A Microvascularized Tumor-mimetic Platform for Assessing Anti-cancer Drug Efficacy

Assessment of anti-cancer drug efficacy in in vitro three-dimensional (3D) bioengineered cancer models provides important contextual and relevant information towards pre-clinical translation of potential drug candidates. However, currently established models fail to sufficiently recapitulate complex tumor heterogeneity. Here we present a chip-based tumor-mimetic platform incorporating a 3D in vitro breast cancer model with a tumor-mimetic microvascular network, replicating the pathophysiological architecture of native vascularized breast tumors. The microfluidic platform facilitated formation of mature, lumenized and flow-aligned endothelium under physiological flow recapitulating both high and low perfused tumor regions. Metastatic and non-metastatic breast cancer cells were maintained in long-term 3D co-culture with stromal fibroblasts in a poly(ethylene glycol)-fibrinogen hydrogel matrix within adjoining tissue chambers. The interstitial space between the chambers and endothelium contained pores to mimic the “leaky” vasculature found in vivo and facilitate cancer cell-endothelial cell communication. Microvascular pattern-dependent flow variations induced concentration gradients within the 3D tumor mass, leading to morphological tumor heterogeneity. Anti-cancer drugs displayed cell type- and flow pattern-dependent effects on cancer cell viability, viable tumor area and associated endothelial cytotoxicity. Overall, the developed microfluidic tumor-mimetic platform facilitates investigation of cancer-stromal-endothelial interactions and highlights the role of a fluidic, tumor-mimetic vascular network on anti-cancer drug delivery and efficacy for improved translation towards pre-clinical studies.

Here, channel width = 100 μm for all channel segments.
Circularity and aspect ratio are defined as below: Aspect ratio =

Major axis
Minor axis (4) Elongation length was defined by the maximum distance between two points along the selected cell boundary (Feret's diameter) and was obtained via automated ImageJ analysis of the traced cell boundaries 1 .
Alignment of the hBTECs with shear flow was analyzed as described. The relative alignment of each microchannel segment within the fluorescence image was measured based on a horizontal frame of reference. The Feret's angle of each traced cell was automatically obtained via ImageJ and the relative channel alignment angle was subtracted from the Feret's angle to obtain the absolute value of the cellular alignment angle. The frequency of each cellular alignment angle was tabulated, grouped and plotted in MATLAB.
Analysis of cancer-fibroblast co-culture morphology: Cancer cells co-encapsulated with BJ-5ta fibroblasts within the PEG-fibrinogen hydrogel matrix in the microfluidic devices were imaged via phase contrast microscopy every 7 days over the 28 day culture period morphological quantification was conducted via ImageJ software. In the case of MCF7 cells, colonies or clusters of cells were manually traced and their area, circularity and aspect ratio were automatically extracted via the ImageJ software. The colony diameter was calculated from the major and minor axes of the traced regions according to Formula (1).

MCF7 colony density was defined as:
Colony density = Number of clusters/colonies in field of view Area of central tumor chamber in the field of view Elongated MCF7 colonies were those distinguished by irregular protrusions in the traced colony boundaries and were characterized as having an aspect ratio > 1.5.
In the case of MDA-MB-231 cells, individual cells were manually traced and cellular area, circularity and aspect ratio was automatically extracted via the ImageJ software. The cellular diameter was calculated in similar manner as above based on Formula (1). The cellular elongation length was considered as the longest distance between two points on the traced cell boundary (Feret's diameter) 1 .

Computational modeling of shear flow profiles in microfluidic chips: A general-purpose
Computational Fluid Dynamics (CFD) code, CFD-ACE+, 2 based on the Finite Volume Method (FVM) was used to discretize and solve the governing equations. Briefly, a three-dimensional computational mesh was created using CFD-GEOM, the grid generation module of CFD-ACE+. Steady state fluid flow was described by the conservation of fluid mass and momentum (Naviér-Stokes) equations similar to our previous studies 3,4 with inlet flow rates matching the experimental situation. The simulations results were analyzed using CFD-VIEW to obtain the shear maps presented in Figure 4A,B.

Analysis of diffusion gradients within the microfluidic chips:
Post-perfusion of TRITC-dextran through the vascular channels of the microfluidic chips, the fluorescence images of the dextranperfused regions of the central tumor chamber were acquired and analyzed in ImageJ in order to quantify diffusion gradients. The raw images were imported in ImageJ and LUTs corresponding to thermal heatmaps were applied to the images to generate visual gradients of the perfused dextran. Selected perfusion directions were measured for each chip design as denoted by arrows (Fig. 4E, F) and the 'Plot Profile' function of ImageJ was used to obtain quantitative values of the fluorescence intensity along each line. The values were normalized to the fluorescence intensity value of the TRITC-dextran in the surrounding vascular channel.

Analysis of drug-treatment effects in tumor-mimetic chips:
Post-drug treatment and live/dead image acquisition of cells within the tumor-mimetic chips, images were analyzed in ImageJ software to evaluate the two parameters: viable cell density and viable tumor area. The number of viable cells (stained green) were manually counted within each field of view and reported in terms of these two parameters.
Viable cell density is defined as the relative percentage of viable cells present within the field of view. The number of live cells counted in each field of view were compared to the control condition and expressed as a percentage to obtain this value.
Viable tumor area is defined as the relative area in the field of view occupied by the viable cells. The overall area occupied by the viable cells within the central tumor chamber was assessed, compared to the control condition and reported as a percentage to obtain this value.
For endothelial cells in the vascular channels and for cancer cells within static 3D cultures, the viable cell density was estimated in a similar manner as described above.
Analysis of MDA-MB-231 cell dynamics within tumor-mimetic chips: MDA-MB-231 cells co-encapsulated with fibroblasts in the PF hydrogel matrix within the tumor-mimetic chips were observed over 28 days in dynamic culture. MDA-MB-231 cells were observed to intravasate from the primary tumor chamber into the adjoining vascular channels as was quantified as the relative intravasation. The number of cancer cells present in the vascular channels at a specific time point was manually counted from phase contrast images and normalized to the number of cells at the initial time point to obtain the relative intravasation value.
The intravasated cancer cells further invaded into the adjacent secondary tumor chamber through long-term culture and the extravasated cell density was quantified by manually counting the number of cancer cells in the secondary tumor chamber and dividing it by the area of the chamber in the field of view. The invasion distance was calculated by measuring the distance covered by the invading cell front after crossing over into the secondary chamber.