ARTICLE IN PRESS

0142-9612/$ - se

doi:10.1016/j.bi

�CorrespondE-mail addr

Biomaterials 28 (2007) 2281–2293

www.elsevier.com/locate/biomaterials

Comparative response of epithelial cells and osteoblasts tomicrofabricated tapered pit topographies in vitro and in vivo

Douglas W. Hamilton, Babak Chehroudi, Donald M. Brunette�

Department of Oral, Biological and Medical Sciences, University of British Columbia, 2199 Wesbrook Mall, Vancouver, BC, Canada V6T 1Z3

Received 31 October 2006; accepted 23 January 2007

Available online 2 February 2007

Abstract

Microfabricated tapered pits in vivo can stimulate connective tissue and bone attachment to percutaneous devices, secondarily

preventing epithelial migration, and promoting long-term implant survival. Epithelial cells, which form a seal with a dental implant,

acting as a barrier, and osteoblasts, which form bone, can come into contact with the same implant topography. To investigate whether

the phenotypic characteristics of each cell type influenced cell response to micro-topography, we compared the response of the two cell

types to the same dimensions of tapered pits, in vitro, and in vivo. Increased spreading, mature FAs, and restricted migration

characterized individual PLE cell response to tapered pits. In contrast, osteoblasts were highly migratory, formed smaller, punctate

adhesions and mineralized. Epithelial sheets formed from high-density PLE cultures demonstrated that tapered pits did not inhibit

migration of the PLE sheets in vitro, similar to in vivo observations. In vitro, PLE sheet migration correlated with increases in vinculin,

tyrosine phosphorylation, cytokeratin and ERK 1/2 phosphorylation. The findings of this study show that tapered pits stimulate

osteoblast mineral deposition in vitro and in vivo, but do not prevent epithelial sheet migration. In vitro results suggest that epithelial sheet

migration could involve altered FA mediated signal transduction.

r 2007 Elsevier Ltd. All rights reserved.

Keywords: Substratum topography; Dental implant; Osteoblasts; Epithelium; Osseointegration; Epithelial downgrowth

1. Introduction

Upon placement in the jaw, a dental implant mustsuccessfully interface with at least three distinct cellpopulations, osteoblasts (bone), fibroblasts (connectivetissue), and epithelium [1,2], at different regions along thelong axis of the implant. Each cell type is specialized toperform a specific physiological function. Osteoblastssecrete matrix and mineral to form bone, and fibroblastsproduce collagen rich connective tissue, with both tissueshelping to anchor the implant. Epithelial cells attach at thedorsal interface of the tissue and implant, ideally forming atight seal between the tissue and the implant surface.Inappropriate, as well as inadequate, adhesion of thesethree cell populations has been implicated in poor implantsurvival [2–5]. Failure of bone and connective tissue toform a tight seal to the implant can result in epithelial

e front matter r 2007 Elsevier Ltd. All rights reserved.

omaterials.2007.01.026

ing author. Tel.: +604 822 2994; fax: +604 822 3562.

ess: brunette@interchange.ubc.ca (D.M. Brunette).

downgrowth, which can lead to deep pocket formation aswell as possible avulsion of the implant, whereas aninadequate epithelial seal can lead to infection [3,6–8].Therefore, identifying the surface features that enhancebone and connective tissue attachment to the implant,whilst, discouraging epithelial downward migration wouldbe advantageous for optimizing implant longevity [2,3,7,9].Advances in implant design have often focused on

altering the micro-topography and chemistry of theimplant surface [2,10–12]. In particular, implant surfacetopography has been identified as a potent modulator ofbone and connective tissue attachment in vivo, as well asshowing success in inhibiting epithelial downgrowth [4,6,7].Chehroudi et al. [4], identified that both grooved substrataand tapered pits (TPs) could promote connective tissueattachment in vivo, which in turn reduced the level ofepithelial downgrowth [4]. However, not all groove depthsformed an absolute barrier to epithelial downward migra-tion, and inhibition of this downgrowth for times greaterthan 2 weeks seemed related more to the level of

ARTICLE IN PRESSD.W. Hamilton et al. / Biomaterials 28 (2007) 2281–22932282

attachment of bone and connective tissue to the implantsurface, than a direct effect of the topography on theepithelial cells themselves [4,13].

That different cell types demonstrate varying degrees ofadhesion to different types of topographies opens thepossibility of optimizing adhesion of defined cell popula-tions on specific regions of the implant surface bymodulating the surface topography along the length ofthe implant. However, in vitro investigations of cellbehaviour have shown that it is not a simple task, becausetopographies have diverse effects on cell behaviour.For example, microfabricated discontinuous edged topo-graphies significantly enhance osteoblast mineralization in

vitro [14], but also promote the attachment, spreading andmigration of both fibroblast and epithelial cells [15]. Incontrast, rough surfaces produced through processes suchas acid etching support excellent fibroblast and osteoblastattachment [16,17], but discourage epithelial attachmentand spreading [18].

As osteoblasts and epithelial cells have markedlydifferent cell migratory behaviours and both interface withdental implants [2], we investigated whether the behaviourof the two cell types in vitro and in vivo on TPs would be

Fig. 1. Diagrammatical representation of the substratum employed in this

study, which comprised of parallel arrays of microfabricated, (A) inverted

pyramids and (B, C) tapered pits.

Table 1

Dimensions of the inverted pyramid and tapered pit topographies used in the

Surface Depth (mm) Pitch (mm) T

30IP185S 30 185 4

30TP50S 30 50 4

60TP105S 60 105 8

120TP185S 120 185 17

120TP280S 120 280 27

related to the innate phenotypic properties that the cellsexhibit in tissues in vivo. TPs and inverted pyramids (IPs)within the range of 30–120 mm in depth were used due totheir potential to promote tissue ingrowth, which isessential for implant integration [6,19]. For this study, theresponse of rat calvarial osteoblasts (RCOs) and porcineperiodontal ligament epithelial (PLE) cells to TP providedthe in vitro model, and a one or two stage rat percutaneousmodel described by Chehroudi et al. used for the in vivo

model [4]. The results from this study demonstrate that TPtopographies are osteoconductive in vitro and in vivo, buttheir influence on epithelial migration is limited.

2. Materials and methods

2.1. Surface fabrication

The TPs and IPs were etched on silicon substrates, using microfabrica-

tion techniques originally developed by Camporese et al. [20]. The

substrata were produced in the laboratories of the Center for Microelec-

tronics, Department of Electrical Engineering, at University of British

Columbia. The depth and spacing of the TPs and IPs are controlled by the

time of etching and the crystalline orientation of the silicon wafer. The TPs

and IP employed in the study are depicted diagrammatically in Fig. 1A–C,

and the dimensions of the TPs and IP are shown in Table 1. The depth of

surface ranged from 30 to 120mm to observe a wide range of depth effect.

The TP structures had a flat surface at the bottom of the pit for cell

spreading and adhesion. The ridge was constant at 10mm amongst all TP

surfaces. The 30 mm IP contained a wide ridge and was designed

specifically to ensure the same number of pits were present on all the

surfaces, irrespective of pit size.

2.1.1. Fabrication and preparation of in vitro substratum replicas

For in vitro experiments, epoxy replicas were fabricated as previously

described [15]. Briefly, epoxy replicas were fabricated from negative

impressions of the silicon wafers, baked at 60 1C for 4 days and sputter-

coated (Randex 3140 Sputtering System, Palo Alto, CA) with 50 nm of Ti.

Prior to use, the replicas were cleaned by ultrasonication in a detergent

[7X (ICN Biomedicals, Inc., Costa Mesa, CA], Argon-gas glow discharge

treatment [21], performed prior to seeding with PLE or RCOs, completed

cleaning and sterilization procedures.

2.1.2. Fabrication and preparation of in vivo implants

Fabrication, and preparation of the implants were performed as

previously described by Chehroudi et al. [4]. In brief, implants were made

from two components; a percutaneous and a subcutaneous component

that had microfabricated and control smooth surfaces. The titanium

coated implants were cleaned by ultrasonication for 20min in 7X

detergent, followed by 20 washes in sterile deionized water. The implants

were dried overnight in a laminar flow hood and argon-gas glow discharge

treatment performed prior to implantation.

study

op of pit (mm) Bottom of pit (mm) Ridge (mm)

5 N/A 140

5 10 5

5 10 10

5 10 10

0 110 10

ARTICLE IN PRESSD.W. Hamilton et al. / Biomaterials 28 (2007) 2281–2293 2283

2.2. In vitro cell isolation and culture

Epithelial cells from porcine periodontal ligament, and osteoblast cells

from newborn rat calvaria were isolated and cultured as previously

described [22,23]. The two cell types were cultured on tissue-culture plastic

in alpha minimal essential medium (Stem cell Technology, Vancouver, BC,

Canada) supplemented with antibiotics (penicillin G, 100mg/ml (Sigma, St.

Louis, MO); gentamicin, 50mg/ml (Sigma); Fungizone, 3mg/ml (Gibco,

Grand Island, NY)) and 15% fetal bovine serum (Flow, McLean, VA) at

37 1C in a humidified atmosphere of 95% air 5% CO2. Both cell types were

trypsinized (0.25% trypsin, 0.1% glucose, citrate-saline buffer pH 7.8)

counted electronically (Cell Counter, Coulter Electronics Limited, Luton,

England) prior to seeding on topographical cues. For low-density cultures,

cells were seeded onto the substrata at 20,000 cells per cm2, and for high-

density cultures, 100,000 cells per cm2. During experiments, osteoblast

media was supplemented with 31.5mg/ml Na-b-glycerol phosphate,

0.58mg/ml L-ascorbic acid phosphate magnesium salt n-hydrate, and

5mg/ml of Alizarin Complexone (Sigma), with media changes performed

twice a week in experiments over 3 days in length to ensure sufficient

nutrient supply.

2.3. In vivo implant procedure and specimen collection/processing

Male inbred Sprague Dawley rats, weighing between 350 and 600 g

were intubated and halothane anesthesia performed according to the

requirements of the Canadian Council of Animal Care, under University

of British Columbia Animal Care certification. Implants were either

placed as a one stage or two-stage procedure [4]. In one-stage procedure,

the two sub and percutaneous components were attached together and

secured with a titanium miniature pin at the parietal areas of the rats. One

stage implants were harvested 3-weeks post-implantation. For two stage

implants, the subcutaneous component was placed 8 weeks prior to the

addition of the percutaneous component to allow connective tissue

ingrowth. The whole implants were then harvested 3 weeks after

placement of the percutaneous component. Animals were euthanized with

an overdose of sodium pentobarbital, and whole body perfusion

performed using 2.5% glutaraldehyde prior to implant removal. Speci-

mens were processed using protocols developed by Chehroudi et al. [4].

2mm sections were cut on a Sorvall MT2 microtome, and stained with

toluidine blue and viewed on a Zeiss light microscope (Zeiss Microscopes,

Oberkochen, Germany). The samples were analysed for the presence of

epithelial downgrowth and/or bone formation.

2.4. Time-lapse microscopy

Time-lapse microscopy was performed using Normarski Differential

Interference contrast as previously described [24]. AVI movies were

produced using Image J software (National Institutes of Health, Bethesda,

MD, US) and Quicktime version 4.0 (Apple Computers, Cupertino, CA,

US), and montages from selected movies created using Adobe Photoshop

7.0 (Adobe Corp, San Jose, CA, US).

2.5. Scanning electron microscopy

Samples designated for SEM analysis were prepared in triplicate and

processed as previously described [14]. Following this, they were viewed

using a Cambridge 260 stereo scanning electron microscope at 6 kV

accelerating voltage.

2.6. Immunocytochemistry

Antibody labeling and F-actin staining were performed as previously

described [14]. The primary antibodies used in this study were anti-mouse

phosphotyrosine (PY99) for phosphorylated tyrosine residues (Santa Cruz

Biotechnology, Santa Cruz, CA), anti-human FA Kinase (Tyr 397)

(Upstate Biotechnology, Charlottesville, VA), rabbit anti-mouse

phospho-MAP Kinase 1/2 (ERK 1/2) (Chemicon, Temecula, CA),

mouse anti-human vinculin (clone V284, Chemicon), and mouse

anti-human pan cytokeratin (Clone AE1:AE3, Chemicon). All primary

antibodies were verified to be cross-reactive with porcine and rat tissue,

and a dilution series was performed for each antibody to select the optimal

working concentration. For observing PY99, FA kinase, ERK 1/2,

vinculin and pan cytokeratin, all samples were fixed in 4% paraformalde-

hyde, and cell membranes permeabilized with 0.5% Triton X-100 (Sigma)

for 5min. Blocking of non-specific binding was performed using 1%

bovine serum albumin (BSA, Sigma) and the samples were incubated with

the appropriate primary antibody, at a pre-determined concentration for

90min. This was followed by three 10min washes in 1% BSA in PBS.

The secondary antibody was either a sheep—anti-rabbit (For ERK 1/2,

FA Kinase) or a goat anti-mouse (For PY99, vinculin, pan cytokeratin),

IgG conjugated to Alexa Fluor 596 (Molecular Probes Inc., Oregon, WA)

at a concentration of 1:200 for 1 h. All antibodies were diluted in PBS

containing 0.1% BSA. Labeling was visualized on a Zeiss Axioscope 2 or a

Nikon Eclipse C1 confocal microscope using up to 63� objectives, under

TRITC optics using a Princeton Pentamax CCD camera, and images

captured using Nikon Eclipse software. Images were imported to

Photoshop 7.0 for analysis and sizing.

2.7. Immunoblotting

Immunoblotting was performed as described previously [25]. In short,

protein levels were quantified using the BCA protein assay kit (Pierce,

Rockford, Ill), and 30 mg of total protein was mixed in a 1:2 ratio with

Laemelli buffer and 2-mercaptoethanol and boiled for 5min to denature

the protein samples. Samples were then loaded into either 10% or 14%

SDS page gels and the samples run at 175V for 45min. Following this, the

gels were blotted onto PVDF membranes at 100V for 2 h. For the

detection of the specific antigens, the PVDF membrane was incubated

with 1:1000 dilution of the appropriate monoclonal antibody overnight at

4 1C, and then washed with TTBS washing buffer for 1 h. After application

of the biotinylated goat anti-mouse IgG at a dilution of 1:5000 (Bio-Rad,

Mississauga, Ont., Canada), the membrane was developed using enhanced

colometric method. For quantification, densitometric analysis was

performed using the Image J analysis software.

2.8. Statistics

Data were analysed using the software program SPSS (version 11.04 for

Macintosh OS X; SPSS Inc., Chicago, IL). For western blots experiments,

three replicate experiments were performed for each surface (smooth,

120TP280S) and each label (tensin, vinculin, ERK 1/2). Comparisons were

made between smooth and the experimental surfaces by time (4 weeks)

using paired t-tests (po0.05).

3. Results

3.1. Epithelial and bone attachment in vivo

Three weeks post-implantation, epithelium had migratedover the 120TP280S topographies on one stage implants(Fig. 2A). The epithelium was observed to consist ofstratified layers, which had migrated and replaced theconnective tissue attachment on the TP surface of theimplant. On both one stage and two stage implants withsmooth control surfaces, epithelium had either partially orcompletely marsupialized the entire implant body (data notshown; for example see [4]). On two stage implants,connective tissue matured and grew into the TPs on thesubcutaneous component at 8 weeks, and after the additionof the percutaneous component, the epithelium was only

ARTICLE IN PRESS

Fig. 2. Light microscope images showing (A) epithelial peri-migration on a one stage percutaneous implant, (B) inhibition of epithelial migration by

connective tissue attachment on a two stage implant, and (C) bone-like tissue formation, on a subcutaneous pitted surface 8 weeks after placement, with

osteocyte like cells evident close to the base of the TP (D). Epi ¼ epithelium, CT ¼ connective tissue, OC ¼ osteocyte. Arrows indicate the boundary of

connective tissue and epithelium. Stain in all images is toluidine blue. As epithelial tissue remains attached to the bottom of the pits, the space between the

tissue and the pit is likely artefactual produced by sectioning.

D.W. Hamilton et al. / Biomaterials 28 (2007) 2281–22932284

able to migrate until it encountered the connectivetissue/implant interface (Fig. 2B). The proximity of thesubcutaneous part of the two-stage implant to theperiosteum of the calvaria often resulted in bone forma-tion on the textured surfaces, and this was evident in120TP280S at after 8 weeks (Fig. 2C). Mineralizedtissue was observed to completely fill the area of the120TP280S pits (Fig. 2C, D), and mature osteocyte-like

cells were present within the region of mineralized tissue(Fig. 2D).

3.2. In vitro migration of individual PLE and RCO cells

To assess the basic response of individual PLE cells andRCO to TPs and 30 mm IP, time-lapse microscopy wasemployed. On the 60TP105S, 120TP185S 120TP280S, PLE

ARTICLE IN PRESSD.W. Hamilton et al. / Biomaterials 28 (2007) 2281–2293 2285

cells were observed to migrate down onto the slopes of thepits from the ridges, but were noted to be unable to migrateback out, seemingly becoming trapped in the TPs (Fig. 3and Movie 1). On 30TP50S and 30IP185S, where the spaceat the top was narrower, PLE cells had a tendency to coverover the top of the pits, and exhibited little migration, butin contrast, RCO cells were able to traverse between TPsirrespective of the depth, or spacing of the pits (Fig. 3 andMovie 2). Furthermore, RCOs were able to bridge acrossthe top of the pits, and were observed to migrate down theridges, across the base, and back up the slopes of thedeepest pits, 120TP185S, and 120TP280S.

3.3. Cell morphology

Scanning electron microscopy was employed to observethe morphological response of PLE cells and RCOs todifferent dimensions of TPs and IP (Fig. 4). On smoothsurfaces, epithelial cells were well spread with very closeattachment to the surface, but they did not exhibit anypreferred orientation. On 30IP185S, PLE cells were wellspread on smooth areas, but were also observed to coverover the top section of the IP. On 60TP105S, PLE cellsbridged from ridge to ridge covering over the top of thepits. On 120TP280S, PLE cells were spread, spanning overridges, but were also found spread on the 551 slopes of theTPs. RCOs were able to bridge and spread on all IP andTPs tested, as was observed in time-lapse movies. The most

Fig. 3. Time-lapse sequence obtained using reflected Normarski DIC optics of

montages from movie 1 (PLE cells) and movie 2 (RCOs). All movies and imag

the arrows highlight points of attachment of the PLE cells, which highlight tha

indicate the movement of the RCO over the ridge into a neighbouring TP.

common RCO morphology was an elongated, spindleshaped with a well-developed leading edge. RCOs wereable to bridge from one pit across a ridge and down ontothe slope of the neighbouring pit.

3.4. Focal adhesion formation

As PLE cells exhibited restricted migration, and werewell spread, forming close attachments to the pitscompared with RCOs, we hypothesized that the PLE cellsmay differ in the formation of focal adhesions (FA) onthese surfaces, when compared with RCOs. PLE andRCOs were cultured on the various dimensions of TPs and30 mm IP for 24 h and then stained for vinculin, a FAprotein. On smooth surfaces, PLE cells were observed toform large plaque FAs, and in comparison, RCOsexhibited much smaller punctate adhesions, althoughplaque adhesions were evident on leading lamellapodia(Fig. 5). When cultured on 30IP185S, both PLE cells andRCOs that bridged over the pits exhibited large plaqueformation on the smooth areas up to the edge of the pits.Similarly, on 60TP105S, PLE cells covered the entire pits,forming FAs on all ridges at the top of the TPs. On120TP185S and 120TP280S, the largest area of FAformation was localized to the smooth areas adjacent tothe top of the pits, with fewer focal contacts formed in thearea of the cells towards the base of the TPs. On both60TP105S and 30IP1755S, RCO cells that bridged between

(A) PLE cells, and (B) RCOs migrating on 60TP105S. Images shown are

es are representative of PLE and RCO behaviour on all TPs tested. In (A)

t the cells do not move significantly over a 60min period. In (B) the arrows

ARTICLE IN PRESS

Fig. 4. Scanning electron microscope images of the morphology of PLE cells and RCOs on varying dimensions of microfabricated, tapered pits, and

smooth control surfaces. In (A) smooth surfaces, (B) 30IP185S, (C) 60TP105S, and (D) 120TP280S.

D.W. Hamilton et al. / Biomaterials 28 (2007) 2281–22932286

ridges, or between ridges and floors, formed FAs at eacharea of the cell in contact with the surface. RCO cells alsospread on the slopes of the TPs and IP, with prominentvinculin containing leading lamellae.

3.5. Influence of TPs on high-density cultures

When seeded as cultures of individual cells, both PLEcells and RCOs were found to exhibit distinct patterns ofbehaviour. To more accurately mimic the in vivo situation,high-density cultures of both cell types were seeded onsmooth controls, 120TP185S, and 120TP280S to assesshow population density influenced the response of bothPLE cells and RCOs. 120TP185S was selected as the twocell types exhibited different migratory responses to thissubstratum in vitro, and 120TP280S was employed as it had

stimulated different responses from the two tissue typesin vivo.

3.5.1. RCO production of mineralized matrix

The formation of mineralized matrix by RCOs, wasassessed through incorporation of alizarin complexone tothe newly formed mineral deposits [14]. On smoothsurfaces at 2 and 4 weeks, mineral deposits were smalland localized to regions above the cells, with noduleformation rarely observed at 4 weeks (Fig. 6A, C). At2 weeks post-seeding, mineral incorporation on 120TP280Swas significantly more intense than that seen on smoothsurfaces, with nodules extending up from the base of the pit(Fig. 6B). Mineralization took the form of variable sizedglobular deposits. By 4 weeks, large mineral deposits wereevident, that in some cases filled entire pits and extendedover the ridges between neighbouring TPs (Fig. 6D).

ARTICLE IN PRESS

Fig. 5. The arrangement of vinculin in PLE cells and RCOs on (A) smooth surfaces, (B) 30IP185S, (C) 60TP105S, and (D) 120TP280S after 24 h of

culture. Stain in all images is a monoclonal antibody to mouse anti-human vinculin. Images were captured using light (lmax ¼ 543 nm) on an

epifluorescence microscope with a low light-intensity CCD camera.

D.W. Hamilton et al. / Biomaterials 28 (2007) 2281–2293 2287

3.5.2. PLE cell sheets

In vivo, covering and lining epithelium exist as sheets ofcontinuing cells with little extracellular matrix [2]. Tomimic the in vivo condition, PLE cells were seeded at highdensity on the smooth region adjacent to the area of thesubstratum containing either the 120TP185S or 120TP280Ssubstratum, where they formed sheets (Fig. 7A). Observa-tions from time-lapse microscopy revealed that PLE cells at

the edge of the sheet migrated onto the ridges of the pits andseparated from the sheet (Fig. 7B and Movie 3). The PLEcells subsequently became trapped within the pit. Visualiza-tion of the cells using F-actin labeling demonstrated that thePLE cells had partially filled the first row of pits and movedinto the subsequent row of pits as early as 24-h post-seeding(Fig. 7C). At 4-week post-seeding, PLE cells completely filledall pits adjacent to the smooth portion of the substratum

ARTICLE IN PRESS

Fig. 6. Mineralization of osteoblasts at 2 and 4 weeks post-seeding. On (A, C) flat surfaces, the staining localizes at the cellular level at both time-points,

with little evidence of nodule formation evident, even at 4 weeks. On 120TP280S, nodule formation and globular mineral deposits are evident at (B) 2

weeks, and by (D) 4 weeks the pits are completely filled with mineral. In all images, mineralization was assessed through incorporation and visualization of

AC under light (lmax ¼ 543 nm) on an epifluorescence microscope using a low light-intensity CCD camera.

D.W. Hamilton et al. / Biomaterials 28 (2007) 2281–22932288

(Fig. 7D) and scanning electron microscopy demonstratedthat the PLE sheet had extended across a significant numberof TPs in a 4-week period (Fig. 7E).

3.6. Influence of substratum topography on adhesion

associated proteins in migrating PLE sheet cultures

Cells can form two distinct types of adhesions, focal and/or fibrillar adhesions [26]. FAs are characterized by thepresence of vinculin, while fibrillar adhesions, which areconsidered more mature adhesion sites, are rich in tensin.Western blot analysis revealed a significant increase in theamount of vinculin present in PLE cells migrating on120TP280S than on smooth surfaces (po0.05) (Fig. 8A),but no significant difference was found in the amount oftensin (Fig. 8B).

3.7. Influence of substratum topography on cell signaling in

migrating PLE sheet cultures

FAs have been identified as being a main cellular siteof tyrosine phosphorylation, but in contrast, fibrillaradhesions are not associated with such phosphorylation[25,26]. As PLE cells cultured on 120TP280S containedsignificantly more vinculin than cells on smooth surfaces,

western blotting for general tyrosine phosphorylation wasperformed. Overall, PLE cells cultured on 120TP280Sshowed an increase in the molecular weight of thephosphorylated proteins in comparison with cells grownon smooth, although no significant increase in the numberof bands was detected (Fig. 9A). As we have previouslyshown that ERK 1/2 increased in RCOs grown on pitscompared to smooth surfaces [25], PLE cells were probedfor phosphorylation of these MAP kinases. Both ERK 1and ERK 2 were phosphorylated in PLE cells on120TP280S, but in comparison, PLE cells on smooth hada significantly lower level of phosphorylation of ERK 1and 2 (po0.05) (Fig. 9B). Localization of the ERK 1 and 2signals through immunocytochemistry in PLE cells grownon 120TP280S, demonstrated that the majority of thesignal was present in the nucleus, but on smooth surfacewas distributed through the cytoplasm (Fig. 9C, D).

3.8. Influence of substratum topography on differentiation in

migrating PLE sheet cultures

As multiple layers of epithelium which can be indicative ofa differentiating epithelium [27], were observed in vivo andin vitro in PLE sheets growing over pits, immunocytochem-istry was employed to assess qualitatively the levels of pan

ARTICLE IN PRESS

Fig. 7. Influence of 120TP185S and 120TP280S on PLE sheet migration in vitro. The position of cell seeding using cloning rings is shown in (A). After

seeding cells (B) a time-lapse sequence was obtained using reflected Normarski DIC optics (see also movie 3). The migration of the PLE sheet over the TPs

was assessed using F-actin staining at (C) 24 h, and (D) 4 weeks. In (E) scanning electron microscope image of the migrating PLE sheet on 120TP185S, 4

weeks post-seeding. F-actin images were visualized under light (lmax ¼ 543 nm) on an epifluorescence microscope using a low light-intensity CCD camera.

D.W. Hamilton et al. / Biomaterials 28 (2007) 2281–2293 2289

cytokeratins (AE1:AE3) expressed by the cells in vitro.Increased staining intensity for pan cytokeratins wasobserved in PLE sheets grown on 120TP280S for 4 weekscompared to PLE sheets on smooth surfaces (Fig. 10).

4. Discussion

4.1. In vivo, tapered pits enhance bone formation, but do not

inhibit long-term epithelial downward migration

Previous observations with microfabricated topogra-phies suggested that long-term prevention of epithelialmigration down percutaneous implants was related moreto the attachment of connective tissue to the implant, thanthe direct influence of the implant topography itself onepithelial migration, which was effective only in the shortterm [4,6,7,28]. Similar results were obtained in vivo in thisstudy, with epithelial downward migration only impededby the presence of mature connective tissue formed in theTPs on the subcutaneous component of two stage implantsprior to the addition of the percutaneous component.Bone ingrowth to the TPs was marked at 8 weeks on thesubcutaneous component of two stage implants, with

osteocyte like cells present within the mineralized tissue.These in vivo observations suggest that osteoblasts maybemore responsive to the TP topographies or the micro-environment surrounding them, but the migration of thestratified epithelial sheets seems largely unaffected by TPtopographies. In vivo, osteoblasts and epithelial cellsperform physiologically distinct functions, with the formercreating bone and the latter migrating to cover and lineother tissues. Thus the possibility arises that these innatecell type specific functions influence cellular response totopography. As these innate behaviours can be observedin vitro, we examined the response of the two cell types toTPs and IP in vitro.

4.2. In vitro individual RCOs and PLE cells exhibit

differences in their migration, adhesion formation, and

spreading on TPs, IP, and smooth surfaces

Individual PLE cells and RCOs were observed to havedifferent morphologies, as well as exhibit differentmigratory behaviour when cultured on TPs and IP. Thecomparative differences in migration and morphologybetween individual PLE cells and RCOs indicated potential

ARTICLE IN PRESS

Fig. 8. Quantitative western blot analysis of the adhesion associated

proteins (A) 130 kDa vinculin and (B) 215 kDa tensin. Vinculin is

associated with FA, and tensin is associated with fibrillar adhesions.

Statistical analysis was performed using a paired t-test, with n ¼ 3.

*po0.05.

Fig. 9. Qualitative western blot analysis of (A) PY99 general tyrosine

phosphorylation, and (B) quantitative analysis of phosphorylated ERK 1

(44 kDa), and 2 (42 kDa), in PLE sheets, grown on smooth surfaces and

120TP280S. In (A), an increase in the molecular weight of the

phosphorylated targets is evident on 120TP280S in comparison with

smooth as indicated by the arrows. In (B), a quantitative increase in both

ERK1 and 2 is evident on 120TP280S, and the signal primarily localized to

cytoplasm of (C) smooth surfaces and (D) the cell nuclei (indicated by the

arrows) on 120TP280S. The ERK 1 and 2 images were visualized under

light (lmax ¼ 543 nm) on an epifluorescence microscope using a low light-

intensity CCD camera.

D.W. Hamilton et al. / Biomaterials 28 (2007) 2281–22932290

differences in cell adhesion mechanisms, a hypothesisconfirmed by immunolabelling for the FA protein,vinculin. The control of FA formation and cellularmigration is an important area of research focus [28–30],particularly in the area of cell engineering and biomaterials[18,31,32]. It is interesting that individual epithelial cellsshow a reduced migratory ability and tendency to spreadforming numerous large FAs. This would suggest that thein vivo function of PLE, the covering and lining of tissuescould be a primary determinant of their response to TPsand IP in vitro, given their tendency to spread over thesetypes of topographies. If properly controlled, this closeadhesion to the topographies could allow for a tight seal atthe top of the implant, preventing infection.

4.3. PLE sheet migration on TPs in vitro, is characterized

by altered tyrosine phosphorylation, multilayering,

cytokeratin expression and increased ERK 1/2

phosphorylation

In vivo, epithelial cells exist primarily in sheets, not asindividual cells, and cell–cell contacts would be expected tolessen flexibility. Using high-density cultures seeded on the

ARTICLE IN PRESS

Fig. 10. Immunocytochemical staining of PLE sheets on (A) smooth, and

(B) 120TP280S. Note the increased intensity of staining in cells grown on

120TP280S. Stain in all images is a monoclonal antibody to rabbit anti-

mouse pan cytokeratin (clone AE1:AE3). Images were captured using

light (lmax ¼ 543nm) on an epifluorescence microscope with a low light-

intensity CCD camera.

D.W. Hamilton et al. / Biomaterials 28 (2007) 2281–2293 2291

edge of the region of TPs, it was observed that thetopographies did not prevent the migration of the sheet,with multilayering evident, most likely due to proliferationof the PLE cells within the TPs. In addition, the stainingintensity of pan cytokeratins increased compared withsmooth. It is possible that as the PLE cells multilayerwithin the pits, the cells begin to terminally differentiate;the PLE cells switch from low molecular weight to highmolecular weight cytokeratins. However, full analysis ofthe cytokeratin expression will need to be analysed bywestern blot to confirm this.

PLE sheets were observed to migrate up to 1000 mmacross the TPs from the initial seeding zone over a 4-weekperiod. The average downward migration by epitheliumin vivo after 3 weeks in the study by Chehroudi et al. [4],was 4757100 mm on smooth surfaces, and 2657111 mm on120TP280S. The most likely explanation for the differencesbetween the 2D sheets in vitro migration and the 3D in vivo

migration is the presence of connective tissue, platelets,inflammatory cells, cytokines, and fibrin in vivo, whichwould likely strongly influence the ability of the epithelium

to migrate across the implant [4]. Also to be considered isthe influence of cell–cell contacts in vivo where multiple celllayers exist, which may make the sheet less flexible, andtherefore tend to bridge over the topography andessentially by-pass it. However, species variation (rat forin vivo, porcine for in vitro) and the source of epithelial cells(dermal for in vivo, epithelial cell rests of Malassez forin vitro) must also be considered in the interpretation ofthe results. Nevertheless, it is evident the epithelial sheetsare not impeded by the presence of TPs, either in vivo orin vitro.We have previously shown that mineral formation by

RCOs grown on microfabricated TPs is associated with anincrease in the levels of general tyrosine phosphorylationand the MAP kinases, ERK 1/2 [25]. Cellular tyrosinephosphorylation is often associated with vinculin rich FAs[26], and we identified a near 3-fold increase in the amountof vinculin in PLE sheets migrating on 120TP280Scompared with smooth. One possible explanation for thisobservation is that as the PLE cells multilayer in the TPs,they begin to form adhesions on both the dorsal andventral side, which could require an increase in vinculincontent to sustain the increased number of cell–celladhesions. An increase in the molecular weight ofphosphorylated proteins, observed in PLE sheets grownon 120TP280S, could be related to the increase in vinculincontaining FAs. Interestingly, there was no significantincrease in the fibrillar adhesion protein, tensin, betweensmooth and 120TP280S. Fibrillar adhesions are notassociated with increases in tyrosine phosphorylation[25,26], so it is likely that any alterations in signaltransduction would be related to the additional vinculinrich FAs. FA mediated signal transduction is an area ofincreasing importance in the biomaterials field [33], mainlybecause of new ‘‘bio-active’’ materials being designed toenhance cell response to implanted biomaterials [34].To further investigate signal transduction in PLE sheets,

we qualitatively measured the phosphorylation of the MAPkinases ERK 1 and 2, which are sensitive to alterations insurface geometry [25], and has been shown to play animportant role in early adhesion mediated intracellularsignaling [35]. A quantitative increase in both phosphory-lated ERK 1 and 2 was observed in PLE sheets grown on120TP280S, with the signal localizing to the nucleus.Activated ERK 1 and 2 have to translocate to the nucleusto exert many of their effects, although the continuedaccumulation of phospho-ERK 1 and 2 in the nucleusrequires continued adhesion and integrin activation [36].Both ERK 1 and 2 also interact with paxillin and FAkinase both molecules associated with formation of focalcontacts [37]. We have previously shown that a reductionof FA kinase activity mediated by PP2 inhibition of srcreduces the level of ERK 1 and 2 phosphorylation in RCOsgrown on microfabricated grooves, but does not affectmigration [24]. It is therefore possible that increasedadhesions formed by the PLE sheets on 120TP280S isresponsible for the increase in ERK 1 and 2, although

ARTICLE IN PRESSD.W. Hamilton et al. / Biomaterials 28 (2007) 2281–22932292

analysis of integrins will be required to confirm thishypothesis. The observed increases in ERK 1 and 2phosphorylation, in both PLE cells and RCO cells grownon microfabricated substratum topographies, infers thatthis molecule maybe of particular importance in adhesionmediated signal transduction regulation of cell response tosubstratum topography.

5. Conclusions

Using TP topographies of precisely defined geometryproduced by microfabrication, we have demonstrated thatepithelium and osteoblasts respond differently to the sametopographies. Furthermore, individual PLE cells do notrespond in the same manner as PLE sheets. Thecomparative differences are related to adhesion mechan-isms, morphology and intracellular signaling. We suggestthat the inherent functional properties of epithelial cellsand osteoblasts strongly influence how they respond tosurface topography.

Acknowledgements

The authors would like to thank Andre Wong forhis suggestions on SEM and confocal techniques, andDr. J. Douglas Waterfield for critical reading of themanuscript. This work was funded by the CanadianInstitutes of Health (CIHR) Operating Grant to DMBrunette.

Appendix A. Supplementary data

Supplementary data associated with this article can befound in the online version at doi:10.1016/j.biomater-ials.2007.01.026.

References

[1] Steflik DE, Sisk AL, Parr GA, Lake FT, Hanes PJ. Experimental

studies of the implant–tissue interface. J Oral Implantol

1993;19(2):90–4 [discussion 136–7].

[2] Hamilton DW, Ghrebi S, Kim H, Chehroudi B, Brunette DM.

Surface topography and cell behaviour. In: Bowlin GL, Wnek G,

editors. Encyclopedia of biomaterials and biomedical engineering.

New York: Taylor & Francis; 2006. p. 1–15.

[3] Kieswetter K, Schwartz Z, Dean DD, Boyan BD. The role of implant

surface characteristics in the healing of bone. Crit Rev Oral Biol Med

1996;7(4):329–45.

[4] Chehroudi B, Gould TR, Brunette DM. The role of connective tissue

in inhibiting epithelial downgrowth on titanium-coated percutaneous

implants. J Biomed Mater Res 1992;26(4):493–515.

[5] Esposito M, Thomsen P, Ericson LE, Sennerby L, Lekholm U.

Histopathologic observations on late oral implant failures. Clin

Implant Dent Relat Res 2000;2(1):18–32.

[6] Chehroudi B, Brunette DM. Subcutaneous microfabricated surfaces

inhibit epithelial recession and promote long-term survival of

percutaneous implants. Biomaterials 2002;23(1):229–37.

[7] Kim H, Murakami H, Chehroudi B, Textor M, Brunette DM. Effects

of surface topography on the connective tissue attachment to

subcutaneous implants. Int J Oral Maxillofac Implants 2006;21(3):

354–6.

[8] Grosse-Siestrup C, Affeld K. Design criteria for percutaneous

devices. J Biomed Mater Res 1984;18(4):357–82.

[9] Glauser R, Schupbach P, Gottlow J, Hammerle CH. Periimplant soft

tissue barrier at experimental one-piece mini-implants with different

surface topography in humans: a light-microscopic overview and

histometric analysis. Clin Implant Dent Relat Res 2005;7(Suppl 1):

S44–51.

[10] Smith DC. Dental implants: materials and design considerations. Int

J Prosthodont 1993;6(2):106–17.

[11] Le Guehennec L, Soueidan A, Layrolle P, Amouriq Y. Surface

treatments of titanium dental implants for rapid osseointegration.

Dent Mater 2006 August 10 [Epub ahead of print].

[12] Wennerberg A. Implant design and surface factors. Int J Prosthodont

2003;16(Suppl):45–7 [discussion 47–51].

[13] Chehroudi B, Gould TR, Brunette DM. Titanium-coated micro-

machined grooves of different dimensions affect epithelial and

connective-tissue cells differently in vivo. J Biomed Mater Res

1990;24(9):1203–19.

[14] Hamilton DW, Wong KS, Brunette DM. Microfabricated discontin-

uous-edge surface topographies influence osteoblast adhesion,

migration, cytoskeletal organization, and proliferation and enhance

matrix and mineral deposition in vitro. Calcif Tissue Int 2006;78(5):

314–25.

[15] Hamilton DW, Brunette DM. Gap guidance of fibroblasts and

epithelial cells by discontinuous edged surfaces. Exp Cell Res 2005;

309(2):429–37.

[16] Wieland M, Textor M, Chehroudi B, Brunette DM. Synergistic

interaction of topographic features in the production of bone-like

nodules on Ti surfaces by rat osteoblasts. Biomaterials

2005;26(10):1119–30.

[17] Wieland M, Chehroudi B, Textor M, Brunette DM. Use of Ti-coated

replicas to investigate the effects on fibroblast shape of surfaces with

varying roughness and constant chemical composition. J Biomed

Mater Res 2002;60:434–44.

[18] Baharloo B, Textor M, Brunette DM. Substratum roughness

alters the growth, area, and FAs of epithelial cells, and their

proximity to titanium surfaces. J Biomed Mater Res A 2005;74(1):

12–22.

[19] Chehroudi B, McDonnell D, Brunette DM. The effects of micro-

machined surfaces on formation of bonelike tissue on subcutaneous

implants as assessed by radiography and computer image processing.

J Biomed Mater Res 1997;34(3):279–90.

[20] Camporese DS, Laster TP, Pulfrey DL. A fine silicon shadow mask

for inversion layer solar cells. IEEE Electron Device Lett

1981;2:61–2.

[21] Baier RE, Meyer AE. Future directions in surface preparation of

dental implants. J Dent Educ 1988;52(12):788–91.

[22] Brunette DM, Melcher AH, Moe HK. Culture and origin of

epithelium-like and fibroblast-like cells from porcine periodontal

ligament explants and cell suspensions. Arch Oral Biol 1976;21(7):

393–400.

[23] Bellows CG, Aubin JE, Heersche JN, Antosz ME. Mineralized bone

nodules formed in vitro from enzymatically released rat calvaria cell

populations. Calcif Tissue Int 1986;38(3):143–54.

[24] Khakbaznejad A, Chehroudi B, Brunette DM. Effects of titanium-

coated micromachined grooved substrata on orienting layers of

osteoblast-like cells and collagen fibers in culture. J Biomed Mater

Res 2004;70A:206–18.

[25] Hamilton DW, Brunette DM. The effect of substratum topography

on osteoblast adhesion mediated signal transduction and phosphor-

ylation. Biomaterials 2007;28(10):1806–19.

[26] Zamir E, Geiger B. Molecular complexity and dynamics of

cell–matrix adhesions. J Cell Sci 2001;114(20):3583–90.

[27] Nettesheim P, Smits HL, George MA, Grey T, Jetten AM. Squamous

differentiation in normal and transformed rat tracheal epithelial cells.

Carcinogenesis 1989;10(4):743–8.

ARTICLE IN PRESSD.W. Hamilton et al. / Biomaterials 28 (2007) 2281–2293 2293

[28] Chehroudi B, Gould TL, Brunette DM. Effects of a grooved

titanium-coated implant surface on epithelial cell behavior in vitro

and in vivo. J Biomed Mater Res 1989;23(9):1067–85.

[29] Li S, Guan JL, Chien S. Biochemistry and biomechanics of cell

motility. Annu Rev Biomed Eng 2005;7:105–50.

[30] Turner CE. Paxillin and FA signalling. Nat Cell Biol 2000;2:E231–6.

[31] Subauste MC, Pertz O, Adamson ED, Turner CE, Junger S, Hahn

KM. Vinculin modulation of paxillin-FAK interactions regulates

ERK to control survival and motility. J Cell Biol 2004;165(3):371–81.

[32] Goto T, Yoshinari M, Kobayashi S, Tanaka T. The initial

attachment and subsequent behavior of osteoblastic cells and oral

epithelial cells on titanium. Biomed Mater Eng 2004;14(4):537–44.

[33] Hamilton DW, Riehle MO, Monaghan W, Curtis AS. Articular

chondrocyte passage number: influence on adhesion, migration,

cytoskeletal organisation and phenotype in response to nano- and

micro-metric topography. Cell Biol Int 2005;29(6):408–21.

[34] Vogel V, Baneynx G. The tissue engineering puzzle: a molecular

perspective. Annu Rev Biomed Eng 2003;5:441–63.

[35] Schuler M, Owen GR, Hamilton DW, de Wild M, Textor M,

Brunette DM, et al. Biomimetic modification of titanium dental

implant model surfaces using the RGDSP-peptide sequence: a cell

morphology study. Biomaterials 2006;27(21):4003–15.

[36] Howe AK, Aplin AE, Juliano RL. Anchorage dependent ERK

signaling—mechanisms and consequences. Curr Opin Gen Dev

2002;12(1):30–5.

[37] Danilkovitch A, Donley S, Skeel A, Leonard AJ. Two independent

signaling pathways mediate the antiapoptotic action of macrophage-

stimulating protein on epithelial cells. Mol Cell Biol 2000;20:2218–27.