Gene Therapy (2000) 7, 734–739
 2000 Macmillan Publishers Ltd All rights reserved 0969-7128/00 $15.00
www.nature.com/gt
ACQUIRED DISEASES
RESEARCH ARTICLE
Genetic enhancement of fracture repair: healing of an
experimental segmental defect by adenoviral transfer
of the BMP-2 gene
AWA Baltzer1, C Lattermann2, JD Whalen2, P Wooley3, K Weiss2, M Grimm3, SC Ghivizzani4,
PD Robbins4 and CH Evans2,4
1
Department of Orthopaedic Surgery, Heinrich Heine University of Düsseldorf, Germany; 2Department of Orthopaedic Surgery,
University of Pittsburgh School of Medicine, USA; 3Department of Orthopaedic Surgery, Wayne State University, USA; and
4
Department of Molecular Genetics and Biochemistry, University of Pittsburgh School of Medicine, USA
This study evaluated the ability of gene transfer to enhance
bone healing. Segmental defects were created surgically in
the femora of New Zealand white rabbits. First generation
adenoviruses were used as vectors to introduce into the
defects genes encoding either human bone morphogenetic
protein-2 (BMP-2) or, as a negative control, firefly luciferase.
Representative specimens were evaluated histologically
after 8 weeks. Healing of the defects was monitored radiographically for 12 weeks, after which time the repair tissue
was evaluated biomechanically. By radiological criteria, animals receiving the BMP-2 gene had healed their osseous
lesions after 7 weeks, whereas those receiving the luciferase
gene had not. Histologic examination of representative rabbits at 8 weeks confirmed ossification across the entire
defect in response to the BMP-2 gene, whereas the control
defect was predominantly fibrotic and sparsely ossified. At
the end of the 12-week experiment, the control femora still
showed no radiological signs of stable healing. The difference in radiologically defined healing between the experimental and control groups was statistically significant
(P ⬍ 0.002). Biomechanical testing of the femora at 12
weeks demonstrated statistically significant increases in the
mean bending strength (P ⬍ 0.005) and bending stiffness
(P ⬍ 0.05) of the animals treated with the BMP-2 gene.
Direct, local adenoviral delivery of an osteogenic gene thus
led to the healing of an osseous lesion that otherwise would
not do so. These promising data encourage the further
development of genetic approaches to enhancing bone
healing. Gene Therapy (2000) 7, 734–739.
Keywords: bone healing; bone morphogenetic protein; direct gene delivery
Introduction
Massive segmental bony defects often occur following
trauma. A segmental defect fracture generally does not
heal, but instead ends up as a non-union if it is not
treated by extended and complicated surgical methods.
Although conventional bone grafting is currently considered to be the method of choice for the treatment of
segmental defect fractures, the procedure is often unsuccessful.1,2 Furthermore, the donor sites available in the
skeletal system for harvesting autogenous spongious or
corticospongious bone grafts are limited. Additionally,
there are a number of complications related to the surgical procedure of harvesting autologous bone grafts, such
as local infections, paresthesias, and pain at the donor
site.3 Unsatisfactory results after surgical treatment of
post-traumatic segmental bone defects are described in
up to 30% of cases.4,5
Segmental defects after tumor surgery are even more
challenging. The area of bone resection is generally large,
Correspondence: CH Evans at his present address: Center for Molecular
Orthopaedics, Harvard Medical School, 221, Longwood Avenue, BL-152,
Boston, MA 02115, USA
Received 12 October 1999; accepted 14 January 2000
all osteoconductive and osteoinductive material has been
completely removed, and often the only reasonable therapy is an amputation of an extremity because of the
complete lack of bony substance.
Different osteoinductive and osteoconductive methods
are currently under investigation for developing
adequate alternatives to bone grafting. One attractive
approach makes use of the powerful osteogenic properties of certain growth factors. Members of the transforming growth factor-␤ family of growth factors, including the bone morphogenetic proteins (BMPs), have
diverse effects on the growth and differentiation of mesenchymal cells, as well as on their ability to synthesize
matrix.6–9 An interesting feature of some of these growth
factors is their osteoinductivity, the competence to induce
bone formation. This has been demonstrated in vivo by
their ability to induce bone formation at heterotopic sites
and in different bone defect models.7,8,10 These factors are
thus of potential use in enhancing the process of fracture
healing. The osteoinductive properties of certain bone
morphogenetic proteins such as BMPs-2, -4, and -7 have
been established in animal models.7,10,11
The clinical application of such potent osteoinductive
substances may be limited by their short biological halflives. Gene transfer is one promising method being
Gene therapy for bone healing
AWA Baltzer et al
evaluated to overcome this problem. Delivery of genes
encoding growth factors can provide high, sustained concentrations of these factors locally for extended periods
of time. Moreover, endogenously synthesized proteins, in
contrast to exogenous recombinant proteins, may have
greater biological effectiveness.12,13 Transfer of osteoinductive genes to laboratory animals results in local bone
formation.14–20 Promising pre-clinical data have been
reported for the healing of bone by ex vivo14,15 and in
vivo16,17 approaches.
We have previously described a bone segmental defect
model in the lapine femur.21 Preliminary studies using
this model have shown that adenoviral vectors can transfer marker genes into the defect, with high levels of local
transgene expression. Gene expression persisted up to 6
weeks in bone, and up to 4 weeks in the surrounding soft
tissues.21 The current investigation used this model to
determine whether direct, in vivo gene therapy based on
adenoviral vectors carrying the BMP-2 gene could accelerate callus formation, as well as ossification and mineralisation, within a segmental long bone defect, and
improve the biomechanical properties of the newly
formed bone.
Results
Production of biologically active BMP-2 after adenoviral
gene delivery
Recombinant adenoviral vectors carrying a human BMP2 cDNA were generated by standard techniques (see
Materials and methods) and their transducing ability
tested on a murine stromal cell line (see Materials and
methods). Transduction of the cell line with 108 particles
of adenovirus carrying the BMP-2 gene increased alkaline
phosphatase activity about eight-fold. Transduction with
109 particles of this virus increased alkaline phosphatase
activity about 200-fold (Figure 1). Naive cells transduced
with the luciferase gene and cells transduced with less
than 108 viruses showed a baseline expression of alkaline
phosphatase. Transduction with 1010 particles of the virus
carrying BMP-2 resulted in detachment of all cells from
the culture dish and elimination of all alkaline phosphatase production. These cells were presumably dead. This
was not observed when the same amount of an adenoviral vector encoding marker genes was used.
735
Radiological analysis
Segmental defects were surgically created in the femora
of New Zealand white rabbits.21 Adenoviruses carrying
either the BMP-2 or luciferase genes were injected into
the defect and the rabbits monitored radiologically for
12 weeks.
Immediately after surgery the size of the defect
between the distal and the proximal femoral segments
was identical for all groups (Figure 2a, e). All rabbits
developed mineralisation within the defect area. Ossification leading to a connection of the femur segments
could be seen in the defect area after 5 weeks in all but
one of the rabbits in the group of rabbits receiving the
BMP-2 gene (Figure 2b). In contrast, the rabbits in the
control group had only a cloudy mineralization in the
defect area without reconnection between the ends of the
femur segments (Figure 2f). Seven weeks after operation
the defect area was restored in all defects receiving the
BMP-2 gene, with filling of at least 75% of the defect area
with cancellous ossification (Figure 2c). Healed defects
are defined as those with bone bridging at least 75% of
the defect site. By this definition, all defects treated with
the BMP-2 gene healed their gaps by 12 weeks after surgery (Figure 2d), whereas none of the control rabbits did
so (Figure 2h). These differences are statistically significant, using the Fisher’s exact test (P ⬍ 0.002). Even after
12 weeks, all but one of the rabbits in the control group
had only a thin bony connection between the cut ends of
the femur, mostly attached to the implant. The remaining
rabbit did not show any evidence of healing and
developed a pseudarthrosis, which broke during plate
removal.
Figure 1 Increase in alkaline phosphatase (ALP) activity following adenoviral transfer of the BMP-2 gene to a stromal cell line.
Gene Therapy
Gene therapy for bone healing
AWA Baltzer et al
736
Figure 2 Radiological analysis of defects treated with BMP-2 and the luciferase genes. Representative radiographs of defects treated with the BMP-2
gene at the time of operation (a) and 5 weeks (b), 7 weeks (c) and 12 weeks (d) after the operation. Representative radiographs of defects receiving the
luciferase gene at the time of operation (e) and 5 weeks (f), 7 weeks (g) and 12 weeks (h) after the operation.
Histology
Representative rabbits were killed 8 weeks after surgery
and the defects examined histologically. The histology of
the undecalcified femur confirmed that introduction of
the BMP-2 gene led to complete ossification across the
entire defect after 8 weeks. Von Kossa staining showed
complete recreation of cancellous bone and early transformation to cortical bone in the former defect area
(Figure 3a). In contrast, the defect area of the control
femur receiving the luciferase gene was dominated by an
extended central fibrosis, surrounded by cloudy new
bone formation and only a faint connection of the distal
and proximal femoral segment along the plate (Figure
3b).
showed gross motion at the bony gap site with gentle
manipulation, supporting the radiographic findings of a
pseudarthrosis. The mean bending strength of specimens
derived from the rabbits receiving the BMP-2 gene (170.7
± 20.6 Newtons) was significantly higher (P ⬍ 0.005) than
the control specimens (70.8 ± 24.4 Newtons). These findings were supported by the results of the testing of
diaphyseal stiffness. The bending stiffness of the femora
receiving the BMP-2 gene was significantly higher (84.0
± 10.2 Newtons/mm) than the stiffness of femora receiving the luciferase gene (49.5 ± 17.4 Newtons/mm) (P ⬍
0.05) (Figure 4).
Biomechanical testing
At the end of the 12-week experiment, rabbits were killed
and their femora excised and subjected to biomechanical
testing. The results of the roentgenographic analysis were
supported by the biomechanical findings (Figure 4). After
removal of the internal plate fixation and the cerclage
wires, one specimen of the control group had to be
excluded from further analysis, because it broke during
the procedure. All remaining femora, treated either with
the BMP-2 or with the luciferase gene, underwent biomechanical testing. One specimen from the control rabbits
Local delivery of growth factor genes is likely to be more
effective in bone healing than the application of pure
recombinant proteins.12,13 Moreover, the application of
gene therapy vectors, unlike present surgical approaches
to non-unions, does not further destroy the integrity of
the injured tissue, and it may be necessary to administer
the therapeutic vectors only once. The local overexpression of growth factors for 3 to 6 weeks21 might
explain the strong acceleration of new bone formation
seen in our experiments.
The experiments presented here demonstrate the
Gene Therapy
Discussion
Gene therapy for bone healing
AWA Baltzer et al
a
b
Figure 3 Histological analysis of defects 8 weeks post-operatively. Von
Kossa staining of the former segmental defect area. In defects treated with
the BMP-2 gene (a) there is complete bridging with cancellous bone and
early formation of cortical bone 8 weeks after gene transfer. In contrast,
the defect area of the control femur receiving the luciferase gene (b) lacks
cancellous bone. Instead it is dominated by an extended central fibrosis,
surrounded by cloudy new bone formation and only a faint connection of
the distal and proximal femoral segment along the plate.
Figure 4 Effect of gene transfer on the mechanical properties of the femora
12 weeks post-operatively.
ability of the BMP-2 gene delivered locally by an adenoviral vector, to enhance healing of a lapine long bone
defect. Radiographic and histologic results show that
bridging of the segmental defect was gained after 8
weeks, whereas there was only a faint bridging or no
bridging in the control defects. Twelve weeks postoperatively biomechanical testing confirmed that the newly
formed bone in the defects receiving the BMP-2 gene was
significantly stronger and stiffer than controls. Indeed,
one of the control rabbits developed a non-union and
another formed new bone so weak that it fractured upon
plate removal. Thus bone formation can be accelerated in
this model by in vivo gene therapy using first generation
adenoviral vectors.
The first use of gene transfer in promoting bone deposition was reported by Fang et al,16 who implanted naked,
plasmid DNA containing BMP-4 and parathyroid hormone combined with a collagenous scaffold, into segmental defects in rats, and subsequently dogs.17 Our
method, in contrast, is viral and does not require scaffolds. Indeed, in an earlier study using this model, collagen was found to impede in vivo cell transduction by
adenoviral vectors.21 This study also revealed very high
expression of the transgene by the skeletal muscle surrounding the defect. Gene expression persisted locally for
approximately 3 weeks in the soft tissues surrounding
the defect, whereas transgene expression from cut ends
of the bone lasted for 6 weeks. The present data indicate
that this period of time is appropriate for enhancing bone
healing in healthy rabbits.
In view of the high levels of gene expression within
the surrounding muscle,21 it is surprising that we did not
observe more ectopic bone formation in rabbits treated
with the BMP-2 gene than in control rabbits. It is possible
that very high levels of BMP-2 gene expression within
muscle leads to suppression of bone formation in this
microenvironment by mechanisms related to the apparent death of osteoprogenitor cells reported in Figure 1.
Cytotoxicity following infection with high concentrations
of adenovirus carrying BMP-2 is an unexpected, novel
observation. Although the mechanisms underlying this
effect are unknown, they cannot be trivially accounted
for by the high viral load as infection with equally high,
or even higher concentrations of adenovirus carrying
marker genes does not do this. It would be valuable to
know whether equivalent amounts of recombinant BMP2 are also cytotoxic, or whether this only occurs after gene
delivery under conditions where BMP-2 accumulates
intracellularly. Mason et al22 have observed a toxic effect
when infecting periosteal cells with vectors carrying the
BMP-7 gene. According to these authors, cell death
resulted from excessive formation of hydroxyapatite.
This possibility was not explored in the present study.
However, the apparent ability of high expression
of BMPs to cause cell death needs to be taken into
account when developing clinical protocols; more is not
necessarily better.
Lieberman et al14,15 have also used adenoviral vectors
carrying the BMP-2 gene to accelerate healing of a segmental defect, but in this case it was achieved by ex vivo
delivery via bone marrow cells. In a related application,
Boden et al18 transfected bone marrow cells with plasmids
encoding the LIM mineralisation protein and in an ex vivo
fashion, used these to enhance spinal fusion in rats. Ex
vivo methods are far more cumbersome than the in vivo
737
Gene Therapy
Gene therapy for bone healing
AWA Baltzer et al
738
system described here, and their clinical application is
more complicated.
In contrast to most other investigations,6,8,14–17,23 ours
did not rely upon supporting biomaterials, scaffolds, or
matrices, which are generally implanted along with the
growth factors cells or genes. The transduced cells act as
natural, local sites which continuously produce the
growth factors encoded by the transgenes, thereby
enhancing immigration of stem cells, the differentiation
of osteoprogenitor cells and the deposition of bone.24–26
The muscle surrounding the defect chamber might serve
as a natural matrix to guide the bridging of segmental
defects, as well as a possible local source of osteoprogenitor cells. Indeed BMP-2 is known to inhibit myogenic
differentiation and to convert the differentiation of
myoblasts into the osteoblast lineage.26,27 The local
environment, the duration of transgene expression, and
the cellular production of BMP-2 might explain why bone
restoration was possible without using additional osteoconductive materials.
Materials and methods
Study design
An osteoperiostal segmental defect was created in the left
femur of each of 13 New Zealand white rabbits by the
method of Baltzer et al21 and stabilised by plate fixation.
Adenoviral vectors (2 × 1010 particles) were injected into
the site of the defect. Seven rabbits received virus carrying the BMP-2 gene, and six control rabbits received virus
carrying the luciferase gene. Bone healing was monitored
radiographically. After 8 weeks, the defects receiving the
BMP-2 gene had healed radiographically. To confirm this
by independent criteria, representative rabbits were
killed at this time for histological evaluation. All remaining rabbits were killed 12 weeks after surgery and the
site of bone healing evaluated biomechanically.
Vectors
A cDNA encoding human BMP-2 was kindly provided
by the Genetics Institute (Cambridge, MA, USA). This
DNA was inserted into the E1 region of a serotype 5
adenovirus deleted in the E1 and E3 regions of the genome, using cre-lox recombination.28 A human CMV early
promoter was used to drive gene expression. A similar
vector containing the luciferase marker gene was generated by the same method. Recombinant virus was propagated in the permissive cell line 293 and purified by standard cesium chloride banding techniques. Viral titers
were determined by optical density at 260 nm (1 unit =
1012 virions/ml).
The ability of the vector to induce BMP-2-expression
after cell transduction and the biological activity of the
BMP-2 were tested in vitro using a bioassay based on the
responses of a bone marrow stromal cell line.29 BMP-2
increases alkaline phosphatase activity in these cells.29
Stromal cell cultures were established in 24-well plates
and transduced with either 106, 107, 108, 109, or 1010 particles of the adenoviruses carrying either the BMP-2 gene
or, as a control, the luciferase gene. Alkaline phosphatase
activity was determined 48 h after transduction by the
methods described in Ref. 29.
Animal model
The animal model used in these experiments has been
described in detail in Ref. 21. Briefly, a 1.3 cm segmental,
Gene Therapy
osteoperiosteal defect, representing approximately 12%
of the total length of the femur, was surgically created in
skeletally mature New Zealand white rabbits. The defect
was stabilised by plating and cerclage wires. The surrounding muscle was closed around the lesion, producing a closed chamber between the cut ends of the bone.
Radiographic analysis
Antero-posterior and lateral radiographs of the operated
left femur were taken immediately after surgery to
exclude animals with pre-existing and intra-operative
bone pathologies, and to ensure the correct placement of
the screws. Antero-posterior radiography was repeated
at 2, 5, 7, and 12 weeks. The segmental defects were
defined as healed if the extent of the defect filling was
radiologically more than 75%.
Histologic analysis
Because the radiographic data suggested healing in the
group of rabbits receiving the BMP-2 gene at 7 weeks,
two representative rabbits were killed for histological
analysis 8 weeks after operation. Undecalcified sections
were stained with the Von Kossa stain after plastic
embedding with Technovit 7200 VLC (Fa Kulzer, Norderstedt, Germany), and cutting of 300 ␮m sections with a
diamant-coated saw (Exact Fa Messner, Norderstedt,
Germany). The sections were made parallel to the long
axis of the bone, extending over the entire length of the
defect.
Mechanical testing
Rabbits were killed 12 weeks post-operatively, and the
soft tissues stripped from the left femora. All femora
were removed and stored frozen at −80°C immediately
after preparation. The femora were thawed for 24 h in a
refrigerator before testing. Each femur was placed in a
three-point bending system which provided an unsupported length of 4.5 cm. The ends of the femur were not
fixed. The condyles were directed upwards, such that the
anterior surface of the femur was placed in tension
by the test. Load was applied in posterior-to-anterior
direction at the mid-point of the unsupported length,
approximately in the middle of the diaphysis.
Testing was conducted on an Instron 8500 servohydralic testing system (Instron, Canton, MA, USA)
using an Instron 2500 lbf load cell. Deformation was measured using the internal system LVDT. Data were
acquired at a rate of 50 Hz using Instron’s MAX software
and was analyzed using MSExcel. A maximum deflection
of 1 cm was applied at a rate of 0.5 cm/min or
0.0833 mm/s. The test was stopped following fracture.
Statistical analysis
The data for the mechanical testing were examined by
Student’s unpaired t test. The data for the radiographic
analysis of healing were examined statistically by a twoby-two Fisher’s exact test to compare treatment groups
and control rabbits.
Acknowledgements
The authors thank Eric Lechmann, Mr Ringler and Mrs
Weber for technical support, Holger Haak for contributing statistical expertise, and Christopher Niyibizi PhD
Gene therapy for bone healing
AWA Baltzer et al
for the stromal cell line as well as advice concerning the
alkaline phosphatase assays. We are also grateful to
Synthes/Colorado for providing the DCP-plates and
screws and to Genetics Institute for providing the
BMP-2 cDNA.
References
1 Albertson KS, Medoff RJ, Mitsumaga MM. The use of periosteally vascularized autografts to augment the fixation of large
segmental allografts. Clin Orthop Rel Res 1991; 269: 113–119.
2 Zaslav KR, Meinhard BP. Management of resistant pseudarthrosis of long bones. Clin Orthop Rel Res 1988; 233: 234–242.
3 Bostrom M et al. Use of BMP-2 in the rabbit ulnar nonunion
model. Clin Orthop Rel Res 1996; 327: 272–282.
4 Moroni A, Rollo G, Guzzardella M, Zinghi G. Surgical treatment
of isolated forearm non-union with segmental bone loss. Injury
1997; 28: 497–504.
5 Südkamp NP et al. Incidence of non-unions in open fractures:
analysis of 948 open fractures. Akt Traumatol 1993; 23: 59–67.
6 Gerhart TN et al. Healing segmental femoral defects in sheep
using recombinant human one morphogenetic protein. Clin
Orthop Rel Res 1993; 293: 317–326.
7 Einhorn TA. Enhancement of fracture healing. J Bone Joint Surg
1995; 77A: 940–955.
8 Ripamonti U et al. Recombinant transforming growth factor-␤1
induces endochondral bone in the baboon and synergizes with
recombinant osteogenic protein-1 (bone morphogenetic protein7) to initiate rapid bone formation. J Bone Min Res 1997; 12:
1584–1595.
9 Yasko AW et al. The healing of segmental bone defects, induced
by recombinant human bone morphogenetic protein (rhBMP). J
Bone Joint Surg 1992; 74-A: 659–670.
10 Sampath TK et al. TGF-␤ superfamily proteins induce endochondral bone formation in mammals. Proc Natl Acad Sci USA 1993;
84: 7109–7113.
11 Hammonds RG et al. Bone inducing activity of mature BMP-2b
produced from a hybrid BMP-2 a/b percusor. Mol Endocrinol
1991; 5: 149–155.
12 Evans CH, Robbins PD. Possible orthopaedic applications of
gene therapy. J Bone Joint Surg 1995; 77-A: 1103–1114.
13 Niyibizi C et al. Potential role for gene therapy in the enhancement of fracture healing. Clin Orthop Rel Res 1998; 355S: 148–153.
14 Lieberman JR et al. Regional gene therapy with a BMP-2 producing murine stromal cell line induces heterotopic and orthotopic
bone formation in rodents. J Orthop Res 1998; 16: 330–339.
15 Lieberman JR et al. The effect of regional gene therapy with bone
morphogenetic protein-2 producing bone-marrow cells on the
repair of segmental femoral defects in rats. J Bone Joint Surg 1999;
81A: 905–917.
16 Fang J et al. Stimulation of new bone formation by direct transfer
of osteogenic plasmid genes. Proc Natl Acad Sci USA 1996; 93:
5753–5758.
17 Bonadio J, Smiley E, Patil P, Goldstein S. Localized direct plasmid gene delivery in vivo: prolonged therapy results in reproducible tissue regeneration. Nature Med 1999; 5: 753–759.
18 Boden SD et al. Lumbar spine fusion by local gene therapy with
a cDNA encoding a novel osteoinductive protein (LMP-1). Spine
1998; 23: 2486–2492.
19 Alden TD et al. In vivo endochondral bone formation using a
bone morphogenetic protein 2 adenoviral vector. Hum Gene Ther
1999; 10: 2245–2253.
20 Musgrave DS et al. Adenovirus-mediated direct gene therapy
with bone morphogenetic protein-2 produces bone. Bone 1999;
24: 541–547.
21 Baltzer AWA et al. A gene therapy approach to accelerating
bone healing: evaluation of gene expression in a New Zealand
white rabbit model. Knee Surg Sports Traum Arthrosc 1999; 7:
197–200.
22 Mason JM et al. Expression of human bone morphogenetic protein 7 in primary rabbit periosteal cells: potential utility in gene
therapy for osteochondral repair. Gene Therapy 1998; 5: 275–282.
23 Wang EA. Bone morphogenetic proteins (BMPs): therapeutic
potential in healing bony defects. Trends Biotechnol 1993; 11:
379–383.
24 Aspenberg P, Wang E, Thorgren KG. Bone morphogenetic protein induces bone in the squirrel monkey, but bone matrix does
not. Acta Orthop Scand 1992; 63: 619–622.
25 Lind M, Eriksen EF, Bünger C. Bone morphogenetic protein-2
but not bone morphogenetic protein-4 and -6 stimulates chemotactic migration of human osteoblasts, human marrow osteoblasts, and U2-OS cells. Bone 1996; 18: 53–57.
26 Yamaguchi A et al. Recombinant human BMP-2 stimulates
osteoblastic maturation and inhibits myogenic differentiation in
vitro. J Cell Biol 1991; 113: 681–687.
27 Katagiri T et al. BMP-2 converts the differentiation pathway of
C2C12 myoblasts into the osteoblast lineage. J Cell Biol 1993; 6:
1755–1766.
28 Hardy S et al. Construction of adenovirus vectors through CreLox recombination. J Virol 1997; 71: 1842–1849.
29 Balk ML et al. Effect of rhBMP-2 on the osteogenic potential of
bone marrow stromal cells from an osteogenesis imperfecta
mouse (oim). Bone 1997; 21: 7–15.
739
Gene Therapy