Remodeling of Human Osteochondral Defects via rAAV-Mediated Co-Overexpression of TGF-β and IGF-I from Implanted Human Bone Marrow-Derived Mesenchymal Stromal Cells

The application of chondrogenic gene sequences to human bone marrow-derived mesenchymal stromal cells (hMSCs) is an attractive strategy to activate the reparative activities of these cells as a means to enhance the processes of cartilage repair using indirect cell transplantation procedures that may improve the repopulation of cartilage lesions. In the present study, we examined the feasibility of co-delivering the highly competent transforming growth factor beta (TGF-β) with the insulin-like growth factor I (IGF-I) in hMSCs via recombinant adeno-associated virus (rAAV) vector-mediated gene transfer prior to implantation in a human model of osteochondral defect (OCD) ex vivo that provides a microenvironment similar to that of focal cartilage lesions. The successful co-overexpression of rAAV TGF-β/IGF-I in implanted hMSCs promoted the durable remodeling of tissue injury in human OCDs over a prolonged period of time (21 days) relative to individual gene transfer and the control (reporter lacZ gene) treatment, with enhanced levels of cell proliferation and matrix deposition (proteoglycans, type-II collagen) both in the lesions and at a distance, while hypertrophic, osteogenic, and catabolic processes could be advantageously delayed. These findings demonstrate the value of indirect, progenitor cell-based combined rAAV gene therapy to treat human focal cartilage defects in a natural environment as a basis for future clinical applications.


Introduction
Focal articular cartilage lesions remain a particularly challenging clinical problem in orthopaedics due to the failure of the articular chondrocytes, the only cell population present in this highly specialized, avascular tissue, to fully provide self-repair activities in sites of cartilage injury [1][2][3][4]. Equally critical, none of the surgical options currently employed in orthopaedic surgery to mobilize or even supply chondroregenerative progenitor cells, like those from the subchondral bone marrow (microfracture, pridie drilling, stromal cell transplantation), can reliably support the strict reproduction of the original hyaline cartilage in such lesions with a typical proteoglycan-and type-II collagen-based extracellular matrix (ECM) [1][2][3][4][5][6][7]. Instead, a fibrocartilaginous repair tissue mostly composed of type-I collagen is generated in sites of cartilage injury, without the proper mechanical integrity that would allow to withstand mechanical loads over time, and that may progress towards irreversible, degenerative osteoarthritis [1][2][3][4].
Strategies that aim to activate the chondroreparative properties of progenitor cells prior to reimplantation in sites of cartilage damage may provide strong tools to overcome these hurdles, like approaches using gene therapy procedures that support the specific and long-term targeting of cellular and biological pathways involved in cartilage repair [8,9]. In this regard, vectors derived from the human adeno-associated virus (AAV) vector are particularly well suited to achieve these goals from a translational perspective [10], as recombinant AAV (rAAV) constructs can safely and durably transduce bone marrow-derived mesenchymal stromal cells (MSCs) at very high efficiencies (about 100% for several weeks) [11][12][13][14][15][16][17]. While the rAAV-mediated transduction of MSCs from human patients (hMSCs) has been attempted in a number of trials in vitro by delivering chondrotherapeutic genes coding for the basic fibroblast growth factor (FGF-2) [13], transforming growth factor beta 1 (TGF-β1) [11,15], insulin-like growth factor I (IGF-I) [16], sex-determining region Y-type high-mobility group box 9 transcription factor (SOX9) [14,17], little is known on the benefits of rAAV for the application of genetically modified hMSCs in human focal cartilage defects in a native microenvironment.
In the present study, we focused on co-delivering TGF-β and IGF-I, two potent factors supporting cartilage repair, in hMSCs via rAAV as improved cell regenerative platforms capable of enhancing the repair activities in a human osteochondral defect (OCD) model system ex vivo, and, in light of our previous work in vitro, showing the feasibility of co-applying these two candidates in settings aiming at directly rejuvenating resident MSCs within a lesion [18]. For the first time, and to our best knowledge, the current work demonstrates that the simultaneous overexpression of TGF-β and IGF-I can be effectively achieved in rAAV-targeted hMSCs implanted within human OCDs, promoting the sustained remodeling of the injured human articular cartilage (cell proliferation, hyaline ECM deposition) over a prolonged period of 21 days while containing unwanted hypertrophic, osteogenic, and catabolic processes relative to independent gene treatments and control condition (reporter lacZ gene transfer). These findings show the value of using combined cell-and gene-based therapy via rAAV as a significant means to enhance human cartilage repair in environmental conditions closely resembling those in patients in vivo.

Cell and Osteochondral Defect Cultures
Bone marrow aspirates were collected from the distal femurs of patients undergoing total knee arthroplasty (n = 3) with informed consent and according to the Helsinki Declaration. The study was approved by the Ethics Committee of the Saarland Physicians Council . Firstly, bone marrow-derived hMSCs were isolated and expanded as previously described [13][14][15][16][17][18][19]. Briefly, freshly isolated hMSCs were washed, centrifuged in Dulbecco's modified Eagle's medium (DMEM) and the pellet obtained was resuspended in a mixture of red blood cell lysing buffer and DMEM in equal ratios. The mixture was washed, pelleted, and resuspended in DMEM containing 10% fetal bovine serum with 100 U/mL penicillin and 100 µL/mL streptomycin. Cells were plated in T75 flasks and kept incubated overnight at 37 °C under 5% CO2, and the medium was removed and replaced by growth medium with recombinant FGF-2 (1 ng/mL), with a medium change every 2-3 days. Cells were tested at passages 1-2 in order to avoid a shift in cell phenotype. Human bone marrow-derived mesenchymal stromal cells MSCs were prepared as high-density aggregates (2 × 10 5 cells/aggregate) in defined chondrogenic medium overnight, transduced with the various rAAV shuttles (TGF-β/IGF-I, TGF-β/lacZ, IGF-I/lacZ: 40 µL each vector/aggregate; lacZ: 80 µL vector/aggregate), and directly applied to human osteochondral defects (OCDs; 6-mm diameter) for further culture in defined chondrogenic medium for 21 days when the OCDs were processed to monitor transgene expression and for biochemical, immunohistochemical/histological, and real-time RT-PCR.

Cell and Osteochondral Defect Cultures
Bone marrow aspirates were collected from the distal femurs of patients undergoing total knee arthroplasty (n = 3) with informed consent and according to the Helsinki Declaration. The study was approved by the Ethics Committee of the Saarland Physicians Council (No. 270-17). Firstly, bone marrow-derived hMSCs were isolated and expanded as previously described [13][14][15][16][17][18][19]. Briefly, freshly isolated hMSCs were washed, centrifuged in Dulbecco's modified Eagle's medium (DMEM) and the pellet obtained was resuspended in a mixture of red blood cell lysing buffer and DMEM in equal ratios. The mixture was washed, pelleted, and resuspended in DMEM containing 10% fetal bovine serum with 100 U/mL penicillin and 100 µL/mL streptomycin. Cells were plated in T75 flasks and kept incubated overnight at 37 • C under 5% CO 2 , and the medium was removed and replaced by growth medium with recombinant FGF-2 (1 ng/mL), with a medium change every 2-3 days. Cells were tested at passages 1-2 in order to avoid a shift in cell phenotype.
Human osteoarthritic (OA) cartilage biopsies excluding unaffected and fibrocartilage (n = 9; 6-mm diameter; Mankin score = 7-9) randomly collected from the femoral condyle of patients undergoing total knee arthroplasty were used to create standardized osteochondral defects (OCDs) with a 1-mm biopsy punch [20,21] and briefly kept in growth medium for 2-3 days (to avoid a shift in cell phenotype in the tissue) prior to direct implantation of the aggregates.

Transgene Expression
The expression of the candidate TGF-β and IGF-I gene products was assessed by specific ELISAs. Briefly, the OCDs were washed twice with serum-free medium and kept in 750 µL DMEM for 24 h. Culture supernatants were then collected and centrifuged and tested following the manufacturer's recommendations on a GENios spectrophotometer/fluorometer (Tecan, Crailsheim, Germany) [15][16][17]24].

Histological and Immonohistochemical Analyses
The OCDs were harvested after 21 days, fixed in formaldehyde (4%), washed with PBS, decalcified in 10% sodium citrate, 25% formic acid for two weeks at room temperature, and dehydrated in graded alcohol [21]. Histological and immunohistochemical analyses were performed on paraffin-embedded sections of OCDs (10 µm). Sections were stained with hematoxylin and eosin (H&E) (cellularization) and safranin O (matrix proteoglycans) [21]. The expression of type-II, -I, and -X collagen was monitored by immunohistochemical detection using specific primary antibodies, biotinylated secondary antibodies, and the ABC method with diaminobenzidine (DAB) as a chromogen [21]. Samples were analyzed under light microscopy (Olympus BX 45, Hamburg, Germany).

Histomorphometric Analyses
The transduction efficiencies (% of TGF-β + or IGF-I + immunoreactive cells to the total cell number), the cell densities (cells/mm 2 ), and the intensities of safranin O staining and those of type-II, -I, and -X collagen immunostaining (ratio tissue surface showing positive immunoreactivity for particular collagen/total tissue surface) were measured using the SIS analysisSIS program (Olympus), Adobe photoshop (Adobe Systems, Unterschleissheim, Germany), Scion Image (Scion Corporation, Frederick, MD, USA) at three random sites standardized for their surface or using serial histological or immunohistochemical sections [21]. Sections were also scored to evaluate the cartilage surrounding the OCDs (magnification × 10) and the implanted aggregates (magnification × 20) for uniformity and intensity using a modified Bern Score grading system [19,21]: 0 (no staining), 1 (heterogenous and/or weak staining), 2 (homogeneous and/or moderate staining), 3 (homogeneous and/or intense staining), 4 (very intense staining).

Biochemical Analyses
The OCDs were collected after 21 days of culture and digested in papain to determine the DNA contents by using the Hoechst 33,258 fluorometric assay, the proteoglycan contents by binding to the dimethylmethylene blue dye (DMMB), and the total cellular protein contents for normalization with the Pierce protein assay (Pierce Thermo Scientific, Fisher Scientific, Schwerte, Germany) [21]. Measurements were performed with a GENios spectrophotometer/fluorometer (Tecan).

Real-Time RT-PCR Analyses
Total RNA was extracted from the OCDs containing the aggregates but excluding bone tissue (RNeasy Protect Mini kit, Qiagen, Hilden, Germany) for reverse transcription (1st Strand cDNA Synthesis kit, avian myeloblastis virus (AMV) (Roche Applied Science, Mannheim, Germany). cDNA amplification was performed via SYBR Green real-time RT-PCR (Stratagene, Agilent Technologies, Waldbronn, Germany) maintaining initial incubation (95 Control conditions included reactions using water and non-reverse-transcribed mRNA. The specificity of the products was confirmed by melting curve analysis and agarose gel electrophoresis. The threshold cycle (Ct) value for each gene of interest was measured for each amplified sample using MxPro QPCR software (Stratagene), and values were normalized to GAPDH expression by using the 2 −∆∆Ct method [13][14][15][16][17].

Statistical Analysis
Each condition was performed in duplicate in three independent experiments for each patient. All hMSCs aggregates were tested for all experiments using the human OCDs. Results are expressed as the mean ± standard deviation (SD). A t-test was employed with p < 0.05 considered statistically significant.

Effective rAAV-Mediated Co-Overexpression of TGF-β and IGF-I in Human Osteochondral Defects Upon Implantation of Genetically Modified hMSC Aggregates
The rAAV vectors were first applied to hMSC aggregates (n = 3) and (co-)transduced aggregates were then directly implanted in human OCDs (n = 3) according to the study design ( Figure 1) to test the ability of rAAV to co-overexpress the therapeutic TGF-β and IGF-I candidates (TGF-β/IGF-I) over time (21 days) versus independent gene transfer (TGF-β/lacZ or IGF-I/lacZ) and control (lacZ) condition (endogenous growth factor expression). Effective expression of TGF-β and IGF-I via rAAV was noted in the specific conditions examined, especially in the TGF-β/IGF-I OCDs ( Table 1, Table 2 and Figure 2).

Biological and Chondrogenic Differentiation Activities in Human Osteochondral Defects Mediated by Co-Overexpression of TGF-β and IGF-I via rAAV in Implanted hMSC Aggregates
The hMSC aggregates (n = 3) transduced with the various rAAV vectors were next directly implanted in human OCDs (n = 3) to determine the capacity of the TGF-β/IGF-I combination to

Biological and Chondrogenic Differentiation Activities in Human Osteochondral Defects Mediated by Co-Overexpression of TGF-β and IGF-I via rAAV in Implanted hMSC Aggregates
The hMSC aggregates (n = 3) transduced with the various rAAV vectors were next directly implanted in human OCDs (n = 3) to determine the capacity of the TGF-β/IGF-I combination to enhance the proliferative, anabolic, and chondrogenic activities in sites of transplantation in situ over time (21 days) versus independent gene (TGF-β/lacZ or IGF-I/lacZ) and control (lacZ) gene treatments.

Influence of Co-Overexpression of TGF-β and IGF-I via rAAV Upon the Osteogenic and Hypertrophic Differentiation Activities in Human Osteochondral Defects Upon Implantation of Genetically Modified hMSC Aggregates
The hMSC aggregates (n = 3) transduced with rAAV were then directly implanted in human OCDs (n = 3) to determine the potential influence of the TGF-β/IGF-I combination on the osteogenic and hypertrophic activities in transplanted sites in situ over time (21 days) versus independent gene delivery (TGF-β/lacZ or IGF-I/lacZ) and control (lacZ) treatment.

Discussion
Cell-based strategies using the application of progenitor cells to focal cartilage defects may be of strong therapeutic value to enhance the intrinsic healing processes in sites of cartilage damage [7], especially following modification of such cellular platforms by gene therapy procedures that aim at stably triggering their chondroreparative activities [8,9]. Here, we tested the ability of the clinically adapted rAAV gene transfer vectors to co-deliver the highly chondrogenic TGF-β and IGF-I growth factors [11,15,16] to hMSCs for implantation in a model of human OCD that provides a natural environment similar to that noted in patients in vivo [20,21] as a means to repopulate the defects upon stimulation of the local healing processes relative to independent gene treatments and control (reporter gene vector) condition.
The data first indicate that rAAV is capable of successfully promoting the concomitant, sustained overexpression of TGF-β with IGF-I in hMSCs transplanted in human OCDs over the entire period of evaluation (at least 21 days) relative to single and reporter gene treatments, in agreement with the ability of rAAV to transduce and be maintained in such targets [11,[13][14][15][16], especially in a simultaneous fashion [17]. Of note, the amounts of each respective growth factor produced in the OCDs were significantly higher upon the application of hMSCs co-transduced with TGF-β/IGF-I relative to single, independent candidate gene treatments at similar vector doses, suggesting a possible reciprocal regulation of growth factor expression as previously noted in articular chondrocytes [25]. The production levels of TGF-β and IGF-I raised from implanted TGF-β/lacZ-or IGF-I/lacZ-treated hMSCs were comparable to those measured in our previous work using solely rAAV-hTGF-β or rAAV-hIGF-I in hMSCs in vitro at equivalent MOIs (~70-300 pg/mL at MOI = 8) [15,16] or the current combination in settings in vitro aiming at directly rejuvenating the cells (~50-100 pg/mL at MOI = 8) [18]. Interestingly, transgene (co)-expression was also noted in the cartilage surrounding the defects where the respective candidate treatments were applied, showing the potentially beneficial production of the growth factors via rAAV-modified, transplanted cells at a distance. It remains to be seen whether too high and/or prolonged levels of growth factor co-overexpression may have a detrimental impact on tissue repair in this model system or in vivo-however, such an issue might be addressed by applying various (lower) vector doses or using regulatable (tetracycline-controlled) or tissue-specific promoters (sox9, type-II collagen) [26] instead of the strong CMV-IE regulatory element. While no immune responses were expected in such a model ex vivo as it lacks the full spectrum of immune components, they may occur in vivo [27] although to a low extent as rAAV has been reported for its absence of deleterious effects when directly applied to focal defects in vivo [28].
The results next show that significant, prolonged TGF-β/IGF-I co-overexpression from co-transduced, implanted hMSCs triggered the biological and chondrodifferentiation activities in human OCDs over time (cell proliferation, proteoglycan/type-II collagen deposition) more effectively than when using each individual candidate gene treatment and versus control (lacZ) condition, as a possible growth factor intercooperation [25] and concordant with the properties of TGF-β and IGF-I [5,6,11,29] and with our work using the vectors independently [15,16] or simultaneously in a direct transfer approach [18]. Such effects were also observed in the cartilage surrounding the defects, likely due to the expression of the candidate transgenes at a distance and/or to the paracrine effects of the factors secreted. A degree of variability was noted between samples, reflecting the different features and background of the individual donors, but in good agreement with previous observations by Mackay et al. [6] who reported similar discrepancies between human donors. Of note, as the gene expression profiles were tested in bone-free OCD-implant systems due to the limited size of the cell-confined aggregates, the contribution of the implants to such profiles remains to be determined [20] versus a possible impact of the (slightly less active) chondrocytes in the surrounding cartilage. Despite such results, integration between the implant and the host cartilage and self-repair (even for a small, 1-mm lesion) did not fully occur, in agreement with our previous observations using such a model system [20], an outcome that might be achieved at much longer time points [30,31].
Finally, the data demonstrate the ability of combined rAAV-mediated TGF-β/IGF-I co-overexpression in implanted hMSCs to prevent undesirable hypertrophic, terminal differentiation, and osteogenic differentiation as well as catabolic processes over time in treated OCDs relative to single candidate gene treatments and the control lacZ condition, probably due to higher levels of SOX9 expression, a transcription factor known for its anti-hypertrophic and anti-catabolic activities [14,32], and in agreement with our findings using a direct approach [18]. This is in contrast with our previous findings when applying either rAAV-hTGF-β or rAAV-hIGF-I to hMSCs in vitro [15,16] yet in these studies, a higher MOI was provided to the cells (MOI = 20 while we used an MOI = 8 here, i.e., a 2.5-fold difference), showing that the optimal treatment will be a function of the vector doses applied. Here also, these effects were seen in the cartilage surrounding the defects, again probably resulting from the expression of the transgenes at a distance and/or to paracrine effects of the products.

Conclusions
The present findings show the benefits of cell-and gene-based therapy using multiple therapeutic rAAV gene transfer procedures in order to regenerate human articular cartilage lesions with an adapted phenotype in a natural environment. Work is ongoing to evaluate the potential benefits of this indirect regenerative approach in clinically relevant animal models of (orthotopic) cartilage defects in vivo and to evaluate possible immune responses [33,34]. Overall, the current data provide a basis for the future treatment of focal cartilage defects in human individuals using the patient's self-reparative progenitor cells.