Adipose tissues were obtained by simple liposuction from the abdominal subcutaneous fat of human donors. All donors provided informed consent, and the donor program was approved by the Institutional Review Board of Seoul National University (IRB No. H-0610-015-186). The subcutaneous adipose tissues were digested with collagenase I (1 mg/mL) under gentle agitation at 37℃ for 1 hr. Next, the digested tissues were filtered through a 100 µm nylon sieve to remove cellular debris and centrifuged at 1,500 rpm for 5 min. The pellet was resuspended in DMEM (Invitrogen, Carlsbad, CA, USA)-based media containing 0.2 mM ascorbic acid and 10% fetal bovine serum (FBS). The cell suspension was re-centrifuged at 1,500 rpm for 5 min. The supernatant was discarded and the cell pellet was collected. The cell fraction was cultured overnight at 37℃/5% CO₂ in DMEM-based media containing 0.2 mM ascorbic acid and 10% FBS. Cell adhesion was checked under an inverted microscope the next day. After 24 hr, non-adherent cells were removed and washed with phosphate-buffered saline (PBS). The cell medium was changed to Keratinocyte-SFM (Invitrogen)-based media containing 0.2 mM ascorbic acid, 0.09 mM calcium, 5 ng/mL recombinant epidermal growth factor (rEGF) and 5% FBS. The cells were maintained for four or five days until the cells became confluent, and considered at this point to represent passage 0. When the cells were 90% confluent, they were subculture-expanded in Keratinocyte-SFM-based media containing 0.2 mM ascorbic acid, 0.09 mM calcium, 5 ng/mL rEGF and 5% FBS until passage 3. The immunophenotypes of MSCs were then analyzed using a FACS Calibur Flow Cytometer. Every MSC harvest resulted in a homogenous population of cells with high expression levels of CD73 and CD90 and without expression of CD31, CD34 and CD45. Cell viability evaluated by the trypan blue exclusion method before shipping was > 95%. No evidence of bacterial, fungal, or mycoplasmal contamination was observed in cells tested before shipping. The procedures for MSC preparation were performed under good manufacturing practice (GMP) conditions (RNL BIO, Seoul, Korea).

To assess their MSC character, cells were tested in vitro for mesenchymal multipotency (osteogenic, chondrogenic, adipogenic, neurogenic and myogenic differentiation) before transplantation using lineage-specific stimulation (data not shown).

Porous hydroxyapatite/β-tricalcium phosphate (HA/TCP) ceramic blocks (Zimmer, Warsaw, IN, USA), with macropore size of 300-600 µm for and micropore size of ≤ 10 µm, porosity of 60%-70%, and Ca:P ratio of 1.6 were shaped into cylinders approximately 3 mm in diameter and 5 mm in length with a 1 mm wide central canal. The HA/TCP scaffolds were cleaned by sonication, rinsed in distilled water, and then sterilized by gamma radiation. In order to improve their cell-adherent properties, the scaffolds were coated with human fibronectin (100 µg/mL) prior to cell loading. MSC-loaded implants were prepared by incubating human fibronectin-coated HA/TCP scaffolds in a three density of 7.5 × (10⁵ or 10⁶ or 10⁷) cells/mL suspension of hATMSCs. Two millimeters of cell suspension were added to a 7 mL Vacutainer tube containing the ceramic implants. All loaded carriers were subjected to a vacuum to allow penetration of the MSC suspension. They were then placed in an incubator at 37℃ for 3 hr. The loaded carriers were re-fed with cell media overnight and transported to the operating room. Cell-free implants were prepared in a similar fashion by incubating fibronectin-coated scaffolds in a suspension devoid of cells. This procedure was similar to one that was described previously (11).

Cell attachment on the scaffold was confirmed by SEM as previously described (12). Before delivery, specimens of cell-free scaffolds and various densities of cell-loaded scaffolds were fixed in glutaraldehyde (24 hr), dehydrated in ethanol, coated with gold palladium and photographed by SEM.

All animal experiments were approved by the Institutional Animal Care and Use Committee (IACUC) of the Clinical Research Institute at Seoul National University Hospital (an AAALAC accredited facility), and followed National Research Council (NRC) guidelines for the care and use of laboratory animals (revised 1996). The rat femoral segmental defect model described here is a modification of a previous model used extensively to study long bone repair (11, 13). Twelve-week old adult male athymic nude rats (Harlan, Indianapolis, IN, USA) were used for xenotransplantation experiments. A total of 60 rats were divided into six groups (Table 1). Critical-sized femoral bone defects 5 mm in length were created in the mid-portion of the femoral diaphysis in each rat. In control animals, the bone defects were left empty with cell media. In Group I, the defects were filled with HA/TCP scaffold without any MSCs. In Group II, the defects were filled with hATMSC suspension (cell-loading density: 7.5 × 10⁷ cells/mL) alone to evaluate the therapeutic capacities of MSCs without scaffold. Cell-loaded scaffolds were used to study the effects of initial cell-loading density on bone tissue formation. Three scaffolds were constructed with cell-loading concentrations of 7.5 × 10⁵ cells/mL (Group III), 7.5 × 10⁶ cells/mL (Group IV), or 7.5 × 10⁷ cells/mL (Group V).

Femoral segmental defect surgeries were performed as follows (Fig. 1). Briefly, rats were anesthesized with zoletil/xylazine and prepared aseptically. A longitudinal skin incision was made along the lateral surface of the femoral diaphysis and the overlying periosteum was resected from the defect area. A polyethylene fixation plate (4 × 4 × 23 mm) was secured to the anterolateral aspect of the femur by four 1.1-mm Kirschner wires and two 34G cerclage wires, and a 5 mm transverse segment of the central diaphysis was removed with an oscillating bone saw (Surgic-XT, NSK Nakanishi Inc., Tochigi-ken, Japan) under 0.9% normal saline irrigation. The defect was filled with implants that varied by experimental group (see "Animal model and experimental design", above). Implants were secured by placing two 4-0 Vicryl (Ethicon, Somerville, NJ, USA) sutures around the scaffold and the fixation plate. The muscles were apposed, and the fascia and skin were closed in a routine layered fashion. The animals were allowed full activity in their cages postoperatively.

Postoperative radiographs verified implant placement and served as a baseline for radiographic evaluation. Serial radiographs were taken on the day of operation, and at postoperative 4, 8, 12, and 16 weeks, using a high-resolution radiography system (Faxitron 43805N, Hewlett-Packard, McMinnville, OR, USA). The exposure conditions were 40 kV, 1.2 mAs and 6.3 ms. The percentage areas of the femoral segmental defect occupied by newly formed bone were scored as previously described (14). The area percentage of bone regeneration filling the segmental defect and porous materials were graded from 1 to 4 using radiograph. The grades were defined as: [1] < 25% space of -; [2] 25% to 50% space of - ; [3] 50% to 75% space of - ; [4] > 75% space of - segmental defect and/or porous materials filled with new bone. The means of the scores of each group were calculated and compared.

The representative excised bone specimens were scanned with a Skyscan 1076 (Skyscan, Antwerpen, Belgium) for three-dimensional visualization of new bone formation in the segmental defect. Briefly, the specimens were placed in the holder and the centers of the segmental defects were identified with the scout view window. Femoral defects were scanned using 89 kVp, 112 ms integration time with 0.5 mm filter.

Sixteen weeks after the surgery, the animals were sacrificed and each defect area was evaluated histologically. The femora were dissected and fixed in 10% neutral-buffered formalin, decalcified (6.5% nitric acid), embedded in paraffin, and cut into 5-µm sections. Specimen histology was examined by hematoxylin and eosin (H&E) staining. Stained sections from each group were observed under a light microscope and were scanned using an attached digital camera and NIS-Elements system (Nikon, Tokyo, Japan). The new bone formation, residual HA/TCP scaffold and connective tissue areas were determined and converted to a percentage of total area of the bone defect with Image Software (Leica Application Suite System, Wetzlar, Germany).

We conducted a general toxicity study in accordance with good laboratory practices (GLP) and the OECD Guidelines for the Testing of Chemicals (TG409) with some modifications. During the postoperative period, we observed all animals for changes in behavior, general clinical signs, mortality, body weight, and food consumption.

At 16 weeks after the surgery, hematological analyses were performed with an animal blood analyzer (MS 9-5: Melet Schloesing Laboratories, Osny, France), including red blood cell (RBC) count, mean corpuscular volume (MCV), hemoglobin (Hb), mean corpuscular hemoglobin (MCH), mean corpuscular hemoglobin concentration (MCHC), white blood cell (WBC) count, differential WBC count, platelets (PLT), and hematocrit (HCT). Prothrombin time (PT) and activated partial thromboplastin time (APTT) were measured with plasma samples obtained after mixing sodium citrate and centrifuged at 3,000 rpm for 15 min using an automated coagulation analyzer (ACL-100, Lexington, MA, USA). Serum biochemistries were analyzed with an automated chemistry analyzer (Hitachi 7070, Hitachi Co. Ltd., Tokyo, Japan) for the following parameters: total protein (TP), albumin (ALB), A/G, glucose (GLU), cholesterol (CHOL), triglycerides (TG), total bilirubin (TBIL), urea nitrogen (BUN), creatinine (CRE), alanine aminotransferase (ALT), asparate aminotransferase (AST), alkaline phosphatase (ALP), chlorine (Cl), calcium (Ca), potassium (K), and phosphorus (P).

At sacrifice, we measured the following parameters with a urine analyzer (Miditron Junior II, Roche Diagnostics, Mannheim, Germany): specific gravity, pH, leukocytes, nitrite, protein, glucose, ketones, urobilinogen, bilirubin, and urinary hemoglobin. At the same time points, a veterinarian performed the following ocular examinations: pupillary light reflex (PLR), blink reflex, anterior segment, transparent media, and fundus.

At the end of the study the animals were sacrificed by bloodletting under deep anesthesia. Organ weights of the heart, liver, lungs, spleen, kidneys, adrenal glands, testes, brain and pituitary gland were determined by necropsy. Histopathological examinations were performed on the femoral segmental defect (transplantation site), heart, aorta, lungs, trachea, thyroid, parathyroid and salivary glands, stomach, duodenum, jejunum, ileum, colon, adrenals, testes, epididymes, prostate, urinary bladder, skeletal muscles, sciatic nerves, skin, mammary glands, pituitary, eyes, optic nerves, liver, spleen, pancreas, mesenteric lymph nodes, thymus, kidneys, spinal cord and brain. After fixation in 10% neutral-buffered formalin, all tissues were processed in a routine manner, embedded in paraffin, sectioned, and stained with H&E for light microscopic evaluation.

Data are presented as means ± SD One-way ANOVA, was performed using SPSS 12.0K to detect and compare differences among groups. For data sets that failed normality or equal variance testing, we used a post hoc Tukey's test or Dunnet's method for subsequent pairwise comparisons. P values (P < 0.05) indicated significant differences between the control group and treated groups.

Cell attachment on scaffolds was confirmed by SEM (Fig. 2). Scanning electron micrographs revealed the ultrastructural architecture of the fibronectin-coated HA/TCP scaffolds. The pores and surfaces of HA/TCP scaffolds were filled with hATMSCs in a cell loading density-dependent manner. In low density cell-loaded scaffolds, the attached cells appeared spindle-like on the scaffold surface. In higher density cell-loaded scaffolds, numerous spherical and spindle cells were detected in the pores and on the surface.

High-resolution Faxitron radiographs provided sufficient clarity and detail to detect subtle changes occurring within the implants and the surrounding host bone (Fig. 3). Defects of control (media) and Group II (hATMSCs alone) rats remained unhealed at 16 weeks. Nonunion at the host bone-implant interfaces occurred in Group I (HA/TCP scaffold alone), and the defect areas appeared granular and fractured. The HA/TCP scaffolds showed incomplete resorption within the 16-week follow-up period. Radiographs demonstrated substantially more bone in animals that received hATMSCs-loaded HA/TCP scaffolds than those that received cell-free HA/TCP scaffolds. Unions between implant and host bone were seen in Groups III, IV and V. The defects in these groups were bridged with radiopaque osteoid tissue that was more radiopaque at higher cell-loading densities. At 12 weeks, union was complete and additional bone was evident in the scaffolds. The scores for percentage areas of the femoral segmental defect occupied by newly formed bone were significantly higher in Group V than in the control group. These differences among the groups were first detected at 4 weeks after implantation and were maintained at 16 weeks after implantation. These data suggest that segmental defects that were filled with hATMSC-loaded HA/TCP scaffolds exhibited new bone formation in a cell loading density-dependent manner.

The defects in each group were visualized by three-dimensional micro-CT imaging (Fig. 4). Defects treated with cell media (control) and hATMSCs alone (Group II) did not close. Defects filled with cell-free HA/TCP scaffolds showed minor bone formation and bridging, but union was not complete. Micro-CT images demonstrate substantially greater bone formation in animals that received hATMSC-loaded HA/TCP scaffolds than in those that received cell-free HA/TCP scaffolds. The newly formed bone was more evenly distributed across the gap in Groups III, IV and V than in controls or Groups I and II.

Photomicrographs of representative sections of the implant groups at 16 weeks are shown in Fig. 5. In controls, most defects were filled with loose fibrous tissue and vessels. Small amounts of callus could be seen in the position on the plate. This pattern was also observed in defects treated with hATMSCs alone in Group II. In defects filled with HA/TCP alone, the tissues around the scaffolds were predominantly filled with fibrous tissue even at 16 weeks and most scaffolds remained unabsorbed by 16 weeks. Defects treated with hATMSC-loaded HA/TCP scaffolds in Groups III, IV and V mainly consisted of bone woven into the pores of scaffolds and around the scaffolds. As the amount of loaded cells increased from Groups III to V, more new bone formation was observed and the HA/TCP scaffolds were better absorbed. These signs indicate new bone formation in a cell loading density-dependent manner. These findings were consistent with the results of radiography and micro-CT examinations in these groups. In Group V, the highest density loaded group, large amounts of new bone and marrow were observed in the defects, whereas very little scaffold seemed to be buried in the newly formed osseous tissues. Filling of the pores with new bone and vasculature was evident in regions deeper within the HA/TCP. Despite xenogeneic hATMSC transplantation, no adverse host responses such as lymphocytic infiltration was detected.

Histomorphometric analyses revealed bone in the healed defects of femora treated with low (7.5 × 10⁵ cells/mL), middle (7.5 × 10⁶ cells/mL), and high (7.5 × 10⁷ cells/mL) density hATMSC-HA/TCP scaffolds. The percentages of newly-formed bone were 56.87 ± 23.856 vs 80.72 ± 17.170 vs 89.07 ± 6.518, respectively, indicating patterns of bone formation that increased according to cell-loading density. The bone formation in Groups IV and V was significantly greater than the mean area of bone induced by cell-free HA/TCP scaffolds alone (40.36 ± 18.442), in controls that received only cell media (52.40 ± 17.600), or hATMSCs alone (58.92 ± 15.774). Whereas the residual percentage of HA/TCP scaffolds in Group I was 14.11 ± 7.891, residual percentages of scaffold in Groups III, IV and V were 13.28 ± 8.855, 11.12 ± 8.476, 3.04 ± 3.397, respectively. Percentages of new bone formation and residual scaffold showed an inverse relationship. The percentages of connective tissues in controls and in Group II were 44.15 ± 15.801 and 41.08 ± 15.774, respectively. Whereas the percentage of new connective tissue in Group I was 47.91 ± 19.623, the percentages of new connective tissues in Groups III, IV and V were 26.69 ± 21.620, 8.16 ± 10.746 and 7.89 ± 6.492, respectively.

All follow-up animals survived for the duration of the study. In addition, no adverse clinical signs or symptoms were observed during the experimental period except in the two animals that showed ataxia and dysplasia of femurs mentioned above. There were no significant differences in food and water consumption between control and treated groups (data not shown).

No significant differences in body weights between controls and treated groups were observed during the study period (Fig. 6). No significant differences in hematological parameters were detected (Table 2); hemoglobin, hematocrit, total erythrocyte count, erythrocyte indices (MCV, MCH and MCHC), total leukocyte count, differential leukocyte count, platelet and coagulation time (PT, APTT) were similar across all groups.

Significant increases of BUN and calcium, and significant decreases of phosphorus were observed Group V compared to the control (P < 0.05). Significant decreases of chlorine and sodium were observed in Group III compared to the control (P < 0.05). However, these differences were considered to be minimal or within normal physiologic limits and not of biologic significance (Table 3).

No visible evidence of ocular abnormalities of effects was detected in ophthalmologic examinations. There were no significant differences in the results of urinalysis between control and treated groups (data not shown).

There were no significant differences in the absolute and relative weights of major organs related to detoxification and excretion (liver and kidney), immunity (spleen and thymus), endocrine function (adrenal glands and pituitary gland), circulation (heart), respiration (lung), nervous system (brain) or reproduction (testis) (data not shown) across groups. Results of gross postmortem examinations performed on all animals did not reveal any evidence of visible morphologic abnormalities in any animals used in the study. Histopathologic evaluations showed that the T-cell-dependent regions of thymus, spleen and lymph nodes were reduced or absent of, that were of the lesions seen in the athymic nude rats. There were no indications of any microscopic abnormalities related to the implantations of hATMSCs and HA/TCP scaffolds (data not shown).