2.1 Isolation, cultivation and identification of hUCMSCs

The experiment design and examinations of this study were briefly illustrated (Figure S1). With the informed consents from three volunteers, the umbilical cords were obtained for cell isolation and histological observation. Briefly, after sterilization and thorough washing, the umbilical vein, two arteries and umbilical cord envelope were removed, and the Wharton's jelly was obtained. After cut into pieces, 0.1% type II collagenase (Sigma-Aldrich) was used for digestion overnight at 4°C. Then, the tissue suspension was screened, plated in mesenchymal stem cells medium (SM, Table 1) and incubated in 5% CO₂ at 37°C. After 3 days, the adherent growth of hUCMSCs was observed. Passage 4-8 was chosen for subsequent experiments.

For identification, 6 × 10⁵ hUCMSCs in PBS were incubated with FITC-labelled mouse anti-human IgG1 kappa isotype (eBioscience), mouse anti-human CD73 (eBioscience) and mouse anti-human CD90 (eBioscience), PE-labelled mouse anti-human IgG1 kappa isotype (Santa Cruz), mouse anti-human CD34 (Santa Cruz) and CD45 (Santa Cruz), respectively, on ice for 30 minutes. Flow cytometry (BD) was applied for detection and analysis.

2.2 Co-culture of dual-directional induction of hUCMSCs

For the co-culture of the dual-directional induction of hUCMSCs, a series of culture media were selected (Table 1). When cell confluence reached 70%-80%, the culture medium would be changed from the proliferation medium (PM) to the induction medium. Before sent for co-culture, the os-hUCMSCs (osteogenically induced hUCMSCs) and en-hUCMSCs (angiogenically induced hUCMSCs) were pre-induced for 3 days. A gradient ratio was selected for co-culture study with different culture medium (Table 2). For co-culture, pre-induced os-hUCMSCs and en-hUCMSCs were mixed with different ratios accordingly and incubated for another 3 days before in vitro and in vivo study.

2.3 Determination of osteogenesis and angiogenesis capacity of co-cultured hUCMSCs

ALP activity (n = 4): ALP quantitative and qualitative study was both carried out. According to the manufacturer's instructions of ALP quantitative kit (Nanjing Jiancheng), after cell lysis and centrifugation, 30 μL supernatant of each sample reacted with 50 μL buffer solution and 50 μL matrix solution at 37°C for 15 minutes. Subsequently, 150 μL colour developer was added and immediately sent for the OD reader at 520 nm. The protein concentration of each sample was determined by the BCA protein quantitative kit (Best Bio) for normalization. The ALP activity was expressed as King unit per gram protein. For ALP staining, the diethyl 2,5-di(thiophen-2-yl)terephthalate, 5-bromo-4-chloro-3-indolyl phosphate (BCIP) and nitro blue tetrazolium (NBT) were mixed with the ratio of 300:1:2 to get BCIP/NBT staining working reagent (Beyotime). All the samples were fixed in 4% paraformaldehyde (PFA) for 15 minutes before incubated with the staining working reagent for 20 minutes.

Alizarin red staining (ARS) (n = 4): After sample fixation, the alizarin red solution (0.2%, pH 8.3) was added in for reaction of 10 minutes. Before imaged under the stereomicroscope to visualize the mineralization deposition, the wells were rinsed with distilled water to removing the non-specific staining. For quantification analysis, 10% hexadecyl pyridinium chloride monohydrate (CPC) was used to dissolve the mineralized nodules and then measured the colorimetric absorbance at 562 nm.

Matrigel angiogenesis assay (n = 5): To investigate the angiogenesis capacity, the matrigel angiogenesis assay was performed. Briefly, 10 μL ice-cold matrigel (354230, BD Biosciences) was added into the angiogenesis slides (ibidi) and sent for incubation at 37°C for 30 minutes. After that, 50 μL cell suspension (2 × 10⁴ cells) was dropped into each well and imaged after 2.5 hours incubation for microtubule formation.

Acetylated low-density lipoprotein labelled with Dil (Dil-ac-LDL) phagocytosis test (n = 5): Before testing, the culture medium was changed to EIM containing 10 μg/mL Dil-ac-LDL (Invitrogen). After incubation at 37°C for 4 hours followed by several PBS washes, the nuclei staining was carried out with 5 μg/mL Hoechst (Sigma-Aldrich). The confocal laser scanning microscope (CLSM, Leica) was used for observation.

Immunofluorescent staining (n = 3): After endothelial induction, the cells were fixed before penetrated by the frozen methanol at 20°C for 10 minutes. Blocking buffer (Beyotime) was employed to block the endogenous peroxidase at room temperature for 1 hour. Monoclonal antibody for EphrinB2 (1:100; Abcam) and EphB4 (1:800; CST) against human were incubated at 4°C overnight followed by incubation with secondary antibody (1:750; Biotium) for 1 hour. Later on, 5 μg/mL Hoechst (Sigma-Aldrich) was used for nuclei staining for 15 minutess and photographed by CLSM.

Quantitative reverse transcription-polymerase chain reaction (qRT-PCR) (n = 3): Total RNA was extracted from the target cells by TRIzol (Invitrogen) and quantified by NanoDrop 2000 Spectrophotometer (Thermo Fisher Scientific, USA). About 500 ng of total RNA from each sample was used to synthesize complementary DNA (cDNA) via PrimeScript™ RT Master Mix (Takara). Finally, 10 ng cDNA from each sample was utilized for qPCR with TB Green Premix Ex Taq II (Takara) and synthesized primers (Generay) in real-time fluorescent quantitative PCR (Bio-Rad). The relative gene expression was calculated using 2^−∆∆CT and GAPDH as the housekeeping gene. The primer sequences were listed in Table 3. The genes examined were ALP, runt-related transcription factor 2 (RUNX2), collagen type 1 (COL1), osteopontin (OPN), EFNB2, EPHB4, VEGF, basic fibroblast growth factor (bFGF), SERPINF1, ANGPTL1, SPROUTY1, angiogenin (ANG), CD146, PDGFB and GAPDH.

Western blot (n = 3): For cell lysis, RIPA containing 1% proteinase inhibitor, cocktail (ComWin Biotech) was applied followed by centrifugation. The supernatant was harvested and quantified by the BCA kit before subjected to Western blot analysis. The protein (20 μg) was separated by 10% SDS-PAGE gel electrophoresis, and the targeted protein was transferred to PVDF membrane and blocked with 1%. PBST containing 3% FBS for 1 hour. The primary antibody rabbit anti-human RUNX2 (1:1000; CST), mouse anti-human CD146 (1:1000; CST), rabbit anti-human β-actin (1:1000; CST) were incubated at 4°C overnight followed by the incubation of the secondary antibody goat anti-rabbit (1:2000; CST) or rabbit anti-mouse (1:2000; CST) for 1 hour. Finally, chemiluminescence detection reagents (ECL) was dropped onto the membrane and exposed to X-ray film. The gel imaging system (Bio-Rad) was used for photography, and the final results were analysed by Image J.

2.4 Construct and characterizations of tissue-engineered bone graft

Three-dimensional printing TCP scaffold fabrication: Briefly, β-TCP (Kunshan Chinese Technology New Materials Co., Ltd.) ink was prepared by dispersing the powder in dispersant with distilled water, a proper amount of hydroxypropyl methylcellulose, and polyethylenimine (PEI) to increase agglomeration. A bio 3D printer (Regenovo) was employed for the scaffold's fabrication. The printing parameters were set as 700 μm in distance and the height of 200 μm to get the final scaffold with a diameter of 5 mm and a thickness of 1 mm. After printing, the scaffolds were dried in air for 24 hours and sintered at 1100°C for 3 hours.

For the construct of tissue-engineered bone graft, before cells seeding, the sterile 3D TCP scaffolds were soaked in culture medium overnight and dried in air. About 15 μL cell suspension (6.67 × 10⁶/mL) of the 3:1 group was dropped onto the scaffolds. After 3 hours attachment, the culture medium was refilled.

Scanning electron microscopy (SEM): The morphology of the scaffolds and the attached cells was observed by SEM (S3400N, Hitachi). Before analysis, 2.5% glutaraldehyde (Leagene) was applied for fixation. With gradient alcohol dehydration, dry in air and gold spraying, the samples were mounted in for SEM.

Live/dead staining (n = 3): 3 days after cell seeding, the live/dead staining kit (BestBio) was used for characterization according to the instructions, and the samples were observed under CLSM.

2.5 In vivo study and analysis

With the permission of Guangzhou Medical University Ethics Committee, 12 male Sprague-Dawley rats (12 weeks, weight of 250-300 g) were chosen for the bone repair study. Groups division: blank, scaffold, 4:0 and 3:1 group (4:0 and 3:1 group were chosen based on the above research results) (n = 6). Sodium pentobarbital (30 mg/kg) was used for general anaesthesia by intraperitoneal injection. After shaved and sterilized, the soft tissue was cut along the middle line of the rat skull. The muscles and periosteum were separated bluntly to expose the operation area thoroughly. The full-thickness bone defect with a diameter of 5 mm was made with trephine under saline irrigation continuously. Two critical-sized calvarial bone defects were created per rat. The bone grafts were implanted in accordingly with strict sutures. Penicillin 400 000 U was injected intramuscularly for 3 days after the operation. Four weeks later, the rats received euthanasia with inhalation anaesthesia of isoflurane (Yipin China), and the specimens were harvested and fixed for further studies.

Micro CT: Micro CT (SkyScan 1172; Bruker) was used for radiography analysis with the scanning parameter of 1 mm thick aluminium filter, resolution 2000 × 1332, pixel size 9 μm, voltage 80 kV, current 100 μA. NRecon 1.0 software was then used to rebuild, Dataviewer for ROI selection and CTAn1.13 software for new bone volume, bone trabeculae number and thickness analysis. On account of no bone formed in the centre area of the defect in the blank group, the new bone volume, bone trabeculae number and thickness of the blank group were set zero.

Histology: For detecting the tissue structure of the umbilical cord, the harvested tissue was fixed in 4% PFA for 24 hours. Together with the decalcified calvarial specimens, the harvested umbilical cord tissue was washed overnight, dehydrated gradient, embedded in paraffin and sliced at a thickness of 4 μm for further studies.

For immunohistochemical staining, the umbilical tissue was sent for CD31 and CD34 staining. After pre-treatments of the samples, mouse anti-human CD31 (1:1600; CST) and rabbit anti-human CD34 (1:200; Abcam) were incubated at 4°C overnight, respectively. After rewarming and PBS washing, the secondary antibody rabbit anti-mouse (1:200; CST) and goat anti-rabbit (1:200; CST) were added for a reaction at 37°C for 1 hour. About 10% DAB was used for colouring with haematoxylin to stain the nuclei.

For haematoxylin-eosin (HE) staining, an automatic dyeing machine (Sakura) was used. The staining kit of Masson trichrome was applied for further analysis by the manufacturer's instructions (Servicebio). Briefly, after dewaxing, a series of staining reagents were used with haematoxylin dyeing for 3 minutes, acid fuchsin for 5-10 minutes, molybdenum phosphate acid dipping for 1-3 minutes and aniline blue dyeing for 3-6 minutes. For quantitative analysis, the percentage of the new bone volume was calculated with the new bone area/ tissue area × 100%.

2.6 Statistical analysis

All the presented data were expressed as the mean ± standard deviation. Tukey's multiple comparisons test in one-way ANOVA was used to a comparison of two groups, and Sidak's multiple comparisons test was used to analyse the mean of multiple comparisons tests at multiple time points. The significance level was set at P < .05.

3.1 Characteristics of the umbilical cord and hUCMSCs from Wharton's jelly

The umbilical cord tissue structure was demonstrated by HE staining. Two darkly stained round arteries coupled with one oval-shaped vein were observed clearly in the central part of the pink-stained Wharton's jelly, with a thinner layer epithelium surrounded. Under higher magnification, Wharton's jelly was loose with eosinophilic stained mononuclear cells scattered (Figure 1A). Immunohistochemical staining showed that Wharton's jelly did not contain CD31 and CD34 positive haematopoietic cells which could be observed clearly in the arteries and the vein with aligned brown colour (Figure 1B). The isolated P1 cells from the Wharton's jelly were homogeneous bipolar spindle-shaped and arranged in a whirlpool (Figure 1C). Flow cytometry showed that more than 99% of the cells expressed the surface markers CD73 and CD90 of MSCs, but did not express the haematopoietic markers CD34 and CD45 (Figure 1D).

3.2 Osteogenic differentiation of hUCMSCs

To evaluate the osteogenic differentiation capacity of hUCMSCs, ALP activity of quantitative and qualitative analysis, osteogenesis-related genes expressions and mineralized nodules were studied. After 4, 7 and 10 days of osteogenic induction, ALP staining of hUCMSCs in OM became purple, darker than that in PM, with the darkest colour observed on the 4th day (Figure 2A), which was in accordance with quantitative examination of ALP activity (Figure 2B), showing that the ALP activity increased significantly during the observed period in OM compared with that in PM, and reached the highest level on the 4th day, which was twice and three times higher than that of the 7th and 10th day, respectively. Meanwhile, the mRNA expressions of ALP and OPN increased significantly after osteogenic induction at all detection time points and gradually increased with time, while the expressions of RUNX2 and COL1 genes increased significantly only on the 4th day (Figure 2C). ARS showed scattered calcium nodules at the 3rd week after induction and patches after 5 weeks. No calcium nodules were found in the PM group (Figure 2D).

3.3 Endothelial differentiation of hUCMSCs

Three days after EIM induction, the induced hUCMSCs became elongated and polygonal with a large number of vascular-like structures (Figure 3A). Since tube formation on the matrigel and ac-LDL phagocytosis is the two representational experiments to determine whether the induced MSCs possess endothelial cells’ functions, we examined the morphology and function changes with the two tests. Studies showed that the 4 days’ pre-induced hUCMSCs inoculated on the matrigel for 2.5 hours could form a reticular structure like vessels, while normal hUCMSCs distributed in clusters (Figure 3B). Similarly, pre-induced hUCMSCs could engulf Dil-ac-LDL, with red fluorescent-labelled ac-LDL found around the blue nuclei, while the non-induced MSCs did not possess the same function (Figure 3C). Immunofluorescence staining showed that the cell connected in a circular pattern after induction, expressing both arterial endothelial cells’ marker EphrinB2 and vein endothelial cells’ marker EphB4 in the cytoplasm (Figure 3D). QRT-PCR examination detected that the expressions of angiogenic genes EFNB2, EPHB4, VEGF and bFGF on the 4th, 7th and 10th day after induction, and EFNB2, VEGF and bFGF enhanced on the 4th day. All the tested genes expressions declined on the 7th and 10th day significantly (Figure 3E). In contrast, except for the expression of anti-angiogenic gene SPROUTY1 declined significantly, the expressions of SERPINF1 and ANGPTL1 increasing over the whole period were contrary to that of EFNB2, VEGF and bFGF (Figure 3F).

3.4 Screening of the co-culture medium

When pre-induced os-hUCMSCs and en-hUCMSCs were co-cultured for 17 days, ARS showed that cells co-cultured in PM with the ratio of 3:1, 2:2 and 1:3 also had osteogenic differentiation ability with calcium nodules formation (Figure 4A), which had no statistical difference with the positive control group (Figure 4B). Similarly, there was no significant difference between PM, OM, EMM (2:2) groups and control group. Only in PMM (3:1) culture medium, a large number of scattered calcium nodules formed in 3:1 group (Figure 4A), which was dramatically increased compared with the others with the OD value of about 5 times higher than that of the control group (Figure 4B). Therefore, PMM (3:1) was selected for the subsequent co-culture medium.

3.5 Effects of co-culture on the osteogenesis and angiogenesis of hUCMSCs

After co-culture, the cell proliferation rate and ALP activity of 3:1 group were similar to those of 4:0 in OM, which was the positive control (Figure 5A,B), while the osteogenesis-related genes RUNX2, ALP and protein expression in group 3:1 were higher than that of 4:0 (Figure 5C,D). The proliferation rate of 2:2 and 1:3 group maintained at a higher standard in the overall observation (Figure 5A). Considering the osteogenesis-related genes and protein expression, ALP, RUNX2 and its protein of 2:2 group were the highest (Figure 5C,D). In contrast, ALP activity of 1:3 group on the 1st and 2nd day showed the highest compared with the others (Figure 5B), while ALP gene expression decreased after 3 days’ co-culture, with RUNX2 gene and its protein still higher than that of 4:0 group (Figure 5C,D).

For angiogenesis analysis, compared with 0:4 group, ANG mRNA expression of 3:1 group was the highest on the 3rd day, as well as PDGFB mRNA expression of 2:2 group, VEGF mRNA expression of 1:3 (Figure 5C). All these differences were statistically significant. CD146, the marker of pericyte, related to vascular stability, its gene and protein expression declined from 3:1 group to 0:4 group gradually (Figure 5C,E).

To further illustrate the angiogenetic capacity of the co-culture system, the matrigel tube formation assay was adopted. After 3 days pre-induction (0 day), the more percentage of en-hUCMSCs mixed, the more vascular tubules formed on the matrigel, with a more complete and continuous circular shape (Figure 6). However, after 3 days co-culture, the 0:4 group still possessed good angiogenesis ability, but the tube formation in the other groups predominantly weakened, with only a small amount of branching structures (Figure 6).

3.6 Characterization of the tissue-engineered construct with co-cultured dual-directional differentiated hUCMSCs

SEM showed that the struts of the 3D printed TCP scaffolds were dedicated and uniform with pore size ranging from 350 to 400 μm (Figure 7A). At 3.5 k magnification, microporous and nanoporous structures could be observed (Figure 7B). Live/dead staining results showed that green-stained viable cells attached to the scaffold with hardly visible red-stained dead cells (Figure 7C). Similarly, the hierarchical porous structure was beneficial to the growth and migration of MSCs, suggesting the 3D scaffolds possessed excellent cytocompatibility, and this combination was feasible for bone graft fabrication (Figure 7D).

3.7 Rehabilitation of the critical-sized bone defect

After the operation, all rats recovered soundly. From micro CT, 4 weeks post-operation, only a small amount of new bone grew from the host bone in the peripheral area of the defect in the blank group. With scaffold implantation, less bone formed on the scaffold. When incorporating the os-hUCMSCs, much new bone grew evenly along the scaffold. Moreover, the 3:1 ratio of os-hUCMSCs to en-hUCMSCs could significantly enhance the bone volume, trabecular number and thickness (Figure 8A-B).

For further ensuring the bone formation outcome, HE and Masson trichrome staining were both applied. From the gross view of the newly regenerated tissue in the defect area, amounts of pink-stained connective tissue filled in with slightly new bone regenerated along the host bone. By scaffold implantation, there were observed osteoid in the central part with a few capillaries. With os-hUCMSCs coated scaffold implantation, red-stained mature bone regenerated in the central area with dense, regular connective tissues along the porous structures of the decalcified TCP scaffold. In group 3:1, abundant mature bone scattered with dark pink-stained bone lacuna surrounded by less connective tissue (Figure 8A). With Masson trichrome staining, dark red-stained mature bone and light blue-stained collagen could be observed in the 3:1 group, while less mature bone formed in 4:0 group and no bone regenerated only with or without scaffold. The quantitative study of histology also validated that the 3:1 group possessed a greater bone inductive capacity (Figure 8C).