Recombinant human bone morphogenetic proteins (rhBMPs) have been used in several recent studies to elicit the healing of bone defects. BMP-2, 3 (osteogenin), 4, 5, 6, and 7 (OP-1) each have osteoinductive potential18,19,23,26, and, when delivered with a carrier substance, they have elicited the healing of bone defects in a variety of animal models4-6,10. However, in those models, the adjacent host bone was well vascularized, with an intact soft-tissue envelope. The models do not simulate the conditions that are often encountered clinically. For example, fracture nonunions often are associated with poor bone stock, compromised vascularity, and extensive scar tissue that can inhibit bone repair. In this milieu, a single exposure to an exogenous growth factor may be insufficient to stimulate an adequate osteoinductive response.
The development of a cellular vehicle (that is, BMP-producing bone-marrow cells), created through adenoviral gene transfer, that can provide the sustained local release of BMP may enable the host to respond to this osteoinductive stimulus in a more robust fashion. Sufficient bone must be formed not only to bridge the defect but also to attain the biomechanical strength necessary for normal function.
Human gene transfer can be performed with use of either an ex vivo or an in vivo gene-transfer strategy. With the ex vivo method, the cDNA is transferred to cells in tissue culture and the genetically modified cells are subsequently administered to the patient. An alternative in vivo technique is to transfer the gene directly into the target somatic cells of the patient2,7,8,12,25,27. The type of gene-therapy system that is used clinically will depend on the location and size of the defect, the condition of the surrounding soft tissue, and the number of growth factors necessary to treat a particular clinical problem.
For several reasons, we used an ex vivo gene-transfer system with an adenoviral vector expressing rhBMP-2 to infect autologous rat-bone-marrow cells. First, adenoviral vectors are an appropriate choice if a large amount of protein must be produced for an intermediate time-period (a few days to a few weeks). Second, high titers of virus can be produced and both replicating and nonreplicating cells can be infected. Third, transient expression of protein has been a major limitation of gene therapy for chronic diseases such as cystic fibrosis, but chronic production of protein is not necessary to manage patients who have bone defects1,8,15,16,27. Therefore, the transient gene expression secondary to the episomal location of the expression cassette and the immunological reaction to adenoviral proteins produced by the virus may not be problems in this particular clinical situation2,8,20,21,27. Finally, we used the adherent fraction of bone-marrow cells as a biological carrier because they are easy to obtain and to grow in culture and they are inherently osteoinductive3,10,24. In the current study, we sought to determine whether BMP-2-producing rat-bone-marrow cells, created through adenoviral gene transfer, can be used successfully to heal segmental femoral defects in syngeneic Lewis rats.
*No benefits in any form have been received or will be received from a commercial party related directly or indirectly to the subject of this article. Funds were received in total or partial support of the research or clinical study presented in this article. The funding sources were National Institutes of Health Grant K11 AR01931-02, Orthopaedic Research and Education Foundation Grant H-950225, Veterans Administration Merit Award 015683, a Gene Therapy Seed Grant from the University of California at Los Angeles, the Milken Family Foundation, and a Howard Hughes Medical Institute physician postdoctoral fellowship (L. W.).
†Departments of Orthopaedic Surgery (J. R. L., A. D., P. McA., Y. P. L., J. M. K., and G. A. M. F.), Urology (L. W.), and Microbiology and Molecular Genetics (A. J. B. and O. N. W.), University of California at Los Angeles Medical Center, 10833 LeConte Avenue, Los Angeles, California 90095.
‡Advanced Tissue Sciences, Incorporated, 10933 North Torrey Pines, La Jolla, California 92037.
Preparation of the Adenoviral Vector
The cDNA for the rhBMP-2 (Genetics Institute, Cambridge, Massachusetts) was introduced into an adenoviral vector. A 1.5-kilobase Xba-to-Xba fragment containing the BMP-2 gene was cloned between the cytomegalovirus immediate early promoter, the SV40 splice, and the poly A sites of pACCMVpLpASR. The E1 region from nucleotide 455—3322 was replaced with BMP-2 cDNA. The recombinant BMP-2 adenovirus was constructed by cotransfection of 293 cells (complementing cell-line supplies E1A and E1B functions [Fig. 1]) with pACCMVpLpASBMP-2 and d1309 viral DNA cut with Xba1 and Cla114. The 293 cells are a human embryonic kidney fibroblast cell-line that expresses E1 proteins (key growth-regulatory proteins of the adenovirus) constitutively. Thus, a virus from which the E1 sequence has been deleted must be cotransfected with 293 cells in order to propagate (Fig. 1). Clones of recombinant virus were identified by restriction, digestion, and Southern blot analysis of Hirt DNA from infected 293 cells. The adenovirus containing the BMP-2 cDNA was plaque-purified on 293 cells three successive times and then was grown into a high-titer stock with more than 1 x 1010 plaque-forming units per milliliter assayed by plaque formation on 293 cells13,14 (Fig. 1).
Short-Term Cultures of Rat-Bone-Marrow Stroma
Bone marrow was obtained from the long bones of Lewis rats and was cultured in Dulbecco's modified Eagle's medium and 15 percent fetal calf serum (Fig. 2). The medium was changed on every third day for two weeks, and the nonadherent hematopoietic cells were aspirated from the plates. When the adherent cells were confluent, approximately 5 x 106 adherent bone-marrow cells were infected with 3.1 x 1010 plaque-forming units per milliliter of the BMP-2-containing adenovirus (MOI [multiplicity of infection] = 100) or 3.1 x 1010 plaque-forming units per milliliter of a comparable adenoviral vector expressing Escherichia coli ß-galactosidase instead of BMP-2. The supernatants of the infected bone-marrow cells were assayed for BMP-2 production with use of Western blot analysis14.
In Vivo Bone Formation at an Orthotopic Site
Approval was obtained from the Institutional Animal Care and Use Committee before any animal studies were begun. The rats were anesthetized with twenty milligrams of pentobarbital, administered intraperitoneally, and the hindlimbs were prepared for the operation under sterile conditions. The femur was approached anterolaterally, and the periosteum was incised and then removed circumferentially. A polyacetyl plate, measuring twenty-three by four by four millimeters, was fixed to the femur with four 0.99-millimeter transverse-threaded Kirschner wires and two 1.0-millimeter surgical steel cerclage wires. An eight-millimeter critical-sized full-thickness defect was created in the diaphysis with use of a one-millimeter side-cutting burr (Zimmer, Warsaw, Indiana) under irrigation with saline solution10. The experimental material was implanted in the femoral segmental defect with twenty milligrams of guanidine hydrochloride-extracted demineralized bone matrix, which has limited osteoinductive potential (Fig. 2).
All animals were killed at two months. Radiographs were made at two, four, and eight weeks. The healing of the bone defects was quantitated with use of the plain radiographs that had been made at four and eight weeks and was recorded as the percentage of the total area of the defect that was filled with new bone in each rat. Three blinded, independent observers scored the bone formation in each defect according to a 6-point scale, with 0 points indicating no bone formation; 1 point, bone filling less than 25 percent of the defect; 2 points, bone filling 25 to 50 percent; 3 points, bone filling 51 to 75 percent; 4 points, bone filling 76 to 99 percent; and 5 points, bone filling 100 percent of the defect28.
The defects were filled with one of five different materials. Defects in Group I (twenty-four femora) were filled with 5 x 106 BMP-2-producing bone-marrow cells; and those in Group II (sixteen femora), with twenty micrograms of rhBMP-2; those in Group III (twelve femora [negative controls]), with 5 x 106 ß-galactosidase-producing bone-marrow cells; those in Group IV (ten femora), with 5 x 106 uninfected bone-marrow cells; and those in Group V (ten femora), with guanidine hydrochloride-extracted demineralized bone matrix only. Guanidine hydrochloride-extracted demineralized bone matrix was used as a substrate in all groups because the inherent osteoinductive activity of demineralized bone matrix is removed, which allows the assessment of the true osteoinductive potential of the BMP-2-producing bone-marrow cells.
Histological Techniques
Forty femora were analyzed histologically (Table I). These included fourteen femora in Group I (BMP-2-producing bone-marrow cells), six in Group II (rhBMP-2), eight in Group III (ß-galactosidase-producing bone-marrow cells), six in Group IV (uninfected bone-marrow cells), and six in Group V (demineralized bone matrix only).
At the time that the animals were killed, the quadriceps muscle was removed, radiographs of the specimens were made, and the specimens were fixed in 40 percent ethanol at 4 degrees Celsius. After dehydration and cleaning, the internal fixation devices were removed and the specimens were embedded in polymethylmethacrylate. Serial cross sections, 200 micrometers thick, were cut from the central five millimeters of the defect, mounted on plastic slides, ground to a thickness of 100 micrometers, polished, and surface-stained with toluidine blue. Sagittal sections of the proximal and distal host-defect interfaces also were prepared.
Histomorphometric Analysis
The forty femora (three cross sections from each) also were analyzed histomorphometrically. With use of the Bioquant-IV system (R and M Biometrics, Nashville, Tennessee), the total cross-sectional area that contained any bone was measured by tracing a line around the perimeter of the cross section (the perimeter area) and quantifying the area of new bone that had formed within that area (the bone area). The total area of bone formation was calculated by dividing the bone area by the perimeter area. For each parameter, a mean was calculated from all cross sections of a given specimen, and that mean was subjected to statistical analysis.
Mechanical Testing
The surrounding soft tissue, pins, and polyacetyl plates were removed from ten femora in Group I, ten in Group II, and four each in Groups III, IV, and V; the ends of each femur were then mounted in methylmethacrylate. Each specimen was suspended in a Burstein-Frankel torsional tester. Loading of the specimens was initiated by pendulum impact. The data were recorded in real time on a storage-and-oscilloscope photographer, for subsequent data reduction. The specimens were tested until failure in torsion, and energy to failure, torsional stiffness, and torque to failure were recorded. Nine specimens from age-matched rats in which no segmental defect was created served as controls.
Statistical Methods
The radiographic findings were assessed with a nonparametric Kruskal-Wallis test comparing the distributions of ranked data in the various groups. The radiographs were reviewed by three observers, and, since the findings were based on ranked data, the kappa statistic was calculated as a measure of interobserver reliability. One-way analyses of variance were performed with use of treatment as the independent variable. When a significant F-ratio was found, the post hoc Scheffé test was used to identify significant differences among treatment groups. The results of mechanical testing were compared among all groups with one-way analysis of variance and the post hoc Student-Newman-Keuls test.
This study involved the use of a potentially novel approach for the treatment of bone loss—regional gene therapy. With use of an ex vivo gene-transfer strategy, BMP-producing bone-marrow cells successfully healed a critical-sized segmental defect in twenty-two of twenty-four femora in rats.
Various studies have shown that demineralized bone matrix or rhBMPs can be used to elicit the repair of segmental defects in animal models4-6,10,11,22,28. Both rhBMP-2 and rhBMP-7 have led to the healing of segmental defects in rodents28, rabbits4, dogs5, sheep11, and nonhuman primates6. However, these models may not accurately reflect the clinical situation, where defects are often much larger and healing is hampered by poor vascularity and by scar tissue separating the defect. In addition, the control of the duration of the availability of the growth factor is limited with present-day carrier technology. Therefore, the genetically manipulated bone-marrow cells could serve as a valuable delivery system for recombinant proteins.
As the defect in our model was larger (eight millimeters) than the five-millimeter defect that is often used with this type of model, we increased the dose of rhBMP-2 from eleven micrograms, which is typically used by investigators studying a five-millimeter defect, to twenty micrograms28. Twenty-two of the twenty-four defects that had been treated with BMP-2-producing bone-marrow cells had healed radiographically at two months, and the other two, which had not healed completely, received a score of 4 of 5 points. Both of these defects had been treated with the same batch of adenovirally infected bone-marrow cells. The Western blot assessment of BMP-2 production in these two defects revealed decreased signal (data not shown). This suggests that a threshold of BMP-2 production is required to heal defects of this magnitude. If these two femora are eliminated from the biomechanical analysis, then no significant difference between the defects treated with BMP-2-producing bone-marrow cells and the intact femora (controls) can be detected with respect to either energy to failure or torque to failure.
The use of regional gene therapy to repair bone defects and enhance bone formation is attractive because genes can be delivered to a specific anatomical site and the duration of protein expression can be determined by the selection of a particular vector. We selected the adherent population of bone-marrow cells as our carrier cells because they have inherent osteoinductive potential and are readily available and because transplantation of human bone-marrow cells has been successful in clinical applications17. We envision a system in which bone marrow is obtained from the patient and the cells then are grown in short-term culture, genetically manipulated, and autologously reimplanted at the site of the defect.
Our goal was to determine if regional gene therapy could be used successfully to heal a critical-sized defect. We were not trying to compare the efficacy of this technique with that of treatment with rhBMP-2, so we purposely used a large dose of rhBMP-2 that would heal the defect. However, it is interesting that the pattern of bone formation in the defects that had received BMP-2-producing bone-marrow cells was different from that in the defects that had been treated with rhBMP-2. The typical histological picture in the former group was one of robust, coarse trabecular bone spanning the defect, whereas that in the latter group was one of thin, lace-like bone spanning the defect. This difference may be secondary to the kinetics of protein release or to the use of the rat-bone-marrow cells as a carrier, or both. The osteoinductive stimulus associated with the BMP-2-producing bone-marrow cells may be enhanced because the BMP-2 protein is released continuously.
In our model, the bone-marrow cells were the protein-delivery system, but another alternative is to inject the adenovirus vector or even plasmid DNA directly into the specific anatomical site. Recently, Fang et al.9 reported that a BMP-4 cDNA construct, which was delivered to a five-millimeter-thick segmental bone defect in rats by loading on a gene-activated matrix, provided sufficient BMP-4 to augment healing. Limited biomechanical testing was performed on these specimens. The advantages of this delivery system with use of plasmid DNA are that it is relatively simple, no viruses are used, and it may be more cost-effective than ex vivo gene transfer since it is not necessary to obtain or reimplant cells. However, this technique requires the BMP-4 plasmid to transfect local fibroblasts, which may be more difficult in the typical clinical situation, where the fracture nonunion is associated with scar tissue that needs débridement. In addition, the surrounding soft tissue, muscle, and bone often have compromised vascularity, which could limit the response to the transfected fibroblasts14. Further investigation of this innovative treatment strategy is necessary.
Although the results of regional gene therapy appear promising, a number of questions remain to be answered before this strategy can be applied to humans. The duration and amount of protein production in vivo are unknown. Other concerns include the safety of the adenoviral vector, the potential immunological response to adenoviral proteins, and the fate of the BMP-2-producing bone-marrow cells after implantation. Investigations of these issues are currently under way in our laboratory.
In summary, this study demonstrated that BMP-2-producing cells created through adenoviral gene transfer produce sufficient protein to heal a segmental bone defect. We also established the feasibility of ex vivo gene transfer with use of biologically active autologous short-term cultures of bone-marrow cells. This technology eventually could be adopted to enhance bone repair in humans.
NOTE: The authors thank Frederick Dorey, Ph.D., for his support with respect to the statistical analysis.