Phase 1: Vector Construction and in Vitro
Testing
Construction of Recombinant Adenoviral Vector
Rat LMP-1 cDNA with the human cytomegalovirus promoter was cloned
into a transfer vector and subsequently was transferred into the
adenoviral genome by homologous recombination (Adeno-Quest Kit;
Quantum Biotechnologies, Montreal, Quebec, Canada) (Fig. 1). The adenovirus
DNA used in this kit belongs to the human serotype 5, from which
the E1 and E3 regions have been deleted. Once the expression cassette
was introduced into the transfer vector, the vector was linearized
and cotransfected (with use of CaPO4) with the long arm of the adenoviral
vector, into human 293 cells. Recombination in the 293 cells between
the homologous regions of the linearized transfer vector and the
adenovirus genome resulted in the formation of the complete adenoviral recombinant,
which contained LMP-1 cDNA. Recombinant plaques were identified
by polymerase chain reaction for LMP-1 cDNA and also by functional
assay. Once the recombinant adenovirus had been selected, it was
further amplified for large-scale production. The amplified recombinant adenovirus
(AdLMP-1) was further purified by cesium chloride gradient centrifugation
and titered by plaque assay3.
Virus Concentration
Cesium chloride (CsCl) was removed with Sephadex G25 columns
(Pharmacia Biotech, Piscataway, New Jersey). This step is crucial
since a high concentration of CsCl can interfere with viral infection in
cell or tissue. The virus was stored in phosphate-buffered saline
solution containing 5% sucrose. The titer of the viral
stock was 1 ¥ 1010 plaque-forming
units (pfu)/mL as determined by the tissue-culture infective
dose50 (TCID50) method with use of 293 cells3.
The virion concentration was 1.34 ¥ 1012 viral
particles/mL as determined by ultraviolet spectrophotometry
(conversion factor: one optical density unit OD, A260 corresponds
to 1012 viral particles/mL)4. The virus was divided into small
aliquots and stored at —70°C. The virus was always thawed
on ice prior to use for infections.
Rat Osteoblast Cultures
After approval of the study protocol by the Institutional Animal
Care and Use Committee, fetal Sprague-Dawley rats were removed and
decapitated on the twenty-second day of gestation, and the heads were
submerged in sterile phosphate-buffered saline solution containing
5000 units of 1% penicillin and streptomycin (Gibco BRL,
Gaithersburg, Maryland). The crania were dissected with use of a sterile
technique in a laminar flow hood. Parietal and frontal bones were
dissected free from the sutures and subjected to four collagenase
digestions (type I:type II = 6:1). Cells from the latter
two digestions were pooled to provide a cell suspension enriched
in rat osteoblasts5. The pooled
cells were washed, pelleted, resuspended in minimum essential medium
(MEM) and 2 mM L-glutamine (Gibco BRL)/10% fetal
bovine serum (Hyclone Laboratories, Logan, Utah), counted by hemocytometer,
and seeded into T-75 vented flasks (Corning, Corning, New York)
at 1 ¥ 106 cells/flask.
The rat osteoblasts were grown at 37°C in 5% carbon
dioxide with humidification. The cells were fed at forty-eight hours
and again at ninety-six hours with MEM with L-glutamine and 10% fetal
bovine serum. Seven days after seeding, the primary culture was
trypsinized and passed into six-well plates at 1 ¥ 105 cells/35-mm
well as first subculture cells. Cells were grown for an additional
seven days, at which time they reached confluence (day 0). On day
0, the cultures were infected with AdLMP-1 and the medium was changed
under a laminar flow hood. The culture medium consisted of MEM,
L-glutamine, 10% fetal bovine serum, and ascorbic acid
(50 mg/mL) on days 0 through 7, and it consisted of BGJb
medium (Sigma, St. Louis, Missouri), 10% fetal bovine serum,
and 5 mM beta-glycerol-3-phosphate on days 8 through 14. Bone nodule
formation was assessed on day 14.
Analysis of Bone Nodule Formation
Cultures were fixed overnight in 70% ethanol and stained
with von Kossa silver stain. A semiautomated computerized video
image-analysis system (Optomax 5; Optomax, Hollis, New Hampshire) was
used to quantitate the nodules in each well. The results were expressed
as the number of bone nodules per well. This automated technique
was validated previously against a manual counting technique and
demonstrated a correlation coefficient of r = 0.92 (p < 0.000001)5. All data were expressed as the mean
and the standard error of the mean, calculated from five or six wells
under each set of conditions. Each experiment was repeated at least
two times with use of cells from different calvarial preparations.
Infection of Rat Osteoblast Cultures
Rat calvarial osteoblast cultures were infected for ten, thirty,
or sixty minutes at 37°C with adenovirus containing the cDNA for
either LMP-1 (AdLMP-1) or b-galactosidase (AdBgal) (Quantum Biotechnologies)
at various doses in 150 mL of MEM. Dose was expressed as the multiplicity
of infection (MOI) and was calculated as the number of plaque-forming
units (pfu) per cell—that is, the number of recombinant
adenoviral plaque-forming units to which a single osteoblast was
exposed. Several doses (MOI = 0.0025, 0.025, 0.25, and
2.5 pfu/cell) were tested. After infection, the 500,000
cells in each well were diluted from 150 mL to 2.0 mL of media volume
and then were cultured for seven days as described above.
Phase 2: Pilot Experiments
Design and Surgical Technique
Nine adult, female New Zealand White rabbits, each approximately
one year old and weighing 4.0 to 4.5 kg, were included in a pilot
study with use of bone-marrow-derived buffy-coat cells as the cell source.
The rabbits were housed individually in cages and were maintained
on a diet of rabbit chow and water ad libitum.
They were inspected daily for general health and neurological condition.
A single-level posterolateral intertransverse-process lumbar arthrodesis
of the fifth and sixth lumbar vertebrae was performed, as previously
described, with use of the bilateral paraspinal muscle-splitting
approach6. After decortication
of the transverse processes with an electric burr (Stryker Instruments,
Kalamazoo, Michigan), 2.0 mL of infected cells per side were loaded
onto the carrier material and placed into the transverse process
bed. The rabbits were killed after five weeks; the spines were excised
and evaluated with use of manual palpation, radiography, computed
tomography scanning, and histological analysis of undecalcified
specimens.
Six animals received devitalized rabbit bone matrix (guanidine
hydrochloride-extracted demineralized bone matrix) (Osteotech, Shrewsbury,
New Jersey) as the carrier material, and three animals received
a hemostatic collagen-sponge carrier (Integra LifeSciences, Plainsboro,
New Jersey). One side of the spine received cells infected with
LMP-1 cDNA (MOI = 0.4 pfu/cell), while the contralateral
side received cells infected with b-galactosidase cDNA as a negative
control or a lower dose of adenovirus with LMP-1 cDNA (MOI = 0.04
pfu/cell).
Infection of Bone Marrow Cells
In vivo experiments were performed by aspirating
3.0 mL of autologous bone marrow from the distal part of the femur
of each rabbit and isolating the buffy-coat cells by centrifugation7. The cells were counted, and 8 ¥ 106 cells
were infected with adenovirus at an MOI of 0.4 pfu/cell
in 1.5 mL of alpha-MEM for ten minutes at 37°C. After infection,
the cells were diluted to 2.0 mL of alpha-MEM and delivered to the
operating room for implantation.
Radiographic Analysis
The rabbit spines were evaluated on posteroanterior radiographs
made with a tube-to-plate distance of 90 cm. The radiographs were
viewed in a blinded manner, and the implant sites were graded, by
two observers, as fused or not fused on the basis of the presence
of continuous bone-bridging between the transverse processes. Computed
tomography scans of the lumbar spine were performed on a high-speed
CT-scanner (General Electric Medical Systems, Milwaukee, Wisconsin),
with a 10-cm field of view, a 1-mm interslice gap, and a 1-mm slice thickness,
operating at 150 mA and 100 kV peak.
Histological Analysis of Undecalcified Specimens
The excised lumbar spine of each animal was fixed for twenty-four
hours in 10% neutral buffered formalin and then was placed
in 70% ethanol. After fixation, the specimens were trimmed,
dehydrated in 95% and 100% ethanol, and cleared
in xylene. Undecalcified specimens containing devitalized bone matrix
or the collagen-sponge carrier were embedded in polymethylmethacrylate
and sectioned on a microtome (Jung Polycut E; Leica, Deerfield,
Illinois) to a thickness of 5 mm. One side of each fused level was
sectioned in the sagittal plane; the other was sectioned in the
coronal plane. The spine was considered fused histologically when
there was continuous bridging of new bone across the carrier connecting
the two lumbar segments.
Phase 3: In Vivo Spinal Arthrodesis Experiments
A larger study was undertaken to evaluate the reproducibility
of the bone-formation response and to improve the technique. First,
peripheral blood rather than bone marrow was used as the source
of the buffy-coat cells. Second, the carrier was changed to a collagen-ceramic
composite with 15% hydroxyapatite and 85% tricalcium
phosphate (Integra LifeSciences) to provide slower resorption that
would be more suitable for future primate studies8,9.
In this phase of the study, the rabbits were given cells infected
either with AdLMP-1 (ten rabbits) or with adenovirus without a transgene
as the negative control (ten rabbits). Spinal arthrodesis was performed
as described above except that the rabbits were killed one week
earlier, at four weeks.
Radiographic analysis of each specimen was performed as in Phase
2. Two specimens from each group underwent histological analysis,
and the remaining eight specimens in each group underwent biomechanical
testing to determine the relative strength and stiffness of the
fusion masses as previously described6.
Briefly, uniaxial tensile testing was performed at a displacement
rate of 0.5 cm/min with use of a servohydraulic materials-testing
system (MTS, Minneapolis, Minnesota). Load-displacement data were continually
generated and were recorded digitally (with use of a computer) and
graphically (with use of an x-y recorder). Ultimate tensile strength
was read directly as the peak load to failure. Stiffness was calculated
as the slope of the line joining two points on the load-displacement
curve (at 50% and 75% peak load). The adjacent
unfused motion segment between the fourth and fifth lumbar vertebrae in
each rabbit was biomechanically tested in an identical manner. This
was done in order to obtain an internal control to minimize the
effect of biological variation between animals.
Undecalcified specimens containing the collagen-ceramic carrier
were divided in the midsagittal plane, embedded in methylmethacrylate,
and sectioned coronally or sagittally with use of an automated system
(Exakt Technologies, Oklahoma City, Oklahoma) to a mean thickness
of 25 m. Contiguous sections were stained with 1% methylene
blue and 0.3% basic fuchsin. The specimens were examined
qualitatively for the extent of ingrowth of trabecular bone and
the presence of residual ceramic carrier.
Statistical Analysis
Biomechanical parameters were expressed as the mean and the standard
error of the mean. A one-way analysis of variance was performed,
and the Bonferroni test was used for multiple comparisons. Differences
between groups were considered to be significant when p < 0.01.
Phase 1: Effect of AdLMP-1 on Bone Nodule Formation
in Cultures of Rat Calvarial Osteoblasts
We performed a dose-response experiment with use of rat calvarial
osteoblast cultures to establish the optimal osteoinductive dose
of AdLMP-1 for ex vivo infections. Our data suggested
that the optimal dose for nodule formation in rat calvarial osteoblast cultures
was dependent on both multiplicity of infection (MOI) and infection
time. The lowest dose (MOI = 0.0025 pfu/cell)
required infection for thirty minutes to produce large numbers of
bone nodules by day 14. A higher dose (MOI = 0.25 pfu/cell)
was found to be optimal with only a ten-minute infection time (Fig. 2-A). The MOI
of 0.25 pfu/cell was less effective at longer infection
times, as was the higher MOI of 2.5 pfu/cell. At the ten-minute
infection time, the MOI of 0.25 pfu/cell was most effective
and formed similar or greater numbers of nodules compared with positive
control cultures that were stimulated to differentiate with a glucocorticoid
(Fig. 2-B).
There was minimal nodule formation in cultures of cells that were
infected with the negative control vector AdBgal (data not shown).
Phase 2: Pilot Experiments
One rabbit died from perioperative anesthetic complications,
and the remaining eight tolerated the procedure well and survived
until the completion of the experiment (Table I). All eight sites (six treated
with devitalized bone matrix and two, with collagen) that received AdLMP-1
at an MOI of 0.4 pfu/cell had a bone fusion mass. None
of the six sites (four treated with devitalized bone matrix and
two, with collagen) that received AdBgal at an MOI of 0.4 pfu/cell
had a bone fusion mass, although two of the four that received the
devitalized bone matrix carrier had some spotty bone formation within
the carrier, which may occur spontaneously with this material (guanidine
hydrochloride-extracted demineralized bone matrix) (Fig. 3). Neither of
the two sites that received AdLMP-1 at an MOI of 0.04 pfu/cell
had any bone formation.
The specimens with a complete spinal fusion mass unilaterally
had a solid fusion on manual palpation of the spinal motion segment,
and plain radiographs demonstrated bridging bone on the side with
the fusion mass. Computed tomography scans made through the fusion
masses on both sides of the spine revealed a solid, mature cancellous
bone mass corresponding to the solid fusion mass on the side that had
received AdLMP-1 at an MOI of 0.4 pfu/cell and a noncalcified
soft-tissue mass representing the carrier on the side that had received
cells transfected with the control gene or the lower dose of AdLMP-1
(Figs. 4-A and 4-B).
Histological analysis of nondecalcified specimens revealed normal
trabecular bone formation in the areas seen as fusion masses on
plain radiographs and computed tomography scans (Fig. 5-A and 5-B). Little residual
carrier was evident, and there was remodeling of the fusion mass.
The sites that had received cells without the active LMP-1 cDNA showed
residual carrier with fibrous tissue when a devitalized bone matrix
had been used and nearly complete resorption of the carrier when
collagen sponge carrier had been used (Figs. 6-A and 6-B).
Phase 3: In Vivo Spinal Arthrodesis Experiments
All twenty animals survived until the completion of the experiment.
One rabbit in the AdLMP-1 group had a subclinical unilateral infection,
but a continuous fusion mass still formed. All ten rabbits that had
received cells infected with AdLMP-1 had a solid fusion on manual
palpation. None of the ten rabbits that had received cells infected
with the empty adenovirus had a spinal fusion (Table II).
Plain radiographs revealed abundant new-bone formation in the
rabbits that had been treated with AdLMP-1 compared with that seen
in the controls (Figs. 7-A and 7-B). Because of the radiopaque nature
of the ceramic, which was still present at four weeks, it was difficult
to make a reliable assessment of the fusion status on the basis
of plain radiographs alone. Computed tomography scans clearly showed
that the spines that had been treated with AdLMP-1 had denser fusion
masses as well as evidence of bridging bone between the collagen-ceramic
carrier and the host bone (Figs. 8-A and 8-B). There was little or no bone outside
the confines of the collagen-ceramic-composite sponge carrier.
Biomechanical testing was performed to compare the relative strength
and stiffness of the fused level and the adjacent, uninvolved level
in each spine. The values for the AdLMP-1-treated spines were significantly
different from those for the control spines (F = 12.2569,
p < 0.000001). The mean relative strength was 2.35 0.25
in the AdLMP-1-treated spines compared with 1.22 0.11 in the control
spines (p < 0.01). The mean relative stiffness was 1.92
0.17 in the AdLMP-1-treated spines compared with 0.78 0.09 in
the control spines (p < 0.01).
Histological analysis of the spines from three rabbits that had
received cells infected with AdLMP-1 demonstrated normal bone formation
throughout most of the collagen-ceramic carrier, with good integration
of new bone with the host bone on the transverse processes (Figs. 9-A and 9-B). Higher-power
magnification showed osteoblasts laying down osteoid on primary
bone trabeculae (Fig. 9-C). Spines from two rabbits that
had received cells infected with the empty adenoviral vector demonstrated
mainly fibrous tissue in the porous spaces of the collagen-ceramic
carrier (Figs. 9-D, 9-E, and 9-F).
The most important results to emerge from these experiments were
the demonstration that expression of the novel intracellular protein
LMP-1 could consistently induce new-bone formation in vivo and
that spinal fusion could be attained in immune-competent animals
with the use of local gene therapy. Moreover, we validated a gene-delivery
technique that may be used with cells from peripheral venous blood
(rather than bone marrow) and can be performed easily and rapidly
in a standard operating room. The histological and biomechanical properties
of the fusion mass were as good as those of fusion masses formed
with use of recombinant or extracted bone morphogenetic proteins10,11.
To date, other studies involving adenoviral delivery of osteoinductive
genes, such as bone morphogenetic proteins, to induce spinal fusion
have yielded inconsistent results12-22.
In addition, the studies have been limited by three primary factors:
(1) the use of much higher doses of adenovirus (an MOI of 40 to
500 pfu/cell), which provoke a greater immune response
to the viral proteins12,13; (2)
long viral infection times, ranging from several hours to overnight16,17; and (3) delivery cells that
require several weeks of expansion in culture prior to viral infection
and reimplantation18,19.
In contrast, our preliminary data suggest that, with a very low
dose of virus (an MOI of 0.25 to 0.4 pfu/cell, 5 ¥ 108 viral
particles), no significant immune response to the adenoviral vector
is encountered. This dose is 100,000 times lower than those used
in other gene-therapy protocols, and it greatly reduces the safety concerns
associated with use of an adenoviral vector14,15. Also, our protocol
for ex vivo gene transfer to bone marrow or peripheral
blood cells can be performed intraoperatively and therefore is easier
than other protocols, which require complex cell-selection techniques
or cell expansion in culture for weeks. Several investigators have attempted
direct injection of adenovirus to achieve spinal fusion in
vivo, which resulted in variable response and was likely hampered
by the immune response to the adenoviral proteins20-22.
Our in vivo experiments showed consistent bone
induction with a very short infection time (ten minutes) in a challenging
spinal fusion model, making this method more clinically feasible.
Previous studies in the rabbit model have demonstrated successful
spinal fusion in only 50% to 70% of rabbits treated with
autogenous bone graft23-25. In
addition, substitutes such as certain forms of demineralized bone
matrix gel, coralline hydroxyapatite, and calcium carbonate have
been unable to induce consistent spinal fusions in the rabbit model26-28. We found that the same dose
of AdLMP-1 that induced bone in cell culture was effective in
vivo; this finding was notable because previous investigators
have found that the delivery of growth-factor proteins has required
substantially higher doses in animal experiments than in cell culture5,9,10.
The heterogeneous family of LIM-domain proteins is important
for growth and differentiation in a variety of cell types, but the
precise mechanisms of action of LIM-domain proteins remain poorly understood29-32. It is thought that LMP-1 directly
or indirectly results in the synthesis and secretion of one or more BMPs
and possibly of ancillary proteins that enhance the activity of
BMPs. Since LMP-1 is an intracellular protein, in some applications
it may offer some strategic advantages over administration of extracellular
proteins such as BMPs, whose action may be limited by the low prevalence
of specific BMP receptors on the surface of resting osteoprogenitor
cells. The main difficulties traditionally associated with gene
therapy (poor gene-transfer efficiency and inadequate duration of
protein expression) did not limit the use of LMP-1 for local gene
therapy to generate bone. Efforts to determine how long the gene
is expressed in vivo have been challenging because
of the fact that the extremely low dose of virus used is below the
level of detection of X-gal staining. On the basis of in
vitro studies, LMP-1 seems to start a cascade of events, including
the secretion of osteoinductive growth factors, and therefore we
believe that its expression does not need to persist very long in
vivo1.
In summary, our observations suggest that local gene therapy
with use of adenovirus to deliver the LMP-1 cDNA is feasible and
promising as an alternative method to achieve bone formation for
spinal fusion. The use of delivery cells readily available from
venous blood and the short infection time should allow this technique
to be performed in any operating room. Finally, the use of an ex
vivo gene-transfer protocol with very low doses of adenovirus
should minimize the immune response and toxicity seen in association
with other adenoviral applications.