The ultimate goal of tissue engineering is to obtain a tissue or organ that is native-like at the structural and functional levels[1–3]. Thus far, however, tissue engineering technology is still facing challenges in mimicking the complex structure and function of biological tissues[4–5]. Human body tissues contain more than 200 cell types, and most of these tissues are composed of a variety of cells, which are required to maintain normal tissue function. For this reason, it is currently impossible to construct complex three-dimensional (3D) tissues and organs in vitro [6]. Therefore, developingin vitro cell culture models that are similar to in vivo cell niches in order to investigate cell proliferation, differentiation, and cell–cell interactions is essential for the fabrication of engineered tissues. in vitro models of cell behavior play a key role in mimicking native tissues and organs. Conventional twodimensional (2D) cell culture models such as cell culture plates cannot provide the proper microenvironment for cells to maintain natural morphologies or efficient communications[7–8]. Cells cultured on 2D surfaces are thought to show improved cell proliferation but inhibited cell differentiation [9]. In addition, natural tissues are complex systems consisting of two or more types of cells that communicate with one another through vital movement [10]. Although establishing a coculture system on a cell culture plate can enhance cell–cell interaction and communication, the cocultured cells on 2D surfaces cannot be organized spatially as in an in vivo environment[7,11,12]. Thus, a more appropriately engineered model is needed to create growth conditions and solve problems.

Previously, 3D cell culture models have been developed to offer relatively natural morphologies and cellular functions in the form of scaffold-free or scaffold-based culture systems, depending on tissue structures and functions. Bone tissue is known to be a complex system that contains many types of cells ordered in the extracellular matrix in response to mechanical and physiological stimuli [13]. Scaffold-based culture systems are advantageous in that they provide mechanical support and tissue continuity. Therefore, scaffold-based coculture systems have been established for research on cell–cell interactions and individual or synergistic effects in bone tissue regeneration. For example, the coculture of human bone-marrow-derived mesenchymal stem cells (HBMSCs) and human umbilical vein endothelial cells (HUVECs) on b-tricalcium phosphate (β-TCP) scaffolds exhibits greater alkaline phosphatase (ALP) activity than an HBMSC monoculture [14]. However, most scaffold-based coculture systems mix different types of cells together, making it impossible to distinguish the specific contributions of direct intercellular contact and the paracrine effect, respectively [15]. In addition, the various cells in human tissues usually exhibit a certain non-contact spatial arrangement, so simply mixing cells together does not accurately simulate their in vivo state [7]. For indirect coculture systems, electrospun scaffolds [16] and multilayered hydrogels [17] have been used by means of different types of splicing. However, to the best of our knowledge, non-contact coculture models comprising three or four different kinds of cells have not yet been established. It is difficult to establish a non-contact multicellular coculture system in an integrated scaffold that can permit cells to interact with their surroundings in a 3D environment.

3D printing technology provides a valuable pathway for implementing the fabrication of complex structural scaffolds that can serve as cell carriers to realize precise cell seeding [18]. In this study, we fabricated a β-TCP/hydroxyapatite (HA) biphasic bioceramic cell container with four different pore structures including triangles, squares, parallelograms, and rectangles integrated together by means of 3D printing technology. Bone-related cells including HBMSCs, HUVECs, human umbilical vein smooth muscle cells (HUVSMCs), and human dermal fibroblasts (HDFs) were inoculated in the different pores of the cell containers to investigate their individual or synergistic effects in osteogenesis and angiogenesis. Four kinds of cells inoculated in a 2D cell culture plate were simultaneously set as the control.

To the best of our knowledge, this is the first report of a four-cell non-contact coculture system within a single integrated cell container that mimics the cell niches in bone tissue. This work provides a research model for studying the mechanism of multicellular interactions and indicates a direction for developing a scaffold-based multicellular coculture system with the ultimate goal of tissue regeneration.

β-TCP/HA powders were purchased from Kunshan Chinese Technology New Materials Co., Ltd. (China) and filtered through a 200 mesh sieve. Pluronic F-127 (Sigma-Aldrich, USA) was dissolved to form 20 wt% aqueous solution as the binder. A homogenous printable ink was then prepared by mixing 4 g β-TCP/HA with 2 g Pluronic F-127 solution.

An integral framework divided into four parts with different pore structures including triangles, squares, parallelograms, and rectangles was designed using computer-aided design (CAD). To ensure that the cell containers could hold the four different cell types independently within the corresponding parts, all of the walls, bottom, and boundaries among different microstructure parts were seamlessly connected. The dimensions of the cell containers for studying cell proliferation were 12 mm × 12 mm × 4.8 mm, while a 24 mm × 24 mm × 4 mm geometrically similar cell container was designed for the cell differentiation experiments. The cell container was then fabricated using an extrusion-based BioScaffolder (GeSiM, Germany). To make the extrusion wire rod consistent and coherent, the moving speed of the print head was set at 4–6 mm·s^-1 and the extrusion pressure was set at 2–4 bar (1 bar = 10⁵ Pa) at room temperature. After the printing process, the green bodies were dried at room temperature for 24 h and then sintered at 1100 °C for 3 h with a heating rate of 2 °C·min^-1 to acquire pure ceramic cell containers.

The phase compositions of the β-TCP/HA cell containers were measured by X-ray diffraction (XRD; D8 ADVANCE, Bruker, Germany). The three orthographic views of the cell containers were observed by digital camera (Nikon, Japan), while the four kinds of microstructure were characterized by optical microscopy (S6D, Leica, Germany).

HBMSC and HBMSC culture mediums were purchased from Cyagen Biosciences Inc. (China). HUVECs were isolated according to the method described previously [19] and cultured in endothelial cell medium (Sciencell, USA). HUVSMCs and smooth muscle cell medium were purchased from STEMCELL technologies Inc. (Canada). HDFs were obtained in accordance with previous methods [20] and grown in Dulbecco’s modification of Eagle’s medium (DMEM) (high glucose, with the addition of 10% fetal bovine serum and 1% penicillin–streptomycin). The cell-seeding density was 50 000 cells per scaffold for the cell proliferation experiment and 800 000 cells per scaffold for the quantitative real-time polymerase chain reaction (qRT-PCR) experiment, which are constants for either monoculture or coculture. For two-cell coculture, HBMSCs were seeded on the rectangular and square parts, while HUVECs were seeded on the triangular and parallelogram parts of the scaffold. The seeding ratio of HBMSCs and HUVECs was 1:1. For three-cell coculture, HBMSCs were seeded on the rectangular and square parts while HUVECs and HUVSMCs were seeded on the parallelogram and triangular parts of the scaffold, respectively. The seeding ratio of HBMSCs, HUVECs, and HUVSMCs was 2:1:1. For four-cell coculture, HBMSCs, HUVECs, HUVSMCs, and HDFs were seeded on the square, parallelogram, triangular, and rectangular parts of the scaffold, respectively. The seeding ratio of HBMSCs, HUVECs, HUVSMCs, and HDFs was 1:1:1:1. For the 2D culture model, the cells were monocultured or cocultured on cell culture plates with the same cell-seeding densities and ratios as in the 3D culture model. All of the above cocultured cells were incubated in the mixture medium that corresponded to the cocultured cells in terms of categories and ratios.

Cell proliferation was analyzed using a cell counting kit-8 (CCK-8, Beyotime, China) assay. Monocultured and cocultured cells were seeded on scaffolds for 1, 3, and 7 days. The absorbance was measured at a wavelength of 450 nm in a microplate reader (Epoch microplate spectrophotometer, BioTek, USA). Cell morphology and adhesion were observed after 1 and 7 days in the four-cell coculture model. Cell-seeded scaffolds were rinsed once with phosphate-buffered saline (PBS, Sangon Biotech, China) and fixed with 2.5% glutaraldehyde, followed by dehydration in graded ethanol (30%, 40%, 50%, 60%, 70%, 80%, 90%, 95%, and 100% (v/v)).

To measure the gene expression of each kind of cell motivated by another in the two-cell coculture models, magnetic beads were used to separate the cocultured cells. In the HBMSC–HUVEC coculture model, all cells were trypsinized and centrifuged. The cells were then resuspended in the buffer solution, PBS with 0.1% bovine serum albumin (BSA, Sangon Biotech, China) with Dynabeads (Invitrogen, USA) combined with the cluster of differentiation (CD) 31 antibody. After being incubated for 20 min at 4 °C, the cell suspension was placed in a magnet for 2 min and the supernatant was transferred to a new tube. This separation process was repeated three times to obtain high-purity isolated cells. The separated HBMSCs and HUVECs in the coculture system are referred to herein as Co-HBMSCs and Co-HUVECs, respectively.

To evaluate the messenger RNA (mRNA) expression levels of osteogenic-specific genes (BMP2, Runx2, ALP, OCN, and Col1) and angiogenic-specific genes (VEGF, KDR, eNOS, and VE-Cad) in the monocultured cells and in the separated and non-separated cocultured cells, the total RNA was extracted using TRIzol reagent (Invitrogen, USA) at 3 days. RNA concentrations were detected with a NanoDrop instrument (NanoDrop Technologies, Thermo, USA). Complementary DNA (cDNA) was prepared using the PrimeScript RT Master Mix kit (TaKaRa, Japan) according to the manufacturer’s instructions. Real-time PCR was performed using a SYBR Green qPCR Master Mix (TaKaRa, Japan). The relative gene expression level was calculated using the cycling threshold (Ct) value comparison () method. GAPDH was used as the housekeeping gene. The primer sequences are listed in Table 1.

All experimental data were statistically analyzed using the Student’s t-test analysis software. The results were presented as means ± standard deviations (SDs). Differences between groups were detected by t-test, and p < 0.05 was considered to be a significant difference between groups.

Fig. 1(a) shows that 3D-printed cell containers with different microstructures can be prepared using 3D printing technology. Figs. 1(b) and (c) provide a schematic diagram of an integrated 3D cell container that combines four different pore structures (triangle, square, parallelogram, and rectangle) with or without seeded cells. A non-contact coculture model with four kinds of cells was established by seeding different cells into the four parts of the scaffolds. Using an extrusion-based 3D printer, microstructured scaffolds were fabricated with a well-defined geometry (Figs. 1(d) and (g)–(j)). From the scanning electron microscopy (SEM) micrographs, it is clear that all the walls, bottom, and boundaries of the different microstructured parts are seamlessly connected, such that a sealed cell container was manufactured (Figs. 1(d)–(f)). The XRD composition analysis of the printed cell container is shown in Fig. 2, indicating that β-TCP/HA biphasic scaffolds were obtained after sintering.

The SEM images in Fig. 3 show the cell morphology of the cocultured HBMSCs, HUVECs, HUVSMCs, and HDFs on the prescribed parts of the scaffolds at days 1 and 7. All the cocultured cells spread well, showing affluent pseudopods at day 1 and covering the whole scaffold at day 7. Fig. 4 shows the cell proliferation of the monocultured and cocultured cells in the scaffolds at different times. The results indicate that the non-contact coculture cell models in the 3D printing scaffolds showed a significant advantage in enhanced cell viability. Higher proliferation rate was observed in the cocultured HBMSC/HUVEC/HUVSMC group and in the cocultured HBMSC/HUVEC/HUVSMC/HDF group in comparison with the monocultured groups at days 3 and 7. The cocultured HBMSC/ HUVEC/HUVSMC/HDF group exhibited the highest proliferation efficiency.

To investigate the effects of the non-contact 3D coculture model on the cell–cell interactions, an in vitro coculture system with HBMSCs and HUVECs in 3D cell containers was applied. After coculturing for 3 days, the two types of cells were separated from the cell containers and the expression of osteogenic genes (BMP2, Runx2, ALP, OCN, and Col1) and angiogenic genes (VEGF, KDR, eNOS, and VE-Cad) from the separated cells were measured (Figs. 5 and 6). In general, the expression of osteogenic and angiogenic genes was found to be significantly upregulated in the non-contact 3D culture system in both the monocultured cells and the separated cocultured cells, in comparison with the 2D culture system. More interestingly, in the HBMSC/HUVEC coculture system, BMP2, the bone-related gene, was mainly expressed by the Co-HUVECs, and the BMP2 expressed by the Co-HUVECs in the 3D coculture model was significantly higher than that expressed by the Co-HUVECs in the 2D coculture model. In addition, vascular endothelial growth factor (VEGF), which promotes angiogenesis, was mainly expressed by the Co-HBMSCs, and the expression of VEGF in the Co-HBMSCs in the 3D coculture model was significantly higher than that in the 2D coculture model. Furthermore, the Co-HBMSCs in the 3D scaffolds showed significantly improved levels of Runx2, ALP, OCN, and Col1, while KDR, eNOS, and VE-Cad were mainly expressed by the Co-HUVECs in the 3D scaffolds.

The osteogenic and angiogenic marker gene expression of monocultured (HBMSC, HUVEC) and cocultured (HBMSC/HUVEC, HBMSC/HUVEC/HUVSMC, and HBMSC/HUVEC/HUVSMC/HDF) cells in the 3D cell containers was distinctly higher than that in the 2D cell culture plate (Figs. 7 and 8). In the 3D multicell coculture system, the coculture modes (including two-cell cocultures, three-cell cocultures, and four-cell cocultures) all showed obvious advantages in enhancing osteogenic and angiogenic differentiation, in comparison with the monoculture modes. Expression of the osteogenic genes Runx2, ALP, OCN, and Col1, and of the angiogenic genes VEGF, KDR, eNOS, and VE-Cad, was significantly improved with the addition of HUVSMCs and HDFs in the coculture system, and the four-cell non-contact coculture mode (HBMSC/ HUVEC/HUVSMC/HDF) in the 3D cell container exhibited optimal osteogenic and angiogenic activities.

One possible reason for failure in the construction of the complex structures and functions of tissues and organs is that there are no suitable engineered cell culture models that can mimic the multicellular components of human tissues. The establishment of tissue-specific multicellular coculture systems at a 3D level is critical for the investigation of cell–cell interactions and for further preparation of engineered tissues in vitro. In the past few years, 3D coculture systems have been widely designed and used in mimicking the physiological environment in vivo, and a handful of works have focused on seeding a mixed suspension of cocultured cells into a homogenous porous scaffold or a collagen, fibrin, or silk gel[21–25]. Although previous studies have demonstrated that simply mixing multiple cells together in 3D niches was effective in enhancing cell proliferation and differentiation compared with traditional 2D coculture [26], most cells communicate with each other in a non-contact manner in vivo. Therefore, developing a 3D non-contact multicell coculture container for sufficient cell–cell communication is a new strategy for tissue engineering.

Although numerous methods, such as molding and ice templating, have been developed to fabricate cell-containing scaffolds, the prepared pore structure is not optimal. It is well known that 3D printing is a preferred technology to efficiently prepare complexstructure scaffolds with a variety of pore sizes and morphologies[27–31]. In this study, we successfully fabricated a β-TCP/HA composite scaffold with four different pore structures including triangles, squares, parallelograms, and rectangles by using an extrusion-based 3D printing method (Fig. 1). Previous studies have found that a large number of cells leaked out of the 3D-printed porous scaffold during the cell-seeding operations. The present study fabricated a cell container in which the walls, bottom, and boundaries among the different microstructured parts of the scaffold were seamlessly connected. This method allowed precise control of the cell-seeding density and ratio to be achieved, without cell suspension loss. It has been reported that the porous structure and pore morphology of the scaffold are important in influencing the cell response, including cell attachment and proliferation, nutrient transport, and tissue growth[32–35]. The present study found that the 3D non-contact cell containers with different pore structures not only promoted the proliferation of a single type of cell, but also significantly stimulated the proliferation of cocultured cells. In particular, the coculture of three or four kinds of cells with a 3D scaffold exhibited a significant advantage in stimulating cell proliferation compared with a single cell type, indicating that cell–cell interactions and communications had taken place in the 3D non-contact cell container. There may be multiple potential molecular mechanisms in the scaffoldregulated cell proliferation; we tend to believe that the 3D cellcontainer-like scaffold might, on the one hand, affect the cytoskeleton through structural effects, which activate cellproliferation-related signaling pathways and promote cell proliferation and, on the other hand, enhance cell proliferation by amplifying the paracrine effect of the cocultured cells. Of course, these hypotheses require further confirmation.

To clarify the roles of the 3D cell containers in affecting cell-tocell communications, we first selected a relatively simple two-cell coculture system with HBMSCs and HUVECs. A previous study on a 2D direct-contact HBMSC/HUVEC coculture system demonstrated that the angiogenic growth factor VEGF secreted from the HBMSCs promoted the activation of KDR in the HUVECs, which further enhanced the expression of BMP2 and eNOS to initiate osteogenesis and angiogenesis [12]. Our study demonstrates that the synergistic effect of HBMSCs and HUVECs in osteogenesis and angiogenesis is amplified in a scaffold-based 3D indirect coculture system, as characterized by the remarkable upregulation of the expression of osteogenic and angiogenic genes by the separated cocultured cells in the 3D model compared with the 2D model. Interestingly, our experiments revealed that the angiogenic growth factor VEGF was mainly expressed by the Co-HBMSCs, and that the expression of VEGF by the Co-HBMSCs in the 3D coculture model was significantly higher than that in the 2D coculture model. Simultaneously, the expression of VEGF receptors such as KDR from the cocultured HUVECs in the 3D cell containers was significantly enhanced. In addition, the Co-HUVECs in the 3D scaffolds showed higher upregulation of angiogenic genes such as eNOS and VE-Cad. It has been reported that HBMSCs might promote angiogenesis via the microstructure-sensitive paracrine effect in a coculture of HBMSCs and HUVECs in 3D substrates compared with 2D substrates [36]. It is therefore reasonable to speculate that, in comparison with a 2D cell culture plate, a cell container with a 3D porous structure would enhance the angiogenic effect of the cocultures, probably because it would amplify the paracrine effect of the coculture system. In addition to VEGF, another important growth factor, BMP2, has been found to play an important role in the early stages of bone regeneration[37,38]. Interestingly, our results found that BMP2 was mainly expressed by the Co-HUVECs, and that the BMP2 expressed by the Co-HUVECs in the 3D coculture model was significantly higher than that expressed in the 2D coculture model. It has been reported that the BMP2 secreted by the HUVECs can regulate the osteogenesis of the HBMSCs [12]. Our results also demonstrated that the Co-HBMSCs in the 3D scaffold showed higher upregulation of osteogenic genes such as Runx2, ALP, OCN, and Col1. Based on these findings, it is clear that in the HBMSC/HUVEC coculture system at the 3D level, the VEGF expressed by the Co-HBMSCs can act on the Co-HUVECs: On the one hand, it activates their surface VEGF receptor KDR, which promotes the expression of downstream genes such as eNOS and VE-Cad in the Co-HUVECs to initiate angiogenesis; on the other hand, it stimulates the expression of BMP2 in the Co-HUVECs, which promotes the expression of downstream genes such as Runx2, ALP, OCN, and Col1 in the Co-HBMSCs to initiate osteogenesis. Our results suggest that the prepared cell container provided a good 3D environment for cell–cell interactions and communications via a paracrine pathway.

In the past decades, growth factors for enhanced angiogenesis and osteogenesis have been widely used in bone tissue engineering applications, and many studies have reported that the addition of growth factors BMP2 and VEGF into tissue engineering scaffolds can enhance the osteogenic and angiogenic activities of biomaterials[39–41]. However, the application of exogenous growth factors is limited in clinical use, due to the challenges of activity maintenance and controlled delivery[42,43]. In the present study, the 3D cell containers activated the two major biological events of osteogenesis and angiogenesis during bone tissue regeneration by increasing the expression of BMP2 and VEGF from the cocultured cells. This result indicates that the engineered 3D coculture model created an environment in which cells could communicate with each other and further secrete endogenous growth factors for enhanced tissue regeneration. Our results provide a new strategy for tissue engineering construction.

As bone formation is a vascular-dependent process, angiogenesis plays a key role in bone repair and skeletal development[44,45]. In addition to endothelial cells, smooth muscle cells and fibroblasts are the major cellular components of the blood vessel wall, and these cells interact with each other in a more complex autocrine and paracrine manner to maintain the homeostasis of the blood vessels[46–48]. In this study, we added smooth muscle cells and fibroblasts into the HBMSC/HUVEC two-cell coculture system and constructed three-cell and four-cell coculture systems to study the effects of multicellular coculture modes at the 3D level on the osteogenesis and angiogenesis of the cocultures. Our results show that the three-cell cocultures and four-cell cocultures both showed obvious advantages in enhancing osteogenic and angiogenic differentiation, in comparison with the monoculture and two-cell coculture modes. Expression of the osteogenic genes Runx2, ALP, OCN, and Col1 and the angiogenic genes VEGF, KDR, eNOS, and VE-Cad improved significantly with the addition of HUVSMCs and HDFs to the coculture system. Furthermore, the four-cell non-contact coculture mode (HBMSC/HUVEC/HUVSMC/ HDF) in the 3D cell container exhibited optimal osteogenic and angiogenic activities. Our results demonstrate that the development of a 3D scaffold-based multicellular non-contact coculture system permits the construction of a variety of cell cocultures of physiological tissues, which is very helpful for mimicking the tissue microenvironment in vivo and for further tissue regeneration.

In this study, we successfully fabricated a cell-container-like b-TCP/HA scaffold with four different pore structures including triangles, squares, parallelograms, and rectangles by means of 3D printing technology. Using this scaffold, an engineered 3D coculture model composed of HBMSCs, HUVECs, HUVSMCs, and HDFs with spatially controlled distribution was constructed to mimic the cell microenvironment of human bone tissue. The 3D-printed cellcontainer-like scaffolds demonstrated obvious advantages in enhancing cell proliferation as well as osteogenic and angiogenic differentiation in the coculture modes, in comparison with 2D planar cultures. In the two-cell coculture mode, the interactions between HBMSCs and HUVECs suggested that the 3D cell containers with different porous structures enhanced the angiogenic and osteogenic effects of the cocultures by amplifying the paracrine effect of the coculture system, in which the stimulation of the expression of VEGF and BMP2 plays a key role. Moreover, the four-cell non-contact coculture modes (HBMSC/HUVEC/HUVSMC/ HDF) in the 3D cell containers exhibited optimal osteogenic and angiogenic activities. Our study indicates a new direction for developing a scaffold-based multicellular non-contact coculture system toward tissue regeneration.

The research was supported by the National Key Research and Development Program of China (2016YFB0700803), the National Natural Science Foundation of China (51761135103), Crossdisciplinary Collaborative Teams Program for Science, Technology and Innovation of Chinese Academy of Sciences (JCTD-2018-13), STS Science and Technology Service Network Plan of Chinese Academy of Science (KFJ-STS-QYZD-092), Science and Technology Commission of Shanghai Municipality (17441903700), and the German Research Foundation (DFG, GE1133/24-1).

Xiaoya Wang, Meng Zhang, Jingge Ma, Mengchi Xu, Jiang Chang, Michael Gelinsky, and Chengtie Wu declare that they have no conflict of interest or financial conflicts to disclose.