Study design

All the experiments in this study were planned and performed according to the international regulatory guidelines33,34,35,36,37 for cell therapies and medical devices. Good Laboratory Practice38 and Standard Operating Procedures (SOPs) for all protocols were used. In vitro cytotoxicity assay was done according to ISO 10993-5 (2009 and 2012) guidelines. Assessment of the OAR was done in accordance to the ICRS II score system39.

Production of the nanofibrous compartment of the ARTiCAR

The nanofibrous component of ARTiCAR was obtained via electrospinning of PCL, as previously described24. Briefly, PCL (PURASORB®, PURAC, Corbion, Amsterdam, Netherlands) was dissolved in a 25% (wt/vol) dimethylformamide/dichloromethane solution (3/2, v/v) and delivered at a constant rate of 1 ml/h to the EC-DIG electrospinning device (IME Technologies, Eindhoven, Netherlands), set to at high voltage (20 ± 3 kV). Following electrospinning, PCL membranes were kept in a desiccator at 45 °C, to remove residual solvents, and sterilized by gamma irradiations (25 kGy). Membranes were then dipped alternately in 200 μg/ml BMP2 solution (rh-BMP2, Inductos, Medtronic, France) in 40 mM 4-Morpholinoethanesulfonic acid (Sigma-Aldrich, Saint-Quentin Fallavier, France), 150 mM Sodium Chloride (Sigma-Aldrich), pH 5.5 (MES buffer) and 0.5 mg/ml Chitosan (Protasan UP CL 113, Novamatrix, Sandvika, Norway), for 12 times. Each bath was followed by three washes in MES buffer.

Production of the hydrogel compartment of the ARTiCAR

Twelve mg/ml sodium alginate (Sigma-Aldrich) and 3 mg/ml hyaluronic acid (Lifecore Biomedical, Chaska, USA) were dissolved in 9 mg/ml Sodium Chloride (Sigma-Aldrich). Prior to implant, the hydrogel was mixed with either human or sheep MSCs. After the MSC/hydrogel compartment was applied to fill the defect, gelation was achieved using 102 mM calcium chloride (Sigma-Aldrich).

Cell culture

Human lung fetal fibroblast MRC-5 cell line (ECACC, Sigma-Aldrich) was cultured in 75 cm2-flasks with EMEM (Lonza, Levallois-Perret, France) containing 10% Fetal Bovine Serum (Lonza), 2 mM Glutamine (Lonza) and 1% Non Essential Amino Acids (Lonza), under 5% CO2 humidified atmosphere at 37 °C. Cell culture was performed accordingly to ISO 10993-5:2009 guidelines. The presence of mycoplasma in culture media was tested according to internal SOPs.

Cytotoxicity assessment in vitro

MRC5 cells were plated into 24-well plates. The NanoM1-BMP2 wound dressing was tested side-by-side with polyurethane film containing 0.1% zinc diethyldithiocarbamate (known for inducing cytotoxic effects; Hatano Research Institute/Food and Drug Safety Center, Japan) and high density polyethylene film (negative control; Hatano Research Institute). To assess cytotoxicity, pieces of different size (20, 16, 9, 4, 1 mm2) were placed in contact to the cultured cells, when 70–80% confluence was reached. Cells were cultured in the presence of the membranes for 3 days before being examined microscopically for changes in the general morphology, presence of vacuolization, detachment, lysis and membrane integrity, following the criteria for the qualitative evaluation of cytotoxicity according to ISO 10993 guidelines, part 5 (2009) and part 12 (2012): Class 0, no reactivity (no effects around or below sample); Class 1, slight reactivity (few malformed or degenerated cells); Class 2, mild reactivity (small area of malformed or degenerated cells below the sample); Class 3, moderate reactivity (malformed or degenerated cells in an area larger than the size of the sample but ≤1 cm2); Class 4, severe reactivity (malformed or degenerated cells in an area larger than the size of the sample but >1 cm2). A grade higher than 2 was considered as cytotoxic.

Quantitative cell viability assay

As a quantitative measure of cytotoxicity, cell viability was evaluated. At day 3, membranes were discarded, cells were washed twice with PBS, fed with 1 ml culture medium and 100 µl/well of Cell Viability Reagent WST-1 (Lonza) was added to each well, according to internal SOPs. The cells were incubated for 3 h at 37 °C in 5% CO2, and 100 µl of supernatant were transferred into a 96-well plate. Absorbance was measured at 450 and 620 nm in a Multiskan EX device (Thermo Fisher Scientific, Graffenstaden, France). Data analysis was performed with Ascent 2.6 (Thermo Fisher Scientific). Results were expressed as percentage of viable cells in respect to a blank control. A decrease of 30% viability was considered as cytotoxic.

Experiments with animals

Animal experiments were performed according to the ethical guidelines for animal experiments59. The protocols used, included in the project “Toxicologie Réglementaire”, was authorized by the “Ministère de l’Enseignement supérieur et de la Recherche” No. 01191.02. Seven-weeks-old rats were maintained for at least 5 days in Specific Pathogen Free rooms (authorized by the French Ministries of Agriculture and Research; agreement No. A35 288-1) before the beginning of the study, according to internal SOPs, under controlled conditions of temperature (22 ± 3 °C), humidity (50 ± 20%), photoperiod (12 h light/12 h dark) and air exchange, according to internal SOPs. Animals were housed in standard-size polycarbonate cages (with filter lid), and bedding was replaced twice a week.

Intra-articular implant of ARTiCAR in nude rats

Evaluation of acute toxicity in vivo was achieved via intra-articular implant of ARTiCAR in a model of induced osteochondral defect in RH-Foxn1 rnu/rnu nude rats (Harlan, Gannat, France). Briefly, sterile NanoM1-BMP2 were rinsed in sterile PBS and cut into quarters (1.77 mm2) before implant. Subconfluent human bone marrow MSCs (Promocell, Heidelberg, Germany) were washed and resuspended in hyaluronic acid/alginate mixture, as previously published28,29, to a concentration of 3.0 × 107 cells/ml. Prior to implant, rats were anesthetized with intraperitoneal injection of a solution of 70 mg/kg ketamine and 10 mg/kg xylazine. After shaving and disinfection of right hind leg, round 1.5 mm osteochondral defects were induced with a short drill in the patellar groove of the femur, in the midline of the femoral trochlea, until bleeding of the subchondral bone (approx. 2 mm). The NanoM1-BMP2 membrane was placed at the bottom of the defect, which was in turn filled with hMSCs/hydrogel mix and gelled via drop-wise addition of 102 mM calcium chloride (Sigma-Aldrich), over 5 min. These rats constituted experimental group 1 (ARTiCAR; n = 20 rats; 10 males and 10 females; 3.5 µl of hydrogel containing 105,000 ± 10% cells). Other rats were subject to the same procedure, but implanted with hydrogel only (as vehicle) and constituted group 2 (n = 20 rats; 10 males and 10 females; 3.5 µl of hydrogel). After gelation, the articulation capsule was closed, muscle and skin were sutured and the wound was thoroughly disinfected with povidone-iodine solution. After surgery, rats were kept under observation for post-anesthesia recovery. After recovery, 0.05–0.1 mg/kg buprenorphine was administered by subcutaneous injection. Animals were allowed unrestricted movement for the duration of the study (90 days). Rats were monitored daily for wound healing, leg mobility, morbidity, mortality and evident sign of toxicity.

Blood analysis

At day 7 post implant, five male and five female fasted rats/group (n = 20) were anesthetized with excess isoflurane and ventricular blood was collected either in EDTA-containing tubes or in heparin-containing tubes, for hematological or biochemical analysis, respectively. Between day 7 and 90, the remaining rats were observed and monitored twice a week for any loss of weight. Haematocrit, haemoglobin concentration, erythrocyte count, leukocyte counts, mean corpuscular volume and platelet count were determined in the blood samples on the day of collection by impedance variation and photometry (MINDRAY BC 2800 hematology analyzer, 4M, France). For biochemistry evaluation, plasma samples were prepared according to internal SOPs. Sodium, potassium, chloride, calcium, inorganic phosphate, glucose, urea, creatinine, total bilirubin, total cholesterol, triglycerides, aspartate aminotransferase (ASAT), alanine aminotransferase (ALAT), total proteins, albumin, and albumin/globulin ratio were quantified (Cobas Mira biochemistry analyzer, 4M, France).

Histopathology analysis of nude rats implants

Histopathology analysis was conducted on the fasted animals used for blood test. Briefly, a macroscopic autopsy was performed on freshly euthanized rats. Organs (treated knee, spleen, mesenteric lymph nodes, liver, lungs with bronchi and bronchiole, kidneys and heart) were macroscopically observed, explanted and collected. The right hind paw was sectioned at the epiphyses of both femur and tibia to recover knee joints subject to implant. Spleen, liver, kidneys and heart were weighed and preserved with the other organs at room temperature in 4% formalin (Sigma-Aldrich) until histological analyses. Organs were fixed in 4% paraformaldehyde, dehydrated, embedded in paraffin, sectioned and examined for histopathology.

Tissue harvesting for human DNA qPCR

Ninety days post implant, five male and five female rats/group (n = 20) were euthanized by exsanguination under anesthesia. After a macroscopic autopsy organs (ovaries with oviducts, testes, brain, treated knee, spleen, liver, kidneys, lungs, bone marrow, heart and the skin covering the treated knee joint) were weighed and collected for DNA extraction using the NucleoBond AXG100 kit (Macherey Nagel, Hoerdt, France), following manufacturer’s instructions. Briefly, all tissue samples except knee joints were homogenized in M-tubes (Miltenyi Biotec, Paris, France) containing buffer G2 on a GentleMACS Dissociator (Miltenyi Biotec, Paris, France). Knee joints were homogenized using Ultra-Turrax® dissociator instrument in buffer G2. Following extraction, the DNA pellet was dissolved in molecular biology grade water, and stored at 20 °C. Quantitative PCR (qPCR) with the iTaq Universal Probes Supermix (BioRad, Marnes-la-Coquette, France) was used to quantify human Alu sequences with the TaqMan AluYB8 Probe (Thermo Fischer Scientific). Genomic DNA from the different tissues of the implanted rats was amplified side-by-side with DNA from control rats, spiked-in with variable amount of DNA from U87-MG human cells (22, 7, 0.7 ng; 70, 7, 2.2, 0.7, 0.07 pg or no DNA), to build a standard curve. Samples were run in triplicate for 35 cycles on a CFX System (Bio-Rad). The limit of detection corresponded to the average signal from control rat DNA not spiked-in with human DNA.

Harvesting of MSCs from sheep bone marrow

Four weeks (28 ± 2 days) prior to interarticular surgery, adult sheep females (Rideau Arcott Hybrids strain) were subject to bone marrow aspiration procedure. Briefly, animals were placed in ventral recumbency and anesthetized with a mix of glycopyrrolate, xylazine and ketamine administered intramuscularly (IM). An IV catheter was placed in the appropriate vein. The larynx was sprayed with lidocaine and the animals were intubated with an appropriate sized cuffed orotracheal tube. If intubation was not possible under IM anesthesia, induction was performed using isoflurane in O2 (1–5%) or propofol intravenously. The sheep were then mechanically ventilated with isoflurane in O2. The harvest site was disinfected and a needle was introduced in the iliac crest. A sterile 10 mL syringe was filled with 1 mL of 5000 IU/mL heparin and filled with ~8 mL of bone marrow. The syringe containing bone marrow sample and heparin was sealed with an appropriate sterile cap for mesenchymal stem cells isolation, characterization and preparation for the surgical procedure.

Isolation and expansion of bone marrow-derived sheep MSCs

The MSCs harvested from sheep iliac crest were isolated according to their adherence to cell culture plastic. Bone marrow aspirates were first washed by addition of an equal volume of phosphate buffer saline (PBS; Sigma-Aldrich, France) and centrifuged at 220×g for 5 min. The cell pellets were suspended in Dulbecco’s Modified Eagle Medium (DMEM; Lonza, Germany) containing 10% heat-inactivated fetal bovine serum (Gibco, Thermo Fisher Scientific, France), 50 U/mL of penicillin (Lonza, Germany), 50 μg/mL of streptomycin (Lonza, Germany), 2.5 μg/mL Fungizone (Lonza, Germany), and seeded in a T75 culture flasks, under standard cell culture conditions. The following day, medium was discarded and attached cells were gently washed up several times with PBS to remove non-adherent cells. Flasks were then incubated for several days in DMEM, replaced every 72 h to promote emergence of colonies from adherent cells. When cells finally reached sub-confluence, they were sub-cultured until passage 2, when they were expanded for stemness characterization.

Characterization of sheep MSCs

The MSCs were characterized according to their ability to form colonies and to their multipotency. Colony-forming unit-fibroblast (CFU-F) assays were performed in triplicates (n = 3) with two different ranges of serial dilutions. After 14 days of culture, MSCs were rinsed with PBS, and fixed with 4% (w/v) paraformaldehyde. The colonies were stained using hematoxylin/eosin (Sigma-Aldrich, France) and counted. The potential of isolated sheep cells to undergo trilineage differentiation was demonstrated in vitro. For adipogenic differentiation, sheep cells were seeded at 2.1 × 104 cells/cm2 and cultured for 3 days with proliferation culture medium. After that, they were induced to differentiate by means of 3 alternate adipogenic induction/maintenance cycles. For each cycle, cells were cultured for 3 days in standard induction medium (Lonza, PT-3004), followed by 3 days of culture in standard maintenance medium (Lonza, PT-3004). After three cycles, cells were then cultured for 7 days with maintenance medium, replaced every 72 h. Negative controls for adipogenic differentiation were cultured in DMEM. After 28 days of culture, cells were rinsed with PBS, fixed in 4% (w/v) paraformaldehyde, rinsed with 60% (v/v) isopropanol and finally stained with 0.5% (w/v) Oil Red O solution to detect lipid vesicles. For chondrogenic differentiation, sheep cells were seeded as a pellet at the density of 2.5 × 104 cells and cultured in chondrogenic medium (Lonza, PT-3925) supplemented with 10 ng/mL of TGF-β3 growth factor (Peprotech, France). Negative control for chondrogenic differentiation were cultured in DMEM. After 28 days of culture, pellets were rinsed with PBS, fixed in 4% (w/v) paraformaldehyde, and embedded in paraffin. Paraffin sections were stained with Alcian Blue solution (Sigma-Aldrich, France) to visualize glycosaminoglycans and with Fast Red solution (Sigma-Aldrich, France) to visualize cells nuclei. For osteogenic differentiation, sheep cells were seeded at 2.1 × 104 cells/cm2 and cultured for 3 days in DMEM. After, they were cultured in osteogenic medium (Lonza, PT-3002). Negative controls for the osteogenic differentiation were cultured in DMEM. After 21 days of culture, cells were rinsed with PBS, fixed with ice-cold 70% (v/v) ethanol and stained with Alizarin Red solution (40 mM, pH 4.1; Sigma-Aldrich, France) to detect calcium deposits.

Immunophenotypic characterization of sheep MSCs

For cell surface markers analyses, cells were treated with 0.5% (w/v) bovine serum albumin and double-stained using monoclonal antibodies, including phycoerythrin (PE)-conjugated mouse anti-human CD34 (BD Biosciences, Le-Pont-de-Claix, France) with fluorescein isothiocyanate (FITC)-conjugated mouse anti-human CD90 (Beckman Coulter, Villepinte, France), PE-conjugated mouse anti-human CD166 (Beckman Coulter, Villepinte, France) with FITC-conjugated mouse anti-human CD45 (Dako, Glostrup, Denmark), PE-conjugated mouse anti-human CD105 (Beckman Coulter, Villepinte, France) with FITC-conjugated mouse anti-human CD44 (Beckman Coulter, Villepinte, France), and PE-conjugated mouse anti-human CD73 (BD Biosciences, Le-Pont-de-Claix, France) with FITC-conjugated mouse anti-human leukocyte antigen-antigen D related (HLA-DR) (Beckman Coulter, Villepinte, France). Antibodies were used at a dilution of 1:10 (PE CD34, FITC CD90, PE CD166, FITC CD44, PE CD105, FITC HLA-DR and PE CD73) or 1:20 (FITC CD45). Appropriate isotype-matched control antibodies named FITC or PE mouse IgG1 (Dako, Glostrup, Denmark) were used in each analysis. Cells were then examined by flow cytometry using a BD LSR II flow cytometer (Becton Dickinson Biosciences, San Jose, California). Fluorescence intensity and percentage of antigen positive cells were determined for each surface marker.

Induction of osteochondral defect in sheep

A total of 16 adult sheep underwent surgical induction of osteochondral defect into the medial femoral condyle. Three groups of sheep (ARTiCAR, AG control, NT control) were considered; each sheep was implanted on either the proximal or distal part of the right or left condyle of posterior legs (supplemental table 1). For surgery, the hind limb was flexed to a position at which the medial condyle could be palpated under the skin. A 15 cm medial parapatellar skin incision was performed. After blunt dissection of the subcutaneous tissues, the fascia overlying the vastus medialis muscle was incised just distal to the belly muscle with a small incision parallel to the muscle fibers and the vastus was retracted proximally. Blunt dissection was used to expose the periosteum down to the medial condyle of the femur. The joint capsule and periosteum were incised just proximal to the origin of the medial collateral ligament. Overlying soft tissues were removed from the bone only in the vicinity of the drill holes. Holes were predrilled using a 6-mm drill bit to a depth of 3 mm, except for the AG group, where the hole had a depth of 6 mm.

Intra-articular implant of ARTiCAR in sheep

Following the induction of the defect, the NanoM1-BMP2 was placed, and the defect was filled with MSCs/hydrogel mix (ARTiCAR combined ATMPs, n = 9). In the AG group (n = 10), a bone sample of 6 mm of diameter and 6 mm deep was taken out from the condyle and placed into the defect. In the NT group (n = 13) the defect was neither treated, nor filled. Up to 5 mL 0.25% bupivacaine were infiltrated into the surgical site to achieve local anesthesia and manage pain after surgery. The tissues were repositioned and closed layer-by-layer using appropriate sutures. Postoperative analgesia and antibiotic therapy were performed, 5 mg/kg Excede (IM) was administered during recovery from anesthesia, and 4 mg/kg Carprofen (IM) was administered 3 days after surgery.

Non-invasive monitoring of ARTiCAR via MRI

For the longitudinal analysis of the knee repair, sheep were examined three times via MRI, immediately after surgery, at 15 and 26 weeks, using a Magnetom Verio 3 T (Siemens). For the procedure, sheep were anesthetized with an intravenous injection of 0.05 mg/kg xylazine and 5 mg/Kg ketamine and placed in dorsal decubitus. A total of six sites of surgery were imaged for each group. Proton density-weighted, fat-saturated sagittal sections of the acquisitions were analyzed using the Osirix open-source software.

Three-dimensional micro-CT of explanted femoral condyles

For analysis of the bone mineralization, sheep were anesthetized, weighed and euthanized by a lethal injection of 540 mg/ml Euthanyl rapid IV bolus 26 weeks after surgery. Death was confirmed and recorded by observation of asystole or ventricular fibrillation, either on the electrocardiogram or by auscultation. Femoral condyle from were explanted from euthanized sheep and imaged via 3D micro-CT (Quantum Fx mCT, Julien Becker, ICS, IGBMC, Strasbourg, France). A total of six sites of surgery were imaged for each group. Three-dimensional surface rendering was obtained from micro-CT 2D images using the Osirix open-source software.

Histopathology analysis of implants in sheep

Treated femurs were removed from euthanized animals and subject to macroscopic inspection of the articular surface. The distal femoral epiphysis (with condyles) were individually identified and collected in 10% neutral buffered formalin, after macroscopic examination. Bone blocks were cut in two halves, by sawing in the middle of the sample along its longitudinal axis. Sections were cut through the defect along its deeper axis, from the bone surface to the end of the drill hole producing rectangular-shaped defect half sections. Full-thickness femoral bone-cartilage defect sites underwent undecalcified bone preparation and were infiltrated with methyl methacrylate and polymerized. A single 8 μm section spanning the entire width of the defect was cut along the parasagittal plane from each medial femoral condyle. The sections were stained with safranin o–fast green, for the staining of both cartilage and bone. The femoral defect sites were carefully evaluated and scored according to the ICRS histological score system39 (Supplemental table 2). Also, the tissue underneath and adjacent to the defect was evaluated for a number of parameters (Supplemental table 3) to assess safety and efficacy of the treatments.

Statistical analysis

Results from the WST1 assay were statistically evaluated using one-way ANOVA followed by Tukey post hoc test on Prism 4.03 (GraphPad). A p value ≤ 0.05 was considered significant. One-way ANOVA followed by Bonferroni post-hoc test was used to compare hematological and biochemical parameters in the blood tests, using Prism 4.03. A p value ≤ 0.05 was considered significant. Both SigmaPlot (SYSTAT Software) and Prism 5.0 (GraphPad) were used to compare the ICSR II scores from the in vivo experiments in sheep. Equal variance test and normality tests were performed. Either one- (differences induced by treatment) or two-way ANOVA (differences induced by both treatment and time) followed by Bonferroni post hoc test were used to assess significant differences among the continuous variables of the study groups. If either equal variance test or normality test failed, a Kruskal–Wallis one-way ANOVA with Dunn’s correction was conducted. A p value ≤ 0.1 was considered significant.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.