Cell culture using centrifugal microfluidic platform with demonstration on Pichia pastoris

Abstract

This paper discusses the vortical flow, mixing and cell culture of Pichia pastoris using a centrifugal microfluidic (CM) chamber. The resultant “spiral toroidal vortex” in the chamber is made up of a primary vortex induced from inertial acceleration/deceleration of the chamber superposed by a secondary toroidal vortex due to Coriolis acceleration acting on the primary vortex. A validated numerical fluid-flow model with minimized numerical diffusion effect has been developed to investigate the flow and consequently mixing of two-color liquids through cyclic constant acceleration-and-deceleration in the same rotation direction until homogeneous mixing of the two liquids in the CM chamber has been established. The specific mixing time is found to improve with increase in acceleration/deceleration and angular span of the chamber. An experimental CM platform with three cell-culture chambers of different angular spans has been built and Pichia pastoris cell culture has been successfully demonstrated. Cell growth can be monitored over time on the extracted samples by measuring the optical density at 600-nm wave-length. Comparing with conventional cell culture, Pichia pastoris cultured on CM platform exhibits a very short lag (cell preparation/budding) phase prior to the log phase (cell growth). While it takes 8 to 12 h for the conventional shake flask in the lag phase, it takes only 2 h for the CM platform irrespective of initial cell concentration (8.1 × 10⁴ to 8.1 × 10⁵/ml), acceleration/deceleration (10 to 32/s²) and angular span of the culture chamber (π/12 to π/4), representing significant time reduction. This is largely attributed to better growth conditions due to enhanced mixing and appropriate shear-stress stimulation from the efficient spiral toroidal vortex.

Appendix-Sensitivity study of grid size, time step and numerical diffusion

To validate our numerical model (i.e., chamber with r _i  = 35 mm, r _o  = 53 mm, θ = 15° and H = 0.5 mm), the grid size and time step sensitivity studies were conducted by checking the velocity profiles at t = 3 s when Δt = 5 s and dΩ/dt = ±20.9 rad/s², as demonstrated in Fig. 13. It shows that the solution can reach convergence when Δt = 0.05 s and Δx = 0.1 mm. The dependence of numerical diffusion on the order of computational discretization scheme (i.e., backward Euler scheme) and grid size at t = 1 and 2 s respectively when Δt = 5 s and dΩ/dt = ±20.9 rad/s² was investigated in Fig. 14, which demonstrates qualitatively that the numerical diffusion can be minimized using the 2nd order scheme and grid size of 0.1 mm. The same conclusion can be drawn from the quantitative results in Fig. 15, wherein the species concentration distribution profile versus radial position r (z = 0.25 mm, ϕ = 3.75°, see the white dash curve in Fig. 14,) is demonstrated and the curve with the steepest slope represents the condition with minimized numerical diffusion. In the meanwhile, the results are compared with the benchmarking condition when the physical diffusion is disabled (Diffusion coefficient, D = 0 mm²/s). It shows some variance (i.e., reduced diffusion layer thickness) from the optimal case when the 2nd order scheme, grid size of 0.1 mm and D = 0.015 mm²/s were used, indicating the mixing effect in this scenario is completely attributed to the numerical diffusion.

Fig. 13

Time step and grid size validation for circumferential velocity over radial position r (z = 0.25 mm, ϕ = 0) at t = 3 s when Δt = 5 s and dΩ/dt = ±20.9 rad/s²

Fig. 14

Order of computational discretization scheme and grid size effects to numerical diffusion at t = 1 and 2 s respectively when Δt = 5 s and dΩ/dt = ±20.9 rad/s²

Fig. 15

Order of computational discretization scheme and grid size effects to concentration distribution profile over radial position r (z = 0.25 mm, ϕ = 3.75°, see the white dash curve in Fig. A.2) at t = 1 and 2 s respectively when Δt = 5 s and dΩ/dt = ±20.9 rad/s²