The Complexity of Joint Regeneration: How an Advanced Implant Could Fail by Its In Vivo Proven Bone Component

Articular cartilage damage is a major challenge in healthcare due to the lack of long-term repair options. There are several promising regenerative implant-based approaches for the treatment, but the ﬁxation of the implant remains a signiﬁcant challenge. This study evalu-atedthepotentialforrepairofanosteochondralimplantproducedthroughanovelcombined bioprinting-basedchondral-boneintegration, withandwithoutcells, inanequinemodel. Implants consisted of a melt electrowritten polycaprolactone (PCL) framework for the chondral compartment, which was ﬁrmly integrated with a bone anchor. The bone anchor was pro-ducedbyextrusion-basedprintingofalow-temperaturesettingbioceramicmaterialthathad been proven to be effective for osteo-regeneration in an orthotopic, non-load bearing and non-articular site in the same species in an earlier in vivo study. Articular cartilage-derived progenitor cells were seeded into the PCL framework and cultured for 28 days in vitro in the presence of bone morphogenetic protein-9 (BMP-9), resulting in the formation of abundant extracellular matrix rich in glycosaminoglycans (GAGs) and type II collagen. The constructs were implanted in the stiﬂe joints of Shetland ponies with cell-free scaffolds as controls. Clinical signs were monitored, and progression of healing was observed non-invasively through radiographic examinations and quantitative gait analysis. Biochemical and histological analyses 6 months after implantation revealed minimal deposition of GAGs and type II collagen in the chondral compartment of the defect site for both types of implants. Quantitative microcomputed tomography showed collapse of the bone anchor with low volume of mineralized neo-bone formation in both groups. Histology conﬁrmed that the PCL framework within the chondral compartment was still present. It was concluded that the collapse of the osteal anchor, resulting in loss of the mechanical support of the chondral compartment, strongly af-fected overall outcome, precluding evaluation of the inﬂuence of BMP-9 stimulated cells on in vivo cartilage regeneration. process, had a much bigger inﬂuence than in previous studies. When combined with the brittle nature of the material, the implant was unable to withstand the complex loading of the knee joint. The current study aimed at evaluating a cell-laden and a cell-free version of an osteochondral composite scaﬀold for cartilage repair that consisted of an in vivo proven osteogenic bone scaﬀold for the osteal compartment and made use of a novel interface for the connection of the chondral and osteal compartments.


Introduction
Focal cartilage damage is a major challenge in human healthcare since it leads to an increased risk of developing early osteoarthritis (Kloppenburg & Berenbaum, 2020). Most of the available repair approaches are palliative with limited alleviation time, generating fibrous tissue with reduced mechanical strength (Kwon et al., 2019). There are no effective treatments that can fully restore the anatomical structure and function of focal cartilage defects. This unmet clinical need drives the ongoing quest for regenerative medicine and tissue-engineering approaches for articular cartilage repair .
Many new promising technologies (Johnstone et al., 2019;Patel et al., 2019) are currently being developed and tested with the aim of finding an implant that is effective in facilitating regeneration of cartilage. Given the difficulties associated with the fixation of chondral constructs in the joint (Gotterbarm et al., 2008;Mancini et al., 2017), an alternative approach is the use of composite osteochondral constructs composed of distinct osteal and chondral compartments that can be surgically press-fitted into suitably prepared defects, thereby avoiding the risk of dislodgement (Martin et al., 2007).
However, the latter approach still faces many challenges, including design and optimization of the osteal compartment to act as an anchor for the overlying chondral compartment, production of a firm and durable connection between osteal and chondral compartment (Boushell et al., 2017), and optimization of the composition and structure of the chondral compartment (Diekman & Guilak, 2013;Lee et al., 2014).
To address those challenges, biomaterials that hold the potential for facilitating osteoregeneration within the osteal compartment were recently developed and investigated. First, 3D printed brushite-based scaffolds have been shown to be effective in promoting new bone growth after 6 months in an equine model that used the tuber coxae as implantation site (R. . However, these materials are usually processed using aggressive acidic treatments, precluding the direct incorporation of cells and/or some types of polymer during the fabrication phase. Therefore, an apatite-based scaffold that could harden under physiological conditions into a calcium-deficient hydroxyapatite (CDHA) was developed. This material had been shown to be effective in a 7-month long in vivo study, upon implantation in a critical size defect in the tuber coxae of horses. That study was performed to compare two sophisticated architectures with constant and gradient pore size, respectively. The material was shown to facilitate excellent new bone formation, particularly when using the scaffold with constant pore size (Diloksumpan, Vindas Bolaños, et al., 2020). This showed the potential for using such material as an osteal anchor of a tissue-engineered osteochondral graft.
Another major challenge is the connection between the osteal and chondral compartments of the tissue-engineered osteochondral graft when using cellfriendly materials of strongly different mechanical characteristics. Recently, a technique for attaching the chondral compartment to the osteal compartment using melt electrowriting (MEW) was developed (Diloksumpan, de Ruijter, et al., 2020), in which MEW fibers of the chondral compartment were partially incorporated into the slowly setting apatite-based osteal compartment, thereby binding the two compartments together. This strategy allows both for optimizing the mechnical properties of the MEW-reinforced chondral compartment, and for integrating the chondral and osteal compartments.
Regarding the seeding of regeneration-competent cells within the chondral compartment, articular cartilage-derived progenitor cells have been recently identified and characterized in both humans and horses as a distinct cell population that has the potential for cartilage repair (McCarthy et al., 2012;Williams et al., 2010). This potential was further shown to be retained in combination with biomaterials (Frisbie et al., 2015;Levato et al., 2017), making this cell type a promising candidate for a comprehensive regenerative approach. Additionally, it was recently discovered that supplementation with BMP-9 during in vitro culture of ACPCs resulted in higher expression of gene-markers related with hyaline-like extracellular matrix production, compared to supplementation with transforming growth factor (TGF ), a more commonly used growth factor in cartilage tissue engineering (Morgan et al., 2020). This observation sparked interest in further investigation of the potential of BMP-9-stimulated ACPCs for cartilage repair in vivo.
The current study aimed at evaluating an osteochondral composite scaffold for cartilage repair. These constructs were composed of a combination of a previously proven osteogenic CDHA 3D printed scaffold for the osteal compartment, onto which a chondral compartment composed by MEW micro-fibrous meshes is tightly anchored. For the chondral compartment, an experimental group in which the MEW structure was seeded with ACPCs that had been stimulated for 28 days with BMP-9 before implantation (Abinzano et al., 2018), was compared with an implant featuring a non-filled, cell-free MEW cartilage scaffold as control. It was hypothesized that 1) the CDHA scaffold would show comparable performance in the horse when implanted in the subchondral bone as in the tuber coxae in terms of firmly anchoring in the surrounding tissue and inducing bone growth; 2) the novel interface would provide a lasting connection between the osteal and chondral compartments of osteochondral graft; and 3) the engineered chondral compartment of the osteochondral graft containing the stimulated ACPCs would outperform the cell-free structures in terms of in vivo cartilage matrix production (specifically, amount and density of type II collagen and GAGs, with resulting mechanical properties as close as possible to those of native, healthy cartilage).

Experimental design
To assess the performance of integrated 3D printed osteochondral grafts that contained a cell-laden or a cell-free chondral compartment, the constructs were orthotopically implanted in a large animal model. Eight Shetland ponies (female, age 4 − 12 years, weight 149 − 217 (166 ± 29 )) were used and samples were implanted in the medial femoral ridge in the stifle joints. Healing D .
E 3 F I G U R E 1 Flow chart representing timeline of the experiment including health monitoring at each phase of the experiment.
was monitored for 6 months, after which the animals were humanely euthanized.
The study was approved by the ethical and animal welfare body of the Utrecht University (Approval nr. AVD108002015307 WP23).
Ponies were housed in individual boxes and fed a limited ration of concentrates together with hay for maintenance and free access to water. Quantitative gait analysis and radiographic examination of the stifle joints were performed before surgery for baseline values. Post-operatively, the animals were kept stabled for 6 weeks with daily monitoring of vital signs, lameness checks at walk, and examination of the operated joints for swelling or other signs of inflammation. In week 5 and 6, they were hand-walked for 10 minutes twice daily and from week 7 on, they were kept at pasture. Quantitative gait analysis and radiographic exams were performed at 3 weeks, 3 months, and 6 months post-operatively. After 6 months, ponies were humanely euthanized for harvesting samples for both quantitative and qualitative analyses. The timeline of the experiment is represented in Figure 1.

Fabrication of construct
Microfiber meshes were produced from medical-grade PCL (Purasorb® PC 12 Corbion PURAC, The Netherlands) by using MEW technology as previously Finally, sterilization was carried out by immersion of the mesh in 70% ethanol for 15 minutes, followed by air-drying in a sterile cabinet until use.
Printable calcium phosphate (PCaP) paste was prepared as earlier described (Diloksumpan, de Ruijter, et al., 2020 Figure 2). Eighty percent of the mesh thickness was set as the initial height for depositing the first of non-macroporous PCaP layer, as this proved to be the height that did not damage the mesh structure and ensured an optimal integration between the bone compartment and the chondral compartment. The first two layers of PCaP were deposited without macro-spacing, to mimic the subchondral bone plate, and followed by layers with designed macro-spacing of 700 to mimic the cancellous bone section (diameter = 6 , height = 5 ).
After finishing the fabrication process, the osteochondral constructs were allowed to set at 37°C under saturated relative humidity to form a solid, biomimetic bone compartment through conversion of the PCaP composite to CDHA. Finally, the osteochondral constructs were disinfected in 70% ethanol and exposed to UV-light for 1 hour, prior to seeding of cells.
In Vitro pre-culture Allogeneic articular cartilage progenitor cells (ACPCs) were obtained as previously described (Levato et al., 2017;Williams et al., 2010) from animals that were euthanized at the Utrecht University Veterinary Hospital for causes unrelated to disease or impairment of the musculoskeletal system and whose remains were donated for research purposes. Briefly, hyaline cartilage was collected in a sterile fashion, minced, and digested at 37°C with 0.2% ⋅ −1 pronase solution for 2 hours, followed by 12 hours in 0.075% ⋅ −1 collagenase solution. ACPCs were then selected using a fibronectin adhesion assay (Levato et al., 2017). Cells were expanded in culture and stored in liquid nitrogen until further use. After thawing, cells were expanded until passage 3 prior to their use for the experiment.
The constructs made of the combined CDHA and MEW meshes were disinfected in ethanol and exposed to UV-light for 1 hour as mentioned above.
To avoid any pH changes that might affect the cells, the constructs were sub-  The medial femoral ridge of the stifle joint was exposed by arthrotomy and an osteochondral lesion (diameter = 6 , depth = 6 ) was surgically created using a power drill. The surgical area was flushed by saline for cooling and removal of debris. Cell-laden constructs were implanted press-fit in a randomly chosen hind limb, with the cell-free control being implanted in the contralateral limb. After closing the arthrotomy wound in four layers in routine fashion, procaine penicillin was administered (Procapen, intramuscular (IM), 20 ⋅ −1 ). Post-operatively, nonsteroidal anti-inflammatory medication (metacam, per os (PO), SID, 0.6 ⋅ −1 ) was administered for 5 days and opioids (tramadol, PO, BID, 5 ⋅ -1 ) were administered for 2 days.

Gait analysis
The ponies were trained on a treadmill prior to the study using a standard protocol for treadmill habituation. Twenty-eight spherical reflective markers with a diameter of 24 (topline) and 19 (elsewhere) were attached with double-sided tape and second glue to anatomical landmarks ( Figure 3). Kinematic data were collected on a treadmill (Mustang, Fahrwangen, Switzerland) at trot using six infrared optical motion capture cameras (ProReflex, Qualisys, Gothenburg, Sweden) recording at a frame rate of 200 for 30 seconds at each session to obtain a sufficient number of strides.
To process the data, the reconstruction of three-dimensional coordinates

Radiographic examination
Stifles were radiographed in 3 projections: lateromedial, craniolateralcaudomedial oblique and caudo-cranial projection using standard machine settings before surgery (baseline), at 3 weeks postoperatively and at 6 months, just before euthanasia.

Euthanasia and sample harvest
After 6 months, animals were euthanized by induction with Midazolam (IV, 0.06 ⋅ -1 body weight) with ketamine IV, (2.2 ⋅ −1 body weight) and subsequent administration of sodium pentobarbital (IV, 1400 ⋅ −1 body weight). Next, the stifle joint was exposed, and gross assessment of the medial trochlear ridge was performed, focusing on the degree of filling of the defect, the integration of repair tissue with the surrounding native tissue and the surface quality of the repair tissue. Subsequently, the entire osteochondral area containing the constructs was harvested for further analyses with the aid of a surgical bone saw. Harvested tissues were initially kept in sterilized PBS for micro-computed tomography (micro-CT) scanning, biomechanical analyses and for collecting tissue from the chondral compartment of the implant for biochemical analyses. After this, all tissues were fixed in 4% formaldehyde for subsequent histological processing.

Biomechanical evaluation
The compressive properties of the chondral compartment of the defect site, the adjacent surrounding native cartilage and the more distant surrounding native cartilage (5 − 10 from the boundary of the defect) ( = 7 for cellladen constructs and = 7 for cell-free constructs) were evaluated with a dynamic mechanical analyzer (DMA, DMA Q800, TA instrument) equipped with a custom-size compressing probe (diameter = 2 ). A ramp force of 0.250 ⋅ −1 was applied until reaching 2.0 , to limit the deformation of sample to values below 200 . Compression modulus was calculated as the slope of the stress-strain curve in the range between 10 − 12% strain.

Biochemical evaluation
Firstly, biochemical analyses were performed on supplemental pre-implantation constructs ( = 3) that had been prepared in the same batch as the constructs that were later implanted. The chondral compartments of 28-day cultured constructs were removed and freeze-dried. Next, dry samples were digested in papain (Sigma Aldrich) at 60°C overnight. DNA, sulphated glycosaminoglycan (sGAG), and alkaline phosphatase (ALP) content were quantified by performing the Quan-iT-Picogreen-dsDNA-kit assay (Molecular Probes, Invitrogen, Carlsbad, USA), the dimethylmethylene blue assay (DMMB, Sigma-Aldrich, The Netherlands) and the p-nitrophenyl phosphate assay (SIGMAFAST™, Sigma-Aldrich), respectively.
Secondly, tissue fractions that were collected from the chondral compartments of harvested implants ( = 6 for cell-laden constructs, = 7 for cell-free constructs) were kept at −80°C, followed by lyophilization. Collagen content was quantified using an hydroxyproline assay (L-Hydroxyproline, Merck KGaA), and the sGAG and DNA quantification was performed as described above.

Microcomputed tomography
Microcomputed tomography was employed for the quantitative analysis of the bone compartments from the harvested osteochondral lesions ( = 7 for cell-laden constructs, = 7 for cell-free constructs). Six freshly made osteochondral grafts were scanned in a micro-CT scanner (Quantum FX-Perkin Elmer) to quantify the initial volume of PCaP material, pre-operatively. The post-mortem harvested tissue containing the defect area and the surrounding native tissue were similarly scanned (voltage = 90 , current = 200 , voxel size = 30 3 and total scanning time = 3 ). Subsequently, the 3D-reconstructed images were processed and analyzed using image J software (Schindelin et al., 2012) and Bone J plugin (Doube et al., 2010). Twodimensional regions of interest (ROIs) were selected in an axial plane at the boundary between the defect and the surrounding native tissue and interpolated to form a three-dimensional volume of interest (VOI). Thresholding was performed to separately select areas of ceramics and newly formed bone respectively for further calculation. Then, the percentages of mineralized newly formed bone, of non-mineralized tissue and of remaining ceramics, including the percentage of ceramics volume loss, were quantified. D .

Histological evaluation
Firstly, supplemental pre-implantation constructs ( = 3) that had been prepared in the same batch as the ones that later were implanted were fixed in 4% formaldehyde. After decalcification in 0.

Statistical analysis
Normality of distribution of the data was assessed from skewness, kurtosis, and Q-Q plots. Results were reported as mean ± standard deviation. Wilcoxon signed rank tests were used to analyze the biochemical, biomechanical, and micro-CT data. Statistical significance was set at = 0.05. All tests were performed using Matlab (version R2018b, The MathWorks, Inc.).
To evaluate the gait parameters, stride-level data were analyzed with R software (version 3.6.0, R Core Team, 2019), using package NLME (version 3.1-137) for mixed modelling. Dependent variables were investigated for normality using normal probability plotting and examining for skewness and kurtosis. If not normally distributed, data were transformed to permit linear mixed modeling. The random effect was subject and timepoint was the fixed effect. Significance was set at p < 0.05 and p-values were corrected using the false discovery rate method. Residual plots were checked for heteroscedasticity versus the outcome, as well as for normality in Q-Q plots.

In vitro
After 4

Evaluation during surgical implantation
Both cell-laden and cell-free constructs were press-fit implanted into the surgi-

Post-operative clinical monitoring
After surgical implantation, the animals were checked clinically for physical appearance and vital signs on a daily basis. All ponies recovered well from anesthesia after surgery and passed uneventfully through the rehabilitation period without any abnormalities in body temperature or behavior, with good weight-bearing on all operated limbs and no clinical signs of lameness during the entire period, with the exception of a single pony that developed severe lameness at 10 weeks after surgery. This pony was treated with anti-inflammatory medication and examined radiographically, which revealed extensive osteolysis around the created lesion. Because of persistent discomfort, the pony was euthanized at 12 weeks after surgery. Therefore, it was excluded from all analyses.

Gait analysis
Objective gait analysis was used to check for lameness or other signs of dysfunction of the musculoskeletal system. Objective data retrieved before implantation and at the end of the experiment were assessed for relevant parameters, including symmetry parameters and limb parameters.

Symmetry parameters
Front and hind limb lameness were analyzed through evaluation of the symmetry parameters of the head (MinDiff Head ( Figure 6A)) and of the pelvis (MinDiff Pelvis (Figure 6B)). These values reflect the differences in minimal vertical displacement with a negative MinDiff indicating a left-sided asymmetry and a positive MinDiff a right-sided asymmetry. In the treated ponies (except for the case referred to above that was euthanized), for both the head and the pelvis, there was no clear pattern in the direction of the asymmetries between baseline and endpoint and those differences between baseline and endpoint were minimal and statistically not significant. Therefore, symmetry measures could not discriminate between cell-laden and cell-free constructs.
Further, there was also no clear effect of timepoint on pelvis roll and pelvis yaw range of motion (Supplementary Figure 12), however, pelvis pitch range of motion (ROM) ( Figure 6C) decreased for all subjects with almost 20% over time (Supplementary Table 1).

Limb parameters, e ects of time
There was a significant effect of time for the height the toe was lifted from the surface during the swing phase of the limb that decreased significantly in the cell-free treated limbs, but not in the limbs treated with cell-laden constructs

Limb parameters, di erences between cell-laden and cell-free at endpoint
There were no significant differences between any of the cell-laden and cell-free

Radiographic examination
Healing progression within the osteal compartment of the implanted osteochondral constructs was followed up non-invasively through radiographic examination. On the radiographs taken at baseline, 3, and 6 months, no obvious abnormalities in term of the architecture of the surrounding native tissue were detected, other than the defects that had been created. This was with the exception of the pony that developed severe lameness. In that animal, severe osteolysis was noted at the implantation site 3 months after the implantation (Supplementary Figure 13).

Post-mortem macroscopic evaluation of the repair tissue
Macroscopic characteristics, for instance, color, appearance, and filling level of the lesion, were observed and documented before harvesting tissue sample for further analyses. After 6 months, macroscopic evaluation revealed that the defects were filled with repair tissue that in all cases did not fill the entire defect and remained lower than the level of the surrounding native cartilage in both cell-laden and cell-free treatments ( Figure 7A). The color of the repair tissue was variable (from reddish, to yellow and translucent) within the different treatments ( Figure 7B). In some cases, ceramic fragments could be observed within the repair tissue of the chondral compartment.

Biochemical analyses of repair tissue within the chondral compartment
The deposition of GAGs and collagen, the two main elements that compose cartilage extracellular matrix, were quantified within the chondral compartment of the osteochondral graft 6 months after implantation. There were no D .

Micro-CT evaluation of repair tissue within the osteal compartment
Bone healing and integration after was assessed through micro-CT scanning 6 months after implantation. Micro-CT images showed significant bone loss surrounding the implant in both the cell-laden and the cell-free groups, which could be visualized as black areas between the porous bioceramic structure (white) and the surrounding native bone (grey). However, mineralized bone formation could be visualized in some scaffolds from both groups with an integration to neighboring native bone ( Figure 10A). Statistically, there were no significant differences in mineralized bone formation (cell-laden: 6.14% ± 10.09%, cellfree: 4.73%±4.93%) and non-mineralized tissue (cell-laden: 81.38%±15.37%, cell-free: 74.71% ± 12.44%). However, there was a significant difference in the amount of remaining ceramics between the two groups (cell-laden: 12.48% ± 9.75%, cell-free: 20.56% ± 10.54%( = 0.0313)) ( Figure 10B

Discussion
This study aimed to evaluate the efficacy of an engineered osteochondral composite scaffold that was fabricated by combining a proven osteogenic CDHA scaffold for the osteal compartment with a novel interface for the connection between the chondral and osteal compartments. For the chondral compartment, BMP-9 stimulated cell-laden and cell-free constructs were compared. The cell-laden constructs contained in vitro formed tissue that was rich in GAGs and type II collagen, obtained by seeding Articular Cartilage Progenitor Cells (ACPCs) and stimulating them with BMP-9 for 4 weeks prior to implantation.
After implantation in an equine osteochondral defect for 6 months, there was poor chondral repair tissue in both the cell-laden and cell-free implants. The Representative micro-CT images from the middle of the sagittal plane of the constructs (white = ceramics, grey = mineralized tissue, black = non-mineralized tissue) (A). Quantitative analysis from micro-CT reconstruction showing percentage of mineralized bone formation, non-mineralized tissue, and remaining ceramics (B). The volume loss of ceramics was slightly higher in the cell-laden constructs compared to the cell-free ones (C).
best base for facilitating cartilage repair. The exact same bioceramic material tested in this long-term, orthotopic equine study, had previously been shown to successfully guide osteoregeneration in the same species, when implanted in the tuber coxae, an anatomical locus less subject to intense mechanical loads (Diloksumpan, de Ruijter, et al., 2020). Additionally, this previous study also focused on comparing different pore architecture within the 3D printed bone scaffolds. The scaffold architecture that led to the highest rate of new bone formation, consisting of a constant pore size across the sample, was selected for the present study, with the goal of maximizing neo-bone repair (Diloksumpan, de Ruijter, et al., 2020). However, there are two major differences with the use of the material in the current study. First, in the previous study the material was implanted in the tuber coxae, which is an orthotopic area but not representative of the intra-articular environment. Second, in the former study the implant was surrounded by a cylindrical case made of PCL that served to prevent bone ingrowth from the sides. Without such a shell of the mechanically deformable PCL in the current study, the surgeon encountered difficulties during the surgical placement of the implants, provoked by the non-resilience and brittleness of the CaP-based material, combined with some deviation from an ideal cylindrical shape of the CDHA implant. This resulted in fragments breaking off from the bioceramic osteal anchor. In fact, although inadvertently and as a side-effect, this problem of fragment formation was avoided in the former study when using PCL to encase. Polycaprolactone is deformable material, and the encasing will have facilitated the sliding of the ceramic implant into the defect. The duration of both studies was not identical (7 months in the earlier study, 6 months in the current), which makes direct comparisons between the two impossible. However, there were clear histological differences with many more multifocal to coalescing inflammatory reactions in both cell-laden and cell-free implants in the current study, compared to the earlier study, in which there were very few inflammatory cells visible.
This difference is likely due to the chronic irritation caused by fragments of material and to instability resulting from the imperfect fit of scaffold within the defect in the current study.
Some scaffolds from the current study collapsed and showed misalignment of the CDHA structure within the defect, with slightly enlarged defect size after an implantation for 6 months (as evident from the micro-CT analysis).
Bone resorption around the implant was found both in the groups with cellladen and with cell-free chondral compartment, which infer the effect from the osteal anchor rather than from the variable within chondral compartment. The circumstances described above likely resulted in failure to place the implant in a real press-fit fashion and hence, in the creation of (micro)movement, leading to increasing instability under repetitive loading together with possible material degradation over time and ensuing osteolysis, as seen earlier (Albrektsson et al., 2019;Goodman et al., 2019). Additionally, the gap between the implant and surrounding native tissue due to the imperfect fit may have allowed for the intrusion of synovial fluid. Contact of synovial fluid with subchondral bone has been shown to induce osteolysis (Kold et al., 1986). In the few scaffolds that remained in place, the volume of mineralized bone formation was also lower than what was found in the earlier study, both in cell-laden and cell-free D .

E 11
treatments. This is potentially due to the higher and cyclical loads experienced in the articulating joint compared to the tuber coxae. Overall, it was not possible to determine a single cause for the failure of the CDHA scaffolds to act as the anchor of the engineered osteochondral implant, and it is likely that the limited osteointegration is due to a combination of misalignment after surgery, mechanical failure under cyclic loading, and synovial fluid infiltration.
In earlier studies (Bothe et al., 2019;van Susante et al., 1998) similar observations were made. In those studies, fibrous repair tissue was seen in the chondral compartment, together with osteolysis and formation of a fibrous interface surrounding the osteal anchor when tissue-engineered osteochondral grafts were implanted in a load-bearing area for a 12-month long-term study.
It was hypothesized that osteolysis and the fibrous layer surrounding the osteal anchor led to instability that might have caused the degradation of the newly formed cartilage-like repair tissue observed at the early of the experiment.
Stability of the osteochondral graft might be affected by multiple parameters including the alignment of an osteal compartment within the defect and the properties of the materials being used (Heuijerjans et al., 2018;Nosewicz et al., 2014;Schlichting et al., 2008;von Rechenberg et al., 2003), as these might affect stability of the overlying chondral compartment. In the current study, properties (Bowland et al., 2015). It is thus clear that the imperfect implantation had severe repercussions and can be considered a major factor that affected the chondral compartment and hence the outcome of the study. This effect was noticeable to the extent that drawing any conclusions about the effect of BMP-9 seeded ACPCs, which was the principal variable that was to be tested in the study, is not possible. Also, no conclusion could be reached about the interface between the osteal and chondral compartments that was used since delocalized MEW-mesh structures were observed in some scaffolds from both groups. This might be due to misalignments of the osteal compartment as discussed above, to shear forces during loading, or a combination of both.
During the in vivo post-operative monitoring of the animals, the clinical signs were very mild and far from alarming, except for the single pony that developed severe lameness. Clinical examinations were performed routinely by experienced veterinary specialists, however, assessment of locomotion through visual observation alone is subjective and known to have poor repeatability, especially in mild cases. This is partly due to the inability of human visual perception to properly distinguish, notice, and quantify differences in locomotion at high resolution (Serra Bragança et al., 2018). Therefore, quantitative gait analysis was employed as an objective and non-invasive assessment. The gait analysis data did not show many differences with respect to baseline. This may to a certain extent have been related to methodological factors. During the assessment, ponies were put on a treadmill and they were imposed the same belt velocity during both measurements. Therefore, the subjects were forced to trot at the same velocity, ensuring that stride length needed to be maintained.
This might be the reason why there were no differences between timepoints for maximal protraction and retraction (the limb parameters). However, pelvis pitch range of motion (ROM) decreased for all subjects with almost 20% over time. This pattern is often seen in case of dysfunction of the back. The finding may thus be related to earlier observations that bilateral hindlimb lameness may induce back problems in horses (Alvarez et al., 2008;Alvarez et al., 2007;Greve et al., 2017). Toe dragging of the lame limb, in which the hoof is lifted less high off the ground, is another sign of pain (Buchner et al., 1995). Nevertheless, the overall impact of the bilateral lesions in the stifle joints was low, as evidenced by the fact that there was no sign of load redistribution from the hind to the front limbs. If that had been the case, the subjects would have compensated by displacing their center of mass more to the front, resulting in more negative angles for forelimb fetlock extension, as fetlock hyper extension correlates with peak ground reaction force (GRFPeak) (Crevier-Denoix et al., in a large animal experiment (R. . Functional joint repair remains a huge challenge that has not been addressed to some satisfying extent during the last decades, despite many promising approaches.
It is likely that the quest for a real solution will go on for some time by trial and error with more errors to come. Those errors are inevitable and need to be made but they should take the least possible toll on experimental animals.

Conclusion
This study presented the results from the evaluation of a cell-laden and cell- F I G U R E 1 2 Gait analysis: symmetry parameters. Pelvis roll and yaw joint angles (A,B) showed no significant di erences between baseline and 6 months a er implantation and neither did the kinematic hind limb parameters (C, D, E) between baseline and 6 months a er induction. Limb height of the hind limbs (F) decreased for both hindlimbs, but only significantly for the cell free group, though there was no di erence in limb height between cell-laden and cell-free groups at 6 months a er implementation.

F I G U R E 1 3
Representative radiographic images (latero-medial and craniolateral-caudomedial oblique projections) of the stifle of ponies before implantation, 3 months a er implantation and 6 months a er implantation. Red arrows indicate the implantation sites. No radiographic abnormalities were noted in any of the ponies (1 st row), except for the pony that became severely lame at 10 weeks. In this animal extensive osteolysis was observed (2 nd row).