The PDOs used in this experiment were derived from a patient that underwent resection for CRCLM at the University of California, San Francisco in 2016. The research protocol was approved by the institutional review board of UCSF (IRB#10-05031). The tumor specimen obtained during surgery was used to generate a mouse xenograft as previously described (Kondo et al., 2019). Animal experiments were carried out by members of the UCSF Preclinical Therapeutics Core Facility in accordance with the University of California San Francisco animal care and use committee (IACUC# AN179937-03A). Briefly, the tumor was minced with a scalpel and digested with Liberase (Sigma) and DNase (Qiagen) for 1 h at 37°C in a Gentle Macs Octo Dissociator (Miltenyi). The digested specimen was then filtered sequentially through 500 μm, 250 μm, and 100 μm filters to remove undigested tissue. Cell clusters were collected using a 40 μm filter and transferred to ultra-low attachment plates and cultured in organoid media (DMEM/F12 Glutamax (Gibco), 1× PenStrep/Glutamine (Invitrogen), 1× STEMPRO hESC SFM (Invitrogen), 0.1 mM beta-mercaptoethanol, 8 ng/mL bFGF (Invitrogen), 1.8% BSA (Invitrogen) for 5 days prior to cryopreservation using CryoStor CS10 (BioLife Solutions).

Organoids were thawed, resuspended in phosphate-buffered saline (PBS) (Gibco, 10010023) to wash away DMSO and pelleted by centrifuging at 300 g for 5 minutes. Pellets were then gently resuspended in 10 mL StemPro CRC media comprised of DMEM/F-12 GlutaMAX (Gibco, 10565-018), 1× Antibiotic-Antimycotic (Gibco, 15240062), 1× StemPro hESC Growth supplement (Gibco, A10006-01), 1.8% BSA (Gibco, A10008-01), 8 ng/mL bFGF (Gibco, PHG0261), and 0.1 mM 2-Mercaptoethanol (Gibco, 21985023). Organoids were cultured for 4 days in ultra-low adhesion 10 cm dishes (Corning, 3,262) prior to use in experiments. After 4 days, organoids were filtered through a 40 μm cell strainer (Fisher Scientific, 087711) to eliminate single cells and counted. Organoids were then cultured in three different conditions (Figure 1): organoids in media, in static gel, or in the multi-well bioreactor.

Once organoids were collected and counted, they were subsequently cultured in either media (StemPro CRC media) as described above or in fibrin-gelatin hydrogels for experimental studies with the chemotherapeutic drug 5-Flurouracil (5-FU). Solutions for the fibrin-gelatin hydrogel were all prepared in PBS without Ca²⁺ and Mg²⁺ (Gibco) and mixed, as described below, resulting in a final gel consisting of gelatin (75 mg/mL, type A, 300 Bloom, Sigma-Aldrich), fibrinogen (10 mg/mL; type I-S, Sigma-Aldrich), transglutaminase (TG; 2 mg/mL; type TI, Modern Pantry), CaCl₂ (2.5 mM, Sigma-Aldrich), and thrombin (1 U/mL; Sigma-Aldrich) (Kolesky et al., 2016; Hynes et al., 2020). 150 mg/mL gelatin stock solutions were prepared in PBS by heating at 70°C for 2 h with gentle stirring until completely dissolved. The gelatin solution was then brought to a pH of 7.5 with NaOH or HCl, then sterile filtered, aliquoted, and stored at 4°C until use. All other materials were prepared the day of experiments, sterile filtered and kept at 37°C until just before mixing.

Fibrinogen, TG, and CaCl₂ were mixed carefully via pipette and allowed to incubate for 5 min at 37°C. Gelatin was then added to the solution, mixed thoroughly, and allowed to incubate for 20 min at 37°C. Finally, the premixed gel solution was added to thrombin and mixed thoroughly before being added to the organoids to obtain a final organoid concentration of 375 organoids/mL. 1 mL of the organoid and fibrin-gelatin hydrogel solution was dispensed into each of the wells of a 12 well plate. After 10 min at room temperature, 1 mL CRC media was overlaid on the hydrogel in each well and the plate was incubated in a cell culture incubator at 37°C for 1 hour followed by 4°C for 20 min to simulate the conditions required for bioreactor fabrication as described below.

Organoids in media and organoids in static hydrogels were cultured for 1 day prior to addition of 5-FU. 5-FU was selected because it is a common chemotherapeutic agent used in the treatment of metastatic CRC (Biller and Schrag, 2021; Morris et al., 2023). 5-FU was prepared at a stock concentration of 50 mM in DMSO and stored at −20°C before adding to media to result in applied concentrations of 0, 1, 5, 25, 100, 150 μM for organoids in media and 0, 2, 10, 50, 200, and 300 μM for the static gel.

Our multi-well bioreactor platform was designed to utilize a standard 12 well tissue culture plate for ease of handling and fabrication (Figure 2A). A custom machined aluminum lid was designed to include fluidic inlet and outlet ports (Nordson stainless steel needles, 20 gauge) for each well and two filling ports per well to allow the user to dispense hydrogel into the wells. A PDMS gasket was fabricated using a custom mold to ensure each well remained fluidically isolated. All bioreactor components (i.e., aluminum base and lid, fluidics, and PDMS gasket) were sterilized by autoclaving prior to use.

Spray dried maltodextrin was molded to form sacrificial channels (Supplementary Figure S1) and sterilized overnight in an oven at 100°C. To assemble the bioreactor, the PDMS gasket was placed onto the underside of the aluminum lid. Sacrificial channels were then attached to the ends of the stainless-steel pins using a 35% (w/w) solution of Pluronic F127 (Sigma) in deionized water. Briefly, the Pluronic solution was loaded into a 1 mL syringe and allowed to solidify at room temperature for 30 min prior to use. A small bead of Pluronic was extruded onto the ends of each pin and maltodextrin channels were gently placed on top using tweezers (Supplementary Figure S2). The Pluronic beads were then allowed to dry for 30 min before flipping the lid over and inserting into the 12 well plate. The bioreactor base and lid were then clamped together using hex bolts and wing nuts (McMaster Carr) until the PDMS gasket was sufficiently and evenly compressed.

Organoids were mixed with the fibrin-gelatin hydrogel at a concentration of 375 organoids/mL as previously described and 5 mL were dispensed into each well to encapsulate the sacrificial channels. The bioreactor was then placed in a humidified incubator with 5% CO₂ for 1 h at 37°C followed by 20 min at 4°C to allow the Pluronic beads to liquefy before flushing the sacrificial maltodextrin from the channels. Upon completion of the incubation steps, the sacrificial maltodextrin and Pluronic were flushed from each of the wells using 3 mL warm CRC media and the wells were connected to flow.

A peristaltic pump (MP2, Elemental Scientific) was used to deliver flow to each of the wells from media reservoirs (14 mL tube, Corning) filled with 5 mL of CRC media in a series of fluidic connections (Figure 2B). Media was drawn from the reservoir using 1/16″ PFA fluidic tubing (0.01” ID, IDEX) that was connected to peristaltic tubing (0.13 mm ID, Elemental Scientific) attached to the peristaltic pump, followed by another section of PFA tubing that directly connected to the inlet of the well with a union and luer adapter. Media was returned to the media reservoir from the outlet of the well using fittings and a short section of PFA tubing. The pump was set to 30 rpm which corresponded to a flow rate of 22 μL/min and was verified using a flow sensor (MFS3, Elveflow). 5-FU was added to the media reservoirs after 1 day of culture at nominal concentrations of 0, 4, 20, 100, 400, and 600 μM in 5 mL CRC media. The media and drug solutions were recirculated in each of the wells for the duration of the 7-day experiment.

Simulations of chemotherapeutic transport in the bioreactors were conducted using COMSOL Multiphysics 5.5a. The bioreactor was modeled in COMSOL’s native CAD software; Free and Porous Media Flow (fp) and Transport of Diluted Species in Porous Media (tds) modules were used to solve the Navier Stokes and chemical diffusion equations respectively, while the Global ODEs and DAEs (ge) module was employed to model the recirculating inlet and outlet of the bioreactor. Standard no-slip boundary conditions were assumed and drug consumption by tumor organoids was considered negligible. Diffusivity of 5-FU was determined through previously described methods (Hsu et al., 2018) (Supplementary Section 1.2). Hydrogel porosity of 0.9 was determined via the Archimedes Method (Ho and Hutmacher, 2006) and permeability values for the hydrogel regions were taken from our previous work (Alonzo et al., 2015).

A parametric sweep of initial inlet concentrations ranging from 0–600 μM for the 7-day experimental time course was performed. Average drug concentrations for the static gel condition and regions near (region 1) and far (region 2) from the bioreactor channel were evaluated for different time points in COMSOL post-processing to inform the applied 5-FU concentration for each experimental platform with the goal of achieving similar cumulative dose values.

An area under the curve (AUC) metric was calculated by integrating the average concentration in each platform (media, static gel, bioreactor region 1 and 2) simulation over the treatment duration (0–7 days) using a trapezoidal Reimann sum. This metric provided the total 5-FU dose that was delivered to the organoids in each culture condition during the weeklong experiment. For the organoid in media condition, it was assumed that the concentration remained constant (i.e., 5-FU was neither consumed nor added) throughout the duration of the experiment.

To determine whether 5-FU concentration influenced organoid size or morphology, organoids in each of the culture conditions (media, n = 3; static gel, n = 30; bioreactor, n = 10) were imaged in brightfield on day 7 of the experiment using an Olympus IX83 microscope fitted with a ×10 objective. Organoid area and perimeter were measured manually using Fiji v1.53s (Schindelin et al., 2012). Circularity was calculated using the following equation c i r c u l a r i t y = 4 π ∗ A r e a p e r i m e t e r 2 .

After 7 days of exposure to 5-FU, cells from each culture modality were collected and assessed for viability using the CellTiter-Glo 3D assay (Promega, G9681) according to the manufacturer’s instructions.

For organoids cultured in media, organoids were spun down and resuspended in StemPro CRC media, 100 μL per well for 96-well. CellTiter-Glo^® Reagent was added to each well and contents were mixed for 2 min on an orbital shaker to induce cell lysis. The plate was incubated at room temperature for 10 min to stabilize the luminescent signal and luminescence was recorded using a plate-reading luminometer (SpectraMax M5e, Molecular Devices).

For organoids cultured in the static gel, a sterile scalpel and sterile forceps were used to mince and transfer the cell laden gels into a sterile 5 mL syringe barrel. The gels were then fully homogenized by pushing the gel through the syringe (no needle attached) into a 15 mL conical tube. A 0.2% collagenase I (Gibco, 17100-017) solution was added to the conical tube and allowed to agitate on a shaker for 1 h at 37°C and 200 rpm to fully dissolve the gel. The solution was then spun down and the pellet collected for use in the CellTiter-Glo assay as previously described.

For the bioreactor samples, the bioreactor gel was cut into equal volumes near and far from the channel (region 1 and 2 respectively). The gel was first cut along the length of the channel to make two-half circles, then each piece was cut parallel to the channel at a distance of 4.5 mm from the channel. Each of the pieces was further cut into two along the symmetry axis to create a total of 4, equally weighted replicates for each region. Once cut, the gels were homogenized and dissociated using the protocol described above for the static gel conditions.

All the CellTiter-Glo results were plotted against the average concentration on day 7 for each condition and IC₅₀s calculated using GraphPad 8.0 (Prism) using nonlinear regression. All IC₅₀ curves were normalized using the lowest and highest concentration of drug.

Samples (100 μL) from the supernatant of organoids cultured in media and static gels or from the media reservoirs of the bioreactor were taken at days 0, 2, 4, and 7. The LactateGlo assay (Promega, J5021) was performed on these samples according to the manufacturer’s instruction. Briefly, 50 μL of collected media sample or lactate standard was transferred into a 96-well plate and 50 μL of Lactate Detection Reagent was added. The plate was shaken for 30–60 s and subsequently incubated for 60 min at room temperature. Luminescence was recorded using a plate-reading luminometer (SpectraMax M5e, Molecular Devices) and converted to lactate concentrations using the standard curve. Lactate concentrations were normalized to the lowest and highest 5-FU concentrations and EC₅₀ curves were calculated using GraphPad 8.0 (Prism) using nonlinear regression.

Multiple linear regression was used to test if culture platform and 5-FU dose correlated with organoid area or circularity. Multiple linear regression was also used to test if viability analysis with the CellTiter-Glo assay correlated with lactate analysis using the LactateGlo assay (Supplementary Figure S6). For all Lactate-Glo results and organoid area/circularity for bioreactor regions 1 and 2 (Supplementary Figure S4), a two-way ANOVA was performed with Tukey post hoc comparisons. IC₅₀ curves were compared using their confidence intervals. If confidence intervals did not overlap, the results were considered statistically significant. In the case that they did overlap, an extra sum-of-squares F test was used to determine if IC₅₀ values differed between groups. All statistical tests were performed with a p-value of 0.05.

Our multi-well bioreactor platform was designed around a standard 12-well cell culture plate to minimize the number of custom parts and enable facile assembly. This layout provides 12 independently fed wells that can be used to study multiple drug concentrations or combinations in one experiment. To further advance our goal of facile assembly, we developed a method to precisely mold sacrificial maltodextrin, eliminating the need for microfabrication steps often required to produce fluidic channels. Channels formed using molded sacrificial maltodextrin had highly reproducible diameters (2,272 ± 167 μm) ensuring consistent flow with each experiment (Supplementary Figure S2B). A custom CNC machined aluminum lid was used to easily connect each well of the bioreactor to a peristaltic pump that recirculated flow to each of the wells, enabling long-term studies and sampling from connected media reservoirs (Figure 2).

Drug concentration in the organoid in media condition was assumed to be constant and the effective average concentration equal to the applied concentration for the duration of the experiment (0, 1, 5, 25, 100, 150 μM). To predict the distribution of 5-FU within our platforms and select the applied drug concentrations, the dynamics of transport of 5-FU were simulated in the static gel and bioreactor platforms. In the static gel, 5-FU diffused quickly from media into the gel, reaching steady state at 24 h of culture (Figure 3A). Since the volumes of gel and media were equal in this condition, the resulting steady state concentration is exactly 50% of the source concentration homogenously spread throughout the gel and media. Informed by these results, we applied a concentration 2× higher than the organoid in media condition (0, 2, 10, 50, 200, 300 μM) to achieve a similar cumulative dose for the duration of the experiment.

In contrast to the static gel condition, even after 7 days of culture, drug concentration in the bioreactor had not reached steady state and a gradient in 5-FU concentration remained (Figure 3B) where 5-FU concentration varied with distance from the channel. As expected, region 1, closest to the channel, contained a higher concentration of 5-FU during the entire 7-day period than region 2, further from the channel (Figure 3C). The disparity in drug concentration peaked early then narrowed over time. After 7 days, region 1 reached an average concentration of 52% of the applied 5-FU concentration whereas the average concentration in region 2 was only 38% of the applied concentration. At day 7, the total average concentration in the bioreactor was 45% of the applied concentration. Since the simulation showed diffusion occurred much slower in the bioreactor, we applied a concentration 4× higher than the media condition (0, 4, 20, 100, 400, 600 μM) to achieve a similar cumulative dose between platforms over the duration of the experiment.

To investigate how tumor organoids respond to 5-FU in varying environments, we cultured patient-derived colorectal liver metastasis organoids in three different platforms: media, static gel, and a multi-well bioreactor platform (Figure 4A). Organoids cultured in media formed large irregularly shaped clusters over the duration of the experiment whereas organoids cultured in the static gel or multi-well bioreactor were significantly smaller and remained single and more circular (Figure 4B).

For the organoid in media condition, organoid area tended to decrease with increasing dose (with a statically significant nonlinear regression coefficient of −1.1e-4 mm²/drug concentration, p < 1e-7, Figure 4C) whereas in the static gel and bioreactor conditions, area tended to increase slightly with 5-FU dose (regression coefficients of 1.1e4 and 1.1e4 respectively, p < 1e-6, Figure 4C).

No statistically significant differences were found when comparing organoid area or circularity for region 1 and region 2 of the bioreactor (Supplementary Figure S4).

To evaluate whether the gradient in 5-FU concentration in the bioreactor predicted by the transport simulation would alter organoid response to 5-FU, the bioreactor gel was divided into equally weighted sections on day 7 that were located close to the channel (region 1) and far from the channel (region 2). The CellTiter-Glo assay was performed and IC₅₀ curves were plotted for both regions (Figure 5A). Our modeling showed that region 1 experienced a higher average 5-FU concentration than region 2 and concomitantly, we observed a statistically significant lower IC₅₀ value in region 1 (Figure 5D). The IC₅₀ for the organoid in media condition (3.387 μM) was an order of magnitude lower and statistically significantly different than the static gel condition and both regions of the multi-well bioreactor (Figures 5B, C). The IC₅₀ for the static gel condition (33.96 μM) was comparable to region 1 of the bioreactor (22.6 μM, p = 0.1413) but was significantly lower than region 2 (69.8 μM, p = 0.0184).

The temporal response of organoids to 5-FU was evaluated using the non-destructive Lactate-Glo assay. As expected, at time 0, there was no statistically significant difference in lactate secretion for the range of 5-FU concentrations (Figure 6A) in the multi-well bioreactor. However, changes in metabolic output could be seen starting at day 2, with significantly less lactate secretion in either of the two highest 5-FU dosed conditions compared to the two lowest 5-FU conditions (p < 0.05), with increasing separation over time (p < 0.001, days 4&7). Lactate secretion continued to increase for all 5-FU concentrations in the bioreactor between days 4 and 7 of the experiment, however differences were not statistically significantly (400 μM, p = 0.059; 600 μM, p = 0.18). Overall lactate levels were lower in the bioreactor than in the media and static gel conditions. Using the Lactate-Glo data taken at day 7 and a standard curve, it was possible to derive an EC50 curve which resulted in an EC50 of 54.33 μM (Figure 6D; Supplementary Figure S5).

Lactate levels in the organoid in media and static gel conditions (Figures 6B, C) decreased significantly with increasing dose at day 7 but remained similar at day 2 with inhibition seen only for the two highest concentrations on day 4 (p < 0.05). Interestingly, on day 7, lactate levels for the organoids in media at a 5 μM applied concentration separated further from the lower concentrations (0 μM and 1 μM) than the static gel at a 10 μM applied concentration despite experiencing similar cumulative doses for most of the experiment indicating improved cell inhibition (Figures 6B, C). The EC50 derived from Lactate-Glo measurements for organoids in media (2.87 μM), static gel (13.65 μM), and the multi-well bioreactor (54.33 μM) were all statistically significant from one another (Figure 6D).

Regression analysis showed that there is a significant correlation between CellTiter-Glo (C) and LactateGlo (L) data for the media condition (9 units L/C, p < 1e-5), but not static gel conditions (4 L/C, p = 0.07) or the bioreactor condition (−6 L/C, p = 0.07).

To compare drug transport more directly between the three culture conditions, the area under the curve (AUC) was calculated by integrating the average concentration from the transport simulations over the duration of the experiment. Figure 7A shows that each of the culture conditions have comparable cumulative doses for each set of applied concentrations. For the lowest applied concentrations (media—1 μM, static gel—2 μM, bioreactor—4 μM), the AUCs were 168 μM*hr, 165 μM*hr, 267 μM*hr, and 135 μM*hr for media, static gel, bioreactor regions 1 and 2 respectively. Bioreactor region 1 had a higher AUC than region 2 since it was closer to the perfused channel. At the same applied concentration (100 μM), the organoid in media condition had a higher AUC than bioreactor region 1 and 2 showing that the same applied concentration results in different cumulative doses between platforms, highlighting the need to account for transport differences between platforms when comparing drug response.

Figure 7B showed that the static gel and bioreactor region 2 have overlapping normalized viability curves when plotted against AUC. Bioreactor region 1 had a lower viability curve due to its higher cumulative dose as evidenced by a higher AUC. Despite having similar AUC values, the media condition had a lower normalized viability curve than the static gel and bioreactor conditions.