Osteochondral tissue coculture: An in vitro and in silico approach

Abstract Osteochondral tissue engineering aims to regenerate functional tissue‐mimicking physiological properties of injured cartilage and its subchondral bone. Given the distinct structural and biochemical difference between bone and cartilage, bilayered scaffolds, and bioreactors are commonly employed. We present an osteochondral culture system which cocultured ATDC5 and MC3T3‐E1 cells on an additive manufactured bilayered scaffold in a dual‐chamber perfusion bioreactor. Also, finite element models (FEM) based on the microcomputed tomography image of the manufactured scaffold as well as on the computer‐aided design (CAD) were constructed; the microenvironment inside the two FEM was studied and compared. In vitro results showed that the coculture system supported osteochondral tissue growth in terms of cell viability, proliferation, distribution, and attachment. In silico results showed that the CAD and the actual manufactured scaffold had significant differences in the flow velocity, differentiation media mixing in the bioreactor and fluid‐induced shear stress experienced by the cells. This system was shown to have the desired microenvironment for osteochondral tissue engineering and it can potentially be used as an inexpensive tool for testing newly developed pharmaceutical products for osteochondral defects.


| INTRODUCTION
Osteoarthritis of the synovial joint is a common cause of osteochondral defects. Osteoarthritis of the knee accounts for 83% of total osteoarthritis burden and affects around 250 million people globally (Vos et al., 2012). Injured cartilage does not heal spontaneously due to limited access to progenitor cells and scarce blood supply (Redman, Oldfield, & Archer, 2005).
Osteochondral tissue engineering aims to restore tissue that is functionally and mechanically comparable to native hyaline cartilage and its subchondral bone (Nukavarapu & Dorcemus, 2013). Given the distinct difference in structure and microenvironment of the two tissue types, osteochondral tissue engineers often employ bilayered scaffolds and bioreactors to provide different microenvironments to bone and cartilage layers and to facilitate nutrient and waste transport. Previously, our group cocultured chondrocytes and osteoblasts on a hyaluronate/β-tricalcium phosphate (β-TCP)  (Kuiper, Wang, & Cartmell, 2014). It demonstrated that the bioreactor was able to maintain the respective osteoblast and chondrocyte phenotype in each layer. However, lower mechanical strength and permeability of the scaffold were expected, as its chondral and osseous layers were manufactured independently before joining (Mano & Reis, 2007).
One way to improve scaffold mechanical stability is to produce a gradient structure through additive manufacturing techniques (Giannitelli, Accoto, Trombetta, & Rainer, 2014;Yousefi, Hoque, Prasad, & Uth, 2015). Additive manufacturing is a scalable process that can create complex and tuneable scaffolds from CAD models. It has been shown that the discrepancy between the CAD and the actual manufactured geometry can cause a significant change in the microenvironment inside a bioreactor through finite element analysis (FEA) (Hendrikson, van Blitterswijk, Verdonschot, Moroni, & Rouwkema, 2014). By combining microcomputed tomography (μCT) and FEA, the culture microenvironment of an actual manufactured scaffold can be studied.
Various immortalized cell lines have widely been used for osteochondral tissue engineering because they exhibit specific cell behavior observed in primary chondrocytes or osteoblasts with low cost and ease of use. Murine osteoblastic MC3T3-E1 cells exhibit an osteoblast-like developmental sequence, from proliferation to mineral deposition in vitro (Quarles, Yohay, Lever, Caton, & Wenstrup, 1992;Wang et al., 1999). For cartilage tissue, ATDC5 cells are often used as an in vitro model for skeletal development as they show a sequential chondrocyte differentiation process (Newton et al., 2012;Yao & Wang, 2013).
In this study, we aimed to describe an in vitro osteochondral perfusion coculture system-a novel additive manufactured bilayered scaffold inside a coculture bioreactor. The new bilayered scaffolds were designed to have improved integrity and permeability compared to the previous scaffolds. The coculture system was investigated in vitro through coculturing ATDC5 and MC3T3-E1 cells on the respective chondral and osseous layers of the scaffold, as well as in silico through FEA of the microenvironment inside the scaffold (i.e., flow velocity, fluidinduced shear stress, and differentiation media mixing) during the perfusion. The microenvironment inside the actual manufactured scaffold from µCT was compared with the CAD, and the effective microenvironment for osteochondral tissue engineering was discussed.

| Cell line and culture media
Mouse chondrogenic cell line ATDC5 and mouse osteoblastic cell line MC3T3-E1 were purchased from Public Health England.
ATDC5 and MC3T3-E1 cells were maintained in the respective cartilage and bone growth media. The cartilage growth medium was composed of DMEM/F12 with 5% FBS and 1% A/B; and the bone growth medium was composed of α-MEM with 10% FBS, 2 mM L-glutamine and 1% A/B.
Chondrogenic and osteogenic media were prepared. More exactly, for the chondrogenic medium, the cartilage growth medium was supplemented with 0.2% ITS premix and 50 µg/ml ascorbic acid. For the osteogenic medium, the bone growth medium was supplemented with 10 mM β-GP and 50 µg/ml ascorbic acid. to improve the surface wettability. The collagen suspension was produced according to a previously established method (Liu, Shen, & Han, 2011;Tamaddon, Walton, Brand, & Czernuszka, 2013). Briefly, a dispersion of 1% bovine Achilles tendon collagen in 0.05 M acetic acid solution (pH = 3.2) was homogenized on ice and degassed using centrifugation. The bi-layered scaffold was then produced by casting the collagen dispersion into custom-made 3D printed cylindrical resin molds (15 mm diameter, 10 mm height, Figure 1b

| Cell seeding
The scaffolds were sterilized using 70% ethanol three times for 15 min, washed twice with DPBS for 5 min and were stored in α-MEM in a humidified incubator at 37°C before use. XUE ET AL.

| 3113
Before cell seeding, the scaffolds were dabbed with sterile tissue to remove excess liquid. A total of 500,000 ATDC5 cells in 100 µl chondrogenic medium were placed onto the top of the scaffold, followed by 2 hr incubation at 37°C. The scaffolds were then inverted and 500,000 MC3T3-E1 cells in 200 µl osteogenic medium were placed onto the bottom of the scaffold, followed by 2 hr incubation at 37°C.

| Perfusion coculture
The dual-chamber perfusion bioreactor used in the study was described previously (Kuiper et al., 2014). Briefly, the cell-seeded scaffold was put into the bioreactor. 30 ml chondrogenic medium and 30 ml osteogenic medium was added to the reservoirs connected to the respective top and bottom part of the bioreactor. Each bioreactor and the two reservoirs were then connected to a peristaltic pump equipped with 1.02 mm tubing (U205, Watson Marlow). Chondrogenic and osteogenic media were perfused through the respective top and bottom part of the coculture bioreactor at 0.5 rpm (~0.02 ml/min or 0.41 mm/s; Figure 1c). The differentiation media were changed every 3 days.
The scaffolds were harvested after 7 days of perfusion culture.
Each scaffold was cut into two equal half-cylinders with a surgical blade (Swann-Morton) and placed in α-MEM before immediate analysis. Each half-cylindrical scaffold was considered as one sample.

| Live/dead assay
To make the working solution, equal volumes of the LIVE/DEAD Cell Imaging Kit assay solution and cell culture medium were mixed.
Samples were washed and then incubated in the working solution for 15 min at the room temperature. Fluorescence micrographs were taken at 488/515 and 570/602 nm excitation/emission wavelength for the respective viable and dead cells with a confocal laser scanning microscope (CLSM, Leica).

| Resazurin assay
The resazurin assay was used to quantify the cell metabolic activity and proliferation in the whole scaffold, including both the bone and the cartilage section. Samples were washed with DPBS. After washing, each sample was incubated in 4 ml α-MEM containing 10% resazurin assay solution at 37°C for 4 hr. Next, 100 µl culture medium was collected and its fluorescence intensity at excitation/emission 560/ 590 nm was measured with a Microplate Reader (BMG Labtech).

| DiO and DiD cell tracking
DiO and DiD dyes were used to label the respective ATDC5 and MC3T3-E1 cells before cell seeding to study the cell

| Scanning electron microscopy
Samples were fixed with 1.5% glutaraldehyde solution at 4°C for 30 min, followed by dehydrating through ascending grades of ethanol (from 50% to 100%). Dehydrated samples were further dried by evaporation of the HMDS. Next, samples were mounted onto aluminum pin stubs (Agar Scientific) with Adhesive Carbon Tabs (Agar Scientific). Samples were sputter-coated with Au/Pd before imaging with Phenom Pro desktop SEM (Phenom-World) at approximately 500× magnification.

| Finite element analysis
The chamber geometry (Figure 2a) of the coculture bioreactor was generated in COMSOL Multiphysics (COMSOL). To model the PLA scaffold, its geometry was obtained from either the CAD or the μCT scan of the actual manufactured scaffold.
For the model of the CAD scaffold, the CAD file was imported to COMSOL multiphysics and physics-controlled normal mesh with boundary layers disabled was used to mesh the geometry (Figure 2b).
A total of 468234 tetrahedral elements were generated with size ranging from 0.411 to 2.74 mm. For the model of the actual manufactured scaffold, scaffolds were μCT-scanned using a Nikon XT H225 at 80 kV and 125 μA. A total of 3142 projections were captured with a 2000 × 2000 pixel detector, leading to a voxel size of 7.9 μm. μCT data were reconstructed with CT Pro 3D (Nikon) with beam hardening and center of rotation automatically calculated. Reconstructed data then was smoothed with bilateral filter and segmented with automatic thresholding in Avizo (FEI). As collagen had a very low attenuation under X-ray illumination, the geometry from PLA was reconstructed.
The finite element volume mesh was generated in a specialized meshing software Simpleware (Synopsys). More precisely, segmented μCT data of the scaffold generated from Avizo and the chamber geometry generated from COMSOL Multiphysics were imported to Simpleware.
The scaffold and the chamber were aligned and a volume mesh with F I G U R E 2 (a) Volume mesh of bioreactor chamber. (b) Volume mesh of scaffold geometry obtained from CAD or μCT, the collagen layer is not shown. CAD, computer-aided design; μCT, microcomputed tomography [Color figure can be viewed at wileyonlinelibrary.com] were then imported to COMSOL Multiphysics (Figure 2b).
In COMSOL multiphysics, the material properties of the perfusion media, namely, the dynamic viscosity of 1 × 10 −3 Pa·s, the density of 1000 kg/m 3 , and diffusion coefficient of 2.907 × 10 −9 m 2 /s (Holz, Heil, & Sacco, 2000) were used in the model.     From literature, the average fluid-induced shear stress on the scaffold was reported to be in the range of 0.05-100 mPa depending on the scaffold geometry (e.g., porosity and pore size) and inlet velocity (Boschetti, Raimondi, Migliavacca, & Dubini, 2006;Maes et al., 2012;Porter et al., 2005;Zhang, Yuan, Lee, Jones, & Jones, 2014;Zhao, Vaughan, & Mcnamara, 2015;Zhao, Vaughan, & McNamara, 2016). It is worth noting that for the top (i.e., chondral) layer of the scaffold, the collagen was not captured by μCT due to  Results showed that, in general, the different media were well contained in their respective sections. Various biochemical growth factors were often supplemented in the differentiation media to facilitate phenotype development (Vater, Kasten, & Stiehler, 2011).

| Cell attachment to the scaffold
As bone and cartilage tissues require different growth factors to promote their respective phenotype development in vitro, a coculture system with minimal differentiation media mixing is desired for osteochondral tissue engineering (Alexander, Gottardi, Lin, Lozito, & Tuan, 2014;Vater et al., 2011).
The results with different scaffold orientation ( Figure S1) showed that distributions of the flow velocity, FSS, and media mixing did not change significantly with scaffold rotation. Quantification data further confirmed the findings (Table S1).
Collagen and PLA are biodegradable materials with different degradation rate (García-Gareta, Coathup, & Blunn, 2015); and the degradation process will likely cause a change in the microenvironment during the perfusion. However, the current FEM did not consider the degradation process. In the future, the FEM can be improved by incorporating the materials degradation profile through a time-dependent study.

| Discrepancy between the CAD and the actual manufactured scaffold
Comparing the model created with the CAD to that with the μCT image, the latter generally resulted in reduced flow velocity except for the mean velocity at the osseous section which saw a 130% increase (Table 1). For the FSS, increased mean magnitude but decreased maximum values were observed in the μCT model. For the media concentration, results showed a slight increase for the chondral layer and 17% increase for the osseous layer. The differences can be linked to the less homogeneous structure caused by common additive manufacturing methods including part accuracy, shrinkage, surface finish, and so on (Leong, Cheah, & Chua, 2003 et al., 2000). Pazzano et al. (2000) also showed that the flow perfusion was able to maintain the pH gradient throughout the scaffold leading to increased DNA content. However, other researchers found that the flow perfusion led to downregulation of SOX9, GAG, and collagen II expressions, indicating reduced chondrogenic and increased osteogenic differentiation (Guo et al., 2016;Kock, Malda, Dhert, Ito, & Gawlitta, 2014;Mizuno, Allemann, & Glowacki, 2001 seeded on a decellularized trabecular bone (DTB) after 7-day perfusion culture (Cartmell, Porter, Garcia, & Guldberg, 2003;Porter et al., 2005). Shear stress of 0.05 mPa resulted in high-cell viability and proliferation; 1 mPa led to high osteogenic gene expression, and 5 mPa resulted in significant cell death. Zhao, Chella, and Ma (2007) perfusion cultured human MSC on polyethylene terephthalate scaffolds for 20 days and found that appproximately 0.01 and 0.1 mPa shear stress led to increased proliferation and osteogenic expression, respectively. Similarly, whereas maintaining the mass transport (flow rate), Li, Tang, Lu, and Dai (2009) showed that the lower shear stress (5 mPa

| CONCLUSION
In conclusion, this study demonstrated that the current osteochondral culture system supports the coculture of ATDC5 and MC3T3-E1 cells on a novel additive manufactured scaffold with regard to cell viability, proliferation, distribution", and attachment.
The microenvironment inside the bioreactor during the perfusion culture including flow velocity, fluid-induced shear stress, and media mixing was studied using FEA. This system was shown to be viable in vitro osteochondral model due to its desirable microenvironment. It can be readily used as a platform for the cytotoxicity test or drug delivery study. For more clinically relevant applications like drug efficacy tests for osteoarthritis, the cell lines used can be easily replaced by primary cells or mesenchymal stem cells.
F I G U R E 9 Bone and cartilage tissue engineering conditions for cells seeded on a 3D porous scaffold inside a perfusion bioreactor. For each study, cell type, scaffold material, and experiment outcome are presented as "Cell type/Scaffold material: experiment outcome", followed by reference number. ADSC, adipose-derived stem cell; ALP, alkaline phosphatase; BAC, articular chondrocyte; CPBTA, chitosan poly(butylene terephthalate adipate); ECM, extracellular matrix; PCL, polycaprolactone; PGA, polyglycolic acid; PLLA, poly L-lactic acid