Human adipose tissue samples were collected from patients after abdominal liposuction (n = 5; average age, 39.4 ± 6.9 years). Informed consent was obtained preoperatively. Liposuctions were washed in phosphate buffer saline (PBS), centrifuged, and transferred into a 50-ml syringe (opening diameter: 2 mm; area: 3.1 mm²; Becton Dickinson). The syringe was then connected to another empty 50-ml syringe via a three-way stopcock (Medex). Adipose tissue was then fractionated by passing it manually for 30 cycles through the stopcock to generate nanofat, as described in more detail elsewhere (Huang et al., 2020). 60% hydroxyapatite/40% tricalcium phosphate granules (400–1,000 μm, BoneCeramic, Straumann) were mixed with the nanofat [1:8 (V/V) granule/adipose ratio]. The mixture was seeded onto agarose-coated6-well plates (2 ml/well) and cultured for 3 weeks with proliferation medium, consisting of alpha-minimal essential medium supplemented with 10% fetal bovine serum, 1% HEPES, 1% sodium pyruvate, 1% penicillin streptomycin glutamine (all from Gibco), 0.01 mM ascorbic acid, 10^–7 M dexamethasone (both from Sigma-Aldrich), 5 ng/ml fibroblast growth factor-2 (FGF-2, R&D), and 20 ng/ml platelet-derived growth factor (PDGF, R&D). The generated constructs are referred to as Adiscaf (Guerrero et al., 2018).

Adiscaf disks were then generated by using a biopsy punch (4 mm diameter). The 2- to 3-mm-thick disks were placed in 12-well plates for chondrogenic and hypertrophic induction. For chondrogenic induction, the Adiscaf disks were cultured for 4 weeks in a serum-free culture medium composed of Gibco Dulbecco’s Modified Eagle Medium (DMEM), 1% HEPES, 1% sodium pyruvate, 1% penicillin streptomycin glutamine (all from Gibco), ITS (Sigma), and 1.25 mg/ml human serum albumin. The serum-free culture medium was further supplemented with 10^–7 M dexamethasone, 0.01 mM ascorbic acid, 10 ng/ml transforming growth factor-β3 (TGF-β3, R&D), and 10 ng/ml bone morphogenetic protein 6 (BMP-6, R&D). The disks were then cultured in hypertrophic induction medium for two more weeks. The hypertrophic induction medium is composed of serum-free culture medium, 10 mM β-glycerophosphate, 10^–8 M dexamethasone, 50 mM L-thyroxin, and 50 pg/ml interleukin1-β (IL-1β, all from Sigma-Aldrich). Hypertrophic Adiscaf disks were then analyzed or implanted in a rat mandible defect model, as described below.

The rat mandibular bone defect model used in this study has been previously described elsewhere (Pigeot et al., 2021). Fourteen 12-week-old male adult nude rats (300–350 g) were assigned to this study (cantonal permission BS3005) and divided into three experimental groups: hypertrophic Adiscaf (n = 8, generated from two independent adipose tissue donors, four replicate constructs per donor), Bio-Oss® (n = 4) and blank (n = 2). The number of replicates for each experimental group was defined based on the following: only two blank defects, to confirm the size criticality of the model already demonstrated in a previous study (Pigeot et al., 2021), eight hypertrophic Adiscaf replicates to take into account the interdonor and intradonor variabilities, and four Bio-Oss replicates given the homogeneity of the material. Buprenorphine (0.1 mg/kg/dose) was administered subcutaneously to the rats 1 h before surgery. The operation was performed under anesthesia by isoflurane inhalation, using a mixture of oxygen (0.6 L/min) and isoflurane (1.5–3 vol%). A 20-mm skin incision was made, and the mandible angle was exposed after detaching the masseter and pterygoideus medialis muscle. Electrocautery was used to cauterize superficial venous structures. A 5 × 5×3-mm segmental defect was created in the mandible using a piezo-electric saw (Figure 1A). The mandibular segment was removed without perforation of the oral mucosa. In the control group, Bio-Oss® (Geistlich) was used to fill the defect (Figure 1B), and a collagen membrane (Geistlich Bio-Gide®) was used to wrap the defect region (Figure 1C). In the experimental group, one Adiscaf disk was used to fill the defect and a Bio-Gide collagen membrane was used as well. In the blank control group, only a Bio-Gide collagen membrane was used to wrap the defect. Then, the extra-oral defect was closed respecting tissue layers using 3.0 and 4.0 Vicryl® resorbable sutures (Johnson & Johnson, St. Stevens-Woluwe, Belgium). Buprenorphine (0.1 mg/kg/dose) and Meloxicam (5 mg/ml, by s.c. injection of 1 mg/kg) were administered after surgery.

After in vitro culture, Adiscaf disks were fixed in 4% paraformaldehyde for 24 h at 4°C. Mandible samples after in vivo implantation were fixed in 4% paraformaldehyde for 48 h at 4°C, and then an 8 mm × 10 mm osteal block including the defect region was excised. These mandible blocks were decalcified by 15% EDTA for 2 weeks and embedded in paraffin. 5-μm-thick sagittal histological sections were generated by using a microtome and stained with Safranin O (Sigma-Aldrich), hematoxylin and eosin (H&E), Alcian Blue (Sigma Aldrich), or Masson’s trichrome (RAL Diagnostics, Martillac, France), as previously described (Huang et al., 2020). For immunostaining, the sections were rinsed and blocked with 10% normal goat serum to reduce nonspecific background staining and then incubated overnight at 4°C with the following primary antibodies: rabbit anti-human collagen II (Col II, ab34712), mouse anti-human collagen X (Col X, ab49945), rabbit anti-human Bone sialoprotein (BSP, ab52128), rabbit anti-human Osteocalcin (OCN, ab 1857, all from Abcam, Cambridge, United Kingdom), and mouse anti-human leukocyte antigen (HLA, ab70328). After sequential incubation with a biotinylated secondary antibody and an ABC-alkaline phosphatase complex, the specific staining was revealed by using Fast Red (Dako, Bollschweil, Germany).

Sections from each sample were subjected to quantitative analysis of bone tissue by computerized bone histomorphometry. H&E Sections were excitated using a 488 nm laser. Autofluorescence images of sections from each construct were acquired and used to measure the area covered with bone tissue formed within the implanted graft. Signal quantification was done by using the software Fiji, an open-source image processing package based on ImageJ2.

After in vitro culture, Adiscaf disks were digested with proteinase K (1 mg/ml proteinase K, 50 mM TRIS, 1 mM EDTA, 1 mM iodoacetamide, and 10 μg/ml pepstatin-A; Sigma Aldrich) for 16 h at 56°C. GAG content was measured as previously described (Barbosa et al., 2003). Briefly, digested Adiscaf were incubated in dimethylmethylene blue (DMMB) solution on a shaker at room temperature for 30 min. Precipitated DMMB-GAG complexes were centrifuged and dissolved in decomplexation solution. The samples were measured at 656 nm and the GAG concentrations were calculated with a standard curve prepared with purified bovine chondroitin sulfate. DNA content was measured using the CyQuant^® Cell Proliferation Assay Kit (Molecular Probes) according to the manufacturer’s instructions.

To detect repetitive human-specific Alu sequences in the samples, paraffin sections were subjected to chromogenic ISH as previously described (Osinga et al., 2016). To improve the binding of the probe, the sequence was altered by one base as follows: 5′-cga​ggc​ggg​tgg​atc​atg​agg​t-3′, reverse 5′-ttt​ttt​gag​acg​gag​tct​cgc-3′ to 5′-cga​ggc​ggg​tgc​atc​atg​agg​t-3′, reverse 5′-ttt​ttt​gag​acg​gag​tct​cgc-3’.

After mandible blocks’ inclusion in Histogel (Thermo Fisher Scientific, Waltham, MA, United States), micro-computed tomography (micro-CT) data were acquired using a high-resolution scanner (SkyScan1172, Skyscan, Belgium) and 0.5-mm aluminum filtered X-rays (applied voltage 50 kV; current, 200 mA) as previously described (Huang et al., 2020). Transmission images were acquired during a 360°rotational scan with an incremental rotation step size of 0.25°. Reconstruction was performed using a modified Feldkamp algorithm at an isotropic voxel size of 10 μm. Three-dimensional rendering of the structures was performed using VG Studio MAX 2.2 software (Volume Graphics, Heidelberg, Germany). The density values of the mandible block were analyzed. The range from 19,000 to 40,000 was regarded as densities in which bone tissue is to be expected, while granules containing mostly anorganic mineralization were located above 40,000. Densities below 12 k indicate no mineralization, which corresponds to soft tissues.

All data are presented as the means ± SD. The significance of differences was evaluated by Student’s t-tests. p values <0.05 were considered statistically significant. The data were visualized and analyzed using GraphPad Prism.

After 3 weeks of proliferation in vitro, 4 weeks of chondrogenic induction and two additional weeks of hypertrophic induction, the formation of hypertrophic cartilage tissue in the Adiscaf disks was demonstrated by the presence of large chondrocytic cells inside lacunae and by the deposition of glycosaminoglycans (GAG) (Safranin O staining, Figures 2A,B and Supplementary Figure 1), proteoglycans (Alcian blue staining, Figures 2C,D), and Collagen type II in the matrix (Figures 2E,F). After hypertrophy induction, chondrocytes within the constructs acquired hypertrophic phenotype, which was demonstrated by the presence of the hypertrophic marker Collagen type X in the matrix (Figures 2G,H). While a cartilaginous shell had formed at the periphery of the disks, some remnants of adipose tissue were still present in the center. The GAG/DNA ratio in the hypertrophic disks was 31.5 ± 7.2 μg/ng, similar to the ones found in a previous study (Huang et al., 2020).

Eight hypertrophic Adiscaf disks (Figures 2I,J), engineered using cells of two different donors, were implanted into a mandible defect in rats and compared to Bio-Oss^® granules, as described in the Materials and Methods section above and in Figures 1A–C. After 12 weeks, the jaws of the rats were harvested (Figure 1D). Macroscopically, both Adiscaf and Bio-Oss^® treatments resulted in a significant restoration of the geometry of the mandible (Figures 3A,B). In particular, while Adiscaf disks were able to restore the overall bone geometry of the mandible, and there was still 0.2- to1.0-mm spaces in the surgical defects filled with Bio-Oss^®. In contrast, empty defects were minimally repaired after 12 weeks, confirming the critical size of the experimental defect (Figure 3C).

Mandible blocks containing the defect region were then scanned by micro-CT. The original defect region was distinguished based on CT values, and osteogenesis from the incised edge was defined in each experimental group. CT pictures showed good osseointegration, that is, bridging of the construct to the adjacent bone, between Adiscaf disks and the mandible in all eight samples (one representative picture in Figure 3D), with a continuous network of mineralized tissue all over the bone defect. Although three out of four Bio-Oss^® samples remained in the mandibular defect, one Bio-Oss^® sample stuck to the surrounding muscle tissues and therefore failed to fill the mandibular defect. In the Bio-Oss^® group, osteoconduction from the border of the defect was evidenced without interaction of the granules with the host bone (Figures 3E, 4C). In the empty defect group, limited new bone formation was observed at the surface of the bone in the defect (Figure 3F). A quantitative analysis of mineralized volume per total volume inside the defect zone (Figure 3G) showed a trend for Adiscaf hypertrophic disks to generate more mineralized tissue volume than Bio-Oss^® granules (51 ± 15% vs. 38 ± 7%). Both Adiscaf and Bio-Oss^®, however, had significantly more mineralized tissue volume than the blank group (value for the blank group: 16 ± 4%, p = 0.01 and 0.03, respectively).

In histology, a round-shaped bone formation indicated that Adiscaf kept its shape after 12 weeks of in vivo implantation. Bone tissue formation by Adiscaf was confirmed histologically by H & E and Masson’s trichrome staining in the mandible regions, with characteristic osteoblasts located in the bone matrix and showing positive BSP immunostaining (Figures 4A,B,F,G,K,L and Supplementary Figure 2). There was no obvious transition between Adiscaf and the original mandible defect limit, in accordance with the corresponding micro-CT images. The new bone formation from Adiscaf constructs had tight connection with the mandibular bone and mostly restored the integrity of the mandible. In contrast, there was no new bone tissue formation observed in the defect region with Bio-Oss^®. Granules surrounded by the matrix but without BSP-positive osteoblastic cells were found in the mandible defect region (Figures 4C,D,H,I,M,N). The newly formed bone inside the defect region was observed and quantified on histology sections by using fluorescence images of the histological sections (Figure 5). A histomorphometric analysis of bone formation inside the bone graft demonstrated that Adiscaf disks generated significantly more bone tissue than Bio-Oss^® granules (18.9 ± 3.4% of surface area with bone tissue vs. 3 ± 0.7%, p < 0.01). In the blank control group, no newly formed bone was observed (Figures 4E,J).

To investigate the direct contribution of Adiscaf cells to bone formation, OCN and HLA-ABC, an osteoblastic and a human cell marker, respectively, were immunostained in two consecutive histological slides. Areas with mature osteoblasts, that is, positive for OCN, co-localized with areas containing HLA-positive (human) cells (Figures 6A–D). In situ hybridization for human-specific ALU sequences confirmed the presence of human cells after 12 weeks in vivo (Figure 6E), especially at osteocyte locations. In the Bio-Oss^® group, OCN-positive areas were found around the granules, but since no human cells were implanted in this experimental group, no HLA- or human-ALU–positive cells could be found (Figures 6H–J).