed the effect of 28 treatment conditions on a panel of 8 primary HCC tumors and 2 cell lines. Approximately 1200 HCC PDchips were grown under perfusion flow, exposed to treatments, and subsequently assessed for viability, tumor-associated vasculature responses and chemokine and cytokine changes. Although the SoC therapeutics sorafenib and lenvatinib reduced culture viability and produced profound changes in the organization of the vascular beds, they did not affect the tumor cell population in these cultures. Atorvastatin, a 3-hydroxy-3-methylglutaryl-coenzyme A (HMG-CoA) reductase inhibitor, reduced PDChips viability but did not affect vascular bed organization. Sorafenib, lenvatinib and atorvastatin also affected chemokine and cytokine release. Tocilizumab, galunisertib, and vactosertib decreased the level of IL6, a relevant prognostic marker for HCC, while IL6 was increased by halofuginone. In conclusion, HCC PDChip models enabled a detailed evaluation of drug-induced responses in the tumor and associated microenvironment, highlighting their importance in preclinical research for understanding diseases and developing new drugs.

Fig. 1. Generation and characterization of HCC PDChips.

A Schematic representation of HCC tissue processing and PDChips generation using the OrganoPlate Graft. DTCs, CAFs, and HUVEC were mixed in fibrin and dispensed into the chamber of the OrganoPlate Graft microfluidic platform where they were cultured for 6 days, followed by exposure to a selection of compounds for 3 additional days. To assess compound responses, we evaluated cell viability, vascular network organization and chemokine and cytokine levels. B Phase contrast images of tumor compartment and lining vasculature on day 1 and 6 of HCC PDChips, these cultures were generated using samples from 8 primary HCC tumors and 2 HCC cell lines. B Also shows the bottom right quadrant of the tumor compartment of a HCC3 PDChip, images show that cells are poorly organized on day 1, while on day 6 culture shows tumor aggregates and an organized vascular network. C Confocal microscopy of HCC5 PDChip shows the presence of tumor cells (albumin, in green) and vasculature (VE-Cadherin, in red). D Confocal microscopy of a HCC6 PDChip shows the presence of CAFs and vascular network, vasculature is immunostained for CD31 (red), CAFs are immunostained for aSMA (green) and nuclei are labeled with Hoechst (blue). Overlay image shows that CAFs seem to line vascular structures. E Shows the presence of AFP, IL6 and CCL2 (ng/mL) in HCC patient-derived and HCC cell lines models, HCC patients (N = 8) and HCC cell lines (N = 2; Huh7 and HLE). AFP, IL6 and CCL2 were measured in the supernatant collected from DMSO controls for all donors and cell lines on day 9. Statistical significance was determined using unpaired Welch-corrected t-test to compare AFP, IL6, and CCL2 levels between HCC patient cultures and HCC cell lines. Adjusted p-values are expressed as follows: ****p < 0.0001, ***p < 0.001, **p < 0.01, *p < 0.05.

Fig. 2. HCC PDChip model viability in response to CAF targeting and SoC.

A Scatterplot shows replicates reproducibility (r = 0.8236) of PDChips viability results. DMSO control viability distribution across plates is shown in (B). B Shows DMSO control viability percentage (values normalized to average of DMSO control viability/plate, n = 6). C Heatmap showing viability after treatment of the tumor compartment of HCC models. D Shows viability after treatment of the vascular compartment of HCC cultures. Heatmaps represent viability percentages (normalized to average of DMSO controls). Heatmaps include n = 12 for DMSO control; n = 4 for IgG1 control; n = 4 for each treatment condition. For HCC4 n = 6 for DMSO control; n = 2 for IgG1 control; n = 2 for each treatment condition were included. E Representative confocal image of Albumin and Hoechst staining and respective segmentation are shown. Albumin and Hoechst stained cultures were imaged on an ImageXPress Micro Confocal XLS (Molecular Devices) and images were analyzed using INCarta image analysis software (Molecular Devices, version 2.1). F, G Show the quantification of albumin positive cells (tumor cells) in HCC PDChips in response to drug panel, (n = 24, for DMSO control; n = 8, for IgG1 control; n = 8, for each treatment condition).

Fig. 3. Treatment-induced vascular response in HCC PDChips.

A HCC PDChip models associated-vasculature organization in response to sorafenib and lenvatinib. HCC cultures were immunostained for CD31 (red), and confocal imaged. B Vasculature of HCC cultures was immunostained for CD31 and imaged using confocal microscopy (CD31 immunostaining for all cultures and treatment conditions can be found in Supplementary Fig. 2) (n = 30, for DMSO control; n = 10, for IgG1 control; n = 10, for each treatment condition). Abbreviation of drug names: Gal 0.1 µM and 1 µM (Galunisertib), Vac 0.1 µM and 1 µM (Vactosertib), LY 0.1 µM and 1 µM (LY2090314), WZ 1 µM and 10 µM (WZ811), Ator 5 µM (Atorvastatin), SH 0.1 µM and 1 µM (SH-4-54), Sora 0.1 µM and 1 µM (Sorafenib), Atez 5 µg/mL (Atezolizumab), Beva 5 µg/mL (Bevacizumab), Halo 0.1 µM (Halofuginone), UNB 10 µM (UNBS5162), Lenv 0.1 µM and 1 µM (Lenvatinib), Toci 5 µg/mL (Tocilizumab), Cri 0.1 µM and 1 µM (Crizotinib), Atez-Beva (Atezolizumab 5 µg/mL-Bevacizumab 5 µg/mL), Vac-SH (Vactosertib 0.1 µM -SH-4-54 0.1 µM), UNB-SH (UNBS5162 10 µM -SH-4-54 0.1 µM), WZ-SH (WZ811 1 µM - SH-4-54 0.1 µM), UNB-Halo (UNBS5162 10 µM -Halofuginone 0.1 µM), WZ-LY (WZ811 1 µM - LY2090314 0.1 µM). DMSO (Dimethyl sulfoxide 0.1%), and IgG1 (Immunoglobulin G1 10 µg/mL) were used as vehicle controls. Atezolizumab 5 µg/mL, Bevacizumab 5 µg/mL and Tocilizumab 5 µg/mL, also contains IgG1 5 µg/mL. Images were processed and 15 vascular organization-related descriptors were extracted. C Unsupervised linear (PCA), and D non-linear (t-SNE) reduction of the data show a clustering of sorafenib and lenvatinib. E–H Shows sorafenib, lenvatinib, LY2090314 and atorvastatin-induced changes in total vessel density, total vessel area, total vessel length and branching index. E–H include HCC patient-derived and HCC cell line culture results. Statistical significance was determined using either the Kruskal–Wallis test followed by Dunn’s post hoc test or ANOVA followed by Tukey’s post hoc test, depending on whether assumptions for parametric tests were met, with p-values adjusted using the Bonferroni correction. Adjusted p-values are expressed as follows: ****p < 0.0001, ***p < 0.001, **p < 0.01, *p < 0.05.

Fig. 4. Immunomodulatory response in HCC PDChip models.

Supernatant of HCC cultures was collected and analyzed for expression of a panel of 13 cytokines and chemokines and IL6 by Luminex. A Significant changes in cytokine and chemokine production in HCC PDChip models in response to treatments. B Heatmap shows an overview of cytokine and chemokine responses of HCC cultures in response to treatments. (n = 48, for DMSO control; n = 16, for IgG1 control; n = 8–24, for treatment conditions). Statistical significance was determined using Wilcoxon tests, with p-values adjusted using the Benjamini–Hochberg correction. Adjusted p-values are expressed as follows: ****p < 0.0001, ***p < 0.001, **p < 0.01, *p < 0.05.