Optimal treatment of cancer requires diagnostic methods to facilitate therapy choice and prevent ineffective treatments. Direct assessment of therapy response in viable tumor specimens could fill this diagnostic gap. Therefore, we designed a microfluidic platform for assessment of patient treatment response using tumor tissue slices under precisely controlled growth conditions. The optimized Cancer-on-Chip (CoC) platform maintained viability and sustained proliferation of breast and prostate tumor slices for 7 days. No major changes in tissue morphology or gene expression patterns were observed within this time frame, suggesting that the CoC system provides a reliable and effective way to probe intrinsic chemotherapeutic sensitivity of tumors. The customized CoC platform accurately predicted cisplatin and apalutamide treatment response in breast and prostate tumor xenograft models, respectively. The culture period for breast cancer could be extended up to 14 days without major changes in tissue morphology and viability. These culture characteristics enable assessment of treatment outcomes and open possibilities for detailed mechanistic studies.

Significance:

The Cancer-on-Chip platform with a 6-well plate design incorporating silicon-based microfluidics can enable optimal patient-specific treatment strategies through parallel culture of multiple tumor slices and diagnostic assays using primary tumor material.

Cancer treatment faces major challenges in defining the optimal therapy for each individual patient. While moving toward personalized treatment regimens, better stratification of patients is becoming a major hurdle. Although several molecular biomarker-driven treatment strategies have been employed for cancer therapy, they cannot reliably predict individual therapy response to chemotherapy in most cases. Therefore, ex vivo bioassays that can predict response of the patient would strongly aid selection of the most effective treatment for each patient aiming for optimal life expectancy and quality of life. Although cancer cell lines and animal models have been used in the past to study efficacy of chemotherapy (1–5), these models cannot easily be adapted to probe sensitivities for individual patient tumors. These preclinical models are mainly used to study general characteristics for a certain tumor type or disease stage. However, the heterogeneity of cancer and lengthy procedures to establish these models limit their use to predict individual treatment responses.

With the advent of cutting-edge molecular diagnostics tools, molecular profiling of tumors is expected to help select optimal therapies for individual patients. For example, patients with (breast) cancer with pathogenic BRCA1/2 mutations are homologous recombination (HR)-deficient, which makes them eligible for PARP inhibitor (PARPi) therapy (6). However, in most cases interpretation of molecular profiles for chemotherapy sensitivity prediction is not yet possible.

Direct assessment of drug response for patient stratification and personalized medicine in primary tumor tissue slice cultures may be a solution, as they contain all tumor cells, including immune cells, and maintain tissue architecture (7–9). For example, ex vivo culture of primary tumor tissue slices has been employed for functional analysis of HR proficiency in patients with breast cancer who may be eligible for PARPi therapy (8, 10). We have also shown that ex vivo treatment of prostate patient-derived xenograft (PDX) tissue slices with drugs faithfully recapitulates the in vivo response (11). However, long-term ex vivo culture (beyond 7 days) is still a challenge, probably due to mechanical stress on the tissue slices that may lead to disruption of tissue integrity and nonphysiologic behavior and/or suboptimal culture conditions (e.g., oxygen gradients that may occur in air–liquid interface cultures). Therefore, it is important to design a more physiologic ex vivo tissue slice culture system that allows long-term culture of tumor slices under precisely controlled conditions.

Organ-on-chip (OoC) systems encompass a combination of microfluidics, microfabrication, and tissue engineering for controlled culture of cells or tissue specimens (12, 13). It provides perfusion of media with continuous supply of nutrients to the cells, removal of waste, and the option of repeated sampling of the outflow for analysis. OoC systems have been used for preclinical studies with engineered tissue from layered cancer cell lines, spheroids, and organoids (13–15). Recently, several groups have shown that various types of primary tissue can also be maintained in a microfluidic OoC platform under controlled flow of media (16–20).

Chemotherapy sensitivity assessment of (breast) tumors requires development of a reproducible culture system for the biological material. Ideally, the assay should use live material that closely resembles the original tumor and probes sensitivity in a reproducible way. Three-dimensional tissue culturing with continuous perfusion in a controlled environment appears to be the best solution for this purpose. The challenge is to design the optimal Cancer-on-Chip (CoC) platform, sufficiently representative for the in vivo cancer tissue that needs to be treated and allowing long-term culture of complex 3D tumor tissue without gross change in viability or tissue characteristics. Ideally, such a system should allow direct evaluation of treatment response by microscopic imaging and liquid sampling-based analysis. In this study, we describe a novel microfluidic CoC platform, which is based on an easy-to-use 6-well plate design supporting silicon-based microfluidic chips. These chips are more flexible than culture systems using glass because they allow easy integration of sensors for detection of pH, metabolite screening, and oxygen sensing. Moreover, a silicon-based design allows parallelization as it is based on semiconductor technology improving scalability, reproducibility, and possibility for cost-effective large-scale production. One of the applications of our novel CoC platform is in personalized medicine by enabling culturing of tumor tissue slices in vitro under precisely controlled conditions to predict in vivo tumor response of individual patients to therapy. We here show that the novel CoC platform allows tumor-slice culturing using PDXs and that this culture system faithfully mimicked the in vivo response to cisplatin therapy for breast cancer and apalutamide treatment for prostate cancer.

Design of the CoC platform

We have developed a microfluidic chip that consists of microfluidic channels embedded in a polydimethylsiloxane (PDMS) film (Fig. 1A; Supplementary Fig. S1). The PDMS film with the microfluidics is supported by a silicon (Si) frame, which includes a well facing the PDMS layer. The frame also includes two openings, an inlet and outlet, to the channels in the film (Fig. 1B). Each channel (400 μm wide) is fully isolated and confined in the PDMS film and in contact with the open well via an array of micropores with 4 μmol/L pore size and 4% porosity (4.54 × 105 pores cm–2). The organotypic tissue slices were cultured in a closed version of a BIOND plate, the COMPlate (Fig. 1C and D). This plate enables connection between multiple microfluidic chips with the tissue slices and the perfused media provided through a pumping system with controlled flow (Fluigent Deutschland GmbH, Germany). The plate can accommodate up to 6 chips in parallel in a 6-well format. Each chip is sealed between the bottom and top plates with its own inflow and outflow channels, allowing the possibility to take out one of the chips individually during the test without interfering with the others (Fig. 1D). The top plate with 4-mm x 4-mm width and 3.5-mm height provides the compartment to flow the media from the top channel to preserve the viability of the multilayered tissue. Additionally, the top plate provides an interface between the inlet and outlet of the chip to enable the diffusion of media through the membrane that supports the tissue. The top plate has four standard 1/4″ microfluidic fittings to connect to external pumping systems (Fig. 1D). The bottom part is designed to guarantee microscope compatibility and oxygenation through the PDMS of the window openings underneath the chips (Fig. 1E; ref. 21). Detailed description on fabrication of the BIOND microfluidic chip can be found in the supplementary method. From now on, we refer to the combination of microfluidic chip and tissue slice culture interface as CoC platform.

Figure. 1.

Microfluidic CoC platform design and overview. A, Top view of the microfluidic chip illustrating its components: the PDMS film in which the microfluidics are embedded, the silicon frame, which includes the inlet and outlet to the channels in the film. Scale bar, 2.5 mm. B, Vertical cross-section of the microfluidic chip. The PDMS film with the microfluidic channel is supported by a silicon frame, which includes a well facing the PDMS layer. The microchannel and the well are separated by a microporous PDMS membrane (4-μm pore size). Scale bar, 100 μm. C, Representation of the CoC platform. The platform, which consists of a bottom and a top plate, houses the microfluidic chip and allows for its connection to external fluidics. The ring is used to seal the system, to maintain adequate pressure for the controlled flow within the fluidic channel, and to minimize leakage. Scale bar, 2 cm. D, Accessories of the CoC platform with two different top plate designs. We have used double flow for this study. E, Cross-section of CoC illustrating the diffusion and perfusion toward the tissue slice. First the tissue slice is added to the microfluidic chip, which is in turn sandwiched between the top and bottom plates and connected to the external pump. Breast PDX tissue slices were perfused with an inlet flowrate of 5 μL/minute through the top and bottom channels. F, CoC platform connected to Fluigent Microfluidic flow control system that was further connected to flowrate sensors (Fluigent FLOW UNIT-S) using Fluigent MAT for the entire culture period.

Figure. 1.

Microfluidic CoC platform design and overview. A, Top view of the microfluidic chip illustrating its components: the PDMS film in which the microfluidics are embedded, the silicon frame, which includes the inlet and outlet to the channels in the film. Scale bar, 2.5 mm. B, Vertical cross-section of the microfluidic chip. The PDMS film with the microfluidic channel is supported by a silicon frame, which includes a well facing the PDMS layer. The microchannel and the well are separated by a microporous PDMS membrane (4-μm pore size). Scale bar, 100 μm. C, Representation of the CoC platform. The platform, which consists of a bottom and a top plate, houses the microfluidic chip and allows for its connection to external fluidics. The ring is used to seal the system, to maintain adequate pressure for the controlled flow within the fluidic channel, and to minimize leakage. Scale bar, 2 cm. D, Accessories of the CoC platform with two different top plate designs. We have used double flow for this study. E, Cross-section of CoC illustrating the diffusion and perfusion toward the tissue slice. First the tissue slice is added to the microfluidic chip, which is in turn sandwiched between the top and bottom plates and connected to the external pump. Breast PDX tissue slices were perfused with an inlet flowrate of 5 μL/minute through the top and bottom channels. F, CoC platform connected to Fluigent Microfluidic flow control system that was further connected to flowrate sensors (Fluigent FLOW UNIT-S) using Fluigent MAT for the entire culture period.

Close modal

Simulation of fluid-flow and analysis of shear stress

The microfluidic chip and COMPlate were modelled using the finite element method–based software COMSOL Multiphysics. The modelling was specifically made for solving the Navier–Stokes equations under steady-state conditions given the pressure-driven system used in practical experiments.

PDX models

Breast cancer PDX tumors have been described before (2). Mouse experiments were approved by the Netherlands Cancer Institute animal experiments committee. We have used a breast cancer PDX model established from a triple-negative breast tumor with a HR deficiency caused by BRCA1 mutations (c.2210delC/2329delC) resulting in cisplatin-sensitive tumors (2).

A cisplatin-resistant subline was created by repeated cisplatin treatment (T250), as described previously (2). The prostate cancer PDX model0 PC82 was established from a radical prostatectomy specimen of an otherwise untreated patient previously (22). The PDX is androgen dependent and antiandrogen therapies, such as apalutamide treatment, cause immediate tumor regression (22, 23).

Tumor tissue slice culture in CoC platform

To establish the tumor tissue slice culture protocol in our CoC platform, breast and prostate cancer PDX tumors were used. For each experiment, breast PDX tumor was sliced in 300-μm–thick slices using a Leica VT1200S vibratome and placed directly in 3 mL customized breast medium within 3 hours after the tumor was removed from the mouse, as described previously (8). Breast PDX tissue slices were either cultured in 3 mL customized culture medium in 6-well standard plates on an orbital shaker at 60 rpm (referred to as ex vivo culture) or in the CoC platform in duplicates (Supplementary Fig S2A). Prostate PDX tumor PC82 tissue slices were processed as per previously established protocol (11) and cultured using serum-free aDMEM/F12 K medium with synthetic androgen R1881 in ex vivo culture and in the CoC platform. For the CoC platform, breast and prostate PDX tissue slices were attached to the PDMS membrane using Mebiol hydrogel (Cosmobio), a biocompatible thermo-reversible copolymer of poly(N-isopropylacrylamide) and poly(ethylene glycol) providing additional protection of the tumor tissue slice from shear force and allowing gas and nutrient exchange. The complete system was assembled and connected to a FLPG Plus pumping system for continuous perfusion (Fig. 1F). The flow rate was maintained by a pressure-driven microfluidic flow control system (Fluigent MFCSTM-EZ) and monitored by flowrate sensors (Fluigent FLOW UNIT-S) using Fluigent Microfluidic Automation Tool (MAT) for the entire culture period. PDX tissue slices were perfused with an inlet flowrate of 5 μL/minute through the top and bottom channels. Both ex vivo and CoC tissue culture were performed in a humidified 5% CO2 atmosphere at 37°C and medium was changed every 3 days for up to 2 weeks. Tissue slices were removed from the chip at various time points (day 7 and day 14), kept on prechilled medium at 0°C for 5 minutes to liquify the hydrogel and transferred to fresh medium containing 3 μg/mL 5‐ethynyl‐2′‐deoxyuridine (EdU; Invitrogen) during the last 2 hours before fixation. Subsequently, the tissue slices were fixed in 10% neutral buffered formalin for at least 24 hours at room temperature. Tumor slices were embedded in paraffin and 4-μm sections were made for microscopic analysis.

Treatment of PDX tissue slices in CoC platform

Cisplatin response was evaluated using cisplatin-sensitive breast PDX tumors and cisplatin-resistant breast PDX tumors. Breast tumor tissue slices from cisplatin-sensitive breast PDX tumors (n = 3) and cisplatin-resistant breast PDX tumors (n = 3) were cultured in ex vivo 6-well plates and the CoC platform, with or without cisplatin (5 μg/mL) in medium for 72 hours. Subsequently, tissue slices were harvested and processed as described above (Supplementary Fig. S2B).

In addition, the CoC platform was tested for the treatment response of prostate PDX against the antiandrogen apalutamide. PC82 tumor (n = 1) tissue slices were cultured in duplicates using serum-free aDMEM/F12 K medium with synthetic androgen R1881 with or without apalutamide at a final concentration of 1 μmol/L for 7 days in CoC platform. After 7 days, PC82 PDX tissue slices were harvested and processed for staining and image analysis as described above (Supplementary Fig. S2B).

Tissue slice staining and analysis

Hematoxylin and eosin staining was performed as described previously (8). To study cell proliferation, thymidine nucleoside analog EdU incorporation was assessed. EdU (3 μg/mL; Invitrogen) was added to the culture medium 2 hours before formalin fixation and its incorporation was analyzed as described previously (8). Tissue slices that had been cultured in the CoC device, were removed from the microfluidic chamber and incubated for the last 2 hours under ex vivo conditions with EdU as described above. Cell viability was determined using the terminal deoxynucleotidyl transferase–mediated dUTP nick end labelling (TUNEL) assay (Roche Life Sciences; ref. 8). To quantify EdU and TUNEL staining, 10 random images (×400 magnification) from different locations of the section were captured from each tumor tissue slice section that was cultured in ex vivo 6-well plate or CoC platform using a Leica fluorescence microscope (DM4000b) to represent the tumor tissue slice heterogeneity. After fixation, paraffin embedding, and sectioning, sections were deparaffinized in xylene followed by rehydration in graded alcohols. Antigen retrieval was performed with target retrieval buffer, pH 6 (Dako). For diaminobenzidine (DAB) staining, endogenous peroxidase activity was blocked with 3% hydrogen peroxide solution in methanol for 20 minutes at room temperature. Primary antibody anti-Pan Cytokeratin (AE1/AE3; Santa Cruz Biotechnology, sc-81714) was diluted in blocking buffer (1/500 in 5% BSA in PBS) and applied to the sections for 90 minutes at room temperature and a secondary Alexa Fluor 488 antibody conjugate was used to detect the first antibody. The Ki67 antibody (ab16667, Abcam) was diluted in blocking buffer (1/200 in 5% BSA in PBS) and applied to the sections at 4°C overnight. Ki67 was detected with horseradish peroxidase (HRP)‐conjugated anti‐rabbit IgG secondary antibody (Dako) at a 1:100 dilution for 1 hour at room temperature. Ki67-positive cells were visualized using DAB staining kit (Agilent) followed by counterstaining with hematoxylin. TUNEL, EdU, and DAPI images were analyzed automatically using previously published methods (10). Quantitative image-based cytometry single-cell analysis (QIBC) was performed for immunofluorescence analysis of cell cycle using anti-geminin (Proteintech Group 10802–1-AP, 1:400) in PDXs cultured in ex vivo and CoC platform conditions. The analysis of the imaging data has been conducted using customized pipelines in Cell Profiler (24). Nuclei were detected using the DAPI signal and filtered for intensity, size, perimeter, and solidity to exclude nontumor and stromal cells. The intensity in 594 channel was corrected with the background intensity. The selection of geminin-/DAPI-positive cells were performed based on the mean intensity of the geminin signal for individual nucleus. For the quantification of 53bp1 foci per nucleus, a mask was generated in the 53BP1 channel using the detection of local intensity maxima, with a threshold set for spot radius and intensity. Clumped spots were split using the shape features. The desired quantified values for each nucleus were exported to the Tibco Spotfire software and to the GraphPad Prism 8 for generation of scatter diagrams and bar chart respectively.

For prostate PDX tumor slices, primary androgen receptor (AR) antibody (M4074, 1/200, SPRING Bioscience), diluted in blocking buffer, was applied to the sections at 4°C overnight. Subsequently, slides were incubated with HRP-conjugated anti-rabbit IgG secondary antibody (Dako) at a 1:100 dilution for 1 hour at room temperature. AR-positive cells were visualized using DAB staining kit (Agilent) followed by counterstaining with hematoxylin. Imaging of AR staining was performed using a light microscope (Olympus). For quantification of AR expression, four fields (×200 magnification) from each prostate PDX tumor slice section were captured and analyzed using Image J software as described previously (11). To assess DNA damage repair, 53BP1 (anti-rabbit, 1/1000, Novus Biologicals) staining was performed in PDX tumor tissue slices treated with cisplatin as described previously (25).

RNA isolation and sequencing

Whole transcriptome analysis was performed using RNA sequencing (RNA-seq) using three independent breast cancer PDX tumors. RNA was isolated from formalin-fixed, paraffin-embedded (FFPE) tissue using RNeasy FFPE kit (Qiagen) according to the manufacturer's protocol. Quality and quantity of the RNA was assessed using a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific). RNA libraries for mRNA sequencing were prepared from FFPE tissue using Kapa mRNA Hyperprep kit (Roche Life Sciences) with unique dual indices from Integrated DNA Technologies (IDT). Sequencing was performed using 2 × 150 bp paired-end method on the Illumina Novaseq 6000 platform.

Bioinformatics analysis

The FastQ files obtained were analyzed for QC using the FastQC program. Adapter trimming and removal of homo-polymers were performed using FLEXBAR and reads less than 25 bases were discarded (26).The trimmed and processed FastQC files were separately aligned to human and mouse genomes (GRCh38 and GRCm38 versions) downloaded from GENCODE respectively using HISAT2 (a splice aware aligner; ref. 27). Samtools and PICARD programs were used on the aligned files to obtain the mapping statistics and other RNA-seq metrics (28). We used XenofilteR, an R-package to separate mouse reads from the aligned human sequence reads (29). Expression analysis was performed for both human and mouse mapped reads using StringTie using the reference only approach and the gene abundance tables were obtained (30). We used the feature Counts program to generate the count table for differential expression analysis (31). Bioconductor package edgeR was used for differential expression analysis (32). Gene ontology and pathway analysis of differentially expressed genes were performed using metascape (www.metascape.org), Ingenuity Pathway Analysis (IPA) for canonical pathway analysis.

Pathway analysis

qRT-PCR–based pathway analysis tests were performed using three independent breast PDX tumors with each cultured at day 0, up to day 7, and up to day 14 in ex vivo and CoC platform. Functional activity of the estrogen receptor (ER), AR, PI3K, and MAPK growth factor pathways, Hedgehog (HH), Notch, and TGFβ cell signaling pathways was measured using qRT-PCR–based pathway analysis tests using FFPE derived RNA from each breast PDX tumor group cultured in ex vivo and CoC platform (33). Functional pathway activity of individual pathways was analyzed based on the interpretation of mRNA levels of target genes of the pathway-associated transcription factors (www.Philips.com/Oncosignal). Pathway activity scores are presented on a normalized 0 to 100 scale, as described previously (34–36).

Statistical analysis

Results are expressed as the mean ± SEM in bar graph. Mann–Whitney test was used to analyze the differences between two groups. Statistical analysis and generation of graphs was performed using GraphPad Prism 8.0. P value < 0.05 was considered statistically significant.

CoC workflow

Development of methods to predict chemotherapy sensitivity of tumors requires a robust and reproducible culture system for the biological material. Ideally, the assay should use live material that closely resembles the original tumor and probes sensitivity. Three-dimensional (3D) tissue culturing with continuous perfusion in a controlled environment appears to be the best solution for this purpose. Therefore, we designed a novel platform with a microfluidic chip and tissue culture interface for OoC cultures with 3D tissue specimens (Fig. 1A and B). The platform was designed to accommodate tissue slices that can be used for cancer (chemo)therapy response assays. Tissue slices were immobilized using a thermoreversible hydrogel that allows easy removal of the biological specimen after cooling on ice. The tissue culture interface was made in a 6-well plate format, which hosts the microfluidic chips with tumor tissue slices (Fig. 1C). The design of the COMPlate allows individual tumor slices to be cultured and analyzed independently. The top plate is designed to provide option for single flow or double flow of the media (Fig. 1D). Each well provides enough space to accommodate the tissue slice and allows growth over time. The tissue slices can be maintained while removing waste products and replenishing nutrients through constant perfusion and diffusion from the microchannel and the top part of the interface (Fig. 1E). Moreover, oxygenation of the tissue slices is augmented by gas exchange through the PDMS layer of the optical window (Fig. 1E). The entire CoC platform was connected to a Fluigent Microfluidic flow control system to monitor fluid flow rate inside each individual well during the entire culture period (Fig. 1F).

Modeling of shear stress in the CoC platform

The CoC geometry consists of three microfluidic channels (90 μmol/L thick, 400 μmol/L wide) with a porous membrane (4 μmol/L pore size), the well (4 mm × 4 mm, 3.5-mm height) and the corresponding inlets and outlets for perfusion of media (Supplementary Fig. S3A shows the situation for one microfluidic bottom channel). As the precise characteristics were not available for all components, we modelled the microfluidics set up such that maximally expected shear force in the most unfavorable case was determined. The tissue slice was modelled as a porous material with permeability 3 × 10–12 m2 (37). The hydrogel used to fix the tissue slices was modelled as a water permeable rectangular region (Supplementary Fig. S3B; ref. 38). The medium was modelled as an incompressible fluid with a density of 1,000 kg/m3 and a viscosity of 1.0 mPa-s. The inlet flow rate on top and bottom was set to 5 μL/minute. The outlets were set as open boundaries at environmental pressure. The boundaries defining the microchannel, pores and well were set as walls with a no-slip condition. The domains were then meshed with a highly dense (3.5 million elements) and average quality (0.8%) mesh. The model was solved in static conditions to obtain the velocity field and shear stress distributions close to the tissue. The velocity and stress fields shown correspond to the well and areas where the tissue is located during experiments (Supplementary Fig. S3C and S3D). Simulations showed that under the conditions set for the inlet flow (5 μL/minute), a laminar flow was obtained with the average velocity (0.4–1 μm/second) around the tissue within the range of physiologic interstitial flow (37, 39). The highest shear stress experienced by the tissue would be very low (approximately 10–5 dyn/cm2), even below physiologic shear stresses detected in lymphatic capillaries (40). Further simulations were performed to show the fluid flow across the membrane inside the CoC platform (Supplementary Fig. S4). We conclude that this low level of shear stress will most probably not influence biological functions in a nonphysiologic way.

Treatment response of PDX tumors in the CoC platform

In order to use the CoC platform for drug sensitivity assessment, the first important question is whether treatment responses in the CoC platform are predictive for therapy response in vivo. We tested response to cisplatin treatment using cisplatin-sensitive and cisplatin-resistant breast cancer PDX tumors (three biological replicates of each). Cytokeratin staining showed that most cells in the breast PDX tumor tissue slices were tumor cells with minimal numbers of stromal cells at the start of the treatment, suggesting tissue responses are mainly caused by tumor cell sensitivities (Supplementary Fig S5). Untreated tumor slices at day 0 and at day 7 in the CoC platform showed Ki67-positive tumor epithelial cells and negative mouse stromal cells, suggesting that the CoC platform culture method retains the tumor-associated cell morphology and proliferative capacity, which is essential for treatment response analysis (Supplementary Fig. S6). Tissue slices cultured under regular ex vivo conditions and in the CoC platform were treated with cisplatin for 3 days and evaluated for proliferation and cell death. Cisplatin-sensitive tumor slices in the CoC platform showed a significant increase in TUNEL-positive (apoptotic) cells and a significant drop in EdU-positive (replicating) cells upon cisplatin treatment (Fig. 2A and B). As expected, the cisplatin-resistant PDX tissue slices did not show any significant change in EdU and TUNEL signal when compared with untreated controls (Fig. 2C and D). Cisplatin treatment for 72 hours caused a much higher accumulation of 53BP1 foci in cisplatin-sensitive PDX than in cisplatin-resistant tumor slices (Fig. 2E and F), presumably as a result of failed repair of the DNA damage in the sensitive model. Cultures in the CoC platform showed a response to cisplatin treatment that correlated well with the known tumor response in vivo (2), as well as in ex vivo cultures (Fig. 2B and D; Supplementary Fig. S7A and S7B). Interestingly, breast PDX cultured in the CoC platform showed a stronger response to cisplatin treatment than the ex vivo culture method, suggesting that the CoC platform provides more optimal drug delivery into the tumor slices than the ex vivo culture method (Supplementary Fig. S8).

Figure 2.

Prediction of therapy response using cisplatin-sensitive and -resistant PDX in ex vivo and CoC platform. Three cisplatin-sensitive and three cisplatin-resistant breast PDX tumors were used in three independent experiments to study cisplatin response in ex vivo 6-well plate culture and CoC platform. Cisplatin-sensitive and -resistant PDX tumor tissue slices were exposed to 5 μg/mL cisplatin for 3 days and evaluated for cell proliferation (EdU incorporation, red nuclei) and apoptosis (TUNEL staining, green nuclei). DAPI stains all nuclei blue. A, Representative EdU and TUNEL staining of cisplatin-sensitive breast PDX. B, Quantification of the fraction of EdU-positive and TUNEL-positive cells showing breast PDXs (n = 3) were sensitive to cisplatin. C, Representative EdU and TUNEL staining of cisplatin-resistant breast PDX. D, Quantification of the fraction of EdU-positive and TUNEL-positive cells showing breast PDXs (n = 3) were insensitive to cisplatin therapy, thereby validating the application of CoC for therapy response for patient tumors. Ten fields of view were quantified from each breast PDX slice. Each point in the graph represents one breast PDX sample and SEM is indicated for all three tumors in the three independent experiments. Scale bar, 50 μm. E, Analysis of DNA damage response in cisplatin-sensitive and -resistant PDX treated with cisplatin. Cisplatin treatment induced more double-strand breaks in cisplatin-sensitive PDX than in cisplatin-resistant PDX. Scale bar 50 μmol/L. F, Scatter plot showing 53BP1 foci count per cell in cisplatin-sensitive and -resistant PDX. Averages and SEM are indicated. Cis, cisplatin; cis res, cisplatin-resistant; cis sen, cisplatin-sensitive; H&E, hematoxylin and eosin. NS, not significant; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

Figure 2.

Prediction of therapy response using cisplatin-sensitive and -resistant PDX in ex vivo and CoC platform. Three cisplatin-sensitive and three cisplatin-resistant breast PDX tumors were used in three independent experiments to study cisplatin response in ex vivo 6-well plate culture and CoC platform. Cisplatin-sensitive and -resistant PDX tumor tissue slices were exposed to 5 μg/mL cisplatin for 3 days and evaluated for cell proliferation (EdU incorporation, red nuclei) and apoptosis (TUNEL staining, green nuclei). DAPI stains all nuclei blue. A, Representative EdU and TUNEL staining of cisplatin-sensitive breast PDX. B, Quantification of the fraction of EdU-positive and TUNEL-positive cells showing breast PDXs (n = 3) were sensitive to cisplatin. C, Representative EdU and TUNEL staining of cisplatin-resistant breast PDX. D, Quantification of the fraction of EdU-positive and TUNEL-positive cells showing breast PDXs (n = 3) were insensitive to cisplatin therapy, thereby validating the application of CoC for therapy response for patient tumors. Ten fields of view were quantified from each breast PDX slice. Each point in the graph represents one breast PDX sample and SEM is indicated for all three tumors in the three independent experiments. Scale bar, 50 μm. E, Analysis of DNA damage response in cisplatin-sensitive and -resistant PDX treated with cisplatin. Cisplatin treatment induced more double-strand breaks in cisplatin-sensitive PDX than in cisplatin-resistant PDX. Scale bar 50 μmol/L. F, Scatter plot showing 53BP1 foci count per cell in cisplatin-sensitive and -resistant PDX. Averages and SEM are indicated. Cis, cisplatin; cis res, cisplatin-resistant; cis sen, cisplatin-sensitive; H&E, hematoxylin and eosin. NS, not significant; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

Close modal

The androgen-dependent prostate tumor PC82 was used to assess the performance of the CoC platform in a completely different tumor type. Tumor tissue slices were cultured for 7 days with or without the antiandrogen apalutamide. We found significantly reduced AR expression, increased TUNEL-positive cells, and a significant decrease in EdU-positive cells after apalutamide exposure, very similar to the effect of antiandrogen treatment in vivo (41) and in ex vivo cultures (Fig. 3; Supplementary Fig. S7C). In summary, the CoC cultures recapitulate tumor responses for two different treatments in breast and prostate tumor models with known sensitivities in vivo.

Figure 3.

Response to apalutamide treatment in CoC platform. A, PC82 tumor tissue slices cultured with apalutamide for 7 days and evaluated for AR expression, cell proliferation (EdU incorporation, red nuclei), and apoptosis (TUNEL staining, green nuclei). AR staining for PC82 tissue slice sections at day 0, day 7 ex vivo culture condition, and CoC platform with and without apalutamide treatment. DAPI stains all nuclei blue. H&E, hematoxylin and eosin. Scale bar, 50 μm. B, Quantification of the AR expression in prostate PDX slices cultured in ex vivo and CoC platform. C and D, Apalutamide treatment showed a significant increase in TUNEL-positive cells and significant decrease in EdU-positive cells when compared with untreated tumor slices in our CoC platform. Ten fields of view were quantified per prostate PDX slice. Each point in the graph represents one image field and SEM is indicated for 10 fields. D7, day 7; D0, day 0. ***, P < 0.001; ****, P < 0.0001.

Figure 3.

Response to apalutamide treatment in CoC platform. A, PC82 tumor tissue slices cultured with apalutamide for 7 days and evaluated for AR expression, cell proliferation (EdU incorporation, red nuclei), and apoptosis (TUNEL staining, green nuclei). AR staining for PC82 tissue slice sections at day 0, day 7 ex vivo culture condition, and CoC platform with and without apalutamide treatment. DAPI stains all nuclei blue. H&E, hematoxylin and eosin. Scale bar, 50 μm. B, Quantification of the AR expression in prostate PDX slices cultured in ex vivo and CoC platform. C and D, Apalutamide treatment showed a significant increase in TUNEL-positive cells and significant decrease in EdU-positive cells when compared with untreated tumor slices in our CoC platform. Ten fields of view were quantified per prostate PDX slice. Each point in the graph represents one image field and SEM is indicated for 10 fields. D7, day 7; D0, day 0. ***, P < 0.001; ****, P < 0.0001.

Close modal

Comparison of long-term tumor tissue slice culture in ex vivo and in CoC platform

Ideally, culture time should be extended to allow studies into therapy responses that require incubation times extending beyond 1 week (such as therapy resistance development or clonal outgrowths). One of the main limitations of the ex vivo culture appeared to be the maximum culture time preserving optimal proliferative capacity and tissue architecture. Breast tumor slices from five independent PDX tumors were cultured in the CoC platform for up to 14 days. In parallel, tissue slices from the same breast PDX tumors were cultured in the ex vivo 6-well plate. After 7 days of culture, we observed similar cellular proliferation rate in CoC (fraction of EdU-positive cells) as in day 0 (25.05% vs. 23.06%), but a slightly slower proliferation in the ex vivo condition (23.06% vs. 19.50%; Fig. 4A and B). TUNEL staining showed slightly increased cell death ex vivo than in CoC at day 7 (3.5% versus 2.2%; Fig. 4A and B). Interestingly, this pattern was more pronounced in extended cultures for 14 days, which revealed better preservation of tumor tissue architecture and cell proliferation in the CoC platform than in the ex vivo culture system. We found a much higher fraction of EdU-positive cells in CoC than ex vivo (20.65% vs. 8.06%, P < 0.0001; Fig. 4C). Moreover, analysis of cell death by TUNEL staining showed 17.03% TUNEL-positive cells in the ex vivo condition and only 6% in the CoC platform (P < 0.0001; Fig. 4C). Comparatively, in CoC, we observe very little change in cell proliferation between day 0 (25%) and day 14 tumor slices (20.65%; Fig. 4C).

Figure 4.

Breast PDX tumor tissue slices cultured in ex vivo condition and in CoC platform for up to 14 days. A, We used five independent breast PDX tumors in five independent experiments to establish optimized culture condition for tumor slices in our CoC platform. From each breast PDX tumor, tissue slices were cultured in ex vivo 6-well plate and CoC device for up to 14 days and evaluated for tissue morphology [hematoxylin and eosin (H&E) staining], DAPI (blue nuclei), proliferation (EdU incorporation, red nuclei), Ki67 (brown nuclei), and apoptosis (TUNEL staining, green nuclei). Scale bar, 50 μm. B and C, Quantification of the fraction of EdU-positive and TUNEL-positive cells for 5 breast PDX tissue slices (derived from individual PDX tumors) cultured for up to day 7 (B) and day 14 (C). For each breast PDX tumor and each condition, tissue slices were imaged, and ten random fields of view were quantified from each breast PDX slice. Each data point in the graph represents one image field. Error bar represents the SEM for the five independent tumors performed in five independent experiments. D, Representative image showing breast PDX tumors (n = 3) labeled with geminin (red nuclei) and DAPI (blue nuclei). Scale bar, 50 μm. E, QIBC analysis of three independent breast PDX tumors with more than 3,000 cells analyzed for each are shown in each condition. F, Quantification of geminin-positive cells showed CoC at day 7 had similar cell proliferation profile as in day 0 than ex vivo condition. Error bar, SEM. NS, not significant; D0, day 0; D7, day 7, D14, day 14. NS, not significant; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

Figure 4.

Breast PDX tumor tissue slices cultured in ex vivo condition and in CoC platform for up to 14 days. A, We used five independent breast PDX tumors in five independent experiments to establish optimized culture condition for tumor slices in our CoC platform. From each breast PDX tumor, tissue slices were cultured in ex vivo 6-well plate and CoC device for up to 14 days and evaluated for tissue morphology [hematoxylin and eosin (H&E) staining], DAPI (blue nuclei), proliferation (EdU incorporation, red nuclei), Ki67 (brown nuclei), and apoptosis (TUNEL staining, green nuclei). Scale bar, 50 μm. B and C, Quantification of the fraction of EdU-positive and TUNEL-positive cells for 5 breast PDX tissue slices (derived from individual PDX tumors) cultured for up to day 7 (B) and day 14 (C). For each breast PDX tumor and each condition, tissue slices were imaged, and ten random fields of view were quantified from each breast PDX slice. Each data point in the graph represents one image field. Error bar represents the SEM for the five independent tumors performed in five independent experiments. D, Representative image showing breast PDX tumors (n = 3) labeled with geminin (red nuclei) and DAPI (blue nuclei). Scale bar, 50 μm. E, QIBC analysis of three independent breast PDX tumors with more than 3,000 cells analyzed for each are shown in each condition. F, Quantification of geminin-positive cells showed CoC at day 7 had similar cell proliferation profile as in day 0 than ex vivo condition. Error bar, SEM. NS, not significant; D0, day 0; D7, day 7, D14, day 14. NS, not significant; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

Close modal

This was confirmed by immunofluorescent staining for geminin, as marker for S and G2 phase cells, where we observed a significant decrease in geminin-positive cells ex vivo at day 7 when compared with day 0 PDX tumors (31.65% vs. 19.20%; P = 0.0002), while no significant differences could be observed between day 7 CoC and day 0 (31.65% vs. 27.20%; Fig. 4DF), suggesting slightly slower cellular proliferation of tumors cultured in ex vivo condition in comparison with CoC (Fig. 4DF). These data indicate that the CoC platform provides better culture conditions for extended (beyond 7 days) culture times of tumor tissue slices than the ex vivo system and that there may already be subtle differences between both culture methods even after 7 days.

Gene expression analysis

In order to characterize the CoC culture for more subtle gene expression changes, we performed RT-PCR–based analysis of tumor-specific genetic pathways from breast PDX tumors cultured ex vivo (n = 3) and in the CoC platform (n = 3). We did not observe any statistically significant changes in these pathways (Fig. 5A; Supplementary Table S1), suggesting that both culture conditions did not induce any major changes in growth characteristics of the tumor.

Figure 5.

Gene expression analysis of breast PDX tumor cultured in ex vivo and CoC platform. Three independent breast PDX tumors (P1, P2, and P3) were used for analysis of pathway activity using OncoSignal and RNA-seq for whole mRNA expression analysis. A, Seven major pathways were analyzed (ER, AR, PI3K, and MAPK growth factor pathways, HH, Notch, and TGFβ cell signaling pathways) using OncoSignal. Pathway activity scores are presented on a normalized scale of 0 to 100. Error bars, SEM. Mann–Whitney test of seven pathway activity across PDX tumor slices (day 0, day 7, day 14) in ex vivo and CoC condition did not show any statistically significant change. B, Venn diagram showing differentially expressed human (ex vivo-D7, CoC-D7, and CoC-D14) genes (logFC > 1.5 or logFC ≤ 1.5, FDR < 0.1) in PDX tumors at day 7 and day 14 in ex vivo and CoC platform when compared with day 0 PDX tumors. C, Heatmap diagram showing IFN signaling pathway gene expression at different days in ex vivo and in CoC platform culture condition. D, Network diagram showing top ranked networks significantly enriched in immune signaling and tumor microenvironment (P < 0.05) exclusively in day 7 ex vivo culture condition. Orange, predicted activation of the pathways (Z score > 2); blue, predicted inhibition (Z score ≤2). E, Representative image showing 53BP1 staining in breast PDX tumors cultured at day 0, day 7 ex vivo and CoC condition. DAPI staining is shown as blue nuclei. Scale bar, 50 μm. F, Bar chart showing the percentage of cells with 53BP1 foci. Error bar, SEM for the three independent PDX tumor samples. D0, day 0; D7, day 7, D14, day 14. *, P < 0.05; **, P < 0.01; ****, P < 0.0001.

Figure 5.

Gene expression analysis of breast PDX tumor cultured in ex vivo and CoC platform. Three independent breast PDX tumors (P1, P2, and P3) were used for analysis of pathway activity using OncoSignal and RNA-seq for whole mRNA expression analysis. A, Seven major pathways were analyzed (ER, AR, PI3K, and MAPK growth factor pathways, HH, Notch, and TGFβ cell signaling pathways) using OncoSignal. Pathway activity scores are presented on a normalized scale of 0 to 100. Error bars, SEM. Mann–Whitney test of seven pathway activity across PDX tumor slices (day 0, day 7, day 14) in ex vivo and CoC condition did not show any statistically significant change. B, Venn diagram showing differentially expressed human (ex vivo-D7, CoC-D7, and CoC-D14) genes (logFC > 1.5 or logFC ≤ 1.5, FDR < 0.1) in PDX tumors at day 7 and day 14 in ex vivo and CoC platform when compared with day 0 PDX tumors. C, Heatmap diagram showing IFN signaling pathway gene expression at different days in ex vivo and in CoC platform culture condition. D, Network diagram showing top ranked networks significantly enriched in immune signaling and tumor microenvironment (P < 0.05) exclusively in day 7 ex vivo culture condition. Orange, predicted activation of the pathways (Z score > 2); blue, predicted inhibition (Z score ≤2). E, Representative image showing 53BP1 staining in breast PDX tumors cultured at day 0, day 7 ex vivo and CoC condition. DAPI staining is shown as blue nuclei. Scale bar, 50 μm. F, Bar chart showing the percentage of cells with 53BP1 foci. Error bar, SEM for the three independent PDX tumor samples. D0, day 0; D7, day 7, D14, day 14. *, P < 0.05; **, P < 0.01; ****, P < 0.0001.

Close modal

We extended this analysis to whole transcriptome sequencing using breast PDX tumors (n = 3). Differentially expressed genes at day 7 ex vivo and day 7 and 14 CoC culture conditions were selected based on log fold change (logFC) > 1.5, FDR < 0.1 for significantly upregulated genes and logFC <- 1.5, FDR < 0.1 for significantly downregulated human and mouse genes (Fig. 5B; Supplementary Tables S2 and S3). In this analysis, human genes represent expression in tumor cells, while the mouse genes are derived from stromal cells. We observed 150 human genes differentially expressed in ex vivo day 7 tumor slices that was significantly more than the 30 differentially expressed human genes in CoC day 7 and 14 human genes in CoC day 14 (Supplementary Table S2). These upregulated human tumor genes (logFC > 1.5, FDR < 0.1) in ex vivo culture condition are significantly enriched in several immune signaling pathways (Fig. 5C), one of which is the IFN signaling pathway (Supplementary Table S4). Consistent with this, IPA showed activation of cellular immune response in the ex vivo tumor slice culture condition (Fig. 5D). In line with these findings, gene expression analysis of mouse stromal genes showed downregulation of collagen genes (Col3a1, Col6a2, Col15a1) in the ex vivo culture (Supplementary Table S3), which is a known effect of IFNγ signaling (42).

To investigate the reason for the reduction in cell-cycle progression and the increase in apoptosis, we performed 53BP1 immunostaining to analyze for DNA damage status in tumor slices at day 0 and day 7 under both culture conditions. We observed cells with one large focus, as well as cells with multiple small foci under both ex vivo and CoC conditions (Fig. 5E). However, there was a significantly higher increase in DNA damage in tumors cultured ex vivo condition than in CoC (16.38% vs. 11.34%; P = 0.04; Fig. 5F). We therefore conclude that the ex vivo (but not the CoC) culture condition induced immune activation and that more DNA damage is present after 7 days of ex vivo culture, making the CoC system a more faithful representation of the original tumor and the preferred system for studying therapy responses.

We previously developed methods to culture precision cut breast tumor slices ex vivo (8). Here, we describe the design of a diagnostic CoC platform to grow tumor tissue slices for prolonged periods of time, which permits investigation of chemotherapy responses in a controlled environment. Defining features of this microfluidic system are: (i) relatively large (tumor) tissue slices can be loaded (ii), flow of culture media both above and under the tissue slice to sustain constant delivery of nutrients and gasses and removal of waste products (iii), mainly silicon-based components (iv), standard 6-well format, allowing parallelization, and (v) options for imaging and media outflow analysis to follow tissue responses in time.

These features make this device particularly suitable for diagnostic investigation of fresh tumor biopsies, which requires the option to load a relatively large piece of tumor material in the microfluidic device and relatively constant growth conditions throughout the tissue specimen. The top and bottom flow minimize gradients throughout a relatively thick tissue slice and the fixation of this slice on the membrane using a thermoreversible hydrogel allows both easy loading and disassembly, while also reducing tear forces on the tissue. Previously published similar systems either used only bottom flow (16, 43–45) or perfusion through the slice (18, 20). Our design not only allows efficient perfusion but can also be adapted to mimic a blood vessel (by coating the bottom channel with endothelial cells) or create gradients by using different top and bottom flow media (e.g., by adding the therapeutic agent only in the bottom channel).

Furthermore, silicon-based components reduce absorption of (chemotherapeutic) agents. This is especially important when using PDMS-based devices and hydrophobic compounds (46). In the current version of the device, PDMS is still used for the membrane supporting the tumor slice, but this can also be replaced by other materials, such as Polyethylene terephthalate (PET) and Polystyrene (PS), thermoplastics that are molded with techniques that might be coupled to the last stages of silicon-based fabrication processes, helping to overcome the intrinsic absorption limitations but guaranteeing to maintain manufacturability.

Options for parallelization and analysis of live tumor samples make this device usable for a broad spectrum of experimental designs. The current design allows testing of six tissue slices individually; each unit can be assembled and disassembled without disturbing the other samples, allowing analysis of different time points in the same run. Outflow media are collected individually, allowing analysis of their composition over time, e.g., by isolating extracellular vesicles and investigation of their mRNA content. Furthermore, the CoC device presented here is an evolution of the Cytostretch platform that can be equipped with electrodes and other sensors to allow real time measurement of additional parameters, such as oxygen, pH, and temperature (47, 48).The optical window allows visual (microscopic) detection of tissue samples during the culture period. In the current design, working distance and optical performance of the PDMS window are not optimal for live microscopic imaging. These parameters may be improved in a new version by using alternative materials, such as a glass window instead of PDMS.

The growth conditions in the CoC platform maintain original growth characteristics of the tumor better than the previously described ex vivo culture system. Proliferation was maintained for a longer period of time (14 days compared with 7 days) with less apoptosis induction at day 14. Although proliferation was maintained for 7 days in both systems, careful analysis of gene expression revealed some important changes in the ex vivo condition that were not detected in the CoC cultures. Tumor cells showed increased IFN signaling at day 7 ex vivo, but not in the CoC device, suggesting increased stress levels in the tissue that could translate to changes in therapy sensitivity and nonphysiologic activation of immune responses. In line with the increased stress levels ex vivo, tumor cells showed a subtle difference in cell cycle profile and induction of DNA damage at day 7. Although the induction of DNA damage was also observed in the CoC culture condition, it was more pronounced ex vivo. We observed many cells with a single large accumulation of 53BP1 protein. A similar phenomenon has been observed before in cells and was shown to be due to replication stress (49), suggesting that the culture conditions presented here led to similar replication problems. We suspect that the oxygen concentration may play a role in this (20% ambient oxygen compared with less than 5% in most tissues). One possible solution to reduce the basal DNA damage levels could be lowering the oxygen concentration in the CoC platform. The deterioration of tumor tissue was further enhanced at day 14, when we observed extensive morphologic changes along with increased cell death in PDX tumor slices in the ex vivo condition. The CoC cultures were much less affected and did not show a significant difference in proliferation or apoptosis.

Previously, tissue slice culture methods in air-liquid interfaces have been reported (50–52). These systems are good for mimicking lung or skin tissues. However, it is not desirable for assessment of sensitivities in tumors, which should not be exposed to high oxygen levels. Especially treatment with DNA damaging agents should not be combined with high oxygen levels, as this causes oxidative DNA damage, which may form a gradient from the outer layers of the tumor slice to the center where oxygen levels will be much lower. This would hamper analysis of responses to the DNA damaging agent under investigation. However, also other treatment responses could be influenced by oxidative (DNA) damage in such culture systems. Our observation that 53BP1 foci were increased after 7 days of culture suggests that oxidative damage may indeed be an important factor to be controlled even better than we did in the experiments presented here.

Gene expression analysis showed that immune responses (such as IFN signaling) were affected under ex vivo conditions, but not in the CoC platform. Therefore, we expect that the CoC device will be especially suitable to study immune responses in the tumor. Furthermore, immune cells can be added in a controlled way by inflow through top or bottom channel. The obvious advantage over mouse models is the option to add human immune system components and measure responses in real time.

We developed a microfluidic CoC platform that maintained cell viability, proliferation, and tissue morphology for at least 14 days for breast cancer PDX slices. It faithfully predicted cisplatin therapy response for breast cancer and antiandrogen treatment response for prostate cancer PDX tumor slices. Further validation of its potential as an in vitro diagnostic test for therapy selection will require clinical validation using biopsies of patient tumors that are going to be treated with the same chemotherapy. Although the present study focused on breast and prostate cancer PDX models, the CoC platform may be applied to other solid tumors, as well. With its ease of use and flexibility to operate with minimum footprint, our CoC platform is a versatile tool for ex vivo studies for functional genomics, drug screening, and personalized medicine.

W.F. Quiros-Solano reports a patent for PCT/NL2017/050492 issued; in addition, W.F. Quiros-Solano is cofounder and holds equity in BIOND Solutions B.V. A. Othman is a Field Application Scientist at BIOND Solutions B.V. C. Silvestri reports a patent for PCT/NL2017/050492 issued; in addition, C. Silvestri is cofounder and Chief Executive Officer, and holds equity in BIOND Solutions B.V. N. Gaio reports grants from ECSEL Joint Undertaking during the conduct of the study; in addition, N. Gaio has a patent for PCT/NL2017/050492 issued; and is a cofounder and Chief Technology Officer, and holds equity in BIOND Solutions B.V. R. Kanaar reports grants from Dutch Research Council (NWO) and grants from Gravitation program CancerGenomiCs.nl from the NWO and is part of the Oncode Institute, which is partly financed by the Dutch Cancer Society during the conduct of the study. D.C. van Gent reports grants from Dutch Cancer Foundation (KWF) and grants from Dutch Research Council (NWO) during the conduct of the study. No disclosures were reported by the other authors.

S. Chakrabarty: Data curation, software, formal analysis, validation, investigation, methodology, writing–original draft, writing–review and editing. W.F. Quiros-Solano: Resources, data curation, formal analysis, investigation, methodology, writing–original draft. M.M.P. Kuijten: Data curation, formal analysis, methodology, writing–original draft. B. Haspels: Data curation, formal analysis, investigation. S. Mallya: Data curation, software, formal analysis. C.S.Y. Lo: Software, formal analysis, visualization. A. Othman: Writing–review and editing. C. Silvestri: Writing–review and editing. A. van de Stolpe: Data curation, formal analysis, methodology. N. Gaio: Data curation, formal analysis, investigation, methodology, writing–review and editing. H. Odijk: Data curation, formal analysis. M. van de Ven: Data curation. C.M.A. de Ridder: Data curation. W.M. van Weerden: Resources, formal analysis, investigation, methodology, writing–review and editing. J. Jonkers: Writing–review and editing. R. Dekker: Funding acquisition, writing–review and editing. N. Taneja: Software, formal analysis, writing–review and editing. R. Kanaar: Conceptualization, funding acquisition, project administration, writing–review and editing. D.C. van Gent: Conceptualization, formal analysis, supervision, funding acquisition, investigation, writing–original draft, writing–review and editing.

The authors thank the people from the Preclinical Intervention Unit of the Mouse Clinic for Cancer and Ageing (MCCA) at the NKI for performing the PDX tumor outgrowth.

This work was supported by the KWF grant KWF11011, by the Gravitation program CancerGenomiCs.nl from the Dutch Research Council (NWO), the Oncode Institute, which is partly financed by the Dutch Cancer Society and the NWO (Building Blocks of Life grant 737.016.011). Part of this work is supported by Moore4Medical project funded by the ECSEL Joint Undertaking under grant number H2020-ECSEL-2019-IA-876190 (https://moore4medical.eu/).

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.

Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).

1.
Stebbing
J
,
Paz
K
,
Schwartz
GK
,
Wexler
LH
,
Maki
R
,
Pollock
RE
, et al
.
Patient-derived xenografts for individualized care in advanced sarcoma
.
Cancer
2014
;
120
:
2006
15
.
2.
Ter Brugge
P
,
Kristel
P
,
van der Burg
E
,
Boon
U
,
de Maaker
M
,
Lips
E
, et al
.
Mechanisms of therapy resistance in patient-derived xenograft models of BRCA1-deficient breast cancer
.
J Natl Cancer Inst
2016
;
108
:
djw148
.
3.
Vlachogiannis
G
,
Hedayat
S
,
Vatsiou
A
,
Jamin
Y
,
Fernandez-Mateos
J
,
Khan
K
, et al
.
Patient-derived organoids model treatment response of metastatic gastrointestinal cancers
.
Science
2018
;
359
:
920
6
.
4.
Malaney
P
,
Nicosia
SV
,
Dave
V
.
One mouse, one patient paradigm: New avatars of personalized cancer therapy
.
Cancer Lett
2014
;
344
:
1
12
.
5.
Zhang
Y
,
Wang
L
,
Gao
P
,
Sun
Z
,
Li
N
,
Lu
Y
, et al
.
ISL1 promotes cancer progression and inhibits cisplatin sensitivity in triple-negative breast cancer cells
.
Int J Mol Med
2018
;
42
:
2343
52
.
6.
Rottenberg
S
,
Jaspers
JE
,
Kersbergen
A
,
van der Burg
E
,
Nygren
AOH
,
Zander
SAL
, et al
.
High sensitivity of BRCA1-deficient mammary tumors to the PARP inhibitor AZD2281 alone and in combination with platinum drugs
.
Proc Natl Acad Sci USA
2008
;
105
:
17079
84
.
7.
Seo
YD
,
Jiang
X
,
Sullivan
KM
,
Jalikis
FG
,
Smythe
KS
,
Abbasi
A
, et al
.
Mobilization of CD8+ T cells via CXCR4 blockade facilitates PD-1 checkpoint therapy in human pancreatic cancer
.
Clin Cancer Res
2019
;
25
:
3934
45
.
8.
Naipal
KA
,
Verkaik
NS
,
Sanchez
H
,
van Deurzen
CH
,
den Bakker
MA
,
Hoeijmakers
JH
, et al
.
Tumor slice culture system to assess drug response of primary breast cancer
.
BMC Cancer
2016
;
16
:
78
.
9.
Martin
SZ
,
Wagner
DC
,
Hörner
N
,
Horst
D
,
Lang
H
,
Tagscherer
KE
, et al
.
Ex vivo tissue slice culture system to measure drug-response rates of hepatic metastatic colorectal cancer
.
BMC Cancer
2019
;
19
:
1030
.
10.
Meijer
TG
,
Verkaik
NS
,
Sieuwerts
AM
,
van Riet
J
,
Naipal
KAT
,
van Deurzen
CHM
, et al
.
Functional ex vivo assay reveals homologous recombination deficiency in breast cancer beyond BRCA gene defects
.
Clin Cancer Res
2018
;
24
:
6277
87
.
11.
Zhang
W
,
van Weerden
WM
,
de Ridder
CMA
,
Erkens-Schulze
S
,
Schonfeld
E
,
Meijer
TG
, et al
.
Ex vivo treatment of prostate tumor tissue recapitulates in vivo therapy response
.
Prostate
2019
;
79
:
390
402
.
12.
Jeon
JS
,
Bersini
S
,
Gilardi
M
,
Dubini
G
,
Charest
JL
,
Moretti
M
, et al
.
Human 3D vascularized organotypic microfluidic assays to study breast cancer cell extravasation
.
Proc Natl Acad Sci U S A
2015
;
112
:
214
9
.
13.
Shirure
VS
,
Bi
Y
,
Curtis
MB
,
Lezia
A
,
Goedegebuure
MM
,
Goedegebuure
SP
, et al
.
Tumor-on-a-chip platform to investigate progression and drug sensitivity in cell lines and patient-derived organoids
.
Lab Chip
2018
;
18
:
3687
702
.
14.
Ayuso
JM
,
Virumbrates-Munoz
M
,
McMinn
PH
,
Rehman
S
,
Gomez
I
,
Karim
MR
, et al
.
Tumor-on-a-chip: a microfluidic model to study cell response to environmental gradients
.
Lab Chip
2019
;
19
:
3461
71
.
15.
Holliday
DL
,
Brouilette
KT
,
Markert
A
,
Gordon
LA
,
Jones
JL
.
Novel multicellular organotypic models of normal and malignant breast: tools for dissecting the role of the microenvironment in breast cancer progression
.
Breast Cancer Res
2009
;
11
:
R3
.
16.
Hattersley
SM
,
Sylvester
DC
,
Dyer
CE
,
Stafford
ND
,
Haswell
SJ
,
Greenman
J
.
A microfluidic system for testing the responses of head and neck squamous cell carcinoma tissue biopsies to treatment with chemotherapy drugs
.
Ann Biomed Eng
2012
;
40
:
1277
88
.
17.
Astolfi
M
,
Peant
B
,
Lateef
MA
,
Rousset
N
,
Kendall-Dupont
J
,
Carmona
E
, et al
.
Micro-dissected tumor tissues on chip: an ex vivo method for drug testing and personalized therapy
.
Lab Chip
2016
;
16
:
312
25
.
18.
Riley
A
,
Green
V
,
Cheah
R
,
McKenzie
G
,
Karsai
L
,
England
J
, et al
.
A novel microfluidic device capable of maintaining functional thyroid carcinoma specimens ex vivo provides a new drug screening platform
.
BMC Cancer
2019
;
19
:
259
.
19.
Cao
X
,
Ashfaq
R
,
Cheng
F
,
Maharjan
S
,
Li
J
,
Ying
GL
, et al
.
A tumor-on-a-chip system with bioprinted blood and lymphatic vessel pair
.
Adv Funct Mater
2019
;
29
:
1807173
.
20.
Kennedy
R
,
Kuvshinov
D
,
Sdrolia
A
,
Kuvshinova
E
,
Hilton
K
,
Crank
S
, et al
.
A patient tumour-on-a-chip system for personalised investigation of radiotherapy based treatment regimens
.
Sci Rep
2019
;
9
:
6327
.
21.
Firpo
G
,
Angeli
E
,
Repetto
L
,
Valbusa
U
.
Permeability thickness dependence of polydimethylsiloxane (PDMS) membranes
.
J Membr Sci
2015
;
481
:
1
8
.
22.
van Weerden
WM
,
de Ridder
CM
,
Verdaasdonk
CL
,
Romijn
JC
,
van der Kwast
TH
,
Schröder
FH
, et al
.
Development of seven new human prostate tumor xenograft models and their histopathological characterization
.
Am J Pathol
1996
;
149
:
1055
62
.
23.
Navone
NM
,
van Weerden
WM
,
Vessella
RL
,
Williams
ED
,
Wang
Y
,
Isaacs
JT
, et al
.
Movember GAP1 PDX project: An international collection of serially transplantable prostate cancer patient-derived xenograft (PDX) models
.
Prostate
2018
;
78
:
1262
82
.
24.
Lo
CSY
,
van Toorn
M
,
Gaggioli
V
,
Dias
MP
,
Zhu
Y
,
Manolika
EM
, et al
.
SMARCAD1-mediated active replication fork stability maintains genome integrity
.
Sci Adv
2021
;
7
:
eabe7804
.
25.
Zhang
W
,
Liao
C-Y
,
Chtatou
H
,
Incrocci
L
,
van Gent
DC
,
van Weerden
WM
, et al
.
Apalutamide sensitizes prostate cancer to ionizing radiation via inhibition of non-homologous end-joining DNA repair
.
Cancers (Basel)
2019
;
11
:
1593
.
26.
Dodt
M
,
Roehr
JT
,
Ahmed
R
,
Dieterich
C
.
FLEXBAR-flexible barcode and adapter processing for next-generation sequencing platforms
.
Biology (Basel)
2012
;
1
:
895
905
.
27.
Kim
D
,
Paggi
JM
,
Park
C
,
Bennett
C
,
Salzberg
SL
.
Graph-based genome alignment and genotyping with HISAT2 and HISAT-genotype
.
Nat Biotechnol
2019
;
37
:
907
15
.
28.
Li
H
,
Handsaker
B
,
Wysoker
A
,
Fennell
T
,
Ruan
J
,
Homer
N
, et al
.
The sequence alignment/map format and SAMtools
.
Bioinformatics
2009
;
25
:
2078
9
.
29.
Kluin
RJC
,
Kemper
K
,
Kuilman
T
,
de Ruiter
JR
,
Iyer
V
,
Forment
JV
, et al
.
XenofilteR: computational deconvolution of mouse and human reads in tumor xenograft sequence data
.
BMC Bioinformatics
2018
;
19
:
366
.
30.
Pertea
M
,
Pertea
GM
,
Antonescu
CM
,
Chang
TC
,
Mendell
JT
,
Salzberg
SL
.
StringTie enables improved reconstruction of a transcriptome from RNA-seq reads
.
Nat Biotechnol
2015
;
33
:
290
5
.
31.
Liao
Y
,
Smyth
GK
,
Shi
W
.
featureCounts: an efficient general purpose program for assigning sequence reads to genomic features
.
Bioinformatics
2014
;
30
:
923
30
.
32.
Robinson
MD
,
McCarthy
DJ
,
Smyth
GK
.
edgeR: a Bioconductor package for differential expression analysis of digital gene expression data
.
Bioinformatics
2010
;
26
:
139
40
.
33.
Canté-Barrett
K
,
Holtzer
L
,
van Ooijen
H
,
Hagelaar
R
,
Cordo'
V
,
Verhaegh
W
, et al
.
A molecular test for quantifying functional Notch signaling pathway activity in human cancer
.
Cancers
2020
;
12
:
3142
.
34.
Inda
MA
,
Blok
EJ
,
Kuppen
PJK
,
Charehbili
A
,
den Biezen-Timmermans
EC
,
van Brussel
A
, et al
.
Estrogen receptor pathway activity score to predict clinical response or resistance to neo-adjuvant endocrine therapy in primary breast cancer
.
Mol Cancer Ther
2020
;
19
:
680
9
.
35.
Verhaegh
W
,
van Ooijen
H
,
Inda
MA
,
Hatzis
P
,
Versteeg
R
,
Smid
M
, et al
.
Selection of personalized patient therapy through the use of knowledge-based computational models that identify tumor-driving signal transduction pathways
.
Cancer Res
2014
;
74
:
2936
45
.
36.
van de Stolpe
A
.
Quantitative measurement of functional activity of the PI3K signaling pathway in cancer
.
Cancers (Basel)
2019
;
11
:
293
.
37.
Shim
S
,
Belanger
MC
,
Harris
AR
,
Munson
JM
,
Pompano
RR
.
Two-way communication between ex vivo tissues on a microfluidic chip: application to tumor-lymph node interaction
.
Lab Chip
2019
;
19
:
1013
26
.
38.
Fujiyabu
T
,
Li
X
,
Shibayama
M
,
Chung
U
,
Sakai
T
.
Permeation of water through hydrogels with controlled network structure
.
Macromolecules
2017
;
50
:
9411
6
.
39.
Munson
JM
,
Shieh
AC
.
Interstitial fluid flow in cancer: implications for disease progression and treatment
.
Cancer Manag Res
2014
;
6
:
317
28
.
40.
Pisano
M
,
Triacca
V
,
Barbee
KA
,
Swartz
MA
.
An in vitro model of the tumor-lymphatic microenvironment with simultaneous transendothelial and luminal flows reveals mechanisms of flow enhanced invasion
.
Integr Biol (Camb)
2015
;
7
:
525
33
.
41.
van Weerden
WM
,
van Steenbrugge
GJ
,
van Kreuningen
A
,
Moerings
EP
,
de Jong
FH
,
Schröder
FH
.
Assessment of the critical level of androgen for growth response of transplantable human prostatic carcinoma (PC-82) in nude mice
.
J Urol
1991
;
145
:
631
4
.
42.
Diaz
A
,
Jiménez
SA
.
Interferon-gamma regulates collagen and fibronectin gene expression by transcriptional and post-transcriptional mechanisms
.
Int J Biochem Cell Biol
1997
;
29
:
251
60
.
43.
Chang
TC
,
Mikheev
AM
,
Huynh
W
,
Monnat
RJ
,
Rostomily
RC
.,
Folch A. Parallel microfluidic chemosensitivity testing on individual slice cultures
.
Lab Chip
2014
;
14
:
4540
51
.
44.
Horowitz
LF
,
Rodriguez
AD
,
Dereli-Korkut
Z
,
Lin
R
,
Castro
K
,
Mikheev
AM
, et al
.
Multiplexed drug testing of tumor slices using a microfluidic platform
.
npj Precision Oncology
2020
;
4
:
1
13
.
45.
Rodriguez
AD
,
Horowitz
LF
,
Castro
K
,
Kenerson
H
,
Bhattacharjee
N
,
Gandhe
G
, et al
.
A microfluidic platform for functional testing of cancer drugs on intact tumor slices
.
Lab Chip
2020
;
20
:
1658
75
.
46.
Shirure
VS
,
George
SC
.
Design considerations to minimize the impact of drug absorption in polymer-based organ-on-a-chip platforms
.
Lab Chip
2017
;
17
:
681
90
.
47.
Gaio
N
,
Van Meer
B
,
Quirós Solano
W
,
Bergers
L
,
Van de Stolpe
A
,
Mummery
C
, et al
.
Cytostretch, an organ-on-chip platform
.
Micromachines
2016
;
7
:
120
.
48.
da Ponte
RM
,
Gaio
N
,
van Zeijl
H
,
Vollebregt
S
,
Dijkstra
P
,
Dekker
R
, et al
.
Monolithic integration of a smart temperature sensor on a modular silicon-based organ-on-a-chip device
.
Sens Actuators, A
2021
;
317
:
112439
.
49.
Lukas
C
,
Savic
V
,
Bekker-Jensen
S
,
Doil
C
,
Neumann
B
,
Sølvhøj Pedersen
R
, et al
.
53BP1 nuclear bodies form around DNA lesions generated by mitotic transmission of chromosomes under replication stress
.
Nat Cell Biol
2011
;
13
:
243
53
.
50.
Muruganandan
S
,
Fan
X
,
Dhal
S
,
Nayak
NR
.
Development of A 3D tissue slice culture model for the study of human endometrial repair and regeneration
.
Biomolecules
2020
;
10
:
136
.
51.
Schopow
N
,
Kallendrusch
S
,
Gong
S
,
Rapp
F
,
Körfer
J
,
Gericke
M
, et al
.
Examination of ex-vivo viability of human adipose tissue slice culture
.
PLoS One
2020
;
15
:
e0233152
.
52.
Cao
X
,
Coyle
JP
,
Xiong
R
,
Wang
Y
,
Heflich
RH
,
Ren
B
, et al
.
Invited review: human air-liquid-interface organotypic airway tissue models derived from primary tracheobronchial epithelial cells—overview and perspectives
.
In Vitro Cell Dev Biol Anim
2021
;
57
:
104
32
.
This open access article is distributed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0) license.

Supplementary data