Controlled pharmacokinetic anti ‐ cancer drug concentration profiles lead to growth inhibition of colorectal cancer cells in a microfluidic device

We present a microfluidic device to expose cancer cells to a dynamic, in vivo-like concentration profile of a drug, and quantify efficacy on-chip. About 30% of cancer patients receive drug therapy. In conventional cell culture experiments drug efficacy is tested under static concentrations, e.g. 1 μM for 48 hours, whereas in vivo, drug concentration follows a pharmacokinetic profile with an initial peak and a decline over time. With the rise of microfluidic cell culture models, including organs-on-chips, there are opportunities to more realistically mimic in vivo-like concentrations. Our microfluidic device contains a cell culture chamber and a drug-dosing channel separated by a transparent membrane, to allow for shear stress-free drug exposure and label-free growth quantification. Dynamic drug concentration profiles in the cell culture chamber were controlled by continuously flowing controlled concentrations of drug in the dosing channel. The control over drug concentrations in the cell culture chambers was validated with fluorescence experiments and numerical simulations. Exposure of HCT116 colorectal cancer cells to static concentrations of the clinically used drug oxaliplatin resulted in a sensible dose-effect curve. Dynamic, in vivo-like drug exposure also led to statistically significant lower growth compared to untreated control. Continuous exposure to the average concentration of the in vivo-like exposure seems more effective than exposure to the peak concentration (Cmax) only. We expect that our microfluidic system will improve efficacy prediction of in vitro models, including organs-on-chips, and may lead to future clinical optimization of drug administration schedules.

. Bottom channel concentration would be at 48% of the concentration at the start in the top channel, while top channel concentration after 16 hours would still be 80% of the starting concentration, which is due to diffusion time and the difference in volume between the top and bottom channel (Supplementary figure 11). Concentrations in the bottom channel, based on the calibration plots in b-f, are 10 µM fluorescein and 300 µM oxaliplatin, which is a similar ratio as the loaded solutes in the top channel, which implies similar diffusion coefficients, . The 20x lower concentration in the bottom channel compared to the top channel is due to 10x dilution of the bottom channel volume for achieving the minimum volume needed for analysis, and ~2x dilution due to diffusion. (b,c,d) calibration graphs for oxaliplatin at 250 nm (Rsq=0.99), fluorescein at 490nm (Rsq=0.99), and fluorescein at 250nm (Rsq0.99). (e,f ) Absorbance profiles for different dosages of oxaliplatin and fluorescein for comparison of the absorbance profile of (a).

Supplementary Fig. 2 Oxaliplatin absorption on chip.
No absorption detectable at 31 µM Oxaliplatin flown through the top channel at 10 µl/min as the absorbance profile is ~equal to 31 µM Oxaliplatin of the calibration sequence that was not flown through a chip. Fig. 3 Timing difference in oxaliplatin concentration between administration at the Y splitter and outlet collection due to dead volume lag. (a) Inlet tubing and top channel volume combined is 120 µl, giving rise to an initial (*) dead volume lag of 12 minutes at a flow of 10 µl min -1 , which increases to 40 minutes (**) at 3 µl min -1 . To match samples with the administered concentration steps, sample timing is adjusted accordingly. (b) Concentration in outlet samples collected matches outlet sample measured well.

Supplementary description of simulations
For the simulation of the diffusion of oxaliplatin, a 2D, time dependent model was made in which the diffusion-equation was solved numerically. The bottom channel was simulated with the following geometry:

Supplementary Fig. 8 Bottom channel geometry used in simulations.
The equation solved was: • 0 1 With , the oxyplatin concentration, , the time and , the diffusion coefficient. A no-flux condition is set at the walls and at the cells; • 0 2 Although the cells consume part of the medicine, this is considered to be neglectable compared to the available oxaliplatin within the channel and therefore not taken into account in this simulation. And a flux condition in the middle to simulate the flux through the membrane: With , the concentration in the top channel, , the permeability constant of the membrane and the thickness of the membrane. The flow velocity in the top channel was considered to be sufficiently high to neglect the local depletion in the top channel just above the membrane. A triangular mesh was used with the following parameters; Maximum element size: 0.0255 mm, minimum element size: 0.001 mm, maximum element growth rate: 1.15, curvature factor 0.3, resolution of narrow regions: 1.
To determine the permeability factor, the average concentration in the square area between the membrane and the cells was taken and fitted to experimental fluorescence data, with as the fitting parameter. Where was set at zero and the concentration in the bottom channel over time was evaluated. This gave a permeability factor of = 0.2 ( Supplementary  Fig 3). A graphical depiction of the simulated concentration wash out with a membrane porosity of 0.2 is shown in Supplementary Fig 3. Membrane Cells Supplementary Fig. 9 Comsol simulated wash out of fluorescein out of bottom channel through a membrane of porosity 0.2. Starting concentration is set at 1. Diffusion coefficient for fluorescein is 4.2×10 -6 cm 2 s -1 .
For the simulation of the administering of variable Oxaliplatin doses, a variable was given, and the concentration was determined by using the mean values, at the bottom of the channel over the entire length of the membrane over where the cells are located as a function of time. Supplementary Fig. 10 Numerical simulation of diffusion out of bottom channel for different membrane porosities (k) plotted next to diffusion for different flow rates. 0.2 porosity has a good fit with the diffusion experiments.