1. Introduction
Osteochondral tissue is formed by two main tissue types: the articular cartilage, which functions as a low-friction and wear-resistant surface, and the subchondral bone, which plays a crucial mechanically supportive role [
1]. Without effective and timely interventions, damage to this osteochondral unit, caused by traumatic injury or disease, may progress to osteoarthritis [
2,
3]. Bi-layered biomaterial scaffolds have been developed to restore the structural and physiological properties of the entire osteochondral unit and, thus, to support chondrogenesis and osteogenesis simultaneously [
4,
5]. As a biologically derived protein, collagen is an efficient biomaterial to support cellular activities and promote osteochondral repair [
6,
7,
8]. The addition of hydroxyapatite (HAp) can improve the osteogenic potential of a collagen-based scaffold in vivo, and magnesium ions (Mg
2+) induce osteogenic differentiation and osteoblast differentiation [
9]. Preclinically, bi-layered collagen/collagen-magnesium-HAp (Col/Col-Mg-HAp) scaffolds have successfully reconstructed the articular cartilage and the subchondral bone in animal models [
10,
11,
12]. Clinical cohort studies also demonstrated the excellent stability of this scaffold and clinical improvement in knee function [
13,
14,
15]. However, subchondral bone repair remained suboptimal in some clinical follow-ups in comparison to the cartilage repair capacity of this scaffold [
15], which may lead to altered biomechanical properties of the osteochondral unit and thereby affect the long-term survival of the neo-cartilage [
16]. This might subsequently lead to renewed osteochondral damage and joint disease [
17,
18].
Incorporating factors that stimulate bone formation could be a promising approach to overcome this limitation in bone regeneration capacity [
19]. Bone morphogenetic protein 2 (BMP-2) has a vital role in osteogenesis and osteoclastogenesis [
20,
21] and is approved by the Food and Drug Administration (FDA) as an osteogenic protein. The recruitment of stem cells or osteoblasts is necessary for osteogenic initiation, and BMP-2 was reported to facilitate cell ingrowth [
22]. Platelet-derived growth factor (PDGF) is potent in stimulating cell ingrowth, angiogenesis, and osteogenesis [
23,
24,
25,
26]. The delivery of BMP-2 and PDGF onto biomaterials has been shown to provide an improvement in osteoblast function and bone integration [
27,
28,
29]. Therefore, it might be promising to adsorb BMP-2 and PDGF-BB onto this Col/Col-Mg-HAp scaffold to improve bone healing.
This study aimed to evaluate the osteogenic effectiveness of BMP-2 or PDGF-BB adsorption onto a Col/Col-Mg-HAp scaffold on bone repair in osteochondral defects. We first assessed the release profiles of BMP-2 and PDGF-BB from the two layers of a Col/Col-Mg-HAp scaffold in vitro. Next, an ex vivo osteochondral culture model was used to investigate the added effect of growth factors on cell ingrowth from adjacent tissues. Then, we investigated the effect of BMP-2 or PDGF-BB incorporated into a Col/Col-Mg-HAp scaffold in an in vivo semi-orthotopic mouse model for the early phases of tissue repair. Finally, the effect of BMP-2 incorporated into a Col/Col-Mg-HAp scaffold was investigated in both weight-bearing and non-weight-bearing locations of the knee joint in an established preclinical caprine osteochondral defect model.
2. Materials and Methods
2.1. Scaffold Fabrication and Characterization
Col/Col-Mg-HAp scaffold is a biomimetic scaffold that has a porous, 3-dimensional composite structure. The scaffold is composed of two layers: the cartilaginous layer, consisting of Type I collagen, to maintain joint congruence and the bone layer consisting of a combination of Type I collagen (60%) and magnesium-hydroxyapatite (40%). Each layer of the scaffold is synthesised separately by a standardised process from an atelocollagen aqueous solution (1% w/w) in acetic acid, isolated from equine tendon. The upper non-mineralised chondral layer of the scaffold is obtained by dissolving an acetic solution of Type I collagen in bi-distilled water by adding NaOH. The bone layer of the scaffold is obtained by nucleating nanostructured hydroxyapatite into self-assembling collagen fibres, as occurs in the natural biological neo-ossification process. To stabilise the scaffold, the fibrous structures were chemically cross-linked for 16 hours at room temperature. After chemical cross-linking, the two layers were superimposed and afterwards they are freeze-dried.
The morphology of the scaffold was evaluated by Scanning Electron Microscopy (SEM) performed on a SEM-LEO 438 VP (Carl Zeiss AG, Oberkochen, Germany). The samples were sputter coated with gold prior to examination. The mineral content of the bone layer was evaluated by thermogravimetric analysis (TGA), performed in alumina crucibles in an air atmosphere with a flow rate of 80 mL/min, between 25 and 700 °C (Mettler Toledo DT-TGA/DSC1 Star System, Columbus, OH, USA). The elemental composition of the mineral phase (magnesium-hydroxyapatite) was determined using inductively coupled plasma-optical emission spectrometry (ICP-OES, Thermo Scientific iCAP 7400, Waltham, MA, USA). In particular, the composition is expressed as Ca/P, (Ca+Mg)/P, Mg/Ca% molar ratios. The bone layer was dissolved in hot nitric acid (65
v/v%) in order to completely destroy the collagen matrix and solubilise the inorganic phase. The content (ppm) of magnesium, calcium, and phosphorous in the samples is determined by comparison with a predetermined standard curve: Ca/P = 1.5 ± 0.1%; (Ca+Mg)/P = 1.6 ± 0.1%; Mg/Ca% = 1.5 ± 0.4%. The total porosity of the osteochondral scaffold was determined using Archimedes’ principle. The exterior volume (V
s) of the sample was measured using a Vernier calliper. The sample was then immersed in a pycnometer containing 96% ethanol solution. The actual volume (V
a) of the sample is calculated using the formula:
W
w is the weight of the ethanol and the pycnometer; W
o is the dry weight of the pycnometer; W
t is the combined weight of the ethanol, the pycnometer, and the plug sample; W
p is the combined weight of the dry pycnometer and dry plug sample; and 0.789 g/cm
3 is the density of ethanol solution. The porosity of the scaffold was then determined using the following formula:
2.2. BMP-2 and PDGF-BB Release from the Different Layers of a Col/Col-Mg-HAp Scaffold
To investigate the release kinetics of BMP-2 and PDGF-BB from the different layers of the Col/Col-Mg-HAp scaffold, a time course study was performed (
Figure S1A). A quantity of 35 μL (28.5 μg/mL) BMP-2 or PDGF-BB (Sigma, Saint Louis, MI, USA) was absorbed into the separated layers (either collagen-only layer or Col-Mg-HAp layer) of a Col/Col-Mg-HAp scaffold (Osteochondral scaffold, Finceramica, Italy) in a low-affinity binding plate at 37 degrees Celsius for 30 min. The concentration of growth factors was determined according to our previous studies [
30,
31]. After absorption, the medium was harvested from the plate, and the scaffolds were transferred to a new low-affinity binding plate. A quantity of 800 μL alpha-Minimum Essential Medium (α-MEM, Gibco, Waltham, MA, USA) was added to each scaffold-containing well. At each time point (6, 24, 48, 72, 96, 120, 144, 168 h, 336 h only for BMP-2), the medium was collected and replaced by fresh medium. The collected medium was analysed for BMP-2 by recombinant human BMP-2 (Peprotech, Cranbury, NJ, USA) or PDGF-BB by a recombinant human PDGF-BB DuoSet ELISA kit (R&D Systems, McKinley Place N.E., Minneapolis, MN, USA) according to the manufacturer’s instructions.
2.3. Cell Recruitment Capacity of BMP-2 or PDGF-BB in an Ex Vivo Osteochondral Defect Culture Model
To study the effect of BMP-2 and PDGF-BB adsorbed onto a bi-layered Col/Col-Mg-HAp scaffold on cell recruitment capacity, an ex vivo osteochondral defect culture model, previously developed and validated in our laboratory, was used [
32] (
Figure S1B). Briefly, osteochondral defects were created in bovine osteochondral biopsies (8 mm diameter, 5 mm height) harvested from metacarpal-phalangeal joints of 6- to 8-month-old calves (LifeTec, Eindhoven, The Netherlands), in which a 4 mm wide and 4 mm high defect was created. The osteochondral plugs were kept overnight in Dulbecco’s Modified Eagle Medium high glucose (DMEM, 4.5 g/L glucose, Gibco, Waltham, MA, USA) supplemented with 10% fetal bovine serum (FBS, Gibco, Waltham, MA, USA), 50 μg/mL gentamycin (Gibco, Waltham, MA, USA), and 1.5 μg/mL fungizone (Gibco, Waltham, MA, USA). The following day, the Col/Col-Mg-HAp scaffolds (diameter: 4 mm, height: 4 mm) with or without adsorbed growth factors (
n = 4 for each condition) were inserted into the osteochondral defects. Either 4 μg (57.1 μg/mL) BMP-2 solution or 100 ng (1.4 μg/mL) PDGF-BB was adsorbed onto the scaffold. Each construct was cultured in 3.5 mL medium in a new 12-well plate at 37 °C and 5% CO
2. The medium was refreshed on the first day and subsequently every two days. After 3 weeks, the osteochondral constructs were harvested and fixed in 4% formalin for 1 week with subsequent further histological analysis.
2.4. In Vivo Osteochondral Defect Model in Mice
To assess the effect of BMP-2 and PDGF-BB in the Col/Col-Mg-HAp scaffolds on bone repair, an established in vivo subcutaneous mouse model developed previously in our laboratory was used (
Figure S1C) [
33]. Osteochondral biopsies were harvested, and the defects were created as described previously (see under
Section 2.3) and kept overnight in α-MEM supplemented with 10% FBS, 50 μg/mL gentamycin, and 1.5 μg/mL fungizone until implantation. Eleven 12-week-old NMRI-Fox1nu mice (Taconic, Albany, NY, USA) were used for this study. The animals were randomly assigned and housed under specific-pathogen-free conditions with a regular day/night light cycle and allowed to adapt to the conditions of the animal facility for 7 days. Food and water were available ad libitum. Before implantation, 70 µL of saline solution, or 70 µL of saline solution containing BMP-2 (57.1 μg/mL, 4 μg) or PDGF-BB (28.5 μg/mL, 2 μg or 1.4 μg/mL, 100 ng) was added dropwise onto the Col/Col-Mg-HAp scaffolds. All osteochondral plugs were covered with a circular 8 mm Neuro-Patch membrane (Braun, Melsungen, Germany) to prevent the ingrowth of host cells from the top. The osteochondral plugs were randomly implanted in subcutaneous pockets on the back of the mice under 2.5–3% isoflurane anaesthesia (1000 mg/g, Laboratorios Karizoo, Maharashtra, India). One osteochondral plug was implanted per pocket, and four osteochondral plugs were implanted per mouse. The incisions were closed with staples (Fine Science Tools, Vancouver, BC, Canada). At 1 h before surgery and at 6–8 h after surgery, 0.05 mg/kg body weight of buprenorphine (Chr. Olesen & Co, Gentofte, Copenhagen, Denmark) was injected subcutaneously to ensure pre- and postoperative analgesia. Mice received a subcutaneous prophylactic antibiotic injection of 25 mg/kg body weight of Amoxicillin (Dopharma, Raamsdonksveer, The Netherlands).
After 4 or 8 weeks, mice were euthanised by cervical dislocation under 2.5–3% isoflurane anaesthesia, and the osteochondral plugs were harvested. All samples were fixed in 4% formalin for 1 week for further analysis. This animal experiment was approved by the Ethics Committee for Laboratory Animal Use (AVD101002016991; protocol #EMC 16-691-05).
2.5. In Vivo Osteochondral Defect Caprine Model
A validated bilateral osteochondral defect caprine model was used to assess the osteochondral defect repair capacity of BMP-2-supplemented scaffolds in a preclinical large animal model (
Figure S1D). An experimental unit of 11 skeletally mature female Saanen goats (age: 3 years, weight: 37.9 ± 7.3 kg) was subjected to a bilateral arthrotomy under general anaesthesia as described before [
34,
35,
36]. In short: all animals received a prophylactic antibiotic injection with amoxycillin clavulanic acid 8.75 mg/kg intramuscular (Noroclav, Norbrook, Ireland) and were intravenously sedated with butorphanol (0.2 mg/kg, Butador, Chanelle Pharma, Ireland) and diazepam (0.2 mg/kg, Diazemuls; Accord Healthcare, UK). A lumbosacral epidural block with lidocaine (2 mg/kg, Lidocaine HCl 2%, B. Braun Medical Inc., EU, Melsungen, Germany) and morphine (0.2 mg/kg, Morphine Sulphate 10 mg/mL, Kalceks, Latvia) was performed with the animal in sternal recumbency. Anaesthesia was induced with propofol IV to effect (max. 6 mg/kg, Propofol-Lipuro 1%, B. Braun Medical Inc., Melsungen, German) and was maintained with isoflurane (Vetflurane, Virbac Animal Health, Suffolk, UK) in 100% oxygen via a circle rebreathing system. All animals received analgesia with meloxicam IV (0.5 mg/kg, Rheumocam, Chanelle, Galway, Ireland); and morphine IV (0.2 mg/kg, Morphine sulphate, Mercury Pharmaceuticals, Dublin, Ireland) 90 min after the epidural block.
An arthrotomy of each stifle joint was performed in dorsal recumbency using a lateral parapatellar approach. Under constant irrigation with saline, a pointed 6 mm drill bit was used to drill an approximate 3–4 mm deep non-weight-bearing defect in the transition of the distal 1/3 to the middle 1/3 of the trochlear groove and in the weight-bearing part of the medial femoral condyle. Subsequently, a custom-made flattened drill bit and a depth guide were used to create an exact flat 6 mm deep by 6 mm wide circular critical-sized osteochondral defect in a non-weight-bearing and a weight-bearing location. The joint was flushed with saline to remove any debris, and the defects were press fit with a similar-sized selected scaffold before surgical closure as described before. Each stifle joint was randomly assigned to one of the two treatment groups (
Figure S1D): (1) Col/Col-Mg-HAp scaffold-only (6 mm diameter, 6 mm height, Osteochondral scaffold, Finceramica, Italy), and (2) Col/Col-Mg-HAp scaffold adsorbed with BMP-2 (57.1 μg/mL).
Following surgery, postoperative analgesia was provided (meloxicam 5 days) and goats were housed in indoor pens for daily postoperative welfare monitoring and scoring. Two weeks postoperatively, following the removal of skin sutures, animals were released to pasture or loose housing (weather dependent) for the remainder of the study period with daily health checks. An orthopaedic assessment (
Table S1) was performed on the day of humane euthanasia under sedation with a barbiturate overdose at the predetermined endpoint at 6 months after surgery. Subsequently, all the joints, surrounding joint tissues, and synovial fluids were scored (
Table S2), dissected, and photographed (Body Canon EOS R5, lens: Canon EF 100 mm f/2.8 L Macro IS USM, flash: Macro Ring lite MR-14EX II). Biopsies 1 cm by 1 cm square containing the entire defects were harvested with an oscillating saw.
Ethical evaluation and approval were provided by the Health Products Regulatory Authority of Ireland (AE1898217/P032), the Animal Research Ethics Committee of University College Dublin (AREC-P-12–71) and the Lyons Animal Welfare Board (Health, Husbandry and Monitoring plans).
2.6. Macroscopic Assessment of the Defect Repair in the Caprine Model
The quality of the cartilage repair in the caprine samples was assessed semi-quantitatively using the International Cartilage Repair Society (ICRS) macroscopic evaluation system (
Table S3) [
37] and a macroscopic scoring system (
Table S4) developed by Goebel et al. [
38]. The ICRS scoring system evaluates the macroscopic appearance of cartilage repair tissue as Grade IV (severely abnormal), Grade III (abnormal), Grade II (nearly normal), or Grade I (normal). The Goebel Score describes macroscopic articular cartilage repair with five major evaluation categories. The quality of defect repair was scored blinded on fresh samples by two independent assessors, and the scores were averaged for further analysis. All the samples were fixed in 4% formalin for 10 days after macroscopic assessment for further analysis.
2.7. Micro-Computed Tomography
From the mouse model, the retrieved bovine osteochondral plugs were scanned (Quantum GX, Perkin Elmer, Akron, OH, USA) with the following settings after fixation in 4% formalin: energy 90 KV, intensity 88 μA, 18 mm FOV, 36 μm isotropic voxel size. All the scans above were under an X-ray filter of Cu (thickness = 0.06 mm) and Al (thickness = 0.5 mm) and were calibrated using a phantom with a known density of 0.75 g/cm3, which was additionally scanned before and after each scan. A high-resolution mode was set, and a scan time of 4 min was used.
The caprine samples were scanned with the same settings except for 36 mm FOV, 72 μm isotropic voxel size. Image processing included modest Gauss filtering (sigma = 0.8 voxel, width = 1 voxel) and segmentation using a single threshold. A cylindrical region (4 mm diameter and 5 mm height) within the original defect (6 mm diameter and 6 mm height) was selected as a volume of interest (VOI) for the caprine samples. In this VOI, the following morphometric parameters were measured: bone volume per total volume (BV/TV), trabecular thickness (Tb.Th), trabecular number (Tb.N), and trabecular separation (TB.Sp). Morphological analyses were performed using IPL (Scanco Medical AG, Bruettisellen, Switzerland).
2.8. Histology and Immunohistochemistry
The bovine osteochondral plugs cultured ex vivo were decalcified for 2 weeks using 10% formic acid (Sigma, Saint Louis, MI, USA). After micro-CT scanning, the bovine osteochondral plugs harvested from mice were decalcified using 10% ethylenediaminetetraacetic acid (EDTA, Sigma, Saint Louis, MI, USA) for 4 weeks. The caprine samples were decalcified for 3 weeks using 10% formic acid. Subsequently, all samples were embedded in paraffin and sectioned at 6 µm. Following dewaxing, H&E staining was performed with Hematoxylin (Sigma, Saint Louis, MI, USA) and Eosin Y (Merck, Kenilworth, NJ, USA) to study general cell and tissue morphology. To visualise glycosaminoglycans in the extracellular matrix (ECM), dewaxed sections were stained with Safranin O (Fluka, Buchs, Switzerland) and Light green (Fluka, Buchs, Switzerland). To study the regenerated tissue type in the osteochondral defects, RGB staining was performed (Proteoglycans/hyaline cartilage appears blue, mineralised cartilage matrix appears pink/greenish, collagen fibres/uncalcified bone appears red, and mineralised bone appears green) using Alcian Blue (Sigma, Saint Louis, MI, USA), Fast Green (Sigma, Saint Louis, MI, USA), and Picrosirius Red (Sigma, Saint Louis, MI, USA) [
39]. The cell number in the scaffolds was counted under microscopy. NDP View2 software (version 2.8.24, 2020 Hamamatsu Photonics K.K.) was used to measure the tissue volume in the defect at three sections that were taken at the middle, 0.5 mm, and 1 mm further away for bovine samples or at the middle for caprine samples (
Figure S2). The percentage of the defect covered with newly formed osteochondral tissue (100% indicated that the defect was fully filled with newly formed tissue) was calculated (
Figure S3). All slides were independently scored by two investigators blinded to the experimental condition. The measurements of the two investigators were averaged for each section.
To investigate the presentation of neutrophils in the defect, immunohistochemistry for myeloperoxidase (MPO) was performed on retrieved bovine osteochondral samples from the mouse study. After dewaxing, antigen retrieval was performed by placing the slides with Tris/EDTA (pH9) in a water bath at 95 °C for 20 min. Then, the slides were pre-incubated with 10% normal rabbit serum (NRS, Invitrogen, Waltham, MA, USA) in PBS containing 1% bovine serum albumin (BSA, Sigma, Saint Louis, MI, USA) and 1% milk powder (ELK, Campina, Amersfoort, The Netherlands). The slides were incubated by the first antibody against MPO (Thermo Scientific, Waltham, MA, USA, 1:200 dilution) or rabbit IgG antibody (DakoCytomation, California, USA, 1:10,000 dilution) as the negative control in PBS containing 1% BSA for 1 h. Next, the slides were incubated by biotinylated goat α-rabbit (Biogenex, Fremont, CA, USA, 1:50 dilution in PBS containing 1% BSA and 5% mouse serum of total volume) for 30 min. Then, the reaction was amplified by streptavidin-labelled alkaline phosphatase (Biogenex, Fremont, CA, USA) diluted 1:50 in PBS containing 1% BSA and visualized by subsequent incubation of Neu Fuchsin substrate. Slides were counterstained with Hematoxylin.
To evaluate the infiltration of macrophages in the defect, immunohistochemistry for F4/80 was performed on the bovine osteochondral plugs retrieved from the in vivo mouse study. For antigen retrieval, each dewaxed slide was treated with 300 μL proteinase K (20 µg/mL, Thermo Scientific, Waltham, MA, USA) solution and incubated at 37 °C for 30 min. Then, the slides were pre-incubated with 10% NRS in PBS containing 1% BSA and 1% milk powder. The following steps were similar to the immunohistochemistry for F4/80, with the first antibody against F4/80 (eBioscience, San Diego, CA, USA, 1 μg/mL) or rat IgG2a (eBioscience, San Diego, CA, USA, 1 μg/mL) as the negative control in PBS containing 1% BSA; the second antibody: biotinylated rabbit anti-rat IgG (6 μg/mL in PBS containing 1% BSA and 5% mouse serum of total volume), and third antibody: streptavidin-labelled alkaline phosphatase (Biogenex, Fremont, CA, USA) diluted 1:50 in PBS containing 1% BSA. To distinguish between pro-inflammatory (M1) and anti-inflammatory/tissue-repair (M2) macrophages, immunohistochemistry for inducible Nitric Oxide Synthase (iNOS, as an indicator for pro-inflammatory M1 macrophages) was performed. The steps were similar to the immunohistochemistry for MPO, with the first antibody against iNOS (2 μg/mL, Abcam, Cambridge, UK).
The slides were ranked according to the positive degree of immunohistochemical staining, and all negatively stained sections were ranked 0. Only areas that were also stained for F4/80 were taken into account when the iNOS staining was ranked.
2.9. Statistical Analysis
All statistical tests were performed using SPSS software 28.0 (SPSS Inc., Chicago, IL, USA). The repair tissue volume was expressed as mean ± standard deviation (SD). The rankings of immunohistochemical MPO, F4/80, and iNOS staining were presented as column plots in graphs. Multiple comparisons between scaffold-only, BMP-2, and PDGF groups in bovine osteochondral plug samples were analysed by a Kruskal–Wallis test. Statistically significant differences between scaffold-only and scaffold + BMP-2 groups or between trochlear groove and femoral condyle groups in caprine samples were determined by a Mann–Whitney U test. A p value ≤ 0.05 was considered statistically significant.
4. Discussion
In this study, we evaluated the effectiveness of growth factor adsorption onto a bi-layered Col/Col-Mg-HAp scaffold in osteochondral defect repair. In vitro release results showed that the Col-Mg-HAp (bone) layer retained more growth factor than the collagen-only (cartilage) layer and that BMP-2 was retained much better than PDGF-BB. In an ex vivo osteochondral defect model, cell ingrowth into the scaffold was enhanced by BMP-2 and by PDFG-BB. In a semi-orthotopic non-weight-bearing osteochondral defect mouse model representing the early phase (four weeks) of defect repair in vivo, the addition of growth factors resulted in fewer pro-inflammatory cells and better tissue repair, with BMP-2 showing the most favourable results. Therefore, BMP-2 addition was taken forward for testing in an established preclinical large animal osteochondral defect model to study scaffold enhancement by BMP-2 in a physiological environment with different loading conditions. After six months in goats, both the scaffold-only and the BMP-2-adsorbed scaffold induced good osteochondral defect healing. Surprisingly, the addition of BMP-2 led to worse bone repair in the weight-bearing femoral condyle osteochondral defects, whereas this negative effect of BMP-2 was not seen in the non-weight-bearing trochlear groove osteochondral defect location.
Our approach was based on the sequential use of osteochondral defect models: the first experiments were performed in an ex vivo bovine osteochondral explant model, followed by a semi-orthotopic non-weight-bearing model where we implanted bovine osteochondral explants subcutaneously in mice. This took into account the effects of the innate immune system and blood vessel invasion. We then selected the most promising condition to be further evaluated in a preclinical caprine osteochondral defect model, where we tested both weight-bearing and non-weight-bearing repair conditions simultaneously. All models were developed and used previously to evaluate new osteochondral repair approaches [
34,
40]. In vivo animal models can closely resemble the human osteochondral microenvironment in the context of the presence of immune cells and tissue repair factors. The mouse model allows the screening of four conditions in one animal, whereas the caprine bilateral osteochondral defect model is a fully immune-competent model using outbred animals. That model also allows within-animal controls (comparing left and right knees), and provides the opportunity to assess tissue regeneration in both weight-bearing and non-weight-bearing locations within the same joint. These animal models offered the opportunity to investigate the possible effect of incorporating growth factors into a Col/Col-Mg-HAp scaffold for osteochondral repair towards translation into the human patient.
Loading can have a significant effect on tissue repair [
41,
42]. Not much is known about the interaction between mechanical loading and growth factors and their effect upon osteochondral tissue repair [
43,
44]. To study the effect of mechanical stimuli on osteochondral repair during normal ambulation, we used both a non-weight-bearing location (a distal region in the trochlear groove) and a weight-bearing location (a central region on the medial femoral condyle) in the bilateral stifle caprine osteochondral defect model [
45,
46]. Better subchondral bone repair was observed in the non-weight-bearing trochlear groove defects than in the weight-bearing femoral condyle defects when BMP-2 was adsorbed onto the scaffold, while no significant difference in repair was found between locations when the scaffold without additions was evaluated. In the non-weight-bearing trochlear groove defects, there was a trend towards more bone-like tissue being generated in the BMP-2-adsorbed scaffolds compared with the scaffold-only. However, this difference did not reach statistical significance, which might be due to the excellent repair capacity of the scaffold-only. The weight-bearing femoral condyle defects implanted with BMP-2-adsorbed scaffolds appeared to have less bone repair and more fibrous tissues in the bone defect compared with the scaffold-only.
Notably, there was no mechanical loading in the semi-orthotopic mouse model and at the four-week time point, the addition of BMP-2 to the scaffolds seemed beneficial. Our results reveal the interesting hypothesis that, with a specific dose of BMP-2, the beneficial effect of BMP-2 on bone repair is apparent in non-weight-bearing conditions, whilst BMP-2 addition can be detrimental to bone repair in weight-bearing conditions. Although the effect of BMP-2 on osteochondral defect repair in different loading environments is still unknown, the action of BMP-2 on tibia fractures or femoral bone defect healing was demonstrated to be dependent on the mechanical environment in vivo [
47,
48]. This might be due to an interaction between BMP-2 and mechanical loading [
49] and related to the dosage used. Compression and loading affect BMP signalling both immediately and in a long-term manner [
50]. Mechanical loading was shown to increase BMP-2 expression [
51], and the effect of BMP-2 can be strongly potentiated by mechanical forces [
49]. It is possible that the combination of added and locally produced BMP-2 might lead to an overstimulation of BMP-2 signalling, which is shown to cause inflammation, bone resorption, and fibrotic tissue formation [
52,
53,
54]. No previous study has reported this potential side effect nor the relationship with mechanical loading. However, the mechanical loading patterns in long bone defects with fixation and osteochondral defects with scaffolds in the femoral condyle (especially with the presence of synovial fluid) are quite distinct from one another. Therefore, further studies are needed to elucidate if and at which step of the BMP signalling cascade the pathway is modulated and by which type of mechanical stimuli.
BMP-2 is, to date, the only FDA-approved and commercially available osteoinductive growth factor used in clinics. The function and application of BMP-2 in promoting bone regeneration and bone remodelling has been widely investigated preclinically and clinically [
55,
56,
57]. Mg
2+ was incorporated in this scaffold, which might upregulate the bioactivity of BMP-2 upon calcium phosphate cement via enhanced BMP receptor recognition [
58]. Previous studies have demonstrated that BMP-2 accelerated the migration of bone marrow mesenchymal stromal cells (MSCs) in vitro and in vivo [
59,
60]. In the presented study, we also observe the promotion of cell ingrowth by BMP-2 addition in our ex vivo culture model. Although the primary expected functions of BMP-2 in promoting bone repair are to enhance MSC migration to the sites and differentiation into osteoblasts and to enhance the osteogenic capacity of osteoblasts, BMP-2-induced osteogenesis may also involve an immunoregulatory role [
61,
62]. Macrophages act as immune cells and osteoclast precursors and are involved in multiple stages of bone healing [
63]. BMP-2 might diminish the expression of pro-inflammatory phenotypic markers and promote the macrophage transition towards a more tissue repair-like phenotype [
62]. Neutrophils and M1 macrophages participate in tissue repair as effector cells in inducing inflammation, and M2 macrophages are involved in the resolution of inflammation, promoting angiogenesis, and matrix remodelling [
64]. In our semi-orthotopic model, fewer M1 macrophages, as well as neutrophils, were found in the BMP-2 condition than in the scaffold-only condition after four weeks. Therefore, the regulatory effect of BMP-2 in a local osteoimmune environment might be one of the potential mechanisms for promoting bone healing in the early phase.
Although BMP-2 has been clinically applied because of its osteogenic effect, it is still not widely used due to the adverse events associated with implanted supraphysiological high doses [
57]. The most documented side effect is ectopic ossification. Aulin et al. demonstrated that intra-articular injection of a hyaluronan hydrogel containing BMP-2 (150 μg/mL) resulted in excessive ectopic bone formation on the knee joint surface of rabbits (6–7-month-old females) [
65]. In our study, we used 57.1 μg/mL BMP-2 for both 4 mm × 4 mm and 6 mm × 6 mm cylindrical osteochondral defects according to our previous in vivo results [
30]. HAp was reported to have a high affinity for BMP-2 due to the large surface area and functional groups [
29,
66]. Three types of functional groups, −OH, −NH
2, and −COO, allow BMP-2 to adsorb on the HAp surface [
66]. This is not the case for all growth factors, where, for example, no functional groups or only one might be present. In vitro, we demonstrated that BMP-2 was bound and largely retained in both layers, especially in the Col-Mg-HAp layer. In this way, a sustained release was expected, thereby reducing the adverse effects of BMP-2. Over longer periods, however, correlations between the in vitro and in vivo settings become ever more unreliable since the release in vivo will be determined by the degradation of the scaffold as well. In fact, no BMP-2-related ectopic ossification or abnormal inflammation was observed in the knee joints of our experimental goats macroscopically or on micro-CT. Similarly, no ectopic ossification was reported in a publication that investigated the addition of 625 μg/mL of BMP-2 (adsorbed onto Col/HAp scaffolds) in skeletally mature male rabbits [
29].
Consistent with other studies, we found that PDGF-BB is an effective chemoattractant of cells ex vivo [
31,
67]. PDGF-BB was rapidly released from the collagen-only layers within six hours; most of the PDGF-BB was retained in the Col-Mg-HAp layers in vitro. This might be due to the capacity of HAp in adsorbing a large quantity of proteins and drugs. At the early repair phase in the in vivo mouse model, both dosages (1 μg and 100 ng) of PDGF-BB slightly improved tissue formation. Most of the regenerated tissue was in the bone defect area, which might be related to the sustained release from the Col-Mg-HAp layers, which is aimed at the repair of the bone defect specifically. Lee et al. reported that labelled cells migrated towards the osteochondral defect when defects were treated with PDGF-AA or PDGF-BB-loaded heparin-conjugated fibrin [
67]. Overall, the chemotactic ability of PDGF-BB might be one of the mechanisms in inducing tissue repair in the early phase. However, after eight weeks, the defects loaded with PDGF-BB generated slightly less osteochondral tissue compared with scaffold-only. A previous study on osteochondral repair also showed that the addition of 1 μg/mL PDGF-BB worsened the cartilage repair in an in vivo subcutaneous mouse model, although cell recruitment was enhanced in vitro. The short half-life of PDGF might be one of the reasons for the different results between in vitro and in vivo studies. Zhang et al. demonstrated that PDGF-BB overexpression improved the osteogenic and angiogenic abilities of MSCs in a critical-sized rat calvaria defect model [
68].
We also expected the combination of BMP-2 and PDGF-BB to further improve bone repair, since PDGF was reported to modulate BMP-2-induced osteogenesis in periosteal progenitor cells [
69]. However, no significant improvement was found after eight weeks in the semi-orthotopic mouse model in vivo. Therefore, PDGF-BB and the combination of BMP-2 and PDGF-BB were not further evaluated in our large animal osteochondral defect model.
Suboptimal subchondral bone repair of osteochondral defects might lead to damage of the renewed overlying cartilage in the long term [
16,
17,
18]. In our caprine large animal model, a blinded macroscopic evaluation of repaired cartilage tissue indicated the presence of a smooth white cartilaginous layer that was continuous with surrounding naive cartilage, both with control scaffolds and BMP-2-enhanced scaffolds. Histology further confirmed the presence of hyaline cartilage in the superficial layer, with no significant improvement or deterioration observed when BMP-2 was added. This might be due to the fact that tissue repair was assessed at only one, relatively early, six-month time point in the caprine model. A longer 12- or 24-month study might be useful to confirm the potential effectiveness of enhanced bone repair on the long-term survival of the neo-cartilage.