Knee osteoarthritis is a common and frequently occurring disease in the middle-aged and elderly population, including cartilage destruction and subchondral bone thickening (Peat et al., 2001). It ranks first in causing disability in the elderly and seriously affects the quality of life. Large-area cartilage defects can only be treated by joint replacement surgery. In the tissue engineering field, scaffolds play a pivotal role in tissue engineering, which provide a three-dimensional template for cell seeding, temporary mechanical function, and an extracellular matrix environment for tissue regeneration (Karageorgiou and Kaplan, 2005; Hollister, 2006). Osteochondral scaffolds act as an osteoconductive part and serve as delivery vehicles for cytokines like bone morphogenetic proteins (BMPs), thus providing osteoinduction (Groeneveld et al., 1999). Osteogenesis and bone mesenchymal stem cell (BMSC) attachment occur after the implantation of the scaffolds.

Aiming for osteogenesis and cartilage regeneration, an ideal scaffold is needed to provide suitable biomimetic mechanical and biological environments having similar morphology and function like natural osteochondral bone which could optimize integration into neighboring tissues (Healy, 1999; Hubbell, 1999; Sakiyama-Elbert and Hubbell, 2001; Lin et al., 2004; Karageorgiou and Kaplan, 2005; Wu et al., 2017). In general, in vitro and in vivo tests are used to evaluate scaffold bioperformance with high funding cost. An ethical and economical method is needed to evaluate scaffold bioperformance which considers the influence of the material (surface tension and material kinds) and its geometry (pore size, porosity, geometry, and surface area).

Computational analysis is seen as a promising method for scaffold evaluation. Sanz-Herrera et al. analyzed bone growth and formation in a two-dimensional (2D) scaffold by bone remodeling theories which assumed that it is driven by a mechanical stimulus (Beaupré et al., 1990; Adachi et al., 2001; Sanz-Herrera et al., 2008). With the same assumption, Adachi et al. also used the bone remodeling theory with uniform stress hypothesis to express how new bone tissues formed (Cowin, 1993; Adachi et al., 1998; Adachi et al., 2006). However, neither of them involved fluid stimuli in the simulation, which plays an important role in bone tissue regeneration when using an osteochondral scaffold. Moreover, recruiting BMSCs, as the first step of tissue engineering, and simulating BMSC attachment on the scaffold and their distribution is a really significant step in predicting the final bioperformance (Byrne et al., 2007).

Prendergast et al. proposed a theory that mechanical and fluid mechanical stimuli caused stem cell differentiation (Prendergast et al., 1996). And this mechano-regulation method successfully validated tissue differentiation by experimentally surrounding implants (Huiskes et al., 1997). Damian P.B. et al. explored various design parameters of the scaffold in tissue regeneration, and by this method modeling, cell proliferation, migration, and differentiation, he assumed that cells were randomly seeded on the lattice scaffold. However, cells distributed on the scaffold are not random, and it is determined by a lot of complex influence factors—material, surface roughness, geometry, and bone marrow flow speed. After recruiting the scaffold, BMSCs would start proliferating and migrating until they are mature enough to undergo differentiation. The position of cells would finally influence the final tissue formation (Wendt et al., 2006; Santoro et al., 2010). Till now, few researchers have studied cell distribution in the scaffold at the initial stage of the tissue engineering process after surgery since it is impossible to sacrifice animals at the initial stage to observe cell distribution ethically and economically. In that case, computational analysis provided an important solution to investigate this postsurgery process. As it is impossible to observe cell distribution using in vivo tests to validate simulation results, our previous work (Liu et al., 2020) used a numerical cell attachment model with a volume of fluid (VOF) model and discrete phase model (DPM) to investigate the cell seeding process and validated using in vitro tests.

To explore the relationship between BMSC distribution at the postsurgery initial stage and bone formation on the scaffold, the novel model is developed in a more real environment for cell attachment. The bone marrow is set as non-Newtonian fluid calculated by the power law for non-Newtonian viscosity, rather than normal Newtonian fluid like the solution for cell seeding. The model is validated by using in vivo tests in which bone formation and distribution are quantified by a self-designed MATLAB program by analyzing the Micro-CT–scanned image by threshold-based operation. The information of osteogenesis could be obtained not only on the outside of the scaffold but also within the pores of the scaffold. This model is aiming to decrease both in vivo and in vitro tests for optimizing the scaffold design and material properties which could avoid unnecessary time and money investment.

In recent years, cell-based repairing therapeutics has been proved not successful enough for patients who have OA or cartilage defects (Fellows et al., 2016). Lack of blood vessels, low chondrocyte density, and migration ability make cartilage hard to repair (Henrotin et al., 2005). Compared to the autologous chondrocyte implantation (ACI) technology, using scaffolds to recruit BM-MSCs is seen as an effective way for cartilage regeneration and repair.

An ideal three-dimensional scaffold needs to have the same geometrical, mechanical, and biological properties with the host tissue. Our novel bionic osteochondral scaffold is designed by mimicking three main parts of the osteochondral bone which are cartilage, subchondral bone plate, and trabecular bone (Burr and Gallant, 2012). To mimic cartilage’s mechanical and biofunctional properties, we combined PLGA and collagen together with a 1–2 MPa elastic modulus. Although the peak stresses in vivo could be reached 18 MPa in joints during dynamic loads (Hodge et al., 1986), the elastic modulus is typically 0.3–1.5 MPa (Mow et al., 1984; Jurvelin et al., 1988; Arokoski et al., 1999; Nieminen et al., 2000).This phenomenon is caused by highly pressurized interstitial water during dynamic loading as it cannot be squeezed out during the loading process, which could cause high physiological stresses (Jurvelin et al., 1997; Laasanen et al., 2003). The high-porosity inner structure could also achieve the same function. As for the subchondral plate, it plays an important role in supporting the cartilage and also separating the synovial fluid and trabecular bone because osteogenesis could be inhibited by the synovial fluid (Andrish and Holmes, 1979; Hazelton et al., 1990). The PLA layer design is able to prevent the synovial fluid flow into the trabecular bone defect during the surgery and healing process and also provide sufficient mechanical support. The bottom layer mimicking the trabecular bone is made by 3D-printed titanium which is widely used for trabecular bone regeneration because of its excellent biocompability, mechanical property, chemical stability, and suitability to mimic the biomimetic geometry (Long and Rack, 1998). The trabecular bone, with lower resistance to stress (50 MPa) and high resistance to strain (50%), has 50–90% porosity (Kaplan et al., 1994) and 1 mm diameter pore size (Keaveny et al., 2001). As 100–400 μm pore size is seen as optimal for bone regeneration (Hulbert et al., 1970; Schliephake et al., 1991; Bloebaum et al., 1994; Hofmann et al., 1997), a scaffold with 78.6% porosity and nearly 500 μm is designed.

As the porous design could result in diminished mechanical properties, the design needs to have an adequate mechanical stability to enable initial fixation with the host tissue during implantation as well as surface loading after surgery (Knecht et al., 2007). If the implantation detached partially or even completely fails in vivo, patients would feel serious locking or catching at the target area (Nehrer et al., 1999; Peterson et al., 2000). According to the in vivo test analysis by X-ray and SEM, it is found that new regenerated tissue connected to the scaffold with high porosity with similar mechanical property to natural bone tissue is really stable.

Interestingly, we found that even the regenerated bone tissue does not fill the void space of the bone defect after 3 months; more than 50% of the scaffolds’ whole structure is surrounded by new bone tissue, and there is only a slight decrease in osteogenesis from the side to the middle. This phenomenon illustrates that tissues and cells grow based on the scaffold surface which also points out that cell and tissue distribution is really important as good osteogenesis provides high-quality cartilage.

To further investigate cell distribution in the scaffold for in vivo tests, as it is impossible to sacrifice animals at the initial stage (Keaveny et al., 2001), a novel numerical model has been developed to predict cell distribution after the surgery process. For validating the model, we compared the cell distribution results from simulation in 5 s and the regenerated bone distribution of the sheep after 3 months together. According to previous research of Byrne et al., cell proliferation, differentiation, and migration just occur in neighboring areas (Keaveny et al., 2001), and it is found that cell migration does not influence a lot in cell distribution, especially in macroscale (Keaveny et al., 2001). In that case, it is believed that cell distribution at the initial stage is really related to the final bioperformance—regenerated tissue distribution (Keaveny et al., 2001). In other words, it is found that cells remain in their position on the scaffold when they work together and form tissues, as shown in Figure 10. As for the bone tissue distribution, horizontally, the trabecular bone occupied more percentage of the void volume at an external place than in the middle, which showed a same trend as external beams have more cell density after cell attachment. But the simulation results did not show that all external areas have really high cell density as in vivo tests showed. The reason is that cells attaching to the previous defect hole surfaces cannot be avoided. The cells attaching on the surroundings would proliferate and differentiate to build up the tissues. That is why the external place showed a more intensive cell density. In addition, comparing the cell distribution vertically between the experiment and simulation did not show that the bottom area has higher trabecular bone growth than the top. However, the cell density of the simulation only shows the same trend of the experiment results after the second titanium junction layer. It is clearly seen that not many cells were attached to the lower beam during the attaching process. But, a lot of cells were attached to the bottom surface of the hole. The bone percentage of the bottom is larger than that of the top because the bottom surface has much cell attachment, and they can absorb nutrients more easily as their position is much more close to the bottom and surrounding surfaces (nutrients injection position). In that case, the boundary condition that cells still come to the system after the bone marrow fills the void volume is more accurate to simulate the initial stage of tissue engineering—recruiting cells.

After validation and finding the best boundary condition to simulate the cell attachment process, we found that during the whole attachment process, the external area’s cell density was increasing rapidly, but the cell density of internal and middle areas increased extremely slowly. The reason is that initially (first 2.5 s), cells distribute more uniformly as the bone marrow already fills the void space and become steady at that time. When the scaffold starts to absorb cells (after 2.5 s), at the beginning, the cells attaches to the scaffold uniformly. After that, the balance of the cell density in the fluid breaks down, and cells start to move. As there are many beams in the middle providing drags for cells to move in, cells prefer to move outside to somewhere with less drags. That is why the cell density at external beams increases quicker than in the middle.

This novel numerical method for predicting tissue distribution is also available to provide rehabilitation guidance for patients after the surgery as the inner environment of the knee joint like flow speed and temperature could be influenced by rehabilitation devices. By this model, the optimized flow speed and temperature could be explored, and some special medical device should be designed to achieve this environment.

There are some drawbacks to the experiments. First, it is impossible to recognize whether the surrounding areas’ bone tissue is a new regenerate bone tissue or old ones according to the X-ray images. Although the external bone tissue growth is well, we cannot identify whether its growth is mainly based on the scaffold influence or it is just the surrounding bones’ remodeling and regeneration.

Although we could only ensure that the bone tissue in the internal scaffold’s pores are the new tissues, the biomimic scaffold still showed good-quality bioperformance. Second, during the drilling osteochondral defect process, it is unavoidable that the old bone’s powder created by the driller in the surroundings may be pushed into the bottom when the scaffold is pushed into the hole. In that case, the bone percentage in the bottom is higher than the other areas, which may not only be influenced by the cell attachment. Bone resorption and remodeling would also affect the final results.

This study develops a numerical model which could help researchers to optimize scaffold material property and geometry and could also avoid unnecessary in vivo and in vitro tests. As the initial healing process of the sheep could not be observed accurately, this model provides an opportunity for researchers to understand how the cell attached and distributed influenced by the scaffold structure and a proper mechanical and biological environment. The simulation results of cell distribution in the scaffold matched well with the regenerated bone tissue distribution in the in vivo test. The regenerated bone tissue on the surroundings (scaffold–material interface) is only around 15% more than that in the inner structure. This model could achieve an application to design a personalized scaffold and provide a proper mechanical and biological environment for recovery. It is also useful to help surgeons provide rehabilitation guidance for patients after implantation considering their knee joint inner environment.