Expression of human bone morphogenic protein 7 in primary rabbit ...

Gene Therapy (1998) 5, 1098–1104  1998 Stockton Press All rights reserved 0969-7128/98 $12.00 http://www.stockton-press.co.uk/gt

Expression of human bone morphogenic protein 7 in primary rabbit periosteal cells: potential utility in gene therapy for osteochondral repair JM Mason1, DA Grande2, M Barcia1, R Grant3, RG Pergolizzi1 and AS Breitbart3 1

Viral Vector Laboratory, Department of Research, and Divisions of 2Orthopedics and 3Plastic and Reconstructive Surgery, Department of Surgery, North Shore University Hospital–New York University School of Medicine, Manhasset, NY, USA

A commonly encountered problem in orthopedics is bone and cartilage tissue injury which heals incompletely or without full structural integrity. This necessitates development of improved methods for treatment of injuries which are not amenable to treatment using current therapies. An already large and growing number of growth factors which play significant roles in bone remodeling and repair have been identified in the past few years. It is well established that bone morphogenic proteins induce the production of new bone and cartilage. An efficient method of delivery of these growth factors by conventional pharmacological means has yet to be elucidated. We wished to evaluate the use of retroviral vector-mediated gene transfer to deliver genes of therapeutic relevance for bone and cartilage repair. To determine the feasibility of using amphotropically packaged retroviral vectors to transduce primary rabbit mesenchymal stem cells of periosteal origin, primary periosteal cells were isolated from New Zealand white rabbits, transduced in vitro with a retroviral vector bearing both the nuclear localized lacZ marker gene and the neor gene, and selected in G418. We used a convenient model for analysis of in vivo stability of these cells which were seeded on to polymer scaffold grafts and implanted into rabbit femoral osteochondral defects. The nuclear localized ␤-galactosidase

protein was expressed in essentially 100% of selected cells in vitro and was observed in the experimental explants from animals after both 4 and 8 weeks in vivo, while cells transduced with a retroviral vector bearing only the neor gene in negative control explants showed no blue staining. We extended our study by delivering a gene of therapeutic relevance, human bone morphogenic protein 7 (hBMP-7), to primary periosteal cells via retroviral vector. The hBMP7 gene was cloned from human kidney 293 cell total RNA by RT-PCR into a retroviral vector under control of the CMV enhancer/promoter. Hydroxyapatite secretion, presumably caused by overexpression of hBMP-7, was observed on the surface of the transduced and selected periosteal cells, however, this level of expression was toxic to both PA317 producer and primary periosteal cells. Subsequently, the strong CMV enhancer/promoter driving the hBMP-7 gene was replaced in the retroviral vector by a weaker enhancer/promoter from the rat ␤-actin gene. Nontoxic levels of expression of hBMP-7 were confirmed at both the RNA and protein levels in PA317 producer and primary periosteal cell lines and cell supernatants. This work demonstrates the feasibility of using a gene therapy approach in attempts to promote bone and cartilage tissue repair using gene-modified periosteal cells on grafts.

Keywords: gene therapy; bone repair; retroviral vector; orthopedics; BMP-7; tissue engineering

Introduction Insufficient repair of bone and cartilage tissue injuries remains a common bane of orthopedics. Many injuries such as complex fractures, fusion of vertebrae, and damage caused by surgical removal of defects or tumors require considerable amounts of bone grafting. Excessive amounts of bone graft material is often needed, necessitating the use of allogenic and artificial bone grafts in addition to or instead of autogenic bone grafts. Autogenous bone grafting does not supply large amounts of bone for grafts and greatly prolongs the duration of surgical procedures while adding to the post-operative pain and discomfort of the patient. Allogenic bone materials often Correspondence: JM Mason, Viral Vector Laboratory, Department of Research, North Shore University Hospital, 350 Community Drive, Manhasset, NY 11030, USA Received 13 November 1997; accepted 16 March 1998

do not have the desired properties of autogenic bone especially with regard to durability. Artificial bone materials such as calcium carbonate, collagen, coral, glass, and other synthetic ceramics are in development but their value as graft materials has yet to be definitively assessed.1 The past few years have seen the discovery of a large number of growth factors involved in bone tissue formation and repair. These factors have been isolated from many different tissues with the bone matrix serving as their main repository.2 To date, 12 different but related bone morphogenic proteins (BMP 1–12) have been identified.3–7 BMPs act on mesenchymal stem cells causing their differentiation down the osteoblastic and chondroblastic pathways. These factors affect osteoblast and osteoclast metabolism in the bone matrix to modulate the bone tissue repair response after injury. Thus, BMPs are expressed early on in the fracture healing process8,9 and act at the local cellular environment in the bone matrix

Gene transfer to primary rabbit periosteal cells JM Mason et al

to stimulate the healing response after bone injury.10,11 BMP-7, in particular, has been shown to be a powerful agent in the repair of ulna defects in both rabbit and monkey models.12,13 In addition, BMP-7 used with collagen can promote bone repair similar to autologous bone in a spine fusion model in dogs.14 Although BMPs have demonstrated their utility in bone formation and repair in several in vivo models, their clinical utility is limited by the lack of an efficient delivery system for these and other growth factors.15 BMPs do not work well in solution, thereby requiring the use of carriers such as demineralized bone, collagen, polysaccharide matrices and ceramics, for biological activity.16 BMPs have their peak effects when delivered locally in milligram quantities which is difficult to achieve using standard pharmacological delivery methods and, in addition, BMP delivered in this way will be effective for only a relatively short duration.1 The delivery of genes regulating bone tissue repair locally to the site of injury may resolve many of these problems. This article demonstrates the feasibility of delivering and expressing a marker gene using amphotropically packaged retroviral vectors in primary periosteal cells in vitro and in vivo using a convenient model of bone and cartilage repair, the full thickness cartilage defect model. We also report the cloning and expression of RNA and protein of the human BMP-7 gene in retroviral vector producer cell lines and transduced primary periosteal cells as well as the unexpected overproduction of hydroxyapatite in transduced periosteal cells. These results demonstrate the feasibility of further in vivo experimentation toward a gene therapy approach to promote bone and cartilage tissue repair using gene-modified periosteal cells.

Results In vitro and in vivo marker studies We wished to assess the feasibility of using a gene transfer approach in delivering genes to primary rabbit periosteal cells via retroviral vector, seeding the gene-modified cells on to grafts, and detecting expression of these genes in vitro and in explants of these grafts after several weeks in a rabbit osteochondral defect model.17,18 We used the amphotropically packaged retroviral vector, LZ1219 to deliver to primary New Zealand white rabbit periosteal cells the neor gene and the nuclear localized lacZ gene. The LZ12 supernatant had a neo titer of approximately 1 × 105 c.f.u./ml on NIH 3T3 cells. We did not determine the percentage of transduced periosteal cells which were X-gal positive before selection, however, after full selection in G418, essentially 100% of these cells stained positively with X-gal (Figure 1a). None of the control cells which were fully G418 selected after transduction with amphotropically packaged LNCX20 stained positively with X-gal (Figure 1b) demonstrating that primary rabbit periosteal cells can be effectively transduced with retroviral vectors and stably express transgenes in vitro. Fully G418 selected periosteal cells, gene modified with either LNCX or LZ12, were seeded on to polymer scaffold grafts and implanted into distal femoral osteochondral defects in New Zealand white rabbits as previously described.18 Grafts remained in the animals for either 4 or 8 weeks at which time they were removed and stained

1099

Figure 1 Nuclear ␤-galactosidase expression in primary rabbit periosteal cells in vitro and in explanted grafts after 4 and 8 weeks in vivo. All samples have been fixed and X-gal stained. (a) G418 selected primary rabbit periosteal cells transduced with the LZ12 retroviral vector before implant (×100, out of phase to optimize visualization of blue nuclei). (b) G418 selected primary rabbit periosteal control cells transduced with the LNCX retroviral vector showing no nuclear staining (×100, in phase). (c) Explanted graft containing LZ12-transduced periosteal cells after 4 weeks in vivo (×50). (d) Explanted control graft containing LNCX-transduced periosteal cells after 4 weeks in vivo (×50). (e) Explanted graft containing LZ12-transduced periosteal cells after 8 weeks in vivo (×100). (f) Explanted control graft containing LNCX-transduced periosteal cells after 8 weeks in vivo (×100).

with X-gal. Positive staining was observed in LZ12-transduced grafts at both 4 weeks (Figure 1c) and 8 weeks (Figure 1e) while negative control LNCX-transduced grafts from the same animals showed no staining (Figures 1d and f). Approximately 10% of the cells are Xgal positive after both 4 and 8 weeks in vivo. These results demonstrate that in vivo transgene expression occurs for at least 8 weeks in this model, which may be long enough for BMP genes to promote bone tissue repair.

Cloning and expression of hBMP-7 Although its physiological site of action is in bone, the kidney is the main site of synthesis for BMP-721 (also known as osteogenic protein one or OP-1). Therefore, we isolated total RNA from 293 cells, a human kidney cell line, and used it as a template for RT-PCR using oligonucleotide primers NS 30 and NS 32. These primers generated an approximately 1.4 kb DNA fragment which contains the entire coding sequence of hBMP-7 cDNA from the Kozak translation start region to the translation stop codon inclusive. The PCR primers were designed based on the published sequence of human BMP-722 to eliminate additional 5 prime and 3 prime untranslated sequence. The entire hBMP-7 cDNA was sequenced and exactly matched the published sequence. The hBMP-7 gene was cloned into the retroviral vector plasmid LNCX under control of the CMV immediate– early enhancer/promoter. Transiently expressed amphotropically packaged retroviral vector particles were gen-


1100

erated by CaPO4 transfection of PA31723 cells with plasmids LNCX-hBMP-7 and LNCX as control (Figure 2). These transiently produced vector particles were used to transduce primary rabbit periosteal cells. Upon selection in G418, the PA317-LNCX-hBMP-7 cells died while the control LNCX-transfected cells survived suggesting toxicity of the hBMP-7 gene product. This toxicity was confirmed when the periosteal cells were selected in G418. The LNCX-transduced periosteal cells survived while the LNCX-hBMP-7-transduced cells produced a thick coating of material with concomitant loss of adherence from the plastic tissue culture ware (data not shown). Aliquots of these nonadherent cells were washed with PBS which dissolved the thick coating suggesting the material could be hydroxyapatite. Samples of the cells were analyzed and the material confirmed to contain hydroxyapatite by electron radiographic diffraction (data not shown). Because overexpression of hBMP-7 was toxic to cells, we replaced the strong CMV enhancer/promoter with the weaker rat ␤-actin enhancer/promoter24 to generate the LN␤-actin-hBMP-7 retroviral vector plasmid (Figure 2). The entire rat ␤-actin enhancer/promoter from plasmid pJ6Omega was sequenced and although numerous differences from the published sequence were noted, the sequence was clearly that of the rat ␤-actin enhancer/ promoter. CaPO4 transfection of PA317 cells with the LNCX and LN␤-actin-hBMP-7 plasmids followed by selection in G418 generated stable producer cell lines, demonstrating that the level of hBMP-7 produced in these cell lines is well tolerated. Retroviral vector particles were collected from these stable producer cell lines and used to transduce primary rabbit periosteal cells. Once fully selected in G418, the RNA and protein

Figure 2 Schematic representation of retroviral vectors used in this study. All constructs are based on LNCX with the 5′ LTR driving expression of the selectable neomycin resistance gene (Neo) which confers resistance to the neomycin analog G418. CMV is the immediate–early promoter derived from human cytomegalovirus. HindIII and HpaI are the restriction sites in LNCX used for cloning the hBMP-7 gene. hBMP-7 is the human bone morphogenic protein 7 gene. B-ac is the rat ␤-actin promoter derived from plasmid pJ6Omega. Arrows indicate the RNA species and their sizes relevant to this study.

expression of hBMP-7 in these producer and periosteal cells was analyzed. Total RNA was isolated from a G418-resistant PA317 cell population which had been transfected by plasmid LN␤-actin-hBMP-7, as well as from G418 selected periosteal cells which had been transduced by the amphotropically packaged LNCX and LN␤-actin-hBMP-7 retroviral vectors. Both the PA317 and periosteal cell populations containing LN␤-actin-hBMP-7 express the expected approximately 4.7 kb genomic length RNA as well as the approximately 2.3 kb internally promoted hBMP-7 mRNA when probed for hBMP-7 (Figures 2 and 3a). The control lane of LNCX-transduced periosteal cells does not express hBMP-7 RNA demonstrating the specificity of the hBMP-7 probe (Figure 3a). Levels of expression of the two species of hBMP-7 RNA appear to be equal in periosteal cells while in the PA317 cells, the larger genomic length species appears to be more abundant. This suggests that, as expected, the ␤-actin enhancer/promoter is weaker than the retroviral LTR, at least in PA317 cells. It is not uncommon for different promoters to express to different degrees in various cell types and, as opposed to the case with PA317 cells, we observed that the ␤-actin promoter generates the same steady-state level of RNA as does the LTR in periosteal cells. We presume that the CMV promoter generated high steady-state levels of BMP-7 mRNA in LNCXhBMP-7-transduced periosteal cells resulting in overexpression of BMP-7 and cell death. However, we did not verify this by attempting to isolate RNA from these dying cells, but rather continued our studies with the healthy LN␤-actin-hBMP-7-transduced periosteal cells instead. In any event, the ␤-actin promoter worked equally as well as the LTR in periosteal cells. In addition, the RNA from the PA317-LN␤-hBMP-7 cell line contains some spurious bands while RNA from the periosteal cells does not. This is probably due to the fact that the PA317 cells were transfected, which results in integration of multiple copies of rearranged genomes while the periosteal cells were transduced, which generally results in single copy integration of nonrearranged genomes. As a control, we analyzed an analogous blot using a probe against the neor

Figure 3 Northern blot analysis of total RNA from transduced and fully G418 selected populations of periosteal cells and PA317 producer cells. (a) Hybridization with a hBMP-7 probe: lane 1, LN␤-actin-hBMP-7transduced periosteal cells expressing the 4.7 kb and 2.3 kb RNA species; lane 2, LNCX-transduced periosteal cells; lane 3, LN␤-actin-hBMP-7transduced PA317 cells expressing the 4.7 kb and 2.3 kb RNA species. (b) Hybridization with a Neo probe: lane 1, LN␤-actin-hBMP-7-transduced periosteal cells expressing the 4.7 kb RNA species; lane 2, LNCX-transduced periosteal cells expressing the 3.6 kb RNA species; lane 3, LN␤actin-hBMP-7-transduced PA317 cells expressing the 4.7 kb RNA species.


gene (Figure 3b). In this instance, a single band of approximately 4.7 kb is observed in the LN␤-actin-hBMP7 containing cells (lanes 1 and 3), while the LNCX-transduced periosteal cells in lane 2 express the smaller approximately 3.6 kb species. The bands from PA317LN␤-actin-hBMP-7 (lane 3 in Figure 3a and b) are less intense than those for the periosteal cells. The same amounts of RNA were loaded per well on the gels as determined by ethidium bromide staining (data not shown) suggesting that steady-state levels of these RNA species are higher in periosteal cells than in PA317 cells. Protein levels of hBMP-7 were analyzed by ELISA using a monoclonal antibody directed against human BMP-5. Human BMP-5, 6 and 7 are 90% identical in amino acid sequence7 and this antibody is known to have cross-reactivity to hBMP-7 by ELISA (personal communication from Genetics Institute, Immunology Department, Cambridge, MA, USA). Conditioned optimem media collected from fully G418 selected populations of transduced PA317 and periosteal cells were assayed by ELISA for secreted hBMP-7. Results clearly show that the LN␤hBMP-7-transduced PA317 and periosteal cells secrete hBMP-7 into the culture media at levels six- to eight-fold higher than the negative control LNCX transduced cells (Figure 4). Optimal results were obtained using 24 h conditioned media from PA317 cells, while 72 h conditioned media were optimal for periosteal cells. No expression above background could be detected using 24 h conditioned media from periosteal cells or appreciable levels from 72 h PA317-conditioned media (data not shown). This may reflect proteolytic processing of secreted BMP7 precursor which must occur for it to be detected by our antibody. PA317 cells are fibroblasts that secrete proteases which can quickly process precursor to the mature form but which may also degrade the mature form over extended periods. The periosteal cells, on the other hand,

Figure 4 ELISA analysis of hBMP-7 secreted into culture medium by PA317 producer cells and periosteal cells. Conditioned optimem medium was collected from PA317 cells after 24 h in culture and used undiluted, and at 1:2 and 1:4 dilutions in optimem. Conditioned optimem medium was collected from periosteal cells after 72 h in culture and assayed undiluted. All assays were done in triplicate and repeated at least twice giving similar results. Error bars reflect standard errors of the mean.

may not secrete proteases well, thus requiring a few days for fully processed BMP-7 to accumulate in culture media. Unfortunately, no source of purified hBMP-7 is available to us, therefore, we could not quantify the exact amount of hBMP-7 secreted by these cells. Results of in vivo experimentation will determine efficacy of this approach. These results demonstrate that cells gene modified by hBMP-7 are able to produce and secrete BMP-7 into culture media while cells not gene modified by BMP-7 do not secrete appreciable levels of BMP-7.

Discussion This study demonstrates the feasibility of using retrovirus-mediated gene transfer to express transgenes both in vitro and in vivo in an animal model of bone and cartilage repair. Primary rabbit periosteal cells can be gene modified ex vivo, seeded on to grafts, and reimplanted into femoral osteochondral defects with persistence of transgene expression for at least 8 weeks in vivo. Approximately 10% of cells in the defect area continue to express the lacZ gene at high levels as determined by X-gal staining. We presume that the remaining cells either express lacZ at reduced levels from the RSV promoter in LZ12 or that the promoter is shut down as we found no evidence of preferential cell death of engrafted cells and observed no immune response to the transduced cells. Minimal colonization of the tissue construct from surrounding tissue may also occur but not to a great enough extent to account for this lowered percentage of lacZ-positive cells. Regardless, this duration of expression is likely to be long enough for bone morphogenic proteins serving as the transgene to achieve clinical benefit as BMPs are expressed early on and are known to be efficacious in the bone repair process.8,9,14 An added advantage of using a gene delivery approach is that locally high levels of BMPs can be easily delivered to the wound for extended periods of time. This is not possible using traditional pharmacological approaches for delivery of BMP proteins which have additional complicating factors such as the required use of carriers. Other groups studying cartilage repair have recently demonstrated the ability to transfer and express genes in chondrocytes in vivo in cartilage defect models25,26 while another study has used nongene-modified periosteal cells in cartilage repair models.27 In an attempt to develop a gene therapy approach for bone and cartilage repair by delivering a clinically relevant gene, we have demonstrated that retroviral vectormediated delivery of the hBMP-7 gene can be used to transduce primary periosteal cells in vitro. However, our results demonstrate the importance of preventing overexpression of hBMP-7 in target cells due to its apparent toxicity. We do not know if overexpression of the hBMP7 gene from the CMV enhancer/promoter is toxic to periosteal cells per se or whether periosteal cells just do not grow well in suspension. This is because periosteal cells transduced with the CMV-driven hBMP-7 vector secreted hydroxyapatite on their surface causing their release from adherent growth. Regardless of whether or not these periosteal cells could survive in suspension, any realistic clinical utility of this retroviral vector requires that producer cell lines are generated. Our CMV-driven hBMP-7 construct killed the PA317 producer cells,

1101


1102

thereby requiring a modification of the vector to decrease hBMP-7 expression levels. The LN␤-actin-hBMP-7 construct resolved these issues as both retroviral vector producer cell lines and periosteal cells transduced by this vector grow and survive well. The producer and periosteal cells, each transduced with LN␤-actin-hBMP-7, express hBMP-7 at the mRNA level demonstrating that the rat ␤-actin enhancer/ promoter functions well in these cells and equivalently well to the LTR in periosteal cells. In addition, the ␤-actin enhancer/promoter, being from a cellular housekeeping gene, may avoid problems of promoter shut down sometimes observed with the use of viral promoters. Indirect ELISA demonstrated that hBMP-7 is secreted into media by both the PA317 producer cells and periosteal cells at levels several fold over background levels generated from LNCX-transduced negative control cells. The secreted hBMP-7 should accumulate to locally high concentrations in wounds and effect not only the transduced periosteal cells themselves, but also have positive effects on neighboring nontransduced cells, in aiding the wound repair process. Experiments to evaluate the effect of hBMP-7 expression and secretion in vivo on gene-modified periosteal cells and on neighboring cells at the site of injury are currently underway using a rabbit cranial defect model of bone repair. Encouraging preliminary results suggest that bone repair is significantly augmented using the genemodified periosteal cells in this model. Presentation of these in vivo results will await completion of the study.

Materials and methods Cell culture Primary periosteal cells were obtained from male New Zealand white rabbits. Periosteal tissue was harvested aseptically from the proximal anteromedial portion of the tibia using sharp periosteal elevators. Extreme care was used in harvesting the inner cambium layer. Collected tissue was minced into 1 mm2 pieces and used as explants for initiating cultures in six-well dishes. PA317 cells were the base packaging cell line used to produce retroviral vector particles. 293 Cells were the source of RNA used for cloning hBMP-7 cDNA. All cells were cultured at 37°C in 5% CO2 in D10 medium (high glucose DMEM (GIBCO BRL, Grand Island, NY, USA) supplemented with 10% heat inactivated fetal bovine serum (GIBCO BRL) and 2 mm glutamine). G418 (GIBCO BRL) was added to D10 medium to select for transduced cells at 0.6 mg/ml active G418 for periosteal cells and at 0.3 mg/ml active G418 for PA317 cells. Fully G418-resistant cell populations were obtained after 7–14 days in selective medium. Periosteal cells were expanded from 35-mm wells of a six-well dish directly to 100 mm dishes (Becton Dickinson Labware, Franklin Lakes, NJ, USA) using 0.05% trypsin (GIBCO BRL). Other cell lines were routinely passaged using a 1:10 split ratio. Plasmid DNAs, oligonucleotide primers and RT-PCR Oligonucleotide primers NS 30 and NS 32 were manufactured on a Model 394 Oligonucleotide Synthesizer (ABIPerkin Elmer, Foster City, CA, USA) and used according to the manufacturer’s instructions in an RT-PCR, Reverse Transcription System (Promega, Madison, WI, USA) to

generate an approximately 1.4 kb cDNA fragment of the hBMP-7 gene from RNA isolated from 293 cells using the RNeasy Total RNA kit (Qiagen, Santa Clara, CA, USA). The sequences of the oligonucleotide primers were as follow: NS 30: 5′ GCGCGTAGAGCCGGCGCGATGCACGT GCGCTC 3′ NS 32: 5′ CTAGTGGCAGCCACAGGCCCGGACCA CCATGT 3′ The approximately 1.4 kb PCR product was cloned into plasmid pT7Blue (Novagen, Madison, WI, USA) as per manufacturer’s instructions. E. coli strain HB101 was generally used in electrotransformations. Plasmid pT7Blue-hBMP-7 No. 15 was sequenced to verify that the hBMP-7 gene was correct and complete. It was then double digested with HindIII and SmaI and the approximately 1377 bp fragment containing hBMP-7 was isolated. The LNCX-hBMP-7 retroviral vector plasmid was constructed by double digesting plasmid LNCX with HindIII and HpaI and cloning into this site the approximately 1377 bp HindIII–SmaI hBMP-7 fragment. The LN␤-actin-hBMP-7 retroviral vector plasmid in which the rat ␤-actin enhancer/promoter replaces the CMV enhancer/promoter was constructed by digesting LNCX-hBMP-7 with NruI, Klenow filling the ends, and digesting with HindIII to remove the fragment containing the CMV enhancer/promoter sequence. The approximately 7005 bp NruI–HindIII fragment without CMV was isolated. The rat ␤-actin enhancer/promoter was obtained from plasmid pJ6Omega (ATCC No. 37723; Rockville, MD, USA) and double digested with PvuII and HindIII. The approximately 354 bp PvuII–HindIII fragment containing the rat ␤-actin enhancer/promoter was isolated and cloned into the approximately 7005 bp NruI filled-HindIII fragment resulting in plasmid LN␤-actinhBMP-7.

Production of retroviral vector particles The LNCX, LZ12, LNCX-hBMP-7 and LN␤-actin-hBMP7 retroviral vector plasmid DNAs were used to generate retroviral vector particles from PA317 cells. PA317 cells were seeded into 8 ml of D10 medium in 100-mm dishes at 1 × 106 cells per dish and transfected the following day with 30 ␮g of LNCX, LZ12, LNCX-hBMP-7 or LN␤-actinhBMP-7 plasmid DNA using CaPO4 (5 Prime, 3 Prime, Boulder, CO, USA). The following morning, the medium was removed and cells washed with PBS (phosphatebuffered saline, pH 7.3, No Ca++ or Mg++) before replacement with 8 ml of fresh D10. Twenty-four hours postwash, transiently produced retroviral vector particles (supernatants) were collected, filtered through a 0.22 micron syringe filter (Millipore, Bedford, MA, USA), and used to transduce primary rabbit periosteal cells. Stable G418-resistant PA317 producer cell populations were generated later and used as a source of retroviral vector particles which were filtered and used as described for the transient supernatants. The neo titer on NIH 3T3 cells was approximately 5 × 104 c.f.u./ml. Transduction of cells Primary rabbit periosteal cells were plated into six-well dishes and cultured in D10 medium. When cells reached 25–50% confluence, transductions with supernatants were performed by adding 1.6 ml of fresh D10 medium


supplemented with 8 ␮g/ml polybrene (Sigma, St Louis, MO, USA) to each well. Four hundred microliters of supernatant were added to the cells and incubated overnight at 37°C. The following morning, the medium was replaced with 2 ml of fresh D10G600. As the cells became confluent in the six-well dishes, they were trypsinized, transferred to 10 or 15 cm dishes, and cultured in D10G600. When the cells became confluent, they were seeded on to polymer scaffolds for later implant into rabbit femoral osteochondral defects.

Graft seeding and implant Gene-modified periosteal cells were released from the tissue culture dishes by trypsin treatment and cultured on polymer scaffolds in six-well tissue culture dishes. Each scaffold was initially seeded with 4 × 106 cells in a volume of 100 ␮l. Samples were incubated at 37°C, 5% CO2 for 6 h to permit adhesion and entrapment of cells in the polymer scaffold. At 6 h after seeding, 2.5 ml of medium was added, followed by an additional 1.5 ml 24 h later. Medium was replaced every 3 days for the next 10–12 days. The full thickness osteochondral defect was made according to the procedure described previously.17 New Zealand white rabbits (8-month-old males weighing approximately 4.5 kg) were used according to the NIH guidelines for the care of laboratory animals. Rabbits were placed under general anesthesia with xylazine (5 mg/kg i.m.) and ketamine (35 mg/kg i.m.), shaved and scrubbed with betadine. A medial parapatellar arthrotomy was performed bilaterally with the rabbit supine. A pointed 3-mm diameter custom drill bit (Acufex, Mansfield, MA, USA) was used to create a full thickness defect (1–2 mm deep) in the femoropatellar groove (FPG). An attempt was made to extend this defect just through the subchondral plate without violating the subchondral bone. A surgical trephine (Biomedical Research Instruments, Walkerville, MD, USA) was used to core a 4-mm diameter × 2-mm thick piece of PGA matrix, and this was press-fit into the 3-mm diameter defect. The incision was closed in two layers; the fascia was closed with interrupted 4.0 vicryl (absorbable) and the skin was closed with the same interrupted 4.0 vicryl. The right knee received grafts containing cells expressing both the neo and lacZ genes, whereas the contralateral left knee received the control grafts expressing only the neo gene. The knee joints were not immobilized postoperatively, and the animals were allowed free cage activity. Rabbits were killed after 4 and 8 weeks using an overdose of pentobarbital. The protocol used was 1 ml of ketamine i.m., wait 15 min, then 5 ml of Sleep-away i.m. (Fort Dodge Lab, Fort Dodge, IA, USA). Analysis of hBMP-7 RNA and protein expression Total RNA was isolated from four confluent 100-mm or two confluent 150-mm tissue culture dishes of various cell lines using the RNeasy Total RNA kit (Qiagen). Five micrograms per well of total RNA from PA317-LN␤actin-hBMP-7 and PA317-LNCX producer cells, and rabbit periosteal cells transduced with LNCX or LN␤-actinhBMP-7 were separated on 1% formaldehyde gels and Northern blotted to nitrocellulose using standard methods. 32P-labeled random primed DNA probes were prepared from both a 1.2 kb BstEII–BstBI fragment containing the neor gene from plasmid pLNCX and from a

1.4 kb HindIII–SmaI fragment containing the hBMP-7 gene from plasmid pT7-hBMP-7. Northern blots were hybridized with either the hBMP-7 or neor gene-specific DNA probes by standard means. Twenty-four or 72 h conditioned Optimem medium (GIBCO BRL) was harvested from PA317 or periosteal cells, respectively, harboring the LN␤-actin-hBMP-7 or LNCX constructs. Indirect ELISAs were performed by adding 100 ␮l of conditioned medium undiluted and at several two-fold sequential dilutions into 96-well flat bottom maxisorp plates (Nunc, Roskilde, Denmark). Dilutions were done in optimem and all samples were assayed in triplicate. Antigen was bound for 1 h at 37°C, blocked with 200 ␮l PBS-T (PBS with 0.1% Tween 20) for 1 h at room temperature and washed three times with PBS-T. The primary antibody, anti-BMP-5 monoclonal antibody h1b5/1.8.5, was diluted 1:750 in PBS-T and 100 ␮l added per well for 1.5 h at room temperature. Three washes in PBS-T were followed by development using a biotinylated anti-mouse secondary antibody and HRP conjugate from the Elite ABC kit (Vector Laboratories, Burlingame, CA, USA) as per the manufacturer’s instructions. The chromogenic substrate tetramethylbenzidine (TMB) was used as substrate for the HRP and used according to the manufacturer’s instructions (Vector Laboratories). The plates were read at OD450 within 1 h of development using a model 400 ATC ELISA plate reader (SLT Labinstruments, Grodig, Austria). The background values of the unconditioned optimem alone wells were subtracted out before graphing (Figure 4).

Analysis of cells and grafts for nuclear-␤-galactosidase expression Cells cultured on tissue culture plastic, on grafts, or from graft material removed from rabbits after 1 or 2 months in vivo were analyzed for nuclear-␤-galactosidase expression. Cells were fixed for 30 min in 2% (w/v) paraformaldehyde and 0.4% glutaraldehyde in PBS at 4°C. Cells were subsequently washed with PBS, and stained for 2 h at 37°C in a solution of 1 mg/ml X-gal (Molecular Probes, Eugene, OR, USA), 20 mm potassium ferrocyanide, 20 mm potassium ferricyanide and 2 mm MgCl2 in PBS.

Acknowledgements We wish to thank Ryhana Manji, Patricia Conti and Debra Porti for fine technical assistance, Raymond Pica for help with the ELISAs, and members of the North Shore University Hospital core facility, Dr Dorothy Guzowski and Colleen Milan for oligonucleotide synthesis and DNA sequencing. We also thank the Immunology Department of Genetics Institute, for graciously providing antibody h1b5/1.8.5 and Dr William Landis for performing the electron X-ray diffraction work.

References 1 Lind M. Growth factors: possible new clinical tools. Acta Orthop Scand 1996; 67: 407–417. 2 Baylink DJ, Finkelman RD, Mohan S. Growth factors to stimulate bone formation. J Bone Miner Res 1993; 8 (Suppl. 2): S565S572. 3 Wang EA et al. Purification and characterization of other distinct bone-inducing factors. Proc Natl Acad Sci USA 1988; 85: 9484– 9489.

1103


1104

4 Wozney JM et al. Novel regulators of bone formation, molecular clones and activities. Science 1988; 242: 1528–1533. 5 Celeste AJ et al. Identification of transforming growth factor family members present in bone inductive protein purified from bovine bone. Proc Natl Acad Sci USA 1990; 87: 9843–9847. 6 D’Alessandro JS. Purification, characterisation and activities of rhBMP-5. J Bone Min Res 1991; 6: S153 (Abstr.). 7 Wozney JM. The bone morphogenetic protein family and osteogenesis. Mol Reprod Dev 1992; 32: 160–167. 8 Jin Y, Yang L, White FH. An immunocytochemical study of bone morphogenetic protein in experimental fracture healing of the rabbit mandible. Chin Med Sci J 1994; 9: 91–95. 9 Nakase T et al. Transient and localized expression of bone morphogenetic protein 4 messenger RNA during fracture healing. J Bone Miner Res 1994; 9: 651–659. 10 Canalis E, McCarthy TL, Centrella M. Growth factors and the regulation of bone remodeling. J Clin Invest 1988; 81: 277–281. 11 Joyce ME, Jingushi S, Scully P, Bolander ME. Role of growth factors in fracture healing. In: Anonymous (ed). Clinical and Experimental Approaches to Dermal and Epidermal Repair. WileyLiss: New York, 1991, pp 391–416. 12 Cook SD et al. Healing of large segmental defects using rhOP1. Trans ORS 1992; 581 (Abstr.). 13 Cook SD et al. rhOP-1 heals segmental long-bone defects in nonhuman primates. Trans ORS 1993; 484 (Abstr.). 14 Cook SD et al. In vivo evaluation of recombinant human osteogenic protein (rhOP-1) implants as a bone graft substitute for spinal fusions. Spine 1994; 19: 1655–1663. 15 Gerich TG et al. Gene transfer to the rabbit patellar tendon: potential for genetic enhancement of tendon and ligament healing. Gene Therapy 1996; 3: 1089–1093. 16 Lindholm TS, Gao TJ. Functional carriers for bone morphogenetic proteins. Ann Chir Gynaecol 1993; 207 (Suppl.): 3–12.

17 Grande DA et al. Repair of articular cartilage defects using mesenchymal stem cells. Tissue Eng 1995; 1: 345–353. 18 Ben-Yishay A et al. Repair of articular cartilage defects with collagen-chondrocyte allografts. Tissue Eng 1995; 1: 119–133. 19 Galileo DS et al. Neurons and glia arise from a common progenitor in chicken optic tectum: demonstration with two retroviruses and cell type-specific antibodies. Proc Natl Acad Sci USA 1990; 87: 458–462. 20 Miller AD, Rosman GJ. Improved retroviral vectors for gene transfer and expression. Biotechniques 1989; 7: 980–990. 21 Ozkaynak E, Schnegelsberg PN, Oppermann H. Murine osteogenic protein (OP-1): high levels of mRNA in kidney. Biochem Biophys Res Commun 1991; 179: 116–123. 22 Ozkaynak E et al. OP-1 cDNA encodes an osteogenic protein in the TGF-␤ family. EMBO J 1990; 9: 2085–2093. 23 Miller AD, Buttimore C. Redesign of retrovirus packaging cell lines to avoid recombination leading to helper virus production. Mol Cell Biol 1986; 6: 2895–2902. 24 Morgenstern JP, Land H. A series of mammalian expression vectors and characterisation of their expression of a reporter gene in stably and transiently transfected cells. Nucleic Acids Res 1990; 18: 1068. 25 Baragi VM et al. Transplantation of adenovirally transduced allogeneic chondrocytes into articular cartilage defects in vivo. Osteoarthritis Cartilage 1997; 5: 275–282. 26 Kang R et al. Ex vivo gene transfer to chondrocytes in full-thickness articular cartilage defects: a feasibility study. Osteoarthritis Cartilage 1997; 5: 139–143. 27 O’Driscoll SW, Fitzsimmons JS, Commisso CN. Role of oxygen tension during cartilage formation by periosteum. J Orthop Res 1997; 15: 682–687.