A comparative study of tumour-on-chip models with patient-derived xenografts for predicting chemotherapy efficacy in colorectal cancer patients

Abstract

Inter-patient and intra-tumour heterogeneity (ITH) have prompted the need for a more personalised approach to cancer therapy. Although patient-derived xenograft (PDX) models can generate drug response specific to patients, they are not sustainable in terms of cost and time and have limited scalability. Tumour Organ-on-Chip (OoC) models are in vitro alternatives that can recapitulate some aspects of the 3D tumour microenvironment and can be scaled up for drug screening. While many tumour OoC systems have been developed to date, there have been limited validation studies to ascertain whether drug responses obtained from tumour OoCs are comparable to those predicted from patient-derived xenograft (PDX) models. In this study, we established a multiplexed tumour OoC device, that consists of an 8 × 4 array (32-plex) of culture chamber coupled to a concentration gradient generator. The device enabled perfusion culture of primary PDX-derived tumour spheroids to obtain dose-dependent response of 5 distinct standard-of-care (SOC) chemotherapeutic drugs for 3 colorectal cancer (CRC) patients. The in vitro efficacies of the chemotherapeutic drugs were rank-ordered for individual patients and compared to the in vivo efficacy obtained from matched PDX models. We show that quantitative correlation analysis between the drug efficacies predicted via the microfluidic perfusion culture is predictive of response in animal PDX models. This is a first study showing a comparative framework to quantitatively correlate the drug response predictions made by a microfluidic tumour organ-on-chip (OoC) model with that of PDX animal models.

Keywords: 3D culture; PDX (patient derived xenograft); dose response; in vitro; in vivo; microfluidic lab-on-a-chip; organ-on-chip (OoC).

FIGURE 1

Design and characterisation of the Integrated Microfluidic Tumour Array (IMITA) device for multiplexed 3D cell cultures. (A) The IMITA device consisted of two functional blocks (Lin et al., 2017): concentration gradient generator; and (Wei et al., 2017) cell culture array connected by two orthogonal flow circuits for cell seeding and medium perfusion. Scale bar = 1 cm. (B) The cell culture array consisted of 32 cell culture chambers arranged in 8 rows by 4 columns. (C) Each cell culture chamber comprised a cup-shaped micropillar array with 20 µm gaps where the opening faced the seeding flow circuit to trap the incoming cells. A series of micropillars act as cell filters along the medium perfusion direction to prevent cell clogging during cell seeding. Scale bar = 100 µm. (D) CFD simulation showing 8 mass concentrations generated by a linear concentration generator, which were fed into each row of the IMITA device at steady state when operating at 0.02 ml h⁻¹. (E) Simulated relative mass fractions as a function of distance along a single row of cell culture chamber [box indicated in 1(d)] at steady state operating condition. Mass fraction within a single cell culture chamber remained constant while a linear decrease was observed across each chamber in a row. (F) Experimental validation of the concentration gradient generator using rhodamine fluorescent probe. Data are averages of 3 experiment measurements ± standard deviations.

FIGURE 2

Seeding and perfusion culture of PDX derived tumour spheroids within the IMITA device. Transmission images showing: (A) CRC1030 PDX derived tumour cells seeded and trapped by the micropillar array in the cell culture chamber; (B) seeded cells remodelled into a 3D tumour spheroid after 24 h of perfusion culture. Scale bars = 100 µm; (C) Quantification of PDX derived tumour spheroid size distribution across all rows of the IMITA device after 24 h of perfusion culture (One-way ANOVA, p = 0.0891). The spheroid’s equivalent diameter was estimated from the spheroid’s area; (D) Live (CalceinAM, green)-dead (EthD-1, red) staining of the seeded PDX derived tumour cells within the IMITA device after 96 h of perfusion culture. Scale bar = 100 µm.

FIGURE 3

In vivo-in vitro comparative study to evaluate the efficacy of 5 standard-of-care (SOC) chemotherapies for 3 CRC patients. (A) Experimental design. The 5 SOC chemotherapies tested were 5-Fluorouracil (5-FU), Oxaliplatin (OXA), irinotecan (IRI), 5-Fluorouracil + Oxaliplatin (5-FU + OXA or 5-FU/OXA), and 5-Fluorouracil + irinotecan (5-FU/IRI). Figure created with BioRender.com. (B) Calcein-AM fluorescent images overlaid with transmission images showing viability of CRC 1414 tumour spheroids in IMITA devices after 24 h of treatment with 5 SOC chemotherapies at selected concentrations. Magnification = ×10. Scale bar = 250 µm.

FIGURE 4

Treatment responses of 5 SOC chemotherapies for 3 CRC patients (CRC935, CRC1030, and CRC1414) obtained from in vitro (IMITA) and in vivo (PDX) models. (A–C) Dose-response curves obtained after 3 days of drug treatment in the IMITA devices. Fitted dose-response curve obtained using n > 3 devices with shaded area representing 95% confidence interval. Two-way ANOVA analysis yielded p = 0.002 for CRC935, p < 0.0001 for CRC1030 and CRC1414. (D–F) Tumour growth rates in PDX models. 5-Fluorouracil and IRI were administered at a dosage of 50 mg kg⁻¹. OXA was administered at a dosage of 5 mg kg⁻¹. 5-FU + OXA and 5-FU + IRI were prepared with same concentrations. Data are averages of n > 3 animals with ± standard deviation.

FIGURE 5

In vitro-in vivo drug response correlation. (A) TGI values from in vivo PDX models and IC₅₀ values of dose response curves obtained from in vitro IMITA devices for 3 CRC patients. (B) Patient-specific drug efficiency ranking using in vivo PDX models and estimated IC₅₀ values using the IMITA device. Rank 1 denotes most effective drugs (lower IC₅₀ values for IMITA and higher TGI% for PDX). (C) Pearson correlation matrix comparing drug efficacy ranking estimated from the in vivo PDX models to the IC₅₀ values from the in vitro IMITA devices.