Journal of

Ort hopaed ic
Research
ELSEVIER Journal of Orthopaedic Research 22 (2004) 73-79
www.elsevier.com/locate/ort hres

Induced membranes secrete growth factors including vascular

and osteoinductive factors and could stimulate bone regeneration
Ph. Pelissier A.C. Masquelet b, R. Bareille ', S. Mathoulin Pelissier d, J. Amedee
a Service rie Chirurgie Plasiique, Hopital, Pellegrin-Tondu 33076 Borrleaus, Francc,
Service de Chirurgie Orthopidiique, Hopitul Atrirenne, 93009 Bobigny cedex. France
INSERM U-443 Unicersifi Vicror Segalen-Bordeaux 2, 33076 Bordeaux, Frunce
Service de Biostati.stiyues Institut Bergonii. 33076 Bordeaux, France

Abstract
Based on a new concept, a procedure combining induced membranes and cancellous autografts allows the reconstruction of wide
diaphyseal defects. In the first stage of this procedure, a cement spacer is inserted into the defect; the spacer is responsible for the
formation of a pseudo-synovial membrane. In the second stage, the defect is reconstructed two months later by an autologous
cancellous bone graft. The aim of this study was to evaluate the histological and biochemical characteristics of these membranes
induced in rabbits. Histological studies carried out two, four, six, and eight weeks following implantation revealed a rich vascu-
larization. Qualitative and quantitative immunochemistry showed production of growth factors (VEGF, TGFPI) and osteoin-
ductive factors (BMP-2). Maximum BMP-2 production was obtained four weeks after the implantation, and, at this time, induced
membranes favored human bone marrow stromal cell differentiation to the osteoblastic lineage. Should these results be confirmed in
humans, bone reconstruction could be carried out earlier than previously thought and in better conditions than expected, the
membrane playing the role of an in situ delivery system for growth and osteoinductive factors.
0 2003 Orthopaedic Research Society. Published by Elsevier Ltd. All rights reserved.

Introduction by an autologous cancellous bone graft. The mem-

Currently, the most common procedures for the
treatment of extensive diaphyseal bone defects are the
vascularized bone free transfer [2,9,15,24] and the 11-
izarov intercalary bone transport method [I$]. In order
to minimize the sequelae induced at the donor site of
free flaps, reduce the patient's discomfort, and make the
bone reconstruction more reliable, Masquelet proposed
a procedure combining induced membranes and can-
cellous autografts [ 10,12,17]. This technique allows
the reconstruction of wide diaphyseal defects even if
the recipient site has been irradiated or infected, pro-
vided that an envelope is previously created to protect
and revascularize the bone graft. The first stage consists
of the insertion into the defect of a methylmethacry-
late cement spacer that is responsible for the formation
of a pseudo-synovial membrane. The second stage is
the reconstruction of the defect, two months later,
*Corresponding author. Tel.: +33-5-5679-5548; fax: +33-5-5679-
5687.
E-mailaddress: philippe.pelissier@chu-bordeaux.fr (Ph. Pelissier).
0736-0266/$ - see front matter 0 2003 Orthopaedic Research Society. Published by Elsevier Ltd. All rights reserved
doi: 10.1016/S0736-0266(03)00165-7
brane induced by the spacer prevents the graft from
resorption and favors its vascularity and its corticaliza-
tion.
The first role of the spacer is mechanical as it obviates
fibrous tissue invasion of the recipient site. Moreover, as
the spacer behaves as a foreign body, absence of infec-
tion after two months is an excellent witness of adequate
local conditions for bone grafting. The second role is
biological by the induction of the surrounding mem-
brane that will revascularize the bone graft and prevent
it from resorption [10,12,17]. Using this technique,
Masquelet et al. reported reconstruction of bone defects
up to 25 cm in length, with an average of 8.5 months for
normal walking following weight bearing diaphyseal
segment repair.
The aim of this study was to evaluate the histological
and biochemical characteristics of these membranes, so
as to better understand their structure and behavior. A
histological study was first carried out, then an immu-
nochemistry study was performed to assess the presence
of growth factors within the induced membrane. Finally,
influence of the latter on human bone marrow stromal
74 Pli, Pelissic~ret ul. / Jouniul
cells (HBMSC) differentiation and proliferation was
studied.
Methods
Estohlishnient u / induced nimihrune wtockls in ruhhii.y
A 3 cm long incision was made bilaterally on the back of 16 Fauve
de Bourgogne rabbits (average weight: 3.7 kg) obtained from a licensed
vendor (Dombes, Romans, France). Rabbits were anesthetized, and
four subcutaneous tunnels were made from the incision to allow the
insertion of four cylindrical methylmethacrylate spacers (Surgical
Simplex P, Howmedica Intl Inc, Co. Clare, Ireland), 1 cm in diameter
and 5 cm long. The wound was closed with fast absorbable 3/0 cuta-
neous sutures (Vicryl rapide'", Johnson & Johnson. Ethicon-France),
and a protective aluminum spray dressing (Aluspray@,Vetoquinol.
Lure, France) was applied over the wound. Animals were then as-
signed to a study group according to necropsy time (two, four, six, and
eight weeks). In the postoperative period, animals were allowed to feed
and move without restriction. This protocol was conducted in accor-
dance with the recommendations of the French Ministries of Research
and Agriculture on the care and use of laboratory animals.
Four animals were sacrificed at each time point, and the mem-
branes surrounding the implants were removed. On the inidline of the
back of each animal, four samples of subcutaneous tissue were taken to
serve as a control. All the specimens were divided in two parts: the first
was fixed in 10% (vlv) formalin solution, embedded in paraffin, and cut
in 5 pm sections for histological study, and the second was frozen at
-40 "C prior to protein extraction.
Histology uritl ininiiinohbvtuclieniicul unulysi.~
Some of the sections were stained with Hematoxylin-Eosin-Saffron
(HES). Sections were observed under low and high magnification with
an Optiphot 2 (Nikon@)photomicroscope. The specimens were in-
vestigated for acute inflammatory phase reaction (edema, neutrophils,
and vascular congestion), foreign body reaction, and fibrosis. The
observer was blinded to the conditions of the specimens.
Other sections were used for iinmunochemistry, and paraffin was
first removed by sequential washes (toluene 100'%1,10 min; alcohol
100'%1,5 min: alcohol 90'%1,5 min: alcohol 80'%),5 min twice; water 5
min). lmmunochemistry was started with saturation by 1% (wlv) bo-
vine serum albumin in phosphate-buffered saline (0.1 M PBS p H 7.4).
Then slides were overlayed with monoclonal antibody specific for
vascular endothelial growth factor, VEGF (TAKARA Biomedical,
Shiga. Japan), diluted 1/200 in tris-buffered saline (0.05 M TBS, pH 7).
and a streptavidin alkaline phosphatase-based immunoenzymatic
technique was performed according to the manufacturer's protocol
(LSAB2 kit. DAKO, Carpenteria, CA, USA). A red precipitate de-
noted the presence of VEGF.
Protein ertruction ,front induced inemhruneJ
Membranes were crushed with an electric crusher in the presence of
5 ml of 0.1 M PBS pH 7.4 containing a cocktail of protease inhibitors
and then dialyzed overnight at 4 "C in the same buffer supplemented
with 20% (v/v) glycerol. There were four membrane samples and four
control samples for each time point. Concentrations of the proteins
VEGF, transforming growth factor-beta-1 (TGFPI), and bone mor-
phogenic protein-2 (BMP-2) were determined for each sample.
Protein concentration was quantified with the Bio-Rad protein
assay kit (Bio-Rad Laboratory GmbH, Munich, Germany) according
to the manufacturer's protocol. VEGF concentration was measured
with the human VEGF EIA kit (CYT 132, Chemicon International,
Inc., Temecula, CA), and TGFPl concentration was measured with
the human TGFBl BMS249 kit (Bender Medsystems, Vienna, Aus-
tria), both being performed according to the manufacturer's protocol.
The presence and the amount of BMP-2 into the induced inembrane
extracts were detected with an ELlSA technique using a monoclonal
anti-human BMP-2 antibody (MAB355, R&D Systems, Minneapolis,
MN). Recombinant human BMP-2 used for control was graciously
provided by Genetics Institute (Cambridge, MA).
of Orthopuro'ic Resrurch 22 (2004) 73-79
Ccdl prol(fifi.ration o / huniuri hone murroii' .siromul wlls ( H B M S C )
HBMSC were obtained with consent from one patient, who un-
derwent a femoral head replacement. Stromal cells were obtained as
described previously [22] and were cultured in Iscove's Modified
Dubelcco's Medium (IMDM, Gibco, Grand Island. NY) supple-
mented with 10% (vlv) of fetal calf serum (FCS, Gibco, Grand Island,
NY). HBMSC were seeded in 25 cm2 dishes ( 2 x 10' cellslcm2) and
cultured in 0.5 ml of I M D M supplemented with fetal calf serum (FCS)
lo'%)(v/v), pencilin-streptomycin 1/1000' 1 pl/ml, Amphotericin B 10
p l h l , and 10 p1 of specimen protein extracts (subcutaneous tissue
(control) or membrane). The medium was changed every two days.
Two subcutaneous samples and two membrane samples were ran-
domly selected from the groups obtained after two and four weeks. An
additional reference group was created that consisted of bone marrow
stromal cells cultured with I M D M alone as control. Analysis of cell
growth was performed after one, three, six, and nine days in culture,
which consisted in a rapid colorimetric assay for cellular growth and
survival using a tetrazolium salt [13].
Cell diferentiution of HBA4SC
HBMSC were seeded in 75 cm2 dishes (lo4 cells/cm') and were
cultured into the same medium supplemented with the induced mem-
branes (including controls) obtained after two and four weeks. Ana-
lysis of cell differentiation was assessed by the measurement of alkaline
phosphatase (AL-P) activity after one, three, six, and nine days of
culture with Sigma diagnostic kit (85L-2, St Louis, MO). using naphtol
AS-MX phosphate as substrate. Results were expressed in nmole
phosphate transferred/min/pg proteins.
Stutisticul unulysis
Results are presented as means f standard deviations (s.d.). Statis-
tical analysis was performed with SPSS 10 software (SPSS Inc., Chicago,
IL). Statistical comparison of growth factor production was performed
with the Wilcoxon signed-rank test for matched pairs and the Mann-
Whitney, a non-parametric test, for cell proliferation and AL-P activity
analysis. Differences were considered as significant at p i0.05.
Results
Histology and immunostaining of angiogenic ,firetors
Macroscopic examination carried out at two, four,
six, and eight weeks following implantation showed
many capillaries on the outer surface of the membranes;
the mean capillary diameter was 1.2 mm. At two weeks,
microscopic examination demonstrated the forming
fibrous membrane, the inner part of which was a
synovial-like epithelium, the outer part being made of
fibroblasts, myofibroblasts, and collagen (Fig. 1A). A
focal and mild acute inflammatory reaction and diffuse
edema were noticed. Multinucleated giant cells denoted
a foreign body reaction. The membrane appeared to be
richly vascularized by numerous small capillaries present
in all the layers of the specimen, confirmed by immu-
nostaining of VEGF (Fig. IB). At four weeks, the
membrane was similar to that described above, myofi-
broblasts and collagen fibers being mostly arranged
parallel to the spacer surface, but the edematous reac-
tion appeared to be less. Very rarely, villous hyperplasia
was observed. Small capillaries were still present, and
larger vessels began to develop in the outer part of the
Fig. 1. Photomicrographs of representative sections demonstrating the histological changes over the course of maturation of membranes induced
around a cement spacer after two weeks (A, B) and four weeks (C, D) of maturation. Hematoxylin-Eosin-Saffron staining (A, C) and VEGF im-
munostaining ( B , D). Original magnification 40x. Black star indicates the previous location of the cement spacer (the inner part of the membrane);
arrows indicate VEGF immunostaining.

membrane (Fig. IC). Immunostaining performed on the
same sections also showed multiple sites of fixation for
antibodies against VEGF (Fig. 1D). At six and eight
weeks, no major changes occurred, though edema was
almost completely resorbed, and multinucleated giant
cells were less numerous.
Growth fuctor content in membranes
The control group (subcutaneous tissue) demon-
strated a low concentration for VEGF, TGFP1, and
BMP-2 at every time point (Fig. 2). In the induced
membranes, high concentrations of both VEGF and
TGFPl were observed as early as the second week and
were stable during the following weeks (Fig. 2A and B).
However, at six weeks, concentrations of growth factors
were still five times higher in the membranes than in
the controls. The differences between membranes and
controls were significant for TGF-P1 and VEGF (p <
0.001). Interestingly, the concentration of BMP-2 was
initially close to that of the control, but thereafter in-
creased to its highest level at the fourth week and then
16 Ph. Pelisxier et ul. I Jouniul o f Oriholmdic Research 22 (2004) 73-79

0.5 1 -C- Induced membranes

C -----w---Subcutaneous tissue (control)
T
.
cn 0.3
a
C
.. 0.2
LL

0.1

0
4JA
I I I I

2 4 6 8
Weeks

.-C
d 2.5
2
w 2 L
2_
w
c 1.5

2 4 6 8
Weeks

0.4 ,
.$ 0.35 -
c
2
P
0.3 -
0.25 -
\

g 0.2 -

<
n
0.15 -

I 0.1 -
rn
0.05 -
C
0 ’
2 4 6 8
Weeks

Fig. 2. Growth factor levels in the induced membranes. VEGF (A), TGFbl (B), and BMP-2 (C) concentrations in protein extracts arising from
induced membranes and subcutaneous tissue (control) obtained alier two, four, six, and eight weeks were quantified by ELtSA methods. Values
represent the mean k s.d. at each time-point ( n = 4 for both groups). p < 0.001 between membranes and controls.

decreased to the basal level at the last time point.
However, the difference between membranes and con-
trols was significant at four and six weeks (p < 0.001).
Cell prolifkrution
Influence of the membranes on bone marrow cell
proliferation was monitored until the eighth day (Fig.
3). Since infection occurred in the cell cultures, the re-
sults were unavailable after this time point. As expected,
HBMSC did not proliferate in medium without fetal calf
serum. Addition of subcutaneous tissue extract (control
group) maintained the cell growth at the initial level for
six days. In contrast, addition of membrane extracts
(harvested at two and four weeks) stimulated cell
growth. The difference between membranes and controls
was significant (p < 0.001).
Cell diferent iut ion
Differentiation of bone marrow stromal cells was
followed by the measurement of AL-P activity used as
an early bone differentiation marker. Related to the
production of BMP-2 in induced membranes after four
weeks, differentiation of HBMSC to osteoblastic lineage
was observed in the presence of membranes obtained
 0.25 tween control and membranes obtained after two weeks
(p = 0.015), which contained traces of BMP-2 but sig-
0.20 nificant levels of TGFP- 1 and angiogenic factors.

0
5
$ 0.15
Discussion
*
9
W
0
C
9 0.10 The technique for bone reconstruction described by
2n
u)
Masquelet combines the induction of a membrane by
the means of a cement spacer with a later cancellous
0.05
autograft [10,12,17]. The first role of the spacer is me-
chanical as it obviates fibrous tissue invasion of the re-
0.00 cipient site. The second role is biological, by the
1 3 6 induction of the surrounding membrane, the histological
Days in culture characteristic of which has already been studied. Mem-
I 2 3 Cell culture with IMDM without fetal calf serum branes formed at the bone-cement interface of a total
E 3 Cell culture in the presence of subcutaneous tissue (control) hip replacement prosthesis have the histological and
0Cell culture in the presence of two-week-maturation membrane histochemical characteristics of a synovial-like lining.
ICell culture in the presence of four-week-maturationmembrane In our study, the membrane also had the aspect of a
Fig. 3. Cell growth of HBMSC in the presence of protein extracts of
synovial-like epithelium with few inflammatory cells.
either subcutaneous tissue (control) o r induced membrane (after two The surrounding fibrous capsule was mostly arranged
and four weeks of maturation). HBMSC cultured in the presence of parallel to the spacer surface. Periprosthetic capsular
medium without fetal calf serum were used as basal cell growth con- tissues surrounding breast prostheses have similar char-
trol. Values represent the mean f s.d. at each time-point ( n = 4 at each acteristics, but differences exist between the membranes
point for both groups). p < 0.001 between membranes and controls.
formed around smooth and textured implants. Synovial-
like metaplasia, villous hyperplasia, and foreign mate-
rial were more often observed around textured implants
within the first five years [25]. As the cement spacer
surface in our study was smooth, we also observed a
rare villous hyperplasia and very few foreign-body
.- reactions.
.
E
U
0.8 Bone grafts should be covered by healthy tissues,
such as muscles flaps, to be revascularized. According to
2 0.6 Masquelet et al. [12], a cancellous autograft placed in
v)

-
C
?!
B
0.4
such a recipient site behaves as a foreign body and is
subject to resorption. The well-known function of this
-g
aJ
0.2
induced membrane is therefore to protect the graft from
the environment and to vascularize it. This hypothesis is
C
in accordance with the rare multinucleated giant cells
0
1 3 6 9
present at the inner side of the capsule, which could
Days in culture
prevent resorption, and with the rich vascularization
observed as soon as the second week. The small capil-
Cell culture with IMDM without fetal calf serum
laries persisted until the end of the eighth week, but
B Cell culture in the presence of subcutaneous tissue (control)
additional larger vessels developed in the outer part of
0Cell culture in the presence of two-week-maturation membrane the membrane by the fourth week.
ICell culture in the presence of four-week-maturation membrane
Angiogenesis is now recognized as an essential step
Fig. 4. Alkaline phosphatase activity of HBMSC cultured in the in osteogenesis, and microvessel cells are identified as
presence of protein extracts of either subcutaneous tissue (control) o r having a direct relationship to bone formation [ 16,231
induced membrane (after two and four weeks of maturation). HBMSC because of their proximity to osteoblasts and osteopro-
cultured in the presence of medium without fetal calf serum were used
genitor cells at sites of new bone formation. One of the
as basal enzymatic activity control. Data are expressed in nmole
phosphate transferredlminlpg protein as the mean f s.d. ( n = 4 at each most important growth and survival factors for endo-
point for both groups). p = 0.015 between control group and mem- thelium is the VEGF that induces angiogenesis and
branes. plays an important role in regulating vasculogenesis [7].
Most types of cells, but usually not endothelial cells
after four weeks (Fig. 4), which strongly stimulated themselves, secrete VEGF. For example, in chronic
AL-P activity. The difference was still significant be- wounds, VEGF is secreted by wound macrophages and
78 Ph. Pt.1i.vskr et ul. I Journd qf’ Ortl7opardie Research 22 (2004) 73-79
is responsible for stimulation of angiogenesis and pro-
longed edema [6]. Moreover, VEGF plays an important
role in the regulation of bone remodeling by attracting
endothelial cells and osteoclasts and by stimulating
osteoblast differentiation [3]. Our data (immunohisto-
chemistry and quantitative analysis) confirmed that
these membranes continuously secrete this angiogenic
factor, which is able to stimulate bone cell differentia-
tion. Moreover, TGFPI, which plays an important role
in bone metabolism and extracellular matrix protein
synthesis [I 11, is also considered as a putative regulator
of osteoclastic-osteoblastic interaction and is continu-
ously synthesized in these membranes [18]. Both VEGF
and TGFPl could be partly responsible for the cell
growth observed in the presence of these membranes
obtained after two or four weeks. More interestingly
and identified for the first time, the induced membrane
obtained after four weeks exhibited osteoinductive
properties due to the secretion of BMP-2, which exhibits
pleiotropic functions ranging from extraskeletal and
skeletal organogenesis to bone generation and regener-
ation [5,19].
The synergistic or antagonistic evolution of growth
factors is also interesting in that TGFPI and BMP-2 are
known to have an opposite action on osteoblast differ-
entiation [21]. BMP has the ability to induce ectopic
bone, while the action of TGFPl is to stimulate prolif-
eration of osteoblasts and chondrocytes as well as the
production of extracellular matrix. In contrast, TGFPl
in combination with BMP may enhance formation of
ectopic bone [20]. On the other hand, local injection of
recombinant human BPM-2 (rhBMP-2) in various sites
(e.g., intramuscular, subcutaneous, and fat) induces new
bone formation in all sites. The variations of osteoin-
ductive activity of rhBMP-2 observed within the sites
could be due to differences in the blood supply [14].
Moreover, VEGF-A produced by osteoblasts in re-
sponse to BMPs is not involved in osteoblast differen-
tiation, but couples angiogenesis to bone formation [4].
In the model presented here, the membrane seems able
to secrete the growth factor, which could act sequen-
tially to provide adequate vascularization and then bone
formation.
Bone reconstruction may then be carried out earlier
than previously thought and in better conditions than
expected, the membrane playing the role of an in situ
delivery system for growth and osteoinductive factors.
Acknowledgements
We thank Mr. J.P. Chenu from DETERCA (Uni-
versitt Victor Segalen Bordeaux 2) for animal care, and
INSERM and Pole Aquitaine SantC Bordeaux for fi-
nancial support.
References
[I] Cattaneo R, Catagni M, Johnson E. The treatment of infected
nonunions and segmental defects of the tibia by the methods of
Ilizarov. Clin Orthop 1992;280:143-52.
[2] de Boer H, Wood M. Bone changes in the vascularised fibular
graft. J Bone Joint Surg 1989;71B:374-8.
[ 3 ] Deckers M, Karperien M, van der Bent C, et al. Expression
of vascular endothelial growth factors and their receptors
during osteoblast differentiation. Endocrinology 2000; 141:1667-
74.
[4] Deckers MM, van Bezooijen RL, van der Horst G, et al. Bone
morphogenetic proteins stimulate angiogenesis through osteo-
blast-derived vascular endothelial growth factor A. Endocrinology
2002;143:1545-53.
[5] Ducy P, Karsenty G. The family of bone morphogenetic protein.
Kidney Int 2000;57:2207-14.
[6] Feng 1. Hussain M, Constant J. Hunt T. Angiogenesis in wound
healing. J Surg Pathol 1998;3:1-8.
[7] Ferrara N, Davis-Smyth T. The biology of vascular endothelial
growth factor. Endocr Rev 1997;18:4-25.
[ S ] Green S. Jackson J, Wall D, Marinow H, Ishkanian J. Manage-
ment of segmental defects by the Ilizarov intercalary bone
transport method. Clin Orthop 1992;280: 13&42.
Han C, Wood M, Bishop A, Cooney W. Vascularized bone
transfer. J Bone Joint Surg 1992;74A:1441-9.
Klaue K, Knothe U, Masquelet A. Etret biologique des mem-
branes a corps etranger induites in situ sur la consolidation des
greffes d’os spongieux. Rev Chir Orthop Suppl 1995;70e rkunion
annuelle. p. 109-10.
Marie P. Growth factors and bone formation in osteoporosis: role
for I G F l and TGF beta. Rev Rhum Engl 1997;64:4453.
Masquelet A, Fitoussi F. Begue T. Muller G. Reconstruction des
0s longs par membrane induite et autogreffe spongieuse. Ann Chir
Plast Esthet 2000;45:34653.
Mosmann T. Rapid colorimetric assay for cellular growth and
survival: application to proliferation and cytotoxicity assays.
J lmmunol Meth 1983;65:55-63.
Okubo Y, Bessho K, Fujimura K, et al. Osteoinduction by
recombinant human bone morphogenetic protein-2 at intramus-
cular, intermuscular, subcutaneous and intrafatty sites. Int J Oral
Maxillofac Surg 2000;29:62-6.
Pelissier P, Casoli V, Demiri E, Martin D, Baudet J. Soleus-fibula
free transfer in lower limb reconstruction. Plast Reconstr Surg
2000; 105:567-73.
Pelissier P, Villars F, Mathoulin-Pelissier S, Bareille R, Lafage-
Proust M H , Vilamitjana-Amedee J. Influences of vascularization
and osteogenic cells on heterotopic bone formation within a
madreporic ceramic in rats. Plast Reconstr Surg 2003;111(6):1932-
41.
PClissier Ph, Masquelet AC, Lepreux S, Martin D, Baudet J.
Behavior of cancellous bone graft placed in induced membranes.
Br J Plast Surg 2002;55:598-600.
Pfeilschifter J, Die1 1, Scheppach B, et al. Concentration of
transforming growth factor beta in human bone tissue: relation-
ship to age, menopause, bone turnover, and bone volume. J Bone
Miner Res 1998;13:716-30.
Riley EH. Lane JM, Urist MR, Lyons KM, Lieberman JR. Bone
morphogenetic protein-2: biology and applications. Clin; Orthop
1996;324:3946.
Si X, Jin Y, Yang L. Induction of new bone by ceramic bovine
bone with recombinant human bone morphogenetic protein 2 and
transforming growth factor beta. Int J Oral Maxillofac Surg
I998;27:3 1 0 4 .
Spinella-Jaegle S, Roman S, Faucheu C, Dunn FW, Kawai S, et al.
Opposite effects of bone morphogenetic protein2 and transform-
 ing growth factor beta I on osteoblast differentiation. Bone
1998;29(4):323-30.
[22] Vilamitjana-Amedee J, Bareille R, Rouais F, Caplan AI, Har-
mand MF. Human bone marrow stromal cells expressing osteo-
blastic phenotype in vitro. In Vitro Cell Dev Biol 1993;29(9):
699 707.
[23] Villars F, Guillotin B, Amdee Th, Dutoya S. Bordenave I..
Bareille R, et al. Effect of human endothelial cells on human
osteoprogenitor cell differentiation needs heterotypic gap junction
communication. Am J Physiol 2002:282(4):C775-85.
[24] Weiland A. Current concepts review: vascularized free bone
transplants. J Bone Joint Surg 1981:63A:166 9.
1251 Wyatt L, Sinow J, Wollman J, Sami D, Miller T. The influence of
time on human breast capsule histology: smooth and textured
silicone-surfaced implants. Plast Reconstr Surg 1998; 102:1922
31.