(604b) Rapid Prototyping of a Bone Marrow-on-a-Chip Model with in Situ longitudinal Imaging of the Endosteal Niche

In vitro models of the bone marrow (BM) microenvironment can be powerful tools to study BM pathophysiology and test new therapeutics. Realization of these models is difficult due to the complex cellular and structural components of the BM. Furthermore, conventional trabecular bone-mimicking scaffolds are difficult to image due to the tortuous pores that are impermeable to light. We have designed a rapid prototyping approach to overcome these limitations of in vitro BM models by developing a physical microfluidic device-based perfusion bioreactor model of the endosteal niche, along with a digital computational fluid dynamics (CFD) model to enhance cell engraftment by optimizing the design variables of the model (Fig. 1A). We have designed new scaffolds that are permeable to both light and fluid flow and mimic the trabecular thickness (Tb.Th.) and trabecular spacing (Tb.Sp.) of trabecular bone (Fig. 1B). Microfluidic polystyrene devices have been fabricated for cell imaging by injection molding (Fig. 1C). Our hybrid injection molding approach comprises a machined aluminum mold and a scaffold insert mold 3D printed from a high deflection temperature stereolithography (SLA) resin. A digital model was created using the COMSOL CFD module to calculate fluid streamlines in the microfluidic device, and cell trajectories were simulated using the COMSOL particle tracing module. This rapid prototyping approach allows for design variables such as scaffold morphometric properties and fluid flow to easily be optimized to enhance cell attachment and better recapitulate the BM endosteal niche.

The rate of cell interception with trabecular collectors is hypothesized to scale directly with fluid velocity and indirectly with Tb.Th., as predicted by the Stokes number (Stk). Increasing fluid velocity is limited by wall shear stresses which need to stay between 0 and 2 Pa, which represents the physiological range in BM and is optimal for cell attachment. Therefore, we have created a digital model of our microfluidic system with Tb.Th. between 250 and 875 μm, which is representative of human bone and approaches the resolution limits of SLA printing. Collection efficiency was quantified as the percentage of the total number of cells passed through the microfluidic device that adhered to the trabeculae (Tb.). Contrary to the predicted Stk scaling, our simulations show that cell deposition increases with increasing Tb.Th. over the range of 250 to 875 μm (Fig. 1D). This could be due to the fact that Stk predicts particle behavior in a fluid field with one obstacle, whereas we have an array of obstacles with neighboring Tb. possibly affecting the fluid flow and cell trajectories. Therefore, we have fabricated optimized polystyrene microfluidic devices with 875 μm Tb.Th. using an APSX-PIM injection molding machine. Human mesenchymal stem cells (hMSCs, 10⁶ cells/mL) labeled with red fluorescent CM-DiL dye (553/570 nm) were seeded for 24 h and allowed to proliferate for 1 week. hMSCs were then perfused with osteogenic medium for 21 days to promote their differentiation to osteoblasts. Attachment of hMSCs onto the scaffold was observed by confocal microscopy (Fig. 1E) and deposition of osteoid matrix was observed by polarized microscopy (Fig. 1F), indicating that our device can successfully support osteoblast growth and extracellular matrix deposition to model the BM microenvironment.

Our rapid prototyping, microfluidic device-based approach provides a novel opportunity to simultaneously manipulate and measure niche constituents to study BM disease. We have optimized scaffold fabrication to create an in vitro model that recapitulates the endosteal BM niche and is amenable to longitudinal imaging needed to study the spatiotemporal dynamics of cell growth and interactions in the BM microenvironment, as well as the disruptions in this niche that lead to BM pathophysiology. Our approach is anticipated to accelerate microfluidic device development, recapitulate the physiology of BM more accurately than existing models, and reduce the cell culture time for high-throughput applications such as drug screening.