Bioreactor‐manufactured cartilage grafts repair acute and chronic osteochondral defects in large animal studies

Abstract Objectives Bioreactor‐based production systems have the potential to overcome limitations associated with conventional tissue engineering manufacturing methods, facilitating regulatory compliant and cost‐effective production of engineered grafts for widespread clinical use. In this work, we established a bioreactor‐based manufacturing system for the production of cartilage grafts. Materials & Methods All bioprocesses, from cartilage biopsy digestion through the generation of engineered grafts, were performed in our bioreactor‐based manufacturing system. All bioreactor technologies and cartilage tissue engineering bioprocesses were transferred to an independent GMP facility, where engineered grafts were manufactured for two large animal studies. Results The results of these studies demonstrate the safety and feasibility of the bioreactor‐based manufacturing approach. Moreover, grafts produced in the manufacturing system were first shown to accelerate the repair of acute osteochondral defects, compared to cell‐free scaffold implants. We then demonstrated that grafts produced in the system also facilitated faster repair in a more clinically relevant chronic defect model. Our data also suggested that bioreactor‐manufactured grafts may result in a more robust repair in the longer term. Conclusion By demonstrating the safety and efficacy of bioreactor‐generated grafts in two large animal models, this work represents a pivotal step towards implementing the bioreactor‐based manufacturing system for the production of human cartilage grafts for clinical applications. https://doi.org/10.1111/cpr.12625


| INTRODUC TI ON
In a recent phase I clinical trial, we aimed to treat articular cartilage defects with cartilage tissue grafts, which were engineered from nasal cartilage chondrocytes. 1 Given the highly promising clinical data that we acquired in that study, we now aim to address critical manufacturing related issues, which could ultimately impede the translation of this therapy into widespread clinical use. Since the cartilage grafts in our study were produced by conventional manual tissue engineering methods, the production process was lengthy, labour-intensive and would possess inherent variability among operators Figure 1.
A manufacturing system based on these manual processes may be challenging to standardize, thus presenting obstacles towards regulatory compliance, and may ultimately incur high operating costs and challenges for upscaling, thus presenting significant barriers towards economic viability. Alternatively, automated bioreactor-based manufacturing systems have the potential to overcome these limitations, breaking down regulatory and economic barriers, allowing engineered tissue therapies to reach their full clinical potential 2 Figure 1.
We have previously demonstrated that bioreactor-based perfusion of a cell suspension directly through the pores of a 3D scaffold enhances the cell seeding efficiency and cell distribution compared Materials & Methods: All bioprocesses, from cartilage biopsy digestion through the generation of engineered grafts, were performed in our bioreactor-based manufacturing system. All bioreactor technologies and cartilage tissue engineering bioprocesses were transferred to an independent GMP facility, where engineered grafts were manufactured for two large animal studies.

Results:
The results of these studies demonstrate the safety and feasibility of the bioreactor-based manufacturing approach. Moreover, grafts produced in the manufacturing system were first shown to accelerate the repair of acute osteochondral defects, compared to cell-free scaffold implants. We then demonstrated that grafts produced in the system also facilitated faster repair in a more clinically relevant chronic defect model. Our data also suggested that bioreactor-manufactured grafts may result in a more robust repair in the longer term.
Conclusion: By demonstrating the safety and efficacy of bioreactor-generated grafts in two large animal models, this work represents a pivotal step towards implementing the bioreactor-based manufacturing system for the production of human cartilage grafts for clinical applications.
Read the Edito rial for this artic le on doi: 10.1111/cpr.12625 F I G U R E 1 Conventional manufacturing processes used to produce the engineered grafts are based on traditional bench-top manual culture methods. These manual procedures require a large number of labour-intensive manipulations that pose challenges towards regulatory compliance and ultimately result in high manufacturing costs in the long term. As an alternative, bioreactor-based production systems, which automate and control the bioprocesses, have the potential to overcome the limitations associated with conventional manufacturing methods, facilitating regulatory compliant and cost-effective production of engineered cartilage grafts for widespread clinical use to conventional manual methods. 3 We then demonstrated that culturing cell-seeded constructs under perfusion supported the development of a viable and uniform tissue graft during prolonged culture. 4 Chemo-optic microsensors were also integrated into the bioreactor for continuous online measurements of oxygen levels in the perfused culture medium. With the prospect of clinical applications, we also upscaled the perfusion bioreactor system in order to engineer clinically relevant, large-scale cartilage grafts suitable for treating lesions in a human knee. 5 After establishing these fundamental building blocks for a bioreactor-based production system, we then established an innovative and streamlined approach to engineer human cartilage grafts within a single bioreactor unit, from the introduction of primary chondrocytes freshly isolated from a biopsy, through the generation of a mature cartilaginous tissue graft. 6 In this work, we have adapted and integrated our previously described bioreactor technologies and bioprocesses in order to establish an automated manufacturing platform for the production of nasal chondrocyte-based engineered grafts. Cartilage grafts were manufactured in the bioreactor-based system at a centralized GMP facility and were first assessed in a large animal study based on an acute osteochondral defect in sheep. Following the promising results from the acute defect study, we next assessed grafts generated in our manufacturing system in a more challenging and clinically relevant chronic defect sheep model. The results of this work represent a pivotal step towards implementing the bioreactor-based manufacturing system for the production of human cartilage grafts for clinical applications.

| Scaffold preparation
The bilayered biomimetic osteochondral scaffold used in these studies (Fin-Ceramica Faenza SpA) had a cylindrical shape, 25 mm F I G U R E 2 A, Cartilage digestion bioreactor. An automated tissue digest protocol was developed and converted to an automated algorithm using application-specific Octane software. The control system operates a series of valve actuators to open and close valves on the bioreactor cassette and a peristaltic pump to deliver fluids at a range of flow rates. All biological processes are housed within a single disposable cassette. Automated protocol steps included (a) tissue washed with phosphate-buffered saline; (b) delivery and perfusion of digestion enzyme; (c) removal of the digestion enzyme and replacement with complete medium; and (d) cell collection. B, T-CUP perfusion bioreactor. Cell seeding, 3D expansion and differentiation were performed in a single perfusion bioreactor manufacturing module. Culture medium is forced back and forth between the inner chamber and an outer chamber of the vessel and, therefore, is perfused directly through the 3D construct. C, Oxygen and D, pH sensor data monitored throughout 5-week culture period in the T-CUP bioreactor. Spikes in the oxygen plot are artefacts due to the opening of the incubator door. Peaks in the pH plot are due to fluctuations in pH between the introduction of fresh medium (pH ≈ 7.6) until the next medium exchange (pH ≈ 6.8-7.0) in diameter and 5 mm in thickness for bioreactor-generated cellbased grafts and a 6 mm diameter for cell-free scaffold implants.
The top layer of the scaffold (3 mm in thickness) consisted of equine type I collagen, and the bottom layer (2 mm in thickness) was made of a mineralized blend of type I collagen and Mg-HA, mimicking the structure and biochemistry of cartilage and subchondral bone. 7,8 The scaffold was developed through a bio-inspired process, as previously described. 9

| Bioreactor-based manufacturing
The bioreactor-based manufacturing process is comprised of four main phases: • Cartilage tissue digestion phase, in which the cartilage biopsy is enzymatically digested to liberate isolated chondrocytes in a cell suspension.
• Cell seeding phase, in which isolated primary chondrocytes are efficiently and uniformly seeded throughout the volume of the scaffold.
• 3D proliferation phase, in which chondrocytes are extensively expanded in number to colonize the volume of the scaffold.
• Differentiation phase, in which the 3D expanded chondrocytes are re-differentiated.
The system is comprised of two bioreactor units. The cartilage digestion bioreactor automates and controls the cartilage tissue digestion phase Figure 2A, and the T-CUP perfusion bioreactor controls the cell seeding, 3D proliferation and differentiation phases Figure 2B. The digestion bioreactor and T-CUP perfusion bioreactor were installed within a declassified cleanroom at the GMP facility (Holostem Terapie Avanzate; declassified to permit the introduction of animalderived cells). Laboratory technicians were trained for autonomous use of the bioreactor systems as well as all required cartilage tissue engineering bioprocesses. In line with GMP guidelines, a proper quality management system had been organized for production in the GMP facility. Stringent records ensured full traceability of all materials used. The complete manufacturing process was detailed and documented in standard operating procedures and manufacturing protocols. Proper quality controls were established to ensure standardization of the process and product. Additional testing had to be performed on the starting materials prior entering the facility (eg mycoplasma on the biopsy at arrival), and availability of GMP-grade materials and reagents was verified for future clinical applications.

| Nasal cartilage biopsy
Prior to commencing the large animal studies, procedures for the handling, packaging and transportation of biopsies were first established, tested and validated. Additional procedures were established to facilitate tracking of each biopsy as well as a documentation system to streamline the delivery. Six weeks prior to graft implantation, a cartilage biopsy was harvested from the nasal septum of each sheep with an 8-mm-diameter biopsy punch. Perichondrium was scraped from the cartilage tissue with a scalpel, and biopsies were thoroughly washed with saline, blotted with sterile gauze and stored in complete medium at 4°C for transport. After harvesting the nasal septum cartilage in the veterinary clinic in Croatia, biopsies were transported under defined conditions (eg overall duration of transport, temperature of transport vehicle, monitored temperature within the transport container) to the GMP facility in Italy. Animals were kept at the clinic post-operatively and then transported to a family farm until graft implantation. This 6-week time period allowed for the production of the nasal chondrocyte grafts.

| Cartilage tissue digestion
After passing defined acceptance criteria, biopsies were cut into small pieces (≈1-2 mm) and transferred into the digest module of the diges-

| Cell seeding, proliferation and differentiation
Nasal chondrocyte grafts were generated in the T-CUP bioreactor using a streamlined bioreactor-based process as previously described. 6 Briefly, the limited number of primary chondrocytes, freshly isolated from the small cartilage biopsy, was seeded and extensively expanded directly within the 3D scaffold in the bioreac- Chondrocytes were seeded into the bilayered biomimetic osteochondral scaffold 25 mm in diameter and 5 mm in thickness under alternating perfusion flow within the bioreactor at a perfusion rate of 1 mm/second for 16 hours in 50 mL of complete medium. Following the perfusion cell seeding phase, cell-seeded scaffolds remained within the bioreactor, and culture medium was replaced with "proliferating medium" (complete medium supplemented with 1 ng/mL TGFβ1 and 5 ng/mL FGF-2) to expand the cells directly within the scaffold. Constructs were perfused for 3 weeks at a perfusion rate of 100 µm/s with two medium exchanges per week. Following the 3D proliferation phase, culture medium was replaced with "differentiating medium" (complete medium supplemented with 10 ng/mL TGFβ1, 1 IU/mL insulin and 0.1 mmol/L ascorbic acid 2-phosphate) and constructs were cultured in the bioreactor for an additional 2 weeks with medium exchanges three times per week. Throughout the proliferation and differentiation phases, pH and oxygen levels in the medium were monitored (measurements acquired every 10 minutes) with chemo-optic sensors (PreSens GmbH) integrated into the T-CUP bioreactor vessel. Following 5 weeks of production, half of each engineered nasal chondrocyte-based graft was harvested for histological assessments and the other half was transported from the GMP facility to the veterinary clinic using the transportation conditions and procedures established for biopsy transport.

| Acute defect model
The large animal study based on an acute defect model Figure S1

| Chronic defect model
The large animal study based on a chronic defect model Figure S1 • 1st joint surgery-defect creation: Six weeks prior to graft implantation, a cartilage biopsy was harvested from the nasal septum of each sheep as described above. In addition, cartilage defects were also created at this time on the load bearing surfaces of medial and lateral femoral condyles with a 4-mm-diameter biopsy punch.
Special care was taken not to damage the subchondral bone.
• 2nd joint surgery-defect repair: Defects on the medial and lateral femoral condyles, which had chronified over the 6-week period, were converted to osteochondral defects 6 mm in diameter and 5 mm in depth using a standard instrument for mosaicplasty (COR, DePuy Synthes). Evidences of joint inflammation in terms of mild synovial oedema and hyperaemia were observed at the time of scaffold/tissue implantation in the chronic defect model. Prior to implantation, defects were washed with saline. Implantation of BR grafts and CFS implants, post-operative care and harvesting of the explants were similar as described in the acute defect model.

| Anaesthesia
For identification, ear tags were applied and microchips were placed under the skin at the back of the neck between the shoulder blades on the dorsal midline. Animals had food removed 24 hours before surgery and water removed 8-12 hours ahead.

| Graft implantation
Each stifle was physically examined for any abnormalities while anaesthetized. The animal was placed in a dorsal recumbence position, and following surgical preparation, the right stifle joint was opened via a medial parapatellar approach. The Hoffa was incised to facilitate visualization of the joint, and the knee was flexed to make the medial femoral condyle visible. For the visualization and approach to the lateral femoral condyle, a lateral parapatellar mini-arthrotomy was performed. After exposure, osteochondral defects measuring 6 mm in diameter and 5 mm in depth were created with a standard instrument for mosaicplasty (COR, DePuy Synthes) on both medial and lateral femoral condyles. In the BR group, medial and lateral defects were both treated with autologous nasal chondrocyte-based grafts. In the CFS group, medial and lateral defects were both treated with cell-free scaffolds. BR grafts and CFS implants were cut to a cylindrical shape with a 6mm-diameter biopsy punch and implanted into the defects using a press-fit method. No additional fixation was used. After the implantation, the joint was cycled through a range of motion to ensure a satisfactory rim fixation of the implanted graft or scaffold, following which the joint was closed by standard surgical procedures. Post-operatively, animals were allowed to bear full weight, but kept in small pens for 5 days to reduce ambulation.
After 5 days in the clinic, animals were transferred to a family farm with no ambulation restrictions.

| Euthanasia and necropsy
Following 3 months or 12 months, animals were sedated according to anaesthesia protocols with intramuscular administration of ketamine and xylazine and subsequently euthanized with intravenous administration of T61. To harvest the treated defects, condyles were cut with an oscillation saw. From each condyle, an osteochondral tissue block containing the defect and surrounding cartilage was cut to a size of 15 × 10 × 10 mm and subsequently divided into two halves.

| ICRS macroscopic scoring
Before explantation, two photographs of the exposed condyles were taken in situ. After explantation, four additional photographs were taken of each condyle. Three blinded orthopaedic surgeons, experienced in using ICRS macroscopic scoring system, independently scored the photographs.

| Histology and Immunohistochemistry
Osteochondral tissue samples for histology and immunohistochemical analysis were fixed in 4% PFA, decalcified in 15% EDTA, dehy- Sections were incubated with 3′,3′ diaminobenzidine tetrahydrochloride (DAB) and counterstained with haematoxylin. Normal articular cartilage and subchondral bone were used as positive controls.

| Histological quantification
To quantify immunohistochemical staining of collagen type II, slides were scanned with NanoZoomer 2.0-RS. The reference area was set to 32.5 mm 2 , corresponding to the size of the defect. Collagen

| Statistics
Quantitative data are expressed as the mean ± standard deviation (SD). To analyse differences, independent-samples t test was F I G U R E 4 ICRS II histology scores of explants from acute defects treated with bioreactor (BR) manufactured grafts and cell-free scaffold (CFS) implants. A, 3-mo explants. B, 12-mo explants. Data are presented as mean + SD. (*indicates statistically significant difference between BR and CFS) used and Hedges g effect size was calculated using the formula from Ellis, P.D. 11

| Bioreactor-based manufacture
All of the bioreactor-based production runs, for both animal stud-

| Acute defect model
Grafts produced in the established bioreactor-based manufacturing system were first tested in an acute defect model, based on freshly created osteochondral defects. Surgical wounds healed well after arthrotomy, with no signs of oedema, inflammation or wound dehiscence. Upon explantation, there were no signs of delamination or dislocation of any grafts or implants. Unfortunately, three sheep in the BR group died (two in the 3-month group and one in the 12month group). Upon examinations by an independent examiner, reasons of death were found to be unrelated to the surgical procedures or to the implants. BR defects were filled with glossy white tissue, while CFS defects were partially filled with granulation/fibrous tissues. At 12 months, cartilage repair appeared macroscopically better for both groups compared to the 3-month time point. Repair tissue filled the defects and appeared to be integrated with the surrounding cartilage for both groups. The surface of BR defects appeared smoother with less fissures and had significantly higher ICRS macroscopic scores (BR:

| Macroscopic evaluation
9.1 ± 0.8, n = 6; vs CFS: 6.8 ± 1.9, n = 8; P = 0.014) as compared to CFS defects.  Figure S3. In contrast to BR defects, cartilage-cartilage integration was not observed for the CFS group. Although no tidemark could be observed at 3 months for either group, subchondral bone regeneration was more advanced in BR defects than in CFS. Repair tissue in BR defects was integrated with surrounding bone and contained areas of active ossification.

| Histological evaluation
In contrast, the subchondral region in CFS defects was only partially filled with loose granulation tissue, with few areas of active ossification.
The quantified area of collagen type II staining was significantly larger for BR vs CFS in the chondral region of the repair tissues

| Chronic defects
Following the promising results obtained with bioreactor-produced engineered grafts in the acute defect model, we next aimed to assess the grafts in a more challenging and more clinically relevant model of a chronic defect.
F I G U R E 5 Explants from chronic defects treated with bioreactor (BR) manufactured grafts and cell-free scaffold (CFS) implants. A, Macroscopic and histological assessments after 3 and 12 mo. Scale bar indicates 2 mm. B-G, 12-mo explants from chronic defects. B, C, In BR chronic defects, we observed restoration of articular cartilage with chondrocytes (arrows) in zonal organization typical for articular cartilage. E, F, CFS defects healed with fibrous tissue mixed with hyaline cartilage. Both chondrocytes (arrows) and connective tissue cells (arrowheads) are present, but with no tissue organization. D and G Picrosirius-stained images observed under polarized light microscope show excellent cartilage-to-cartilage integration in BR explants, with homogenous fibril organization across the defect. Poor cartilage healing is observed in CFS explants, with thick non-organized fibres.

| Macroscopic evaluation
At 3 months, BR and CFS defects were partially filled with hyaline and fibrous tissues Figure 5A , with similar ICRS macroscopic scores

| D ISCUSS I ON
In this work, we have established a bioreactor-based manufacturing system for the production of chondrocyte-based engineered grafts.
All bioreactor technologies and cartilage tissue engineering bioprocesses were transferred to an independent GMP facility where all engineered grafts were successfully manufactured for two large animal studies. The results of these studies demonstrate the safety and feasibility of the bioreactor-based manufacturing approach.
Moreover, grafts produced in the bioreactor manufacturing system were first shown to accelerate the repair of acute osteochondral defects compared to cell-free scaffold implants. We then demonstrated that grafts produced in the system also resulted in a faster repair in a more clinically relevant chronic defect model. The data also suggest that the bioreactor-manufactured grafts may result in a more robust repair at a longer term point.
We have previously established a number of bioreactor tech- transferability of the production to other manufacturing centres. 13 In addition, non-invasive sensors were integrated into the bioreactor system to monitor oxygen and pH levels throughout the manufacturing process. Sensing and online monitoring offer great advantages to provide non-invasive and quantitative data relevant to the process and the engineered graft, providing meaningful in-process quality controls and data on graft quality. Automated logging of our sensor data along with other key process parameters will increase traceability of the process and will therefore facilitate compliance to strict regulatory guidelines. The integrated sensors could be used in the future not only to monitor, but also to maintain pH and oxygen at predefined levels in a feedback controlled loop. This may not only improve and standardize graft quality, but would considerably improve process robustness and standardization for reduced process and product variability.
One potential limitation of our manufacturing strategy was the approach of utilizing a centralized manufacturing facility for the production of the engineered grafts. This approach imposes significant logistical and regulatory challenges, as well as high costs, simply for the transportation of biopsies and grafts between the central facility and the clinical site. On the other hand, a centralized manufacturing facility allows for highly trained operators to closely oversee the manufacturing process, to conduct in-process controls and to release the final engineered product in accordance with well-defined quality control release criteria. As specific assays will be required to characterize the final engineered product, it may be challenging and cost prohibitive for a de-centralized facility to have available a qualified laboratory with qualified instruments, validated procedures and the trained operators that would be required to perform the necessary assays. In the future, it may be possible to establish a simple to use plug-and-play bioreactor system which could be installed into qualified clinical centres for the de-centralized production of engineered tissues, mitigating complex logistical issues as well as automating aspects of quality control. Nevertheless, this will still require substantial expertise and investment in infrastructure at the production sites. As opposed to the acute defect model, there was no persistence of cartilage tissue within the subchondral bone layer of the BR defect.
As reported in the literature, it is possible that the inflammatory environment in the chronic defect has facilitated the conversion of cartilage into bone. 17 The results of the chronic model are quite impressive considering the mechanical loading and inflammatory conditions that the grafts were subjected to. Since sheep were not immobilized after surgery, the animals could walk freely post-operatively. Our data show that BR grafts, which were implanted into high load bearing regions of the condyle and therefore subjected to mechanical loading soon after implantation, resulted in hyaline repair tissue resembling native cartilage tissue. This supports a previous in vitro study showing that nasal chondrocytes can respond positively to physical forces resembling joint loading, increasing the production of cartilaginous extracellular matrix proteins. 18 Our results are also consistent with another in vitro study showing that nasal chondrocytes have a high capacity to recover from inflammatory conditions, suggesting that nasal chondrocyte-based grafts could have more favourable chances to successfully engraft into the joint and regenerate the articular cartilage surface. 19 While the level of maturation of engineered cartilage grafts, Nevertheless, in this current manuscript we describe that the bioreactor-generated grafts could be safely implanted into high load bearing sites, subjected to mechanical loading soon after implantation, and result in hyaline repair tissue resembling native cartilage. If engineered cartilage grafts can be generated with sufficient properties to meet defined quality criteria without the use of mechanical preconditioning, bioreactor automation requirements can be greatly simplified, thereby facilitating the development of a more compact, user-friendly and cost-effective bioreactor-based manufacturing system-facilitating clinical translation.

| CON CLUS ION
Conventional manufacturing strategies present significant hurdles for the cost-effective translation of cell-based engineered grafts to the clinic. By demonstrating safety of the generated grafts in two large animal models, this work represents a pivotal step towards a regulatory compliant, bioreactor-based clinical manufacturing strategy for human engineered cartilage implants. Moreover, in view of the significant challenges typically associated with treating advanced cartilage defects, the promising efficacy results of our approach in a chronic defect model highlight its potential applicability not only for the treatment of small focal cartilage lesions, but also for a broader range of clinical indications. Ongoing efforts are currently aimed at qualifying the bioreactor technology and validating the associated bioprocesses in preparation for a first-in-human clinical study.

ACK N OWLED G EM ENTS
This project has received funding from the European Union's Seventh Program for research, technological development and demonstration under Grant Agreement No. 278807 (BIO-COMET). We would like to acknowledge Professor Anthony Hollander for contributing his scientific and pre-clinical expertise in this project.

CO N FLI C T O F I NTE R E S T
The authors declare no conflicts of interest.