THE ANATOMICAL RECORD 245426-445 (1996)

In Vitro Modeling of the BoneAmplant Interface

J.E. DAVIES
Centre for Biomaterials, University of Toronto, Toronto, Ontario, Canada M5S 3E3
ABSTRACT Background: The purpose of this review is to examine the
usefulness of cell culture methods to model the mechanisms of bone forma-
tion on the surfaces of candidate implant materials.
Methods: The central objective is to show that in vitro methods are
uniquely valuable in providing an understanding of how new bone is
formed on solid surfaces. It should be emphasized, at the outset, that the
use of cell culture studies as cytotoxicity assays will not be addressed, nor
is it implied that cell cultures can model all the complexities of the in vivo
environment. Nevertheless, by comparison with in vivo data, which are by
nature retrospective, it is shown that primary differentiating osteogenic
cell cultures, derived from bone marrow, illustrate a sequence of extracel-
lular matrix elaboration events that characterize the establishment of the
interface between newly formed bone and solid surfaces. These solid sur-
faces either may be implant materials, or indeed previously formed bone
matrix, which has been resorbed during normal bone remodeling events. In
each case the first biologically derived matrix at these sites is a morpho-
logically distinct collagen fibre-free extracellular matrix, which, in bone
histology has been referred to for >lo0 years as a cement line.
Results: The sequence starts with secretion and adsorption to the sub-
stratum of organic components, of which the major proteins are osteopon-
tin and bone sialoprotein. Mineralization of this matrix occurs by the seed-
ing of nanocrystalline calcium phosphate, which precedes the appearance
of morphologically identifiable collagen fibres. This is clearly contrary to
the dogma that collagen is necessary for mineralization of bone, but is in
agreement with specific cases of other, particularly dental, calcified con-
nective tissues. Although collagen is synthesized by the differentiating os-
teogenic cells that elaborate the cement line interface, it is not adsorbed to
the underlying solid surface. Following the elaboration of the cement line
matrix, collagen fibre assembly occurs and is then mineralized to produce
morphologically identifiable bone matrix.
Conclusions: Key elements of this sequence of events can be seen at the
interface of implants retrieved from in vivo experiments, which indicates
that these in vitro methods not only mimic known in vivo phenomena, but
also provide a mechanistic understanding of bone elaboration at implant
surfaces. However, distinction is drawn between the events of new bone
formation at implant surfaces and other bonehmplant morphologies, which
are unrelated to de novo bone formation at the implant surface. Finally,
this new information emerging from bone marrow cell culture studies de-
mands a re-examination of the concepts of bone-bonding and nonbonding
implant materials. o 19% Wiley-Liss, Inc.
Key words: Implant, Interface, Osteoblasts, Cell culture

Conventional wisdom holds that integrity of the in- plant surfaces would seem to be essential, not only in
terface is critical to long-term clinical success of load- understanding the biological events that take place a t
bearing, bone-interfacing implants. It is assumed that such interfaces but also, ultimately, in dictating the
juxtaposition of bone tissue to the implant surface and,
in imprecisely defined terms, the structure and vitality
of this bone tissue also will have a bearing on the clin-
ical outcome- Thus the development of a mechanistic Address reprint requests to:J.E. Davies, Centre for Biomaterials,
explanation of bone tissue formation at candidate im- University of Toronto, Ontario, Canada M5S 3E3.
0 1996 WILEY-LISS, INC.
 MODELING OF BONE/IMPLANT INTERFACE
materials characteristics necessary to generate specific
biological responses. But, to date, no such mechanistic
explanation has emerged.
The formation of bone on a material surface depends
upon the metabolic and secretory activities of only one
cell type: the osteoblast. Although this dogmatic state-
ment emphasizes the importance of the colonization of
the implant by phenotypically mature osteoblasts, it
does not negate the importance of many other factors,
both implant and biological in nature, which are in-
volved in both wound repair mechanisms and bone re-
modeling phenomena. For this reason we can restrict
ourselves herein to formation of bone matrix at an in-
terface without consideration of either earlier (ion ex-
change and macromolecular adsorption) or later (tissue
remodeling) interfacial events.
Most studies reporting the relationship of bone tissue
to implants have employed histological procedures and
light microscopic observation. This is the method of
choice in many cases, particularly those of pathological
specimens or tissue harvested during implant revision
procedures. In animal studies, more flexibility exists
and specimen preparation following retrieval can be
factored into the design of the experimental protocol, or
indeed the implants employed, and ultrastructural
studies become feasible, many of which are reviewed in
a companion article (Listgarten, 1995). However, his-
tological evidence is retrospective and is unable to pro-
vide a mechanistic explanation of the sequence of
events that results in bone formation.
In other research disciplines related to normal bone
structure and function, metabolic bone diseases and
the affects of pharmacological agents on bone, biolo-
gists have increasingly turned to biochemical, cell, or
molecular biology experimental regimes to model, and
thus understand, aspects of bone cell behaviour. The
implant interface poses unique questions that, to be
correctly addressed, demand an intimate knowledge of
both materials science and also bone cell biology. Thus
it would seem logical to address the exploration of the
bonelimplant interface armed with biochemical, cell,
and molecular biological tools to study the osteogenic
cells that colonize the material surface and the extra-
cellular matrix that they elaborate. The latter is of
particular importance in understanding the behaviour
of connective tissue cells in general since, by definition,
they fabricate a matrix around themselves to construct
the tissue type that is identified histologically. Despite
the apparent logic in this cellular approach to explor-
ing and understanding bone implant interfaces, little
information has been forthcoming on the mecha-
nisms by which materials influence bone tissue devel-
opment.
In summary, therefore, we can make the following
statements: (1)bone is made by osteoblasts, (2) osteo-
blasts are fully differentiated cells of the osteogenic
lineage that are derived from undifferentiated mesen-
chymal cells through an ill-defined number of discrete
differentiation stages, (3) in bone healing sites, al-
though osteoblasts themselves have the ability to mi-
grate (Jones and Boyde, 19771, it is usually less differ-
entiated cells in the osteogenic lineage, or perhaps
undifferentiated mesenchymal cells, that will migrate
to, and colonize, an implant surface, and (4) in order to
create a bone-implant interface, osteogenic cells must
427
colonize the implant surface. These cells must then
make bone tissue.
In spite of the apparent simplicity of this approach,
these statements beg several questions of which the
following are critical: how do cells make bone on for-
eign surfaces? What is the differentiation state of the
osteogenic cells that colonize an implant? How do these
cells adhere to the implant surface in a manner that
permits maturation of the osteoblastic phenotype? Is
there an identifiable sequence of matrix formation
events that characterizes this bone formation at an in-
terface? Is this sequence of events the same that occurs
in normal bone tissue, or does it differ as a result of the
presence of a nonbiological material? In other words,
can the sequence of matrix formation events on im-
plant surfaces be affected in either subtle or overt ways
by the surface properties of the material?
Recently, biochemical and molecular biological tech-
niques have been employed t o monitor differentiation
events and matrix formation during bone healing in
vivo (Jingushi and Bolander, 1991). Although this ap-
proach holds great promise for understanding the biol-
ogy of fracture healing, the resolution of such tech-
niques when applied to tissue recovered from in vivo
experimentation is several orders of magnitude below
that which can be achieved by the use of cell culture
models. Specifically, identifying interactive events
that occur between the biological and material com-
partments within the first nanometer, or indeed the
first micron, of an interface provides an almost insur-
mountable problem.
Before embarking on a detailed review of the infor-
mation in vitro techniques can yield with respect to de
novo bone formation a t solid surfaces, it is salutary to
discuss not only the manner in which bone tissue in-
terfaces with itself during normal remodeling events,
but also the heterogeneity of the interfacial structures
and compositions that can occur a t the surfaces of im-
planted endosseous implants. Only then can we iden-
tify the specific scenarios that can be modeled using
cell culture methods.
BONEBONE INTERFACE
Bone is a dynamic tissue that undergoes remodeling
throughout life. In the remodeling process, new bone is
laid down on the resorbed surface of old bone, and thus
bone tissue creates an interface with itself on a contin-
ual basis. The first description of the bone/bone inter-
face, the boundary lines between secondary osteons and
surrounding interstitial bone matrix, was probably that
of Deutsch (cited by von Ebner, 1875). These were sub-
sequently described in greater detail by von Ebner, who
called them Kittlinien (translation: cement lines). Ce-
ment lines can be visualized by haematoxylin staining
and appear as basophillic bands in both decalcified par-
affin (Weinmann and Sicher, 1955; Pritchard, 1972;
Sokoloff, 1973) and undecalcified plastic sections (Gru-
ber et al., 1985; Villanueva et al., 1986). Hitherto, ce-
ment lines have not been considered primary features
of bone tissue, as they arise during the reversal period
of bone remodelling and demarcate where resorption
was completed and bone formation was initiated (Prit-
chard, 1972;Parfitt, 1983).However, it is a solid surface
that is the essential prerequisite for cement line elab-
oration. Thus, whereas cement lines are not seen in the
428 J.E. DAVIES

Fig. 2. Diagrammatic representation of the tissue interfaces formed

with the endosseous component of a dental implant. I, Implant; C,
cortical bone; M, marrow cavity; T, trabeculae; arrows = areas where
de novo bone is forming on the implant surface.

earliest stages of intramembraneous ossification, they

are seen a t the junction of newly forming bone and
calcified cartilage during endochondral ossification
(McKee et al., 1989).The width of cement lines has been
variously reported as ranging from 0.2 pm to 5 pm
(Villanueva et al., 1986; Philipson, 1965). This vari-
ability would seem to depend on species, section thick-
ness, (Philipson, 1965) individual variation, (Villan-
ueva et al., 1986),and method of examination (Sokoloff,
1973). Early studies showed that silver impregnation
methods for demonstrating fine fibrils are rejected by
cement lines, leading to the usual conclusion that the
cement lines do not contain collagen (Weidenreich,
1930). Later, ultrastructural investigations reported
that the cement line regions appear as a “ground sub-
stance” by both transmission and scanning electron mi-
croscopic examination. The ground substance is rich in
sulphur and can be digested by trypsin, suggesting the
presence of sulphated protein polysaccharide complexes
in cement lines (Frasca, 1981; Frasca et al., 1981). Mc-
Kee and Nanci (1993) have shown elevated expression
of osteopontin a t these sites with respect to the sur-
rounding bone. Information from micrographic studies
and X-ray diffraction analyses of cortical bones have
shown that cement line regions have the same content
of minerals and sulphur as that in surrounding bone
tissue (Philipson, 1965).
The fundamental question of whether the cement
line was a distinct morphological entity or only a geo-
Fig. 1. Scanning electron photomicrographs of the surface of young metrical boundary was still being debated a century
adult rat trabecular femoral bone. A. The bone surface has been re- after their initial description (Sokoloff, 1973). How-
sorbed by an osteoclast and a portion of the resorption lacuna (L)can ever, since the turn of the last decade, ultrastructural
be seen to comprise a microporous collagen bed. On the periphery of evidence from TEM (Baron et al., 1980; Tran Van et al.,
this lacuna, differentiating osteogenic cells have laid down a globular
matrix to cover the collagen (G)and this acts as an anchorage site for 1982) and SEM (Frasca, 1981; Frasca et al., 1981) stud-
new collagen fibres seen a t the bottom of the field of view (0.Field ies has shown that cement lines are real morphological
width = 8.9 pm. B. High magnification of the new collagen being laid structures, although they had remained refractory to
down on the globular matrix which covers the collagen of the old bone
which has been resorbed as described above. Field width = 3.6 pm detailed compositional or structural analysis. A recent
(from the work of H. Zhou). exception to the latter is the elegant work of the Mon-
treal group (McKee et al., 1989,1990,1993; McKee and
Nanci, 1993; Nanci et al., 1994) using ultrastructural
immuno-localization to reveal the composition of hard
tissue interfaces. Therefore, cement lines are clearly
 MODELING OF BONE/IMPLANT INTERFACE
Fig. 3. Cartoon sequence of the appearance of an initial remodeling
event taking place at the bone interface with a threaded implant.
Prior to remodeling the osteonal structure of the peri-implant bone is
oriented horizontally. The remodeling canal is impinging on the im-
plant surface such that when filling of the osteonal structure is ini-
tiated with the formation of a cement line, this interfacial matrix will
also be formed on the implant surface. This sequence is described in
detail in the text and is after Brunski (1991).
integral parts of bone tissue forming a t unique sites
that mark the junctions of old and new bone and devi-
ate from the composition and structure of the bulk bone
tissue, As such, the extracellular matrix they comprise
is distinct from the mineralized collagenous matrix
elaborated by phenotypically mature osteoblasts. This
interpretation is contrary to that which relies on cells
of the osteogenic lineage being nonsecretory before
they express their mature phenotype (Caplan, 1991).
Using the in vitro method, which forms the focus
here, we have shown that osteogenic cells produce a
layer of noncollagenous matrix, a t the surface of non-
biological substrata, similar to that described as the
cement line in vivo. Specifically, this linear mineral-
ized interfacial matrix is elaborated prior to extracel-
lular collagen fibre assembly as individual calcified
globular accretions that fuse to form a continuous layer
(the details of which are described later). This layer is
-0.5 km thick and is rich in calcium, phosphorus, os-
teopontin, and bone sialoprotein (Shen et al., 1993;
Peel, 1995). These in vitro observations led us to hy-
pothesize that the cement line is elaborated in natural
bone tissue as a layer of calcified globular accretions
that should be morphologically distinguishable from
its surrounding tissues and that we have confirmed
experimentally (Zhou et al., 1994) (Fig. 1). The possi-
bility of using this in vitro approach t o investigate both
bone resorption and bone formation in multiple-stage
429
cultures, as first reported in 1991 by Davies et al., al-
though feasible, is not the focus of this report.
TISSUE INTERFACES WITH ENDOSSEOUS IMPLANTS
Prior to implantation of an endosseous implant, the
receptor bone defect is usually created by drilling or
reaming, and newly created bone surfaces are almost
instantly covered with blood. In all cases, whether or
not the hole is tapped to accommodate an implant
thread, the first tissue to come into contact with the
implant surface is blood, although the static blood vol-
ume surrounding the implanted device will vary con-
siderably as a function of implant design and implan-
tation site. In the case of dental implants, irrespective
of implant design, the implant surface will be juxta-
posed to cortical bone in the alveolar ridge. This corti-
cal bone contact may extend, imperfectly, along the
entirety of the buccal and lingual (palatal) aspects of
the implant, whereas the mesial and distal aspects will
be in contact with either trabecular bone or bone mar-
row. Furthermore, even with precisely implanted,
screw-threaded implants, the residual implant/bone
space will be variable, with the tips of the implant
screw threads compressing the neighbouring bone tis-
sue, whereas other areas will be occupied by blood to-
gether with cell and tissue remnants.
Thus at the outset, different surfaces of an endos-
seous implant will be juxtaposed by varying amounts of
different tissues. This has a critical influence on the
structure and composition of the bonehmplant inter-
face in different sites. In the trabecular/marrow (med-
ullary) compartment, blood clot resolution and white
cell infiltration will be followed by invasion of the
wound site by both endosteal cells of osteogenic pheno-
type and undifferentiated mesenchymal cells with the
capacity to differentiate into osteogenic cells. Thus
bone cells will reach the implant surface within a short
time (days) of the implantation procedure.
With respect to the transcortical portion of the im-
plant, initial healing patterns will differ with the im-
plant design. Porous coated implants provide a rela-
tively large peri-implant blood clot volume, which,
during resolution, can provide a three-dimensional net-
work of fibrin and collagen Type I11 through which, by
migration of endothelial and mesenchymal cells, neo-
vascularization and osteogenesis can occur. The nutri-
ent supply for such new tissue formation will be from
both cortical and medullary sources. In the case of
threaded implants, whereas blood and cell remnants
may be trapped between the implant surface and cor-
tical bone, there will be almost no possibility of coloni-
zation of the implant by osteogenic cells. The local cor-
tical vascular architecture will be disrupted by the
surgical procedure, irrespective of implant type. How-
ever, blood clot resolution in the small peri-implant
volume of screw-threaded implants will be slower than
that in the marrow space and will not be readily acces-
sible to invasion by cells of mesenchymal origin. Fur-
thermore, due to the limited peri-implant space in such
designs, together with compression of the juxtaposed
bone by the screw threads, there will be generally in-
sufficient space to establish peri-implant capillary
anastomoses, and blood flow in the peri-implant bone
will stagnate, leading slowly to bone necrosis. Only by
bone remodeling will there be a gradual replacement
430 J.E. DAVIES
Fig. 4. Scanning electron micrographs to show the stages of matu-
rity of extracellular matrix formation in bone nodules. A. The over-
lying cells and matrix of the nodule have been removed in the centre
and bottom left hand corner of the field of view to expose globular
accretions on the underlying polystyrene substrate increasing in in-
tensity from the periphery (bottom left) toward the centre (top right)
of the nodule. Field width = 86 wm. B. Plan view of a bone nodule
(months) of the peri-implant bone with the possibility
of de novo bone formation a t the implant surface (see
below).
The different endosseous implant/tissue interfaces
are illustrated diagrammatically in Figure 2. It can be
seen clearly that, depending upon the site of examina-
tion, the implant/tissue interface could be either old or
remodeled bone in the cortical compartment, or new
bone formation, as a result of trabecular repair or de
novo bone formation from mesenchymal cells, and mar-
row in the medullary compartment. Thus a t first glance,
in vitro modeling of the bonehmplant interface, which
focuses on the elaboration of new bone tissue a t the
implant surface, is applicable only to healing reactions
in the medullary compartment of endosseous devices.
However, the question of cortical remodeling also
should be briefly addressed. Brunski (1991) has ele-
gantly demonstrated the histological sequelae to vas-
cular interruption and bone necrosis in peri-implant
cortical bone. He showed following implantation of
Brhemark-type implants in canine tibiae, that remod-
eling, which is the result of re-establishment of the
peri-implant vasculature, results in the rearrangement
of osteons such that new osteons circle around the im-
plant with their long axes parallel to the implant sur-
face and perpendicular to the long axis of the implant.
The fact that this remodeled bone can extend up to 1
mm from the implant surface (Roberts, 1988)led Brun-
ski (1991)to include this tissue in the definition of the
bone/biomaterial interface. The establishment of one
that has been disrupted to show the underlying substrate. At this
magnification the globular accretions at the substratum interface are
only just visible. However, within the more dorsal levels of the extra-
cellular matrix, collagen fibres can be seen to contain nodular foci of
mineralization. The extracellular matrix forms the most part of the
nodule from this perspective. Field width = 86 pm.
such osteonal system is illustrated in Figure 3. Here, at
implant placement the original osteonal structure is
running horizontally and is interrupted by the implant
placement. A new osteon is established by the osteo-
clastic resorption of bone, which will be dependant
upon the continuously remodeling vascular supply.
The bony tunnel so created will then fill with new bone
lamellae to complete the osteonal system. The new
bone will be separated as usual from the old bone in
this remodeling cascade by a cement line (see above),
which is the initial extracellular matrix created by dif-
ferentiating osteogenic cells (osteoblasts). Where the
resorption tunnel impinges on the implant, as illus-
trated, the cement line matrix also will be deposited
directly on the implant surface. Thus in these precise
locations, which will increase in number with the num-
ber of remodeling events a t the implant surface, de
novo bone formation will occur on the transcortical por-
tion of the implant. It is highly probable therefore that
on transmission electron microscopic examination of
retrieved implants, the ultrastructure of the cortical
bonehmplant interface could include fields of view con-
taining radically different interfacial structures, as de-
scribed above.
BONE CELL CULTURES
Many bone cell culture systems have been described
in the literature, using either cell lines or cells derived
from different sites in different species, from chick pe-
 MODELING O F BONEfIMPLANT INTERFACE
Initial oraanic matrix
Seeding of calcium phosphate crystallites
en calcification and matrix mineralization
Fig. 5. Diagramatic representation of the establishment of an inter-
face between bone tissue and an underlying substrate. The initial
organic matrix (A) becomes seeded with nanocrystals of calcium phos-
phate as shown in B. These crystals grow and initial collagen fibre
assembly takes place on their surfaces (C). In D, the interface is
almost complete where the interfacial layer is the mineralized organic
matrix, which separates the substrate from the overlying collagen
containing compartment,the collagen itself is now beginning to min-
eralize (the thicker fibres), and there are also foci of a mineralization
occurring in the collagen compartment that are independent of the
fibres themselves.
riosteum to human femoral trabecular bone out-
growths, most of which have been adopted by those
investigating cell/biomaterial interactions. These sys-
tems, and the information they could provide, were re-
viewed in detail in Davies (1990). Since then, one cul-
ture system has emerged as a powerful means of
modeling the formation of bone on solid surfaces and
produces matrix interfaces that match those found in
vivo when de novo bone forms on implanted materials.
Since it is only the latter aspect of bone cell behaviour
that forms the focus of this discussion, we briefly de-
scribe the methods used, the morphological and bio-
chemical tools that can be combined with the basic cul-
ture protocol, and demonstrate how this method can be
used to interpret the morphological and immuno-label-
ing studies reported from in vivo studies. However, we
must also address the issue of whether an in vitro assay
bears any relevance to in vivo reality. This is impor-
431
tant in view of the belief expressed by Albrektsson
(1992) that such experiments “reflect the assumed re-
ality of the test tube, [which is1 very different from the
in vivo situation.” Thus examples are provided that
demonstrate that the interfacial structure achieved in
vitro is equivalent to the cement line structure ob-
served both at remodeling sites in natural bone tissue
and on retrieved implants.
BONE MARROW CULTURES
Primary bone marrow cultures can provide a popu-
lation of differentiating osteogenic cells that elaborate
high yields of morphologically and biochemically iden-
tifiable bone matrix in short experimental periods.
This in vitro method, and its critical dependence upon
the presence of differentiating cells at the substratum
interface, has provided not only new information on the
formation of bone on foreign surfaces, but also a radical
re-interpretation of ultrastructural descriptions of
bone biomaterial interfaces previously available in the
literature. A brief description is provided below to-
gether with some explanation of the cells and culture
conditions employed. Following this, some of the meth-
odological tools that have been employed in conjunc-
tion with these cultures are given. It is not the purpose
here to repeat technical details available in the refer-
enced literature but where pertinent, technical infor-
mation is included in the figure captions to facilitate
comprehension of the images.
Adult Rat Bone Marrow (RBM) Protocol
The philosophy underlying the use of primary cells,
rather than cell lines, or clones, has been provided be-
fore in Davies (1990) and is not repeated here. Never-
theless, it is important to emphasize that, using this
adult animal source, only a small fraction of the har-
vested cells are spontaneously osteogenic, and thus the
differentiation of marrow stem cells during the culture
period is a function of the culture conditions. In this
culture protocol, heterogeneous cell populations are
harvested from whole marrow of young adult male
Wistar rats according to the method described previ-
ously (Davies et al., 1991a). The culture medium,
a-Minimal Essential Medium (a-MEM)containing pen-
icillin G, gentamicin, fungizone, and foetal bovine se-
rum, is further supplemented with freshly prepared
ascorbic acid, Na P-glycerophosphate, and dexametha-
sone (DEX). The specific concentrations of these culture
medium additions employed were those first reported
for this system by Maniatopoulos et al. (1988).Aliquots
of the harvested RBM cell suspension can be added to
any culture vessel, the choice of which will depend on
the purpose of the experiment and the geometry of the
candidate material (substratum) on which the cells are
to be grown. Either primary explants or passaged cells
can be seeded onto the substratum, although a medium
change after the first 24 hours is advisable to remove
the large population of red blood cells. The advantage of
using first passage cells is that nonadherent cell pop-
ulations can be almost completely eradicated. We have
maintained such cultures for periods of hours to months,
depending on experimental design, although, most com-
monly, we have employed a 2-week period to generate
high yields of bone matrix. However, much of the most
important information concerning matrix elaboration
432 J.E. DAVIES

TABLE 1. Major noncollagenous proteins of bone (from Peel, 1995)

Protein Expression Some suggested functions
Osteonectin (SPARC) Widespread expression in hard and Modulates cell-matrix interactions
soft tissues
Regulate mineralization
Osteopontin (SPP1, BSP I) Widespread expression Cell adhesion
Tissue remodeling
Nucleate or regulate mineralization
Bone sialoprotein (BSP 11) Primarily localized to osteoblasts, but Nucleator for mineralization
has been identified in platelets and
cancer cells
Osteocalcin (bone Gla protein, BGP) Primarily restricted to mineralized Osteoclast recruitment
tissues, but OC mRNA has been
identified in platelets
Matrix Gla protein Widespread expression in hard and Regulator of mineralization
soft tissues
CS-PG I (biglycan) Chondroitin sulphate form is unique to Tissue remodeling and wound healing
bone, but core protein same as
dermatan sulphate PG biglycan
found in many soft tissues
CS-PG I1 (decorin) Chondroitin sulphate form is unique to Regulate fibrilogenesis
bone, but core protein same as
dermatan sulphate PG decorin found
in many soft tissues
CS-PG I11 Restricted to bone ? Regulate mineralization
Albumin Synthesized in liver and has Regulating mineralization carrier for
widemread distribution small molecules and ions
cY,HS-glycoprotein Synthegized in liver and has Bone remodelling and resorption
widespread distribution

Fig. 6. Immuno gold labelled transmission electron photomicrographs of CS-56 gold conjugated anti-
bodies. A. A cell process exhibiting CS-56 labelling is in contact with the substrate surface, which shows
smaller areas of a fine electron-dense matrix that itself is positively labelling for the antibody. Field
width = 2.9 pm. B. Mineralization has commenced in this early organic matrix as seen by the very fine
needlelike crystallites, and the antibody no longer penetrates but only labels the dorsal surface of this
initial matrix. Field width = 4.4 pm.
 Fig. 7. Fluorescence photomicrographs to demonstrate labelling of
three antibodies. A and B. Osteopontin label at earlier (A) and
slightly later (B) stages of culture. The FITC conjugated antibody
shows green while the yellow/orangecolouration is auto-fluorescence
cause by the fixation procedure. Note that in (a)the osteopontin labels
predominantly a t the ends of cell pseudopods (*) and to a lesser extent
on the substratum surface (arrows). In (b), the label is intense on the
substrate surface (*) and even more so around cell peripheries (ar-
rows). Where the cells have changed in shape, striae are left in the
signal at the substratum level (arrowheads).C. Antifibronectin label-
ling shows a no label attached to the underlying substratum surface,
but a strong signal is coming from cell cell adhesion patterns. D.
Anticollagen Type-I labelling demonstrates that collagen is present
within the cells both at high (arrows) and lower (arrowhead) concen-
trations, but no signal is seen on the substrate surface.
434 J.E. DAVIES

at the substratum surface has been gained from culture acid (EDTA), which permits extraction of mineral-
periods of 3-8 days. bound proteins and the radio labeling identified by au-
Some comment should be made concerning our addi- toradiography. A detailed description of a range of such
tion of p-glycerophosphate to the culture medium. This extraction procedures has recently been provided else-
inclusion in the original culture medium by Maniato- where (Peel, 1995).
poulus et al. (1988) was based on the observations by
Tenenbaum that P-glycerophosphate was necessary to Forms of microscopy
potentiate calcification of a collagenous extracellular Although the different microscopic methods of exam-
matrix by putative osteoblasts (Tenenbaum and Heer- ining the cultures are not reviewed here, it is worth
sche, 1981). More recently the optimal level of p-glyc- noting that the immunogold labeling method for SEM
erophosphate required for calcification of cartilage cal-
cification was fixed by a t 2.5 mM Boskey et al. (19921, was adapted from the procedure described by de Har-
who reported that it was also possible to substitute this ven et al. (1984). Also, one method of transmission elec-
organic phosphate source by 4 mM inorganic phos- tron microscopy, described generically as field emission
phate. We have recently conducted experiments to as- analytical microscopy (FETEM), should be mentioned.
sess the effect of reduction of p-glycerophosphate from This relatively recent addition to the armamentarium
10 mM to zero in 2 mM steps. The level of mineraliza- of the microscopist relies on the design of a new type of
tion decreased with decreasing level of PGP and simi- TEM based on the production of, usually, 200 kV elec-
lar observations were made using inorganic phosphate, trons from a field emission electron gun. The electrons
although ectopic calcification was seen above 8 mM Pi. provide significant advancements over conventional
These findings are contrary to those of Tenenbaum et thermionic electron sources, even those used in high
al. (1989), who reported that, in a chick periosteal os- resolution TEM. The electron source diameter is ap-
teogenesis model, cultures augmented with inorganic proximately three orders of magnitude smaller (subna-
phosphate failed to mineralize, whereas those with 10 nometre), there is an equivalent increase in brightness,
mM $-glycerophosphate added to the culture medium and a beam energy width of approximately one-tenth
mineralized. However, they also noted that PGP-con- that of the thermionic emitters. When combined with
taining cultures exhibited significantly less bone ma- energy dispersive X-ray analysis (EDX) and/or electron
trix, suggesting that p-glycerophosphate has global ef- energy loss spectroscopy (EELS), these characteristics
fects on bone cell metabolism in vitro including its role confer, on FETEM, a capacity of elemental analysis
in mineralization. Thus in discussing the role of addi- from areas of <1 nm, which has enabled the visualiza-
tives to culture media, it is important to take into con- tion and analysis of earlier stages of biological calcium
sideration not only the physical chemistry of calcifica- phosphate mineralization than has hitherto been pos-
tion phenomena, but also the significant differences in sible with conventional "EM (see below).
cell phenotype.
NEW BONE FORMATION AT A SOLID SURFACE:
Additional Tools SEQUENCE OF EVENTS
Overview
Antibody labeling
The general appearance of the cultures is illustrated
Initial characterization of the earliest organic matrix
in Figure 4, and the similarity between Figure 1 (in
secreted by the differentiating osteogenic cells in thisvivd and Figure 4 (in vitro) is immediately apparent
culture system has been achieved using fluorescent or with a globular matrix forming the interface between
gold-labelled antibodies to known matrix components, the underlying substratum (bone and polystyrene, re-
combined with fluorescence, scanning (SEMI and spectively) and the collagen compartment of the new
transmission (TEM) electron microscopy, and immuno- bone. Figure 5 is a cartoon summarizing the sequence
SEM (Shen et al., 1993; Peel, 1995). Although it is of events occurring a t the interface. Briefly, differen-
important that the results of such labelling correlate tiating bone cells a t the substratum surface will secrete
both with our previous morphological observations and a collagen-free organic matrix (Fig. 5A), which pro-
other reports in the literature concerning expression ofvides nucleation sites for the initiation of nanocrystal-
osteoblast phenotype-related gene products, it is criti-line calcium phosphate mineralization (Fig. 5B). This
cal to note that antibody labeling of the substratum/ initial organic phas is to be expected since such a se-
culture interface is possible only by first surgically
dissecting the cultured tissue to provide antibody pen-
f
quence occurs in a1 biomineralization systems where
macromolecules containing acidic groups, sulfates, car-
etration to the substratum surface. In this regard the boxylic acids, or phosphates may be involved (Lee et
permiablization of the culture to allow antibody pene- al., 1977; Weiner, 1979; Tarasevich et al., 1991; Alper,
tration is not suitable when cell and matrix labeling is1992).
to be readily distinguished. Table 1 summarizes the major noncollagenous pro-
teins to be found in bone. Glycoproteins have been as-
Radio labeling signed two roles in calcifiable tissues by de Bernard
Cultures prepared as described above can be labeled (1982). First, they act as structural components in
with 3H-Proline or 35S-Methionine,which is incorpo- association with proteoglycans and provide calcium
rated into cell-synthesized containing proteins, to ex- and/or phosphate binding sites. Second, they modulate
amine secretion into the extracellular compartment. calcium-dependent phosphatase and ATPase activities,
Matrix components can then be resolved by electro- which could indicate a role in initial mineral deposi-
phoresis before and/or after treatment of the cultures tion. In bones and teeth, phosphoproteins are particu-
with compounds such as ethylinediaminetetracetic larly implicated in the seeding of calcium phosphate
 Fig. 8. High resolution FETEM images of the early crystals laid
down in a stage of culture represented by Figure 6B. A. At one million
times magnification,the needlelike crystals are barely visible as dark
linear arrangements approximately traversing the field of view. This
appearance is caused by the in-block uranyl acetate contrasting. Field
width = 90 nm. B. High magnification of the part of the field of view
illustrated in A. Here, the individual crystallites are clearly visible as
lattice fringe patterns distributed in a random fashion throughout the
field of view. Due to the multiple orientations of these small crystal
structures different spacings are seen in varioius locations. The cir-
cled area contains an hexagonal array of spots that illustrate, in this
case, that the electron beam was travelling parallel to the C-axis of
the crystals. Areas (d) and (e) are illustrated a t high magnificationsin
D and E. Field width = 36 nm. C. EDX analysis from an area con-
tained within the crystal lattice fringe patterns using a 1nm spot size
(approximately equivalent to the ringed area in B). The uranium
peaks are due to the in-block contrasting, whereas the signal clearly
shows the presence of both calcium and phosphorus. D and E: Two
examples a t high magnification of the lattice fringe patterns seen in
B. Field width = 7.8 nm.
436 J.E. DAVIES

Fig. 9. Transmission electron photomicrographsto demonstrate two later stages of mineralization than
that shown in Figure 5b. A. The small foci of mineralization are seen beneath the overlying cell indi-
vidual crystals are clearly discernable. Field width = 2.6 pm. B. A similar field of view but after
significant crystal growth. Field width = 1.6 Fm. Note that in each of these photomicrographs, the
apparent needlelike shape of the crystallites is shadowed by a grey rectangular image. The latter would
indicate a plate rather than needlelike morphology.

crystallites (Mann, 1988). Calcium phosphate crystal tecture of the organic molecules that will primarily
growth (Fig. 5C) follows nucleation. It is interesting to influence the calcium phosphate phase nucleated,
speculate on the significance of the reciprocity of the whereas once mineralization proceeds, and crystal
organic and inorganic phases. Initially, it is the archi- growth ensues, it is the inorganic phase, due to the
 MODELING O F BONE/IMPLANT INTERFACE 437
TABLE 2. Comparison of methods in studies that demonstrate bone-substratum interface in vitro
(from Peel, 1995)

~~
C,T,M' Species Age Site2 Isolation Subcultured Medium
Bhagarva No Rat Fetal Stripped Collagenase- Cells were cultured a-MEM 15%
et al., 1986 calvaria 11-V elastase for 24 hours and FCS NO Dex
then subcultured
Bouvier No Rat Neonate Calvaria I-V Collagenase Yes DMEM + F 1 2
et al., 1994 10%FCS NO Dex
Davies et al., Yes, Rat Young Marrow Mechanical Yes a-MEM 15%
1990, 1991a,b reported adult and no _ _ _ with Dex
FCS
de Bruijn Yes, Rat Young Marrow Mechanical Yes a-MEM 15%
et al., 1992 reported adult FCS with Dex
Gerstenfeld Yes, not Chick Fetal Stripped Trypsin- No MEM or BJGb 10%
et al., 1992, reported calvaria V collagenase FCS NO Dex
1993
Melcher Yes, Rat Fetal Stripped Collagenase- Yes a-MEM 15%
et al., 1986 reported calvaria 11-IV elastase FCS NO Dex
Pitaru Yes, not Rat Young Marrow Mechanical Yes a-MEM 15%
et al., 1993 reported adult FCS with Dex
Pockwinse Yes, not Rat Fetal Stripped Trypsin- No MEM or BGJb 10%
et al.. reDorted calvaria IV -
collagenase FCS NO Dex
1992,'1993
Sautier et al., Yes, Rat Fetal Calvaria with Collagenase Yes D-MEM 10%
1991,1992. reported periosteum FCS NO Dex
'Cement line matrix.
'Site = origin ofprimary tissue. I, 11,111,IV, V refer to the population released after various lengths of incubation with enzyme (I, 0-10 minutes,
11, 11-20 minutes, 111, 21-40 minutes, IV, 40-60 minutes, V, 60 + minutes).

highly ordered ionic structure of the crystalline unit ingly, when this initial organic matrix demonstrates
cells, that will orchestrate the organic architecture. the first evidence of mineralization as is seen in Figure
Importantly, in the implant context, it should be em- 6b, the antibody to chondroitin sulphate no longer pen-
phasized that the substratum does not act as an epitac- etrates the matrix but remains on the surface, which is
tic nucleator in this cell-mediated and protein-depen- continually added to by the overlying cells. It is at this
dent, biological mineralization phenomenon. Thus it is stage of mineralization that we have begun to charac-
critical to differentiate this cell-mediated elaboration terize the crystalline nature and composition of the in-
of a hard tissue matrix from the spontaneously forming organic phase (see below).
calcium phosphate layers that can occur on many ma- We have, to date, employed antibodies against osteo-
terials, even in nonbony sites (LeGeros et al., 1991). pontin, fibronectin, bone sialoprotein, osteocalcin, and
Concomitant with crystal growth in the developing collagens Type I and 111,in this culture system. Whereas
interface, there will be initiation of collagen fibre collagen was not found to be adsorbed to the underlying
assembly. Finally, calcification of the collagen com- substratum, the appearance of osteopontin, initially at
partment will occur (Fig. 5D), both in association with the ends of cell attachment processes a t the substratum
individual collagen fibres or in the interfibre compart- interface, was ofconsiderable interest (Shen et al., 1993;
ment. However, in each case it is likely that the same Peel, 1995). Over an increasing culture period, osteo-
noncollagenous protein, whether bound to collagen fi- pontin was adsorbed to the underlying substratum (Fig.
bres or not, will be responsible for providing the cal- 7a,b). It was particularly interesting to note, in the
cium phosphate crystal nucleation sites. Thus the col- osteopontin signal, the unlabeled striae left a t the sub-
lagen compartment of bone will be separated from the stratum surface. These would indicate that no osteo-
underlying substratum by a collagen-free calcified tis- pontin remained below the threshold detection of our
sue layer containing noncollagenous bone proteins. fluorescence probes in these areas that were aligned
This layer is -0.5 pm thick, as are cement lines found with the remaining attachment areas of the neighbour-
in vivo. ing cells. This observation may indicate that as the cells
Each of four stages that have been identified as a change in shape concomitant with differentiation, the
result of in vitro modeling is dealt with in more detail cell processes retract and remove at least some of the
below. osteopontin that was laid down on the substratum dur-
ing an earlier period when they exhibited a more flat-
Early Organic Matrix and Change in Cell Shape tened morphology. This is particularly evident when
Early culture stages, 3-8 days, have demonstrated the left and right sides of the cell in Figure 7b are
the appearance of a fine electron-dense matrix of an compared. These observations correlate closely with our
amorphous nature at the culture dish surface beneath previous description of the changing shape of these cells
putative bone nodules. As shown in Figure 6, this ma- as seen by scanning electron microscopy (Davies et al.,
trix labels positively for chondroitin sulphate as does 1991a). We have also observed a similar early distri-
the exterior surface of the adjoining cell membrane. bution of bone sialoprotein (BSP) at the substratum
Control incubations with serum containing media but surface. However, Peel (1995)did not find BSP localized
without cells showed no such labeling, and thus it can to cell attachment processes, as seen in the case of os-
be concluded that the chondroitin sulphate is being teopontin, although this may be due to employment of
produced by cells at the substratum interface. Interest- a less specific antibody.
438 J.E. DAVIES

The roles of noncollagenous bone proteins have been

thoroughly reviewed in the implant context by Sodek
et al. (1991). Of these osteopontin, which together with
the bone proteoglycans biglycan and decorin is highly
sulphated, deserves special attention. Indeed, we have
observed, using imaging electron energy loss spectros-
copy (imaging EELS) the presence of sulphur in the
cement-like interfacial matrix (Davies et al., 1991b).
Unlike osteocalcin, which is specific to mineralizing
connective tissues, osteopontin is found in nonminer-
alized tissue, although a posttranslational modification
renders that found in bone characteristically sulphated
(Nagata et al., 1989). It contains a (G)RGD(S) cell at-
tachment sequence (Oldberg et al., 1986) and is impli-
cated in osteoblast attachment (Butler, 1989) as our in
vitro results would confirm. Indeed, Somerman et al.
(1987) have shown that guanidine EDTA protein ex-
tracts of bone-enhanced fibroblast cell attachment to
culture substrata, and they implicated osteopontin in
this role (Somerman et al., 1987). Osteopontin also has
a calcium binding domain, although this is more likely
associated with control of calcium phosphate crystal
growth than with mineral nucleation, which has been
attributed to BSP (Hunter et al., 1994a); see below.
Whereas osteopontin is associated with both cell at-
tachment and calcification phenomena, the other cell
adhesion protein examined, fibronectin, demonstrated
a very different distribution. As shown in Figure 7c,
the fibronectin was distributed in a network array be-
tween the cells but was not seen at the underlying
substratum surface as was the case with the osteopon-
tin. This dramatic difference in labeling would indicate
that osteopontin was associated with adhesion of these
cells to the underlying substratum, whereas fibronec-
tin was employed in cell/cell adhesion. Comparison, in
our culture system, of the distribution of osteopontin
and fibronectin and the appearance of striae in the os-
teopontin signal from the substratum surface associ-
ated with change in cell shape would substantiate this
view.
The early appearance of osteopontin in our cultures
and our failure to detect osteocalcin at such early pe-
riods would corroborate the reports of Stein et al.
(1990), who observed that osteocalcin gene expression
was not detectable before 12 days in rat foetal calvarial
cultures, whereas osteopontin expression peaked a t
25% maximal levels at 7 days. They raised the possi-
bility that these early increased levels of osteopontin
were attributable to mRNA transcribed in vivo in os-
teoblasts undergoing matrix mineralization before iso-
lation from the fetal calvaria. However, the authors did Fig. 10. Scanning electron photomicrographs of the initial calcified
not examine the ultrastructure of the earliest mineral globular accretions laid down on the culture substratum by differen-
tiating osteogenic cells. A. Portion of a cell to the left (c) can be seen
formed in their cultures and focused, in their ultra- above the culture substratum (s),which is partially covered by glob-
structural studies, on the mineralization of the colla- ular accretions. These accretions comprise individual crystalites,
gen compartment where osteocalcin may play an im- which, in this first micrograph,measure -150 nm. Field width = 2.9
portant role. An alternative explanation that emerges Fm. B. Layer of globular accretions that has coalesced to form a ce-
ment line structure has been deliberately disrupted during specimen
from our experimental findings, where specific atten- preparation to expose the underlying substrate. Field width =
tion has been paid to matrix elaboration events a t the 5.7 bm.
interface, is that osteopontin is secreted early in cul-
ture and allows differentiating osteogenic cells to at-
tach to the culture substratum. Once attached, and as
they change in shape concomitant with differentiation, calcium binding domain on osteopontin would serve to
the work of Hunter et al. (1994b) would strongly im- limit crystal growth at this early phase and thus osteo-
plicate bone sialoprotein as the source of nucleating pontin would be serving a dual function. Osteopontin
sites for calcification to be initiated. It could be that the expression could then be expected to decrease, as de-
 MODELING OF BONE/IMPLANT INTERFACE 439

Fig. 11. Collagen that has been elaborated by the cells following the production of globular accretions
(g) is seen in the majority of this photomicrograph. The individual collagen fibres are encrusted with
mineral that demonstrates that they are mineralizing along their lengths. Field width = 8.6 pm. B. At
the top leR of the field of view, collagen fibres can be seen to be varied in the underlying mineralized
matrix; to the centre and bottom right, uncalcified collagenous fibres are still present and distinctly
separate from the globular accretions below. Field width = 5.7 pm.

scribed by Stein et al. (19901, until after collagen as- dicating the crystalline structure of these deposits (Fig.
sembly in the extracellular compartment had occurred 8). The earliest detectable crystallites in this organic
and collagen mineralization was initiated. In culture matrix are 1-2 nm in size, which represents one or two
systems that do not produce the mineralized interfacial unit cells of calcium apatite. Indeed, using the FETEM,
matrix but initially express collagen, osteopontin ex- it was possible, using a one nanometer probe size, to
pression would be delayed, as reported by Sodek et al. conduct EDX and EELS within the crystallite bound-
(1991). aries defined by the lattice fringe patterns. Back-
The change in cell shape, evidence for which we have ground signals from areas 3 or 4 nm away, but not
now observed in both scanning electron microscopy containing lattice fringes, could then be subtracted
(Davies et al., 1991a) and the osteopontin labeling at confirming the calcium phosphate nature of these
the substratum surface (Shen et al., 1993), is worthy of nanocrystals. These results clearly show that initial
note in the biomaterials context. Since cellular mor- mineralization events in this differentiating osteogenic
phology is maintained by the cytoskeleton and the lat- cell culture are independent of assembled collagen
components in the extracellular matrix. It is our cur-
ter is linked to intranuclear filaments, Fey et al. (1984)
have implicated the cytoskeleton in important cell rent belief that, following secretion of both specific and
functional roles. It is through this fibre network con- nonspecific bone proteins by the differentiating osteo-
tinuum that gene regulation by extracellular agents genic cells in this culture system, that mineralization
has been hypothesized by Pienta et al. (1989) and Ben- of this matrix is initiated by the expression of the mem-
Ze’ev (1991), and thus observations of change in shape brane glycoprotein alkaline phosphatase (McComb et
of bone-derived cells on substrata of different chemical al., 1979), which is known to be linked to the cell
(Davies et al., 1986) or physical (Bowers et al., 1992) membrane by a phosphatidyl-inositol-glycanstructure
properties may have important implications in bone (Low, 1987).
formation on candidate implant materials. As this interfacial matrix matures, the crystallites
grow in size, both in girth and length, which is illus-
Initial Mineralization Events and Crystal Growth trated by comparison of Figures 6, 8, and 9, although
As illustrated in Figure 5 and discussed above, the they should not be confused with those described by
early organic interfacial matrix undergoes calcification Matthews et al. (19801, which appeared as a result of
at a stage represented in Figure 6B. High resolution administration of calcitonin and were an estimated two
phase contrast images clearly show lattice fringes in- orders of magnitude greater in size. The general ap-
440 J.E. DAVIES

pearance of the earliest crystal structures is unordered

(Fig. 8 and 9), although, as the interface matures, we
have often observed a fan-shape crystal distribution
(Fig. 9A, B). The appearance of these crystals is similar
to the “brush border” of protruding crystallites found in
vivo by Bonucci (1990).
The images in Figure 9 correspond to the SEM ap-
pearance seen in Figures 4A and 1OA where individual
calcified globular accretions are apparent on the sub-
stratum surface and in both SEM and TEM, individual
crystallites could be easily identified. It should be noted
that in Figure 9 the apparent needle-shape crystals in
fact show some indication of a platelike morphology.
Although few other reports in the literature focus on
this mineralized extracellular matrix, it is seen in many
bone cell cultures (see Table 2) and has also been illus-
trated in a rat periodontal cell culture system (Cho et
al., 1992).It should be emphasized that these accretions
appear only under the developing nodules and not
ubiquitously on the culture substratum. We have de-
monstrated their appearance on the surface of many
substrata including ceramics, metals, and polymers
(Lowenberg et al., 1991; Davies et al., 1991a,b; Pilliar
et al., 1991, Peel et al., 1992). It is interesting to note in
Figure 10 the discrete appearance of each accretion. We
have previously described these as -1 p.m in diameter
and 0.5 pm in height (Davies et al., 1991a) an obser-
vation confirmed by de Bruijn et al. (1992a1, but it is
clear that considerable variation in size exists. Indeed,
the individual accretions appear to be composed of
smaller crystallite structures, which would correlate
with the appearance of individual crystallites men-
tioned above.
These globular accretions eventually fuse to form a
complete sheet of calcified material beneath the devel-
oping bone nodule (Fig. 10B). Clearly, the globules are
adherent to the culture substratum; otherwise they
would have been lost during preparation for electron
microscopy. Furthermore, the individual crystallites Fig. 12. Transmission electron photomicrographs to demonstrate
two consecutive stages in the mineralization process in this culture
that they comprise would have to be themselves “held system. A. The initial mineralization events at the interface have
together” by some matrix that is not visualized in the occurred to produce an early cementlike line structure above which
SEM preparations, but is explained by the organic ma- the collagen component is unmineralized. B. The later stage where
trix described above. The cementlike interfacial calci- the collagen compartment is now mineralized, and it is clear that
mineralization is progressing along the individual collagen fibres as
fied matrix does not contain morphologically identifi- well as in the matrix between the collagen fibres. Field widths =
able collagen fibres as is illustrated by the comparison 3.2 p,m.
of Figures 10 and 11. Following the elaboration of the
interfacial matrix clear evidence of collagen fibre for-
mation became apparent and is described below.
Collagen Formation adsorbed to the culture substratum. In the extracellu-
At early stages of elaboration of the interface, colla- lar space, hydroxylated procollagen is assembled into
gen was not seen in the extracellular compartment, fibrils by the introduction of cross-links, nine of which
although isolated individual cells did label positively have so far been identified (Eyre et al., 1984). Kuboki
with collagen Type-I as seen in Figure 7d. This image et al. (1992) have demonstrated, using the osteogenic
would indicate that collagen was being produced by the MC3T3-El cell line, that the appearance of bone colla-
cell but was not assembled into fibres in the extracel- gen cross-links, dehydro-dihydroxylysinonorleucine
lular compartment. The production by these cells of and dehydro-hydroxylysinonorleucineand the ratio be-
collagen has been confirmed biochemically by conduct- tween them is time dependent and is concomitant with
ing collagenase digests of the medium from radio-la- the disappearance of their precursors. The time lag was
beled cultures (Peel, 1995). The mineralized interfacial attributed to the gaining of a threshold level, which
layer is succeeded by collagen fibre assembly, which is permitted precursor-product transition. In the context
indicated in Figure 12A, where the collagen compart- of our observations, the temporal development of cross-
ment is still uncalcified. Thus a clear difference exists links could be one of many mechanisms by which col-
between the proteins that are secreted by the cells into lagen, which is expressed by the cells, fails to assemble
the medium and those cell secreted proteins that are in the cement line structure. Clearly, the culture sys-
Fig. 13.Immuno-SEM of gold conjugated antibodies labelling collagen Type-I. A. Secondary electron image. B. Backscattered electron image
of the same field of view. Field widths = 3 pm.

Fig. 14. SEM photomicrograph of the interface between bone and a retrieved slip-cast hydroxyapatite
ceramic rod implanted in rat femur. On the surface of the individual grains of the hydroxyapatites
ceramic (HA) a globular extracellular matrix can be clearly seen (G)that is morphologicallyidentical to
that illustrated in Figure 1.This globular matrix separates the underlying implant from the overlying
collagen (C) and above this the rest of the bone matrix including cells can be clearly seen. Field width =
29 pm (from the work de Bruijn).
442 J.E. DAVIES

exercised when employing gene expression to interpret

matrix formation a t the implant interface. Further-
more, the collagen precedes mineral concept would
render interpretation of the morphologies observed at
implant interfaces in vivo, and described below, impos-
sible.
Collagen fibre assembly in the extracellular compart-
ment becomes evident above the interfacial mineralized
zone (Figs. 4, 11, 12). The latter acts as a n anchorage
surface for the collagen that becomes mineralized by
two methods: fibre mineralization and extracollagenous
matrix mineralization (Fig. llb). The latter has an
appearance that is similar to that of the interfacial
mineralization and may be due t o random seeding of
crystallites in the noncollagenous proteins that are con-
tinually secreted by osteogenic cells. Thus, whereas se-
cretion of collagen continued, envelopment of the col-
lagen fibres by the underlying mineralized matrix was
evident (Fig. llb). The identity of this newly forming
collagen, as Type-I, was confirmed by immuno-SEM
(Shen et al., 1993) as illustrated in Figure 13.
The obvious conclusion with respect to the bone-im-
plant interface is that, when bone tissue is elaborated
on an implant surface in the absence of disruptive ex-
trinsic mechanical stimuli, the ideal interfacial tissue
will be a calcified matrix, which will have the biochem-
Fig. 15. The mirror image of the surface of a dense, tape-cast, hy- ical composition and morphological structure of bone
droxyapatite ceramic (prepared according to Arita et al., 1995a,b) cement matrix found a t interfaces in natural bone tis-
retrieved following implantation in rat femur. The individual grains sue.
(G) of the ceramic surface are clearly seen above; the fractured bone
surface demonstrates the morphological disparity between the colla-
gen containing compartment below (B) and the cement line at the THE IN VlVO REALITY
implant interface (arrow) (from Dziedzic, 1995).
Whereas the ultrastructures of the bonehmplant in-
terface have been reviewed (Listgarten, 1995), a brief
consideration of the in vivo implant interface structure
tem described could be employed t o investigate such is necessary in light of the structural descriptions pro-
matrix elaboration mechanisms. vided herein. If de novo bone forms on a n implant sur-
face in the manner described and as this biological cas-
Mineralization of the Collagen Compartment cade has been shown to occur on many substrata in
The concept of osteoid formation preceding mineral- vitro, then such an interfacial matrix should be found
ization is stressed in both popular and scientific texts with endosseous implants of all material compositions.
and is soundly based on histological observation, which However, since the pioneering work of Hench (re-
provides overwhelming evidence for this sequence of viewed in Hench and Wilson, 1984), it is generally ac-
events in almost all bone formation sites. Unfortu- cepted that there are two classes of endosseous im-
nately, this “collagen precedes mineral” dogma ignores plants: bone-bonding and nonbonding. Metals such as
some of the most exquisite histological descriptions of titanium are nonbonding; calcium phosphate materials
normal bone tissue that emerged in Europe in the mid- are considered bone-bonding.
nineteenth century and, in particular, how the cement The mechanism for the bone-bonding phenomenon is
line interface is established between old bone and new generally thought to be a chemical interaction that re-
bone a t remodelling sites (von Ebner, 1875). This con- sults in collagen, from the bony compartment, inter-
cept (collagen precedes mineral) also has been adopted digitating with the chemically active surface of the im-
in the more modern specialities of both cell and molec- plant. Clearly, this mechanism is inconceivable if the
ular biology as well as that of biomineralization. Thus first extracellular matrix elaborated by bone cells at
in a review of osteoblast differentiation, Stein et al. the implant surface is a collagen-free cement line. Thus
(1990) defined a developmental sequence, based on if cement line structures are indeed found on both non-
modifications in gene expression of foetal rat calvarial bonding and bonding biomaterials, then a re-evalua-
cells, as having three principle periods, proliferation, tion of the phenomenon of bone-bonding is essential. In
extracellular matrix maturation, including collagen fact, a cement line matrix can be found in vivo on the
production, and then mineralization. Similarly, Christ- surfaces of both bonding and nonbonding materials
offersen and Landis (1991) have described the presence (Davies et al., 1991c, d; Pilliar et al., 1991; Orr et
of morphologically distinct collagen fibres as an essen- al., 1993; de Bruijn, 1993; de Bruijn et al., 1995;
tial prerequisite to mineralization. However, as indi- Muller-Mai et al., 1995). In the case of bioactive mate-
cated by our experiments, the expression of the colla- rials, one of the most striking confirmations of the se-
gen gene is not necessarily equated with extracellular quence of events explained by our in vitro work is that
assembly of collagen fibres, and thus caution should be of de Bruijn (1993) and is illustrated in Figure 14.
 MODELING OF BONElIMPLANT INTERFACE 443
Here, the interfacial globular accretions and the over- like micromotion, could influence these earliest events
lying collagen are easily seen overlying the dense hy- will be understood only when a more detailed mecha-
droxyapatite implant. Indeed, as previously demon- nistic explanation of the “normal” biology is achieved.
strated, the cement line matrix can form an exquisite This is a challenge for bone cell biology community,
mirror image of the implant surface in both nonbond- since many of these fundamental issues are as yet in-
ing (Pilliar et al., 1991) and bonding materials (Orr et completely explored. Nevertheless, cell-substratum as-
al., 1993; de Bruijn et al., 1995). As expected from Fig- says created to model bone cell-biomaterial interac-
ure 2, the in vivo interface has a variable ultrastruc- tions are shedding a new light on matrix elaboration by
ture, as most recently described by Muller-Mai et al. differentiating osteogenic cells and may facilitate our
(1995). The degree to which the cement line matrix can understanding of these complex biological processes.
be visualized on bone-bonding materials will be a func- Only when these are understood will it be possible t o
tion of their chemical surface reactivity, as discussed engineer materials to produce specific biological re-
by Davies and Baldan (1995). In each case bonding of sponses. Indeed, we predict that unravelling such in-
de novo bone will occur by the fusion of the biological teractions will herald a new era of designer biomateri-
cement line matrix with the surface reactive layer of als, for dental and orthopaedic applications that are
the substratum. In other areas, where connective tis- engineered to produce specific biological responses.
sue collagen is in contact with the implant, it will be-
come encased in the substratum surface reaction layer ACKNOWLEDGMENTS
to produce the ultrastructural appearance of collagen This work contains contributions from several ongo-
interdigitation. In the case of dense, relatively nonre- ing projects within our group. I particularly thank my
active, calcium phosphates (Dziedzic et al., 19951, the colleagues Beate Lowenberg, Xue-Ying Shen, Amy
cement line matrix will juxtapose the implant surface Shiga, Bob Chernecky, and Ed Roberts for the cell cul-
without bonding, as illustrated in Figure 14. ture, immunobiology and electron microscopy; and my
CONCLUSIONS students Dilcele Dziedzic, Zhou Hong, and Sean Peel. I
also thank Joost de Bruijn (University of Leiden, The
The experimental approach described can be em- Netherlands), whose work as a visiting student to our
ployed to deconvolute the detailed biological mecha- group is represented in Figure 14. Access to the FE-
nisms that take place a t the bone-implant interface. A TEM was kindly provided by Nissei Sangyo Canada
clear sequence of events representing the initial stages and Hitachi Corporation in Japan. Financial support of
of extracellular formation by differentiating osteogenic the Medical Research Council, Canada, the Ontario
cells on the surface of nonbiological substrata has been Centre for Materials Research, and the Ontario Gov-
described. Thus this in vitro approach provides a pow- ernment through an OntariolRhone Alpes contract is
erful tool to elucidate the detailed biological events gratefully acknowledged.
that take place a t such an interface and has permitted
the compositional and structural characteristics of the LITERATURE CITED
interface to be defined. These methods can be used not
only to unravel interfacial mechanisms but also to in- Albrektsson, T. 1992 Current issues forum. Int. J. Oral & Maxillofac.
Implants, 7:416-417.
vestigate how these could be influenced by the pres- Alper, M., ed. 1992 Biology and materials synthesis. Part 1 MRS
ence of differing nonbiological substrata. The real BulletidOctober, XVZI:24-59.
strengths of the technique described are twofold: (1) it Arita, LH., V.M. Castano, and D.S. Wilkinson 1995a Synthesis and
clearly allows a mechanistic explanation of bone for- processing of hydroxyapatite ceramic tapes with controlledporos-
ity. J Materials Sci. (in press).
mation at the surface of candidate implant materials to Arita, I.H., D.S. Wilkinson, M.A. Mondragon, and V.M. Castano
be established, and (2) it provides a means of unravel- 1995b Chemistry and sintering behaviour of thin hydroxyapatite
ling the details of the mechanisms involved. ceramics with controlled porosity. Biomaterials (in press).
However, this biological sequence of events is by no Baron, R., A. Vignery, and P. “ran Van 1980 The significance of
lacunar erosion without osteoclasts:Studies of the reversal phase
means simple. Understanding the cellular and molec- of the remodelling sequence. In: Bone Histomorphometry. Metab.
ular mechanisms involved depends to a large extent on Bone Dis. Rel. Res. W.S.S. Jee, A.M. Parfitt, eds. Vol. 2S, pp.
the current state of knowledge in the bone cell biology 35-40.
field. Unfortunately, however, the current relative lack Ben-Ze’ev, A. 1991 Animal cell shape changes and gene expression.
BioEssays, 13:207-212.
of understanding of early differentiation events in os- Bhargava, U., M. Bar-Lev, C.G. Bellows, and J.E. Aubin 1988 Ultra-
teogenic cells, early matrix secretion, composition, and structural analysis of bone nodules formed in vitro by isolated
structure, and the mechanisms by which calcium phos- fetal rat calavaria cells. Bone, 9~155-163.
phate crystallites are seeded into proteoglycan-con- Bonucci, E. 1990 The ultrastructure of the osteocyte. In: Ultrastruc-
taining extracellular matrices, mitigate against an un- ture of Skeletal Tissues: Bone and Cartilage in Health and Dis-
ease. E. Bonucci and P.M. Motta, eds. Kluwer. Boston, pp. 223-
derstanding of the bone-biomaterial interface. In spite 237.
of the considerable amount of research that is still Boskey, A.L., D. Stiner, S.B. Doty, I. Binderman, and P. Leboy 1992
needed to solve the problems outlined above, it is clear Studies of mineralizationin tissue culture: Optimal conditionsfor
cartilage calcification. Bone Miner., 16:ll-36.
that we are already in a position to address some of the Bouvier, M., J.M. Martin, P. Exbrayat, E. Rigollet, T. Le Mogne, D.
critical questions listed at the start of this review. Treheux, and H. Magloire 1994 Ultrastructuralstudy of calvaria-
Thus we are now beginning to understand how cells released osteoblasts cultured in contact with titanium-based sub-
make bone on foreign surfaces. These earliest events strates. Cells Materials, 4~135-142.
are orchestrated by the differentiation of the osteogenic Bowers, K.T., J.C. Keller, B.A. Randolph, D.G. Wick, and C.M.
Michaels 1992 Optimization of surface micromorphology for en-
cells and occur before the cells express a mature osteo- hanced osteoblast responses in vitro. Int. J. Oral & Maxillofac.
blastic phenotype. The manner in which either sub- Implants, 7~302-310.
stratum physical chemistry, or other extrinsic factors Brunski, J.B. 1991 Influence of biomechanical factors at the bone-
444 J.E. DAVIES
biomaterial interface. In: The Bone-Biomaterial Interface. J.E.
Davies, ed. University of Toronto Press, Toronto, pp. 391-405.
Butler, W.T. 1989 The nature and significance of osteopontin. Conn.
Tissue Res., 23:123-136.
Caplan, A.I. 1991 Cell-mediate bone regeneration. In: The Bone-Bio-
material Interface. J.E. Davies, ed. University of Toronto Press,
Toronto, pp. 199-204.
Cho, M.-I., N. Matsuda, W.-L. Lin, A. Moshier, and P.R. Ramakrish-
nan 1992 In vitro formation of mineralized nodules by periodon-
tal ligament cells from the rat. Calcif. Tissue Int., 50:459-467.
Christoffersen, J., and W.J. Landis 1991A contribution with review to
the description of mineralization of bone and other calcified tis-
sues in vivo. Anat. Rec., 230~435-450.
Davies, J.E. 1990 The use of cell and tissue culture to investigate bone
cell reactions to bioactive materials. In: CRC Handbook of Bio-
active Ceramics. T. Yamamuro, L.L. Hench, J. Wilson, eds. CRC
Press, Boca Raton, pp. 195-225.
Davies, J.E., and N. Baldan 1995 Scanning electron microscopy of the
bone/bioactive implant interface. J . Biomed. Mat. Res. (in press).
Davies, J.E., B. Causton, Y. Bovell, K. Davy, and C.S. Sturt 1986 The
migration of osteoblasts over substrata of discrete surface charge.
Biomaterials, 7:23 1-233.
Davies, J.E., R. Chernecky, B. Lowenberg, and A. Shiga 1991a Dep-
osition and resorption of calcified matrix in vitro by rat bone
marrow cells. Cells Materials, 1:3-15.
Davies, J.E., P. Ottensmeyer, X. Shen, M. Hashimoto, and S.A.F. Peel
1991b Early extracellular matrix synthesis. In: The Bone-Bio-
material Interface. J.E. Davies, ed. University of Toronto Press,
Toronto, pp. 214-228.
Davies, J.E., N. Nagai, N. Takeshita, and D.C. Smith 1991c Deposi-
tion of Cement-like Matrix on Implant Materials. In: The Bone
Biomaterial Interface. J.E. Davies, ed. University of Toronto
Press, Toronto, pp. 285-294.
Davies, J.E., R.M. Pilliar, D.C. Smith, and R. Chernecky 1991d Bone
interfaces with retrieved alumina and hydroxyapatite ceramics.
In: Bioceramics, Vol. 4. W. Bonfield, G.W. Hastings, K.E. Tanner,
eds. Butterworth-Heinemann, London, pp. 199-204.
Davies, J.E., D.D. Perovic, and X.Y. Shen 1993 Characterization of
early bone mineral crystallites found at the substratum interface
in vitro. 19th annual meeting of the Society for Biomaterials,
April 28-May 2. Birmingham, AL, p. 247.
de Bernard, B. 1982 Glycoproteins in the local mechanism of calcifi-
cation. Clin. Orth. Rel. Res., 162:233-244.
de Bruijn, J.D. 1993 Calcium phosphate biomaterials: Bone bonding
and biodegredation properties, Ph.D. thesis, University of Leiden,
The Netherlands.
de Bruijn, J.D., J.E. Davies, J.S. Flach, K. de Groot, and C.A. van
Blitterswijik 199213 Ultrastructure of the mineralized tissuekal-
cium phosphate interface in vitro. In: Tissue Inducing Biomate-
rials. L. Cima, et al., eds. Mat. Res. Symp. Proc. 252:63-70.
de Bruijn, J.D., J.E. Davies, C.P.A.T. Klein, K. de Groot, and C.A. van
Blitterswijk 1992c Biological responses to calcium phosphate ce-
ramics. In: Bone-Bonding Biomaterials. P. Ducheyne, T. Kokubo,
C.A. van Blitterswijk, eds. Reed Health Care Communications,
Leiderdorp, The Netherlands, pp. 57-72.
de Bruijn, J.D., J.E. Davies, and C.A. van Blitterswijk 1995 Initial
bone matrix formation at the hydroxyapatite interface in vivo. J .
Biomed. Mat. Res., 2989-100.
de Bruijn, J.D., C.P.A.T. Klein, K. de Groot, and C.A. van Blitterswijk
1992a The ultrastructure of the bone-hydroxyapatite interface in
vitro. J . Biomed. Mat. Res., 26:1365-1368.
de Harven, E., R. Leung, and H. Christensen 1984 A novel approach
for scanning electron microscopy of colloidal gold-labeled cell sur-
faces. J . Cell Biol., 99:53-57.
Dziedzic, D.M. 1995 Effects of implant surface topography on osteo-
conduction, Masters thesis, University of Toronto.
Dziedzic, D.M., I.H. Arita, R. Chernecky, D. Wilkinson, and J.E. Dav-
ies 1995 Effects of Apatite ceramic topography on bone healing.
Proc. Canadian Biomaterials Society Annual Meeting, May 15-
16, 1995, Ottawa, ON.
Eyre, D.R., M.A. Paz, and P.M. Gallop 1984 Cross-linking in collagen
and elastin. Ann. Rev. Biochem., 53:713-748.
Fey, E.G., K.M. Wan, and S. Penman 1984 Epithelial cytoskeletal
framework and nuclear matrix-intermediate filament scaffold:
Three-dimensional organization and protein composition. J. Cell
Biol., 98:1973-1984.
Frasca, P. 1981 Scanning electron microscopy studies of ground sub-
stance in the cement lines, resting lines, hypercalcified rings and
reversal lines of human cortical bone. Acta Anat., 109:115-121.
Frasca, P., R.A. Harper, and J.L. Katz 1981 Scanning electron mi-
croscopy studies of collagen, mineral and “ground substance” in
human cortical bone. Scan. Electn. Microsc., 3:339-346.
Gerstenfeld, L.C., S.D. Chipman, C.M. Kelly, K.J. Hodgens, D.D. Lee,
and W.J. Landis 1988 Collagen expression, ultrastructural as-
sembly and mineralization in cultures of chicken embryo osteo-
blasts. J . Cell Biol., 106:979-989.
Gerstenfeld, L.C., A. Riva, K. Hodgens, D.R. Eyre, and W.J. Landis
1993 Post-Translational Control of collagen fibillogenesis in min-
eralizing cultures of chick osteoblasts. J . Bone Min. Res., 8:1031-
1043.
Gerstenfeld, L.C., M. Feng, Y. Gotoh, and M.J. Glimcher 1994 Selec-
tive extractability of noncollagenous proteins from chicken bone.
Calcif. Tissue Int., 55r230-235.
Gruber, H.E., G.J. Marshall, M.E. Kirchen, J. Kang, and S.G. Massry
1985 Improvements in dehydration and cement line staining for
methacrylate embedded human bone biopsies. Stain Technology,
60:337-344.
Hench, L.L., and J. Wilson 1984 Surface-active biomaterials. Science,
226:226.
Hunter, G.K., C.L. Kyle, and H.A. Goldberg 1994a Modulation of
crystal formation by bone phosphoproteins: structural specificity
of the osteopontin-mediated inhibition of hydroxyapatite forma-
tion. Biochem. J., 300:723-728.
Hunter, G.K., and H.A. Goldberg 1994b Modulation of crystal forma-
tion by bone phosphoproteins: Role of glutamic acid-rich se-
quences in the nucleation of hydroxyapatite by bone sialoprotein.
Biochem. J., 304:175-179.
Jingushi, S., and M.E. Bolander 1991 Biological cascades of fracture
healing as models for bone-biomaterial interfacial reactions. In:
The Bone-Biomaterial Interface. J.E. Davies, ed. University of
Toronto Press. Toronto. DD. 250-262.
Jones, S., and A. Boyde 1977The migration of osteoblasts. Cell Tissue
Res.. 184t179-192.
Kuboki, Y., A. Kudo, M. Mizuno, and M. Kawamura 1992 Time-de-
pendent changes of collagen cross-links and their precursors in
the culture of osteogenic cells. Calcif. Tissue Int., 50:473-480.
Lee, S.L., A. Veis, and T. Glonek 1977 Dentin phospho-protein: a n
extracellular calcium-binding protein. Biochemistry, 16:2971-
2979.
LeGeros, R.Z., I. Orly, M. Gregoire, and G. Daculsi 1991 Substrate
surface dissolution and interfacial biological mineralization. In:
The Bone Biomaterial Interface. J.E. Davies, ed. University of
Toronto Press, Toronto, pp. 76-88.
Listgarten, M.A. 1995 The soft and hard tissue response to endosseous
dental implants. Anat. Rec. (this issue).
Low, M.G. 1987 Biochemistry of the glycosyl-phosphatidlinositol
membrane protein anchors. Biochem. J., 244:l-13.
Lowenberg, B., R. Chernecky, A. Shiga, and J.E. Davies 1991 Miner-
alized matrix production by osteoblasts on solid titanium in vitro.
Cells Materials, 1:177-187.
Maniatopoulos, C., J. Sodek, and A.H. Melcher 1988 Bone formation
in vitro by stromal cells obtained from bone marrow of young
adult rats. Cell Tissue Res., 254, 317-330.
Mann, S. 1988 Molecular recognition in biomineralization. Nature,
332:119-124.
Matthews, J.L., R.V. Talmage, and R. Doppelt 1980 Responses of the
osteocyte lining cell complex, the bone cell unit, to calcitonin.
Metab. Bone Dis. & Rel. Res., 2:113-122.
McComb, R.B., G.N. Bowers, and S. Posen 1979 Alkaline Phos-
phatase. Plenum, New York.
McKee, M.D., and A. Nanci 1993 Ultrastructural, Cytochemical and
immunocytochemical studies on bone and its interfaces. Cells
Materials, 3:219-243.
McKee, M.D., A. Nanci, W.J. Landis, L.C. Gerstenfeld, Y. Gotoh, and
M.J. Glimcher 1989 Ultrastructural immunolocalization of a ma-
jor phosphoprotein in embryonic chick bone. Conn. Tiss. Res.,
21:21-29.
McKee, M.D., A. Nanci, W.J. Landis, Y. Gotoh, L.C. Gerstenfeld, and
M.J. Glimcher 1990 Developmental appearance and ultrastruc-
tural immunolocalization of a major 66 kDa phosphoprotein in
embryonic and post-natal chicken bone. Anat. Rec., 228:77-92.
McKee, M.D., M.J. Glimcher, and A. Nanci 1992 High resolution im-
munlocalization of osteopontin and osteocacin in bone and carti-
lage during endochondral ossification in the chicken tibia. Anat.
Rec., 234:479-492.
McKee, M.D., M.C. Farach-Carson, W.T. Butler, P.V. Hauschka, and
A. Nanci 1993 Ultrastructural immunolocalization of noncollag-
enous (osteopontin and osteocalcin) and plasma (albumin and
a,HS-glycoprotein) proteins in rat bone. J. Bone Min. Res.,
8.485-496.
Melcher, A.H., T. Cheong, J. Cox, E. Nemeth, and A. Shiga 1986
 MODELING OF BONE/IMPLANT INTERFACE
Synthesis of cementum-like tissue in vitro by cells cultured from
bone: A light and electron microscope study. J. Perio. Res., 21:
592-612.
Muller-Mai, C.M., S.I. Stupp, C. Voigt, and U. Gross 1995 Nanoapa-
tite and organoapatite implants in bone: Histology and ultra-
structure of the interface. J. Biomed. Mat. Res., 29:9-18.
Nagata, T., R. Todescan, H.A. Goldberg, Q. Zhang, and J. Sodek 1989
Sulfation of secreted phosphoprotein I (SPP1, Osteopontin) is as-
sociated with mineralized tissue formation. Biochem. Biophys.
Res. Commun., 165.234-240.
Nanci, A,, G.F. McCarthy, S. Zalzal, C.M.L. Clokies, H. Warshawsky,
and M.D. McKee 1994 Tissue response to titanium implants in
the rat tibia: Ultrastructural, immunocytochemcialand lectin-cy-
tochemical characterization of the bone-titanium interface. Cells
Materials, 4:l-30.
Oldberg, A., A. Franzen, and D. Heinegard 1986 Cloning and se-
quence analysis of rat bone sialoprotein (osteopontin) cDNA re-
veals an Arg-Gly-Asp cell-binding sequence. Proc. Natl. Acad.
Sci. USA, 83:8819-8823.
Orr, R.D., J.D. de Bruijn, and J.E. Davies 1993 Scanning electron
microscopy of the bone interface with titanium, titanium alloy
and hydroxyapatite. Cells Materials, 2:241-251.
Parfitt, A.M. 1983 The physiologic and clinical significance of bone
histomorphometric data. In: Bone histomorphometry: Techniques
and interpretation. R.R. Recker, ed. CRC Press, Boca Raton, pp.
143-223.
Peel, S.A.F., R.N. Sodhi, T.M. Duc, and J.E. Davies 1992 Polymer
Surface modification causes change in phenotypic expression of
primary bone cells. Mat. Res. SOC.Symp. Proc., 252:71-77.
Peel, S.A.F. 1995 The influence of substratum modification on inter-
facial bone formation in vitro, Ph.D. thesis, university of Toronto.
Philipson, B. 1965 Composition of cement lines in bone. J . Histochem.
Cytochem., 13.270-281.
Pienta, K.J., A.W. Partin, and D.S. Coffey 1989 Cancer as a disease of
DNA organization and dynamic cell structure. Cancer Res., 49:
2525-2532.
Pilliar, R.M., J.E. Davies, and D.C. Smith 1991 The bone-biomaterial
interface for load-bearing implants. MRS BulletidSept 1991,
XVZ:55-61.
Pitaru, S., S. Kotev-Emeth, D. Noff, S. Kaffuler, and N. Savion 1993
Effect of basic fibroblast growth factor on the growth and differ-
entiation of adult stromal bone marrow cells: Enhanced develop-
ment of mineralized bone-like tissue in culture. J . Bone Mineral
Res., 8.919-924.
Pockwinse, S.M., L.G. Wilming, D.M. Conlon, G.S. Stein, and J.B.
Lian 1992 Expression of cell growth and bone specific genes at
single cell resolution during development of bone tissue-like or-
ganization in primary osteoblast cultures. J . Cell. Biochem., 49:
310-323.
Pockwinse, S.M., J.B. Lawrence, R.H. Singer, J.L. Stein, J.B. Lian,
and G.S. Stein 1993 Gene expression a t single cell resolution
associated with development of the bone cell phenotype: Ultra-
structural and in situ hybridization analysis. Bone, 14:347-352.
Pritchard, J.J. 1972 General histology of bone. In: The Biochemistry
and Physiology of Bone, 2nd ed. G.H. Bourne, ed. Academic Press,
New York, pp. 1-20.
Roberts, W.E. 1988 Bone tissue interface. J . Dent Ed., 52:804-809.
Sautier, J.M., J.R. Nefussi, and N. Forest 1991 Ultrastructural study
445
of bone formation on synthetic hydroxyapatite in osteoblast cul-
tures. Cells Materials, 1:209-217.
Sautier, J.M., J.R. Nefussi, and N. Forest 1992 Mineralization and
bone formation on microcarrier beads with isolated rat calavaria
cell population. Calcif Tissue Int., 50.527-532.
Shen, X., E. Roberts, S.A.F. Peel, and J.E. Davies 1993 Organic ex-
tracellular matrix components at the bone cell/substratum inter-
face. Cells Materials, 3.257-272.
Sodek, J., Q. Zhang, H.A. Goldberg, C. Domenicucci, S. Kasugai, J.L.
Wrana, H. Shapiro, and J . Chen 1991 Non-collagenous bone pro-
teins and their role in substrate-induced bioactivity. In: The
Bone-Biomaterial Interface. J.E. Davies, ed., University of Tor-
onto Press, Toronto, pp. 97-110.
Sokoloff, L. 1973 A note on the histology of cement lines. In: Perspec-
tives in Biomedical Engineering. R.M. Kenedi, ed. University
Park Press, Baltimore, pp. 135-138.
Somerman, M.J., M. Perez-Mera, R.M. Merkhofer, and R.A. Foster
1987 In vitro evaluation of extracts of mineralized tissues for
their application in attachment of fibrous tissue. J . Periodontol.,
58:349-351.
Somerman, M.J., C.W. Prince, J.J. Sauk, R.A. Foster, and W.T. Butler
1987 Mechanism of fibroblast attachment to bone extracellular
matrix: Role of a 44-Kilodalton bone phosphoprotein. J. Bone
Min. Res., 2:259-265.
Stein, G.S., J.B. Lian, and T.A. Owen 1990 Relationship of cell growth
to the regulation of tissue-specific gene expression during osteo-
blast differentiation. FASEB, 4:3111-3123.
Tarasevich, B.J., P.C. Reike, and G.L. McVay 1991 Ceramic synthesis
using biological processes. In: The Bone-Biomaterial Interface.
J.E. Davies, ed. University of Toronto Press, Toronto, pp. 139-
146.
Tenenbaum, H.C., and J.N.M. Heersche 1981 Differentiation of osteo-
blasts and formation of mineralized bone in vitro: Requirement of
organic phosphate. In: The Chemistry and Biology of Mineralized
Connective Tissues. A. Veis, ed. Elsevier, Amsterdam, pp. 199-
201.
Tenenbaum, H.C. C.A.G. McCulloch, C. Fair, and C. Birek 1989 The
regulatory effect of phosphates on bone metabolism in vitro. Cell
Tissue Res., 257:555-563.
Tran Van, P., A. Vignery, and R. Baron 1982 An electron microscopic
study of the bone remodelling sequence in the rat. Cell Tissue
Res., 225:283-292.
Villanueva, A.R., C. Sypitkowski, and A.M. Parfitt 1986 A new
method for identification of cement lines in undecalcified, plastic
embedded sections of bone. Stain Technology, 61:83-88.
von Ebner, V. 1875 Ueber den feineren Bau der Knochensubstanz.
S.-B. Akad. Wiss. Wien, math.-nat. Kl., Abt. 111, 72:49-183.
Weidenreich, F. 1930 Das Knochengewebe. In: Mollendorff' Hand-
buch der mikroskopischen Anatomie des Menschen. Verlag
Julius Springer, Berlin, I1 2:391-520.
Weiner, S. 1979 Aspartic acid-rich proteins: Major components of the
soluble organic matrix of mollusk shells. Calif. Tissue Int., 29:
163-167.
Weinmann, J.P., and H. Sicher 1955 Bone and Bones: Fundamentals
of Bone Biology. C.V. Mosby, St. Louis, pp. 18-46.
Zhou, H., R. Chernecky, and J.E. Davies 1994 Deposition of cement at
reversal lines in rat femoral bone. J . Bone & Mineral Res., 9:367-
374.