Biomaterials 19 (1998) 2219—2232

Response of MG63 osteoblast-like cells to titanium and titanium alloyis dependent on surface roughness and composition

J. Lincks!,", B.D. Boyan",#,$,*, C.R. Blanchard%, C.H. Lohmann#,Y. Liu#, D.L. Cochran", D.D. Dean#, Z. Schwartz",#,&

!Wilford Hall Medical Center, Lackland AFB" Department of Periodontics, University of Texas Health Science Centre, San Antonio, TX, USA# Department of Orthopaedics, University of Texas Health Science Centre, San Antonio, TX, USA$ Department of Biochemistry, University of Texas Health Science Centre, San Antonio, TX, USA

% Southwest Research Institute, San Antonio, Texas, USA& Department of Periodontics, Hebrew University Hadassah Faculty of Dental Medicine, Jerusalem, Israel

Abstract

The success of an implant is determined by its integration into the tissue surrounding the biomaterial. Surface roughness andcomposition are considered to influence the properties of adherent cells. The aim of this study was to determine the effect of chemicalcomposition and surface roughness of commercially pure titanium (Ti) and Ti—6Al—4V alloy (Ti-A) on MG63 osteoblast-like cells.Unalloyed and alloyed Ti disks were machined and either fine-polished or wet-ground, resulting in smooth (S) and rough (R) finishes,respectively. Standard tissue culture plastic was used as a control. Surface topography and profile were evaluated by cold fieldemission scanning electron microscopy and profilometry, while chemical composition was determined using Auger electronspectroscopy and Fourier transform infrared spectroscopy. The effect on the cells was evaluated 24 h postconfluence by measuring cellnumber, [3H]-thymidine incorporation into DNA, cell and cell layer alkaline phosphatase specific activity (ALPase), osteocalcin andcollagen production, [35S]-sulfate incorporation into proteoglycan, and prostaglandin E

2(PGE

2) and transforming growth factor-b

(TGF-b) production. When compared to plastic, the number of cells was reduced on the pure Ti surfaces, while it was equivalent onthe Ti-A surfaces; [3H]-thymidine incorporation was reduced on all surfaces. The stimulatory effect of surface roughness on ALPasein isolated cells and the cell layer was more pronounced on the rougher surfaces, with enzyme activity on Ti-R being greater than onTi-A-R. Osteocalcin production was increased only on the Ti-R surface. Collagen production was decreased on Ti surfaces exceptTi-R; [35S]-sulfate incorporation was reduced on all surfaces. Surface roughness affected local factor production (TGF-b, PGE

2). The

stimulatory effect of the rougher surfaces on PGE2

and TGF-b was greater on Ti than Ti-A. In summary, cell proliferation,differentiation, protein synthesis and local factor production were affected by surface roughness and composition. Enhanceddifferentiation of cells grown on rough vs. smooth surfaces for both Ti and Ti-A surfaces was indicated by decreased proliferation andincreased ALPase and osteocalcin production. Local factor production was also enhanced on rough surfaces, supporting thecontention that these cells are more differentiated. Surface composition also played a role in cell differentiation, since cells cultured onTi-R surfaces produced more ALPase than those cultured on Ti-A-R. While it is still unknown which material properties inducewhich cellular responses, this study suggests that surface roughness and composition may play a major role and that the best designfor an orthopaedic implant is a pure titanium surface with a rough microtopography. ( 1998 Published by Elsevier Science Ltd.All rights reserved

Keywords: Osteoblasts; Titanium; Titanium alloy; Surface roughness; PGE2; TGF-b; In vitro

1. Introduction

The morphology of an implant surface, includingmicrotopography and roughness, has been shown to be

*Corresponding author. Tel.: (210) 567-6326; fax: (210) 567-6295;internet: BoyanB@uthscsa.edu

related to successful bone fixation [1, 2]. In addition, themanufacturing process used to achieve the surface tex-ture, either chemical [3] or mechanical [4], also influen-ces clinical success. At present, titanium implants inclinical use vary with respect to surface roughness andcomposition, with consensus being limited to the factthat bone forms more readily on a rough surface whereas

0142-9612/98/$—See front matter ( 1998 Published by Elsevier Science Ltd. All rights reserved.PII S 0 1 4 2 - 9 6 1 2 ( 9 8 ) 0 0 1 4 4 - 6

fibrous connective tissue is found more frequently ona smooth surface [5].

In vitro studies have provided some insight into theresponse of specific cell types to surface properties. It isclear that surface roughness affects cell response. In par-ticular, osteoblast-like cells exhibit roughness-dependentphenotypic characteristics. They tend to attach morereadily to surfaces with a rougher microtopography[6, 7]. Moreover, they appear to be more differentiatedon rougher surfaces with respect to morphology, ex-tracellular matrix synthesis, alkaline phosphatase specificactivity and osteocalcin production, and response to sys-temic hormones such as 1,25-(OH)

2D

3[8, 9]. The degree

of roughness also affects production of local factors suchas transforming growth factor beta (TGF-b) and prosta-glandin E

2(PGE

2) [10], both of which can act on the

osteoblastic cells as autocrine regulators [11, 12], andcan modulate the activity of bone resorbing cells viaparacrine mechanisms [13, 14].

The morphology of the surface also plays a role. A var-iety of cells can orient themselves in the grooves ofmicromachined surfaces [15—17]. Depending on the de-gree of roughness, these cells may actually see the grooveas smooth. On a randomly rough surface as is created bygrit blasting or chemical etching, cells may form differentfocal attachments which result in a phenotype that isdistinct from that seen on the grooved surface with thesame degree of roughness.

Titanium implants which are currently in clinical usein dentistry and orthopaedics, vary with respect to sur-face roughness and composition. In dentistry, commer-cially pure titanium (Ti) has become one of the mostcommonly used implant materials whereas in orthopae-dics Ti alloys have virtually replaced Ti because ofstrength requirements [18, 19]. Both Ti and Ti—6Al—4V(Ti alloy) develop a surface oxide layer due to the naturalpassivation of Ti [20, 21]. However, differences in thecrystallinity of the underlying metal as well as the segre-gation of alloy components, may cause the oxide thatforms on Ti to be quite different from the oxide thatforms on Ti alloys. Several studies have shown that evensubtle differences in surface composition, including Tioxide crystallinity, can modify cell response, even whensurface roughness is held constant [6, 22—28].

We previously showed that when MG63 osteoblast-like osteosarcoma cells are cultured on Ti discs withaverage surface roughness values (R

!) varying from

(0.1 lm (smooth) to 3—4 lm (rough) to'6 lm (veryrough), there are distinct differences in phenotypic ex-pression [8, 10]. For these studies, the smooth surfaceswere obtained by electropolishing following chemicaletching; rough surfaces were obtained by coarse gritblasting; and very rough surfaces were achieved via Tiplasma spray. The results showed that as surface rough-ness increased, expression of a differentiated osteoblasticphenotype increased, including reduced cell number and

DNA synthesis (proliferation), and increased alkalinephosphatase specific activity (ALPase), osteocalcin pro-duction, collagen synthesis, proteoglycan sulfation, andproduction of latent TGF-b and PGE

2. The optimal

surface appeared to be those with R!values around 4 lm;

cell proliferation was reduced but not blocked andphenotypic differentiation was enhanced. In contrast, cellson the smooth surface had high proliferation rates butALPase and osteocalcin production were low, indicativeof a loss of a differentiated osteoblastic phenotype. Todetermine whether the composition of the surface ormicrotopography are more important variables in deter-mining osteoblastic phenotype, we examined the responseof MG63 cells to machined surfaces with smooth R

!values

as well as with rough R!values that were prepared from Ti

and Ti alloy. The results of the present study using ma-chined surfaces were compared to those of our previouswork using grit-blasting to obtain similar R

!values.

2. Materials and methods

2.1. Titanium disk preparation and characterization

2.1.1. Disk preparationTitanium disks (14.75 mm diameter; 0.8 mm thick)

were fabricated from sheets of either commercially puretitanium (Ti: medical grade 2, ASTM F67, ‘unalloyed Tifor medical applications’) or titanium-6 wt% aluminum-4 wt% vanadium alloy (Ti—6Al—4V; Ti-A) obtained fromTimet, Inc. (O’Fallon, MO). Chemical composition wasprovided by the supplier and was not verified prior tosurface preparation. Each sheet was sectioned into onefoot by one foot plates for ease of handling and to ensurea consistent finish. The disks were either polished orground to acquire the desired surface finishes. Polishingto create the smooth surface was performed by lappingwith 18T grit (oil based 500—600 grit aluminum oxide)followed by polishing with 4.0 paper (1200 grit aluminumoxide) by French Grinding Service, Inc. (Houston, TX).The rough surface was prepared by wet sanding usinga carborundum brand zirconium oxide/aluminum oxideresin bonded to a cloth belt by Metal Samples, Inc.(Mumford, AL).

Disks were stamped using an automated metal punchand cleaned in an acetone bath using an ultrasoniccleaner for one hour. The disks were then washed in Jet-Afuel (grade AL-24487-F; Diamond Shamrock, San Anto-nio, TX) in an ultrasonic cleaner for one hour and wasfollowed by four washes with Versa Clean (Fisher Scient-ific, Pittsburgh, PA). Between each wash with VersaClean the disks were rinsed twice with deionized, distilledwater. After the final wash, the disks were rinsed with70% ethanol and then dried in vacuo. Prior to use eachdisk was washed again three times with ethanol andrinsed three times with deionized, distilled water. The

2220 J. Lincks et al. / Biomaterials 19 (1998) 2219—2232

disks were individually wrapped in gauze to preventdamage and then sterilized by autoclaving.

2.1.2. Surface characterizationRepresentative disks from each group were subjected

to surface analysis. The surface microtopography of thedisks was examined using an Amray 1645 cold fieldemission scanning electron microscope (Amray, Bedford,MA) with a nonthermally assisted tip and secondary andbackscattered electron capability. Two samples fromeach group were examined at 100 to 500].

Surface roughness was measured by profilometry us-ing a Taylor—Hobson Surtronic 3 profilometer (Leicester,UK). Average surface roughness (R

!) measurements were

taken at ten different locations on each one foot]onefoot sheet to obtain an accurate assessment. For thesmooth surfaces, measurements were made in all direc-tions, whereas on the rough surfaces, measurements weretaken perpendicular to the machine markings. Followingthe punching operation, four disks from each sheet wererandomly sampled to confirm the R

!values obtained

earlier.Auger electron spectroscopy was performed to analyze

the Ti oxide layer using a Perkin-Elmer Model 595scanning Auger microprobe (Perkin Elmer, PhysicalElectronics Division, Eden Prairie, MN). Spectra wereobtained from two representative disks from the twogroups with a smooth surface (Ti and Ti-A) to determinethe chemical profile of the subsurface layer. Rough diskswere not examined to avoid artifacts associated withrough morphologies; further, the thickness and composi-tion of the surface oxides on the rough and smooth disksfor each material would be expected to be identical sinceall disks were machined and cleaned using the sameprotocol. The spectra were obtained at regular sputteringintervals at a sputtering rate of 400 As min~1. Comparingspectra and relative peak heights at given surface depthsprovided information about the chemistry of the oxidelayer.

Fourier transform infrared spectroscopy (FTIR) wasperformed to determine if an organic residue remainedon the disk surfaces after cleaning. Spectra were obtainedfrom four disks (two from the smooth Ti group and twofrom the smooth Ti-A group) using a Nicolet MagnaFTIR in reflection mode. Spectra were collected using 32scan summations at a resolution of 16 cm~1. FTIR spec-troscopy was not performed on the rough surfaces, be-cause artifactual measurements are obtained on roughsamples.

2.2. Cell culture

MG63 osteoblast-like cells were used for these experi-ments because they were obtained from a human osteo-sarcoma [29] and have been well-characterized. Theydisplay numerous osteoblastic traits that are typical of

a relatively immature osteoblast, including the stimula-tion of alkaline phosphatase activity and osteocalcinsynthesis and inhibition of proliferation in response totreatment with 1a,25-(OH)

2D

3[29, 30]. As a result

they are a good model for examining the early stagesof osteoblast differentiation. However, the cultureconditions under which MG63 cells will mineralize theirmatrix have not been defined, so terminal differentiationcannot be studied using these cells. Despite thislimitation, we selected this model in preference to fetalrat calvarial cells since the latter are derived fromembryonic rat bone which may differ significantlyfrom adult human bone. We recognize that MG63 cellsare not normal osteoblasts and data interpretationmust take this into consideration. MG63 cells wereobtained from the American Type Culture Collection(Rockville, MD).

For all experiments, cells were cultured on disks placedin 24 well plates (Corning, Corning, IL). Controls consis-ted of cells cultured directly on the polystyrene surface ofthe 24 well plate. Cells were plated at 9300 cells cm~2 inDulbecco’s modified Eagle’s medium (DMEM) contain-ing 10% fetal bovine serum (FBS) and 0.5% antibiotics(diluted from a stock solution containing 5000 Uml~1

penicillin, 5000 U ml~1 streptomycin; GIBCO, GrandIsland, NY) and cultured at 37°C in an atmosphereof 100% humidity and 5% CO

2. Media were changed

every 48 h until the cells reached confluence. Becauseof the opacity of the Ti disks, there was no practicalway to assess confluency of the cultures. As a result, whencells reached visual confluence on plastic, cultures on allother surfaces were treated exactly as those grown onplastic.

2.3. Cell morphology

To determine whether cell morphology varied asa function of surface roughness, the cultures were exam-ined by scanning electron microscopy. At harvest, theculture media were removed and the samples rinsed threetimes with phosphate-buffered saline (PBS) and fixedwith 1% OsO

4in 0.1 M PBS for 15—30 min. After fix-

ation, the disks were rinsed with PBS, sequentially incu-bated for 30—45 min each in 50, 75, 90 and 100%ter-butyl alcohol, and vacuum dried. A thin layer of gold-palladium was sputter-coated onto the samples prior toexamination in a JEOL 6400 FEC cold field emissionscanning microscope (JEOL USA, Inc. Peabody, MA).

2.4. Cell proliferation

2.4.1. Cell numberAt harvest, cells were released from the culture surface

by addition of 0.25% trypsin in Hank’s balanced saltsolution (HBSS) containing 1 mM ethylenediamine tet-raacetic acid (EDTA) for ten minutes at 37°C, and this

J. Lincks et al. / Biomaterials 19 (1998) 2219—2232 2221

was followed by addition of DMEM containing 10%FBS to stop the reaction. Previous studies demonstratedthat two trypsinizations are necessary to quantitativelyharvest MG63 cells from rough Ti surfaces [8]. Accord-ingly, a second trypsinization was performed to ensurethat any remaining cells had been removed from thesurface. Cell suspensions from both trypsinizations werecombined and centrifuged at 500]g for 10 min. Cellpellets were washed with PBS and resuspended in PBS.Cell number was determined by use of a Coulter Counter(Coulter Electronics, Hialeah, FL). Cells harvested in thismanner exhibit'95% viability based on trypan blue dyeexclusion.

2.4.2. [3H]-thymidine incorporationDNA synthesis was estimated by measuring [3H]-

thymidine incorporation into trichloroacetic acid (TCA)insoluble cell precipitates as previously described bySchwartz et al. [31]. MG63 cells were cultured on theplastic surface or Ti disks until the cells on plastic reach-ed visual confluence. Media were changed and the incu-bation continued for an additional 24 h. Four hoursprior to harvest, 50 ll [3H]-thymidine (from a 1 lCiml~1

stock solution) was added to the cultures. At harvest, thecell layers were washed twice with cold PBS, twice with5% TCA, and then treated with ice-cold saturated TCAfor 30 min. TCA-precipitable material was dissolved in0.25 ml 1% sodium dodecyl sulfate (SDS) at 20°C andradioactivity measured by liquid scintillation spectro-scopy.

2.5. Cell differentiation

2.5.1. Alkaline phosphatase specific activityAt harvest, either cell layers, as described below, or

isolated cells, as described above, were prepared andtheir protein content determined by use of commerciallyavailable kits (Micro/Macro BCA, Pierce Chemical Co.,Rockford, IL). Alkaline phosphatase [orthophosphoricmonoester phosphohydrolase, alkaline; E.C. 3.1.3.1] ac-tivity was assayed as the release of p-nitrophenol fromp-nitrophenylphosphate at pH 10.2 as previously de-scribed [32] and specific activity determined.

Cell layers were prepared following the method ofHale et al. [33]. At harvest, culture media were decanted,cell layers washed twice with PBS, and then removedwith a cell scraper. After centrifugation, the cell layerpellets were washed once more with PBS and resusp-ended by vortexing in 0.5 ml deionized water plus 25 ll1% Triton-X-100. Pellets were further disrupted byfreeze/thawing three times. Isolated cells were harvestedas described above for the determination of cell number,except that after the cell pellets had been washed twicewith PBS, the cells were resuspended by vortexing in0.5 ml of deionized water with 25 ll of 1% Triton-X-100.

Enzyme assays were performed on both cell and cell layerlysates.

2.5.2. Osteocalcin productionThe production of osteocalcin by the cultures was

measured using a commercially available radioimmuno-assay kit (Human Osteocalcin RIA Kit, BiomedicalTechnologies, Stoughton, MA). Culture media were con-centrated five-fold by lyophilization and reconstituted in100 ll normal rabbit serum, 10 ll rabbit anti-humanosteocalcin antibody, 100 ll [125I]-human osteocalcin,and 200 ll Tris-saline buffer and placed overnight on anorbital platform shaker (approximately 80 rpm) at roomtemperature. Goat anti-rabbit antibody and polyethy-lene glycol (100 ll each) were added to each tube thefollowing morning. After vortexing, the samples wereplaced on an orbital shaker for 2 h at room temperature.One ml of Tris-saline buffer was added to each sample.The solution was then vortexed and centrifuged at500]g for 20 min at 4°C. The supernatant was decantedand the pellet placed in scintillation cocktail andcounted. Osteocalcin concentrations were determined bycorrelating the percentage bound over unbound countsto a standard curve.

2.6. Matrix production

2.6.1. Collagen productionMatrix protein synthesis was assessed by measuring

the incorporation of [3H]-proline into collagenasedigestible (CDP) and noncollagenase digestible (NCP)protein [34]. When the cells reached confluence onplastic, the media in all cultures were replaced with500 ll DMEM containing 10% FBS, antibiotics, and50 lgml~1 b-amino proprionitrile (Sigma, St. Louis,MO), and 10 lCiml~1 of L[G3H]-proline (New EnglandNuclear, Boston, MA). After 24 h, media were discarded.Cell layers (cells and matrix) were obtained by scrapingand resuspending in two 0.2 ml portions of 0.2 N NaOH.Proteins were precipitated with 0.1 ml 100% TCA con-taining 1% tannic acid, washed three times with 0.5 ml10% TCA#1% tannic acid, and then twice with ice-cold acetone. The final pellets from the cell layers weredissolved in 500 ll 0.05 N NaOH.

Digestion of the cell layer pellet was performed usinghighly purified clostridial collagenase (Calbiochem, SanDiego, CA; 138 Umg~1 protein) as described previously[8]. NCP synthesis was calculated after multiplying thelabeled proline in NCP by 5.4 to correct for its relativeabundance in collagen [34]. Percent collagen productionwas calculated by comparing CDP production with totalCDP#NCP production (i.e.: [CDP/(CDP#NCP)]]100). The protein content of each fraction was deter-mined by miniaturization of the method of Lowry et al.[35]. This assay does not take into account any

2222 J. Lincks et al. / Biomaterials 19 (1998) 2219—2232

degradation that may have occurred during the incuba-tion or during sample preparation.

2.6.2. Proteoglycan sulfationProteoglycan synthesis was assessed by [35S]-sulfate

incorporation according to the method of O’Keefe et al.[36]. Previously, we found that the amount of radio-labeled proteoglycan secreted into the media by MG63cells was less than 15% of the total radiolabeled proteo-glycan produced. Because more than 85% of theradiolabeled proteoglycan was in the cell layer, we exam-ined the incorporation of [35S]-sulfate only in the celllayer.

At confluence, 50 ll DMEM containing 90 lCiml~1

[35S]-sulfate were added to the media to make a finalconcentration of 9 lCiml~1. Four hours later, the mediawere discarded and the wells washed one time with 500 llPBS. The cell layer was collected in two 0.25 ml portionsof 0.25 M NaOH. The protein content was determined bythe method of Lowry et al. [35]. To measure [35S]-sulfate incorporation into the cell layers, the total volumewas adjusted to 0.7 ml by the addition of 0.15 M NaCland the sample dialyzed in a 12 000—14 000 molecularweight cut-off membrane against buffer containing0.15 M NaCl, 20 mM Na

2SO

4, and 20 mM Na

2HPO

4at

pH 7.4 and 4°C. The dialysis solution was changed untilthe radioactivity in the dialysate reached backgroundlevels. The amount of [35S]-sulfate incorporated wasdetermined by liquid scintillation spectrometry and wascalculated as dpm mg~1 cell layer protein.

2.7. Local factor production

2.7.1. Prostaglandin E2

The amount of PGE2

produced by the cells and re-leased into the media was assessed using a commerciallyavailable competitive binding radioimmunoassay kit(NEN Research Products, Boston, MA). In this assay,unlabeled PGE

2in the sample was incubated overnight

with radiolabeled PGE2

and unlabeled PGE2

antibody.Antigen-antibody complexes were separated from freeantigen by precipitation with polyethylene glycol.Sample PGE

2concentrations were determined by cor-

relating the percentage bound over unbound counts toa standard curve.

2.7.2. Transforming growth factor-beta (TGF-b)In order to measure the level of total TGF-b produc-

tion by the cells, a commercially available enzyme-linkedimmunoassay (ELISA) kit (Promega Corp., Madison,WI) specific for human TGF-b1 was used. Immediatelyprior to assay, conditioned media were diluted 1 : 10 inDMEM and the 1 : 10 dilution further diluted by addingfour volumes of PBS. The media were then acidified bythe addition of 1 M HCl for 15 min to activate latentTGF-b (LTGF-b), followed by neutralization with 1 M

NaOH. The assay was performed according to the manu-facturer’s directions. Intensity measurements were con-ducted at 450 nm using a BioRad Model 2550 EIAReader (Hercules, CA). Sample concentrations were de-termined by comparing the absorbance value to a knownstandard curve. The amount of TGF-b1 in the cell layerwas not examined because of difficulties associated withquantitatively extracting this cytokine from the matrix.

2.8. Statistical analysis

Experiments were conducted at least twice and thedata shown are from one representative experiment.For any given experiment, each data point representsmean$SEM of six individual cultures. Data were firstanalyzed by analysis of variance; when statistical differ-ences were detected, the Student’s t-test for multiplecomparisons using Bonferroni’s modification was used.P-values(0.05 were considered to be significant.

3. Results

3.1. Disk characteristics

3.1.1. MorphologyWhen the Ti-S and Ti-A-S disks were examined by

scanning electron microscopy, the surfaces were found tobe very similar (Fig. 1A and C). Morphologically, thedisks had small pits (2 lm in diameter) and randomlyoriented scratches from the polishing operation, whichwere only evident at high magnification (data not shown).The Ti-R and Ti-A-R disks also had a similar appearance(Fig. 1B and D) and contained parallel, longitudinalgrooves with both sharp and serrated edges, resultingfrom the grinding operation. Parallel grooves of varyingheights were prominent; in addition, the distance be-tween the grooves varied. On both rough surfaces, curvedsheets of material were observed occasionally at the apexof the grooves. Additionally, the Ti-A-R surface con-tained areas with pits that were 10—20 lm in diameter.

3.1.2. Surface roughnessBased on profilometry (Table 1) the smooth surfaces,

Ti-S and Ti-A-S, had similar R!

values of 0.22 and0.23 lm, respectively. The Ti-R surface was the roughestand had an R

!of 4.24 lm, while the Ti-A-R surface had

an R!of 3.20 lm. Both rough surfaces were significantly

rougher than both smooth surfaces.

3.1.3. Auger electron spectroscopyBoth smooth surfaces (Ti-S and Ti-A-S) were found to

contain Ti, O, and C by Auger electron spectroscopybefore sputtering. In the alloyed surface, Al was alsofound. After 10 s of sputtering, the C signal was virtuallygone at a depth of 67 As in both Ti and Ti-A disks. In

J. Lincks et al. / Biomaterials 19 (1998) 2219—2232 2223

Fig. 1. Scanning electron micrographs of the different disk surfaces used in this study. Panel A: Ti-S; Panel B: Ti-R; Panel C: Ti-A-S; Panel D: Ti-A-R.Bar"200 lm. Original magnification: 100].

Table 1Average surface roughness values for the Ti and Ti-alloy disks used inthis study

Surface R!value

Ti-S 0.22$0.00!

Ti-R 4.24$0.13"

Ti-A-S 0.23$0.00!

Ti-A-R 3.20$0.12

Note: Ti and Ti alloy (Ti-A) disks were prepared with either a smooth(S) or rough (R) surface as described in the Materials and Methods. TheR

!value for each disk type was determined by profilometry. Data

shown in the table represent the mean$SEM for four disks in eachgroup; each disk was measured in four areas.

! P(0.05, smooth vs. rough surface."P(0.05, Ti-R vs. Ti-A-R.

addition to Ti and O, Al was also present in the alloy.Twenty seconds of sputtering to a depth of 134 As produ-ced a continuously decreasing O signal while sputteringthrough the oxide layer, and an increasing Ti signal. Afterone minute, the Ti signal became very strong, and

the O signal virtually disappeared. No evidence ofvanadium was found in the disks.

3.1.4. Fourier transform infrared spectroscopyFTIR analysis of the disks confirmed that no organic

residue was left on the surface of either the Ti-S or Ti-A-Sdisks.

3.2. Cell morphology

The appearance of the cells varied with surface rough-ness and chemical composition of the disks. Cells grownon the Ti-S surface were spread out across the surfaceand grew as a monolayer, but this monolayer was notcontinuous (Fig. 2C and D). The cells had a dendriticappearance, with extensions that were up to 10 lm inlength and had ruffled membranes on their surfaces. Cellscultured on Ti-R (Fig. 2A and B) and Ti-A-S (Fig. 3C andD) disks grew as a continuous, thin monolayer across thesurface. On the Ti-R surface, all cracks and fissures werecovered by a monolayer of cells (Fig. 2A and B). Cultureson the Ti-A-R surface induced the cells to grow asa multilayer (Fig. 3A and B), with many cells producing

2224 J. Lincks et al. / Biomaterials 19 (1998) 2219—2232

Fig. 2. Scanning electron micrographs of MG63 osteoblast-like cells cultured on smooth and rough Ti surfaces. Panel A: Ti-R, magnification:100], bar"10 lm; Panel B: Ti-R, magnification: 500], bar"1 lm; Panel C: Ti-S, magnification: 100], bar"10 lm; Panel D: Ti-S, magnification:500], bar"1 lm.

extensions that covered distances of up to 10 lm. Inaddition, the cells were oriented along the parallel cracksand grooves and grew over the sharp edges, forminga multilayer.

3.3. Cell proliferation

3.3.1. Cell numberCell number was affected by both chemical composi-

tion and surface roughness (Fig. 4). Compared to plastic,cell number was reduced by 36% on Ti-R. Although notstatistically significant, cell number was also reduced by20% on Ti-S. Fewer cells were present on the Ti-Rsurfaces than on Ti-A-R as well. The numbers of the cellsgrown on the Ti-A-S and Ti-A-R surfaces were similar tothat seen on the plastic.

3.3.2. [3H]-thymidine incorporation[3H]-thymidine incorporation was reduced on all

metal surfaces when compared to plastic (Fig. 5). Theeffect was comparable on the alloyed Ti surfaces (49%)and the Ti-R surface (48%). However, the decrease seenon the Ti-S surface was significantly less than on theother surfaces (19%).

3.4. Cell differentiation

3.4.1. Alkaline phosphatase specific activityEnzyme activity varied with surface roughness and

composition (Fig. 6). Cell layers from cells cultured on alldifferent surfaces contained significantly more alkalinephosphatase specific activity than on the plastic control(1.6 fold to 2.2 fold). Activity on Ti-R was 20% greaterthan on Ti-A-R. Activity on the rough surfaces wasconsistently greater than on smooth surfaces. Alkalinephosphatase on Ti-R was 1.5-fold greater than on Ti-S;on Ti-A-R, alkaline phosphatase was 1.3-fold greaterthan on Ti-A-S.

When enzyme activity of isolated cells was measured,similar observations were made (Fig. 7). Cells grown onTi-R surfaces exhibited a 1.8-fold increase in enzymeactivity over that seen on plastic. On Ti-R, the increasewas 1.4-fold, and on the smooth surface disks, there wasa 1.3-fold increase. Activity was greater on Ti-R in com-parison to Ti-S and in comparison to Ti-A-R.

These results also showed that the effects of surfaceroughness and composition on alkaline phosphatase spe-cific activity were primarily due to enzyme present in thematrix. Specific activity of the cell layer was consistently

J. Lincks et al. / Biomaterials 19 (1998) 2219—2232 2225

Fig. 3. Scanning electron micrographs of MG63 osteoblast-like cells cultured on smooth and rough Ti-A-surfaces. Panel A: Ti-A-R, magnification:100], bar"10 lm; Panel B: Ti-A-R, magnification: 500], bar"1 lm; Panel C: Ti-A-S, magnification: 100], bar"10 lm; Panel D: Ti-A-S,magnification: 500], bar"1 lm.

Fig. 4. Number of MG63 osteoblast-like cells released by two trypsin-izations of the Ti disks 24 h after they had reached confluence on theplastic. Values are the mean$ SEM of six cultures. *P(0.05, Ti diskvs. plastic; d P(0.05, Ti-A-R vs. Ti-R. Data are from one of tworeplicate experiments.

Fig. 5. [3H]-Thymidine incorporation by MG63 osteoblast-like cellsduring culture on plastic or Ti disks. When the cells reached confluence onplastic, the media were changed and culture continued for another 24 h.Four hours prior to harvest, [3H]-thymidine was added and incorpora-tion into TCA insoluble cell precipitates measured. Values are themean$SEM of six cultures. *P(0.05, Ti disk vs. plastic; dP(0.05,Ti-S vs. Ti-R. Data are from one of two replicate experiments.two times that of the isolated cells, despite the larger

denomination due to the presence of matrix protein. Thefold-increases noted as a function of either roughness orcomposition were greater when assaying cell layers, re-sulting in significantly greater real enzyme activity thanwas seen in the isolated cells. This was particularly evi-dent for cell layers cultured on Ti-R.

3.4.2. Osteocalcin productionCell cultures grown on the Ti-R surface showed a sig-

nificant increase (1.9 fold) in osteocalcin production com-pared to plastic (Fig. 8). The osteocalcin production by

2226 J. Lincks et al. / Biomaterials 19 (1998) 2219—2232

Fig. 6. Alkaline phosphatase specific activity of cell layers produced byMG63 osteoblast-like cells during culture on Ti disks. After cells hadreached confluence on plastic, cultures were continued for an additional24 h and then harvested by scraping. Enzyme activity was measured inthe cell layer lysate. Values are the mean$SEM of six cultures.* P(0.05, titanium vs. plastic; d P(0.05, Ti-A-R vs. Ti-R;d P(0.05, smooth vs. rough surface of same material. Data are fromone of two replicate experiments.

Fig. 7. Alkaline phosphatase specific activity of trypsinized MG63 os-teoblast-like cells after culture on Ti disks. After cells had reachedconfluence on plastic, cultures were continued for an additional 24 hand then harvested by trypsinization. Enzyme activity was measured inlysates of the cells. Values are the mean$SEM of six cultures.* P(0.05, titanium vs. plastic; d P(0.05, Ti-A-R vs. Ti-R;d P(0.05, smooth vs. rough surface of same material. Data are fromone of two replicate experiments.

Fig. 8. Osteocalcin production by MG63 osteoblast-like cells duringculture on Ti disks. After cells reached confluence on plastic, the mediawere changed and the culture continued for an additional 24 h. Atharvest, the media were collected, and osteocalcin content measured byRIA. Values are the mean$SEM of six cultures. * P(0.05, titaniumvs. plastic. Data are from one of two replicate experiments.

Fig. 9. Percent collagen production by MG63 osteoblast-like cells dur-ing culture on Ti disks. Values were derived from CDP and NCPproduction and are the mean$SEM of six cultures. * P(0.05, tita-nium vs. plastic; d P(0.05, Ti-A-R vs. Ti-R; d P(0.05, smooth vs.rough surface of same material. Data are from one of two replicateexperiments.

cells grown on all the other surfaces was similar toplastic.

3.5. Matrix production

3.5.1. Collagen productionCollagen synthesis was also affected by surface com-

position and roughness (Fig. 9). While collagen synthesiswas unaffected in cells cultured on Ti-R, cells grown onTi-S, Ti-A-R and Ti-A-S surfaces synthesized 14—30%less collagen compared to plastic. The percent collagen

production by the cells was significantly decreased (15%)on rough Ti-A-R surfaces compared to Ti-R surfaces.Moreover, cells on Ti-S produced 31% less collagen thanon Ti-R, and cells on Ti-A-S produced 17% less collagenthan on Ti-A-R.

3.5.2. Proteoglycan sulfationCompared to plastic, [35S]-sulfate incorporation

by MG63 cells was significantly reduced on all disksurfaces examined (35—48%) (Fig. 10). This effect wasleast pronounced in cells grown on the smooth Ti-A-Rsurface. No significant difference in the [35S]-sulfate

J. Lincks et al. / Biomaterials 19 (1998) 2219—2232 2227

Fig. 10. [35S]-Sulfate incorporation by MG63 osteoblast-like cellsduring culture on Ti disks. When the cells reached confluence onplastic, the media were changed and culture continued for another 24 h.Four hours prior to harvest, [35S]-sulfate was added and incorporationinto the cell layer measured. Values are the mean$SEM of six cul-tures. * P(0.05, titanium vs. plastic. Data are from one of tworeplicate experiments.

Fig. 11. Prostaglandin E2(PGE

2) production by MG63 osteoblast-like

cells during culture on Ti disks. After cells reached confluence onplastic, the media were changed and the culture continued for anadditional 24 h. At harvest, the media were collected, and PGE

2con-

tent measured by RIA. Values are the mean$SEM of six cultures.* P(0.05, titanium vs. plastic; d P(0.05, Ti-A-R vs Ti-R;d P(0.05, smooth vs. rough surface. Data are from one of tworeplicate experiments.

Fig. 12. Latent transforming growth factor b (LTGFb) production byMG63 osteoblast-like cells during culture on Ti disks. After the cellsreached confluence on plastic, the media were changed and the culturecontinued for an additional 24 h. At harvest, the media were collected,and LTGFb content measured by ELISA. Values are the mean$SEMof six cultures. * P(0.05, titanium vs. plastic; d P(0.05, Ti-A-R vs.Ti-R; d P(0.05, smooth vs. rough surface. Data are from one of tworeplicate experiments.

incorporation among the different surface roughnessesand compositions was observed.

3.6. Local factor production

3.6.1. Prostaglandin E2

The level of PGE2

production by the cells was affectedby the different surface treatments (Fig. 11). Significantlymore PGE

2was produced by cells cultured on Ti-R

when compared to plastic (3.9-fold) and to Ti-S surface(2.0 fold). Cells on the Ti-A-R surface synthesized 2.9-foldmore PGE

2than those on plastic and 3.3-fold more than

those on the Ti-A-S surface. The levels on both smoothsurface preparations were not significantly different fromplastic.

3.6.2. Transforming growth factor-bThe level of latent TGF-b in the conditioned media

was also influenced by culture on the different surfaces(Fig. 12). Latent TGF-b levels were increased by 1.7-foldon the Ti-A-R and 2.7-fold on the Ti-R surfaces. LatentTGF-b production was greater on the Ti-R surface com-pared to cultures grown on the Ti-A-R surface (1.6-fold)and 2.1-fold greater when compared to Ti-S. There wasa slight, but insignificant, increase in LTGFb levels pro-duced by cells grown on both smooth surfaces comparedto plastic.

4. Discussion

This study confirms previous observations that os-teoblast-like cells respond in a differential mannerto both surface roughness [7, 37—39] and material com-position [25, 40—42]. As noted previously [8—10],MG63 cells grown on Ti-R surfaces exhibited a moredifferentiated phenotype as evidenced by reduced cellproliferation and increased alkaline phosphatase specificactivity and osteocalcin production. Cells grown on Ti-Ssurfaces also exhibited reduced cell proliferation, andthey had elevated alkaline phosphatase in comparisonwith cultures grown on plastic, but the effects were lessrobust than those seen on Ti-R. Moreover, osteocalcinproduction was unaltered in these MG63 cells, indicatingthat they were not as differentiated as those cells grownon Ti-R.

2228 J. Lincks et al. / Biomaterials 19 (1998) 2219—2232

Although [3H]-thymidine incorporation was reducedin cells cultured on Ti-A-R, total cell number wasunaffected. The latter value is a cumulative measure ofthe viable cells in the culture, whereas the former valueis an indication of the rate of DNA synthesis, and there-fore, cell replication during the radiolabeling period,in our case, the last four hours of culture. This indicatesthat the cells grown on Ti-A-R ceased to proliferateand initiated expression of the mature osteoblasticphenotype at a slower rate than cells cultured on Ti-R,since proliferation is negatively correlated with phenotypicexpression [43]. This hypothesis is supported by the factthat alkaline phosphatase activity on Ti-A-R was elev-ated, but to a lesser degree than seen on Ti-R, and theMG63 cells on Ti-A-R did not exhibit elevated osteocal-cin production. Even for the alloy disks, however, thecells cultured on the rougher surfaces were more differen-tiated than the cells cultured on the smoother surfaces.

Other aspects of osteoblast function were sensitive tothe substrate, either with respect to roughness or to thebulk composition of the material. Production of extracel-lular matrix vesicles was affected by the nature of thesubstrate based on differences in cell layer alkaline phos-phatase, where matrix vesicles are present, whencompared to enzyme activity in isolated cells. Alkalinephosphatase is an early marker of osteogenic differenti-ation. While this enzyme activity is present in allcell membranes, it is found in higher levels in cells whichmineralize their matrix such as osteoblasts [44]. Asosteoblasts mature, they produce extracellular matrixvesicles which are enriched in alkaline phosphatase speci-fic activity; because of this specific enrichment, alkalinephosphatase is the marker enzyme for this extracellularorganelle [45]. Matrix vesicles are associated withthe onset of calcification and they contain enzymes neces-sary for matrix modification necessary for crystal depos-ition and growth [46, 47]. The results of the presentstudy show clearly that the effects of surface roughnesswere targeted to the matrix vesicles, whether the cellswere cultured on Ti-R or Ti-A-R, since the fold increasesin enzyme activity in the cell layer were significantlygreater than the fold increases observed in the isolatedcells.

In addition, the effects of material composition werealso found predominately in the matrix vesicle compart-ment, supporting previous in vivo and in vitro observa-tions. Studies examining endosteal healing adjacent tovarious implant materials demonstrate that matrix ves-icle production and function are sensitive to the type ofmaterial used [41, 48, 49]. Similarly, when cells were cul-tured on thin films of various implant materials whichhad been sputtered onto tissue culture plastic, the effectsof material composition were targeted to the matrixvesicles [26].

In comparison to plastic, proteoglycan sulfation wasreduced in all of the cultures to a comparable extent. In

contrast, collagen production was differentially affectedby the nature of the surface topography and the materialused. In general, synthesis on rougher surfaces wasgreater than seen on smoother surfaces, correlating withthe production of latent TGF-b. The expression of thisgrowth factor is associated with the collagen depositionin the extracellular matrix of osteoblasts [50]. Similarly,production of PGE

2was greater on the rougher surfaces,

supporting our previous observation that there is a posi-tive correlation of latent TGF-b and PGE

2production

with increasing surface roughness [10].Both latent TGF-b and PGE

2are produced by osteo-

blasts as paracrine and autocrine regulators of cellfunction and differentiation. Their release by the MG63cells cultured on Ti-S and Ti-A-S was essentially identicalto the basal levels seen on plastic, another smoothsurface. However, on the rougher Ti-R and Ti-A-Rsurfaces, their production was markedly enhanced,although in a material-specific manner, with the greatestproduction being observed in cells grown on Ti-R.This supports the contention that these cells exhibita more differentiated osteoblastic phenotype. Whethermore differentiated cells produce higher levels ofthese local factors, or whether the cells are more differen-tiated because they produce and respond to higherlevels of these factors, is not known. The amountsof PGE

2produced per culture are well within the

limits of prostaglandin known to be osteogenic and notinflammatory [11].

Since all of the TGF-b released into the media wasin latent form, it is difficult to comment on its contribu-tion to the differentiation of the MG63 cells. However,recent studies in our lab [51] and others [52] indicatethat the latent TGF-b which is incorporated into thematrix may be activated locally via the action of matrixvesicles and may regulate the phenotypic expression ofthe cells. There is some indication that this is the casein the present study. In cells cultured on Ti-R surfaces,both ALPase and osteocalcin production were increased,whereas on Ti-A-R surfaces, ALPase was stimulatedand osteocalcin production was not. When osteoblastsare treated with TGF-b, alkaline phosphatase, anearly marker of osteoblastic differentiation, is stimulated[12], whereas production of osteocalcin, a markerof terminal differentiation, is inhibited [12]. WhetherTGF-b is modulating the differential expression of os-teoblastic phenotypic markers in the MG63 cells is cer-tainly not established by this study but the potential forregulation of this type is evidenced by the fact thatproduction of local regulatory factors is sensitive to thematerial used.

The results presented here also support our previousobservation that roughness may play a more importantrole in determining cell response than the type of topo-graphy, as long as the R

!values can be sensed by the cells.

For practical purposes, the distance between peaks

J. Lincks et al. / Biomaterials 19 (1998) 2219—2232 2229

should not exceed the ability of the cell to form focalattachments on two or more peaks; otherwise, the cellwould sense a rough surface as smooth. In the presentstudy, surface roughness was achieved by machining,resulting in parallel grooves, whereas our previous stud-ies used commercially pure Ti disks that were roughenedby grit-blasting and acid-etching, resulting in randompeaks and valleys. In general, the MG63 cells respondedto smooth surfaces in a manner similar to their behavioron tissue culture plastic and to machined Ti-R surfaces ina manner similar to grit-blasted Ti surfaces with compa-rable R

!values. The morphology of the cells on the Ti-R

and Ti-A-R surfaces demonstrates that they have as-sumed a more cuboidal shape with dendritic extensions,similar to the morphology noted on rough cpTi surfacesachieved by grit-blasting, and typical of a more differenti-ated osteoblast. Similar observations have been notedwith chick embryonic osteoblasts [37]. In contrast, cellson the smoother surfaces appear more flattened andfibroblastic.

Our data also show that MG63 cells are sensitive tothe bulk composition of the material, whether the surfaceis smooth or rough. Even though a titanium oxide layerformed on both the Ti and Ti-alloy surfaces, it is unlikelythat the oxides were identical. Certainly mosaicism of thealloy components would result in a more complex sur-face chemistry. This would have a direct effect on thenature of the conditioning film that forms as the materialsurface interacts with the culture medium [53—55]. Inaddition, ions released from the alloy could also modu-late cellular response. Recently, studies using fibroblastcultures demonstrated that locally released vanadiumions from Ti—6Al—4V alloy surfaces negatively impactedcell adhesion [56]. Thompson and Puleo [57] have alsoshown that Ti—6Al—4V ion solutions can inhibit expres-sion of the osteogenic phenotype by bone marrowstromal cells, suggesting that ions released from implantscould also impair normal bone formation. Despitethe differences in cellular response due to material com-position, roughness remains the overriding variable inpromoting osteogenic differentiation. As strength re-quirements of orthopaedic implants necessitate the needfor alloyed titanium preparations, it is essential that theoptimal surface characteristics be determined, potentiallymitigating any negative effects of the bulk material onbone formation and function.

Acknowledgements

The authors gratefully acknowledge the expert assis-tance of Sandra Messier, Monica Luna, KimberlyRhame, and Roland Campos in the preparation of themanuscript. Jack Lincks is a fellow in the Air ForceInstitute of Technology. This work does not necessarilyreflect the views of the United States Air Force. Funding

for this research was provided by the Center for theEnhancement of the Biology/Biomaterials Interface atthe University of Texas Health Science Center at SanAntonio. Support for Dr. Lohmann was providedby a grant from the B. Braun Foundation, Melsungen,Germany.

References

[1] Rich A, Harris AK. Anomalous preferences of cultured macro-phages for hydrophobic and roughened substrata. J Cell Sci1981;50:1—7.

[2] Thomas K, Cook SD. An evaluation of variables influencingimplant fixation and direct bone appostition. J Biomed Mater Res1985;19:875—901.

[3] Schroeder A, Van der Zypen E, Stich H, Sutter F. The reactions ofbone, connective tissue and epithelium to endosteal implants withtitanium sprayed surfaces. J Maxillofac Surg 1981;9:15—25.

[4] Buser D, Schenk R, Steinemann S, Fiorellini J, Fox C, Stich H.Influence of surface characteristics on bone integration of tita-nium implants. A histomorphometric study in miniature pigs.J Biomed Mater Res 1991;25:889—902.

[5] Cochran DL, Simpson J, Weber H, Buser D. Attachment andgrowth of periodontal cells on smooth and rough titanium. IntJ Oral Maxillofac Implants 1994;9:289—97.

[6] Michaels CM, Keller JC, Stanford CM, Solursh M. In vitro cellattachment of osteoblast-like cells to titanium. J Dent Res1989;68:276.

[7] Bowers KT, Keller JC, Randolph BA, Wick DG, Michaels CM.Optimization of surface micromorphology for enhanced osteob-last responses in vitro. Int J Oral Maxillofac Implants1992;7:302—10.

[8] Martin JY, Schwartz Z, Hummert TW, Schraub DM, Simpson J,Lankford J, Dean DD, Cochran DL, Boyan BD. Effect of tita-nium surface roughness on proliferation, differentiation, and pro-tein synthesis of human osteoblast-like cells (MG63). J BiomedMater Res 1995;29:389—401.

[9] Boyan BD, Batzer R, Kieswetter K, Liu Y, Cochran DL, Szmuck-ler-Moncler S, Dean DD, Schwartz Z. Titanium surface rough-ness alters responsiveness of MG63 osteoblast-like cells to 1a,25-(OH)

2D

3. J Biomed Mater Res 1998;39:77—85.

[10] Kieswetter K, Schwartz Z, Hummert TW, Cochran DL, SimpsonJ, Dean DD, Boyan BD. Surface roughness modulates the localproduction of growth factors and cytokines by osteoblast-likeMG63 cells. J Biomed Mater Res 1996;32:55—63.

[11] Raisz LG, Fall PM. Biphasic effects of prostaglandin E2

on boneformation in cultured fetal rat calvariae: interaction with cortisol.Endocrinology 1990;126:1654—9.

[12] Bonewald LF, Kester MB, Schwartz Z, Swain LD, Khare AG,Johnson TL, Leach RJ, Boyan BD. Effects of combining trans-forming growth factor b and 1,25-dihydroxyvitamin D

3on differ-

entiation of a human osteosarcoma (MG63). J Biol Chem1992;267:8943—9.

[13] Akatsu T, Takahashi N, Udagawa N, Imamura K, Yamaguchi A,Sato K, Nagata N, Suda T. Role of prostaglandins in interleukin-1-induced bone resorption in mice in vitro. J Bone Miner Res1991;6:183—90.

[14] Bonewald LF, Mundy GR. Role of transforming growthfactor-beta in bone remodeling. Clin Orthop Rel Res 1990;250:261—76.

[15] Cheroudi B, Gould TRL, Brunette DM. Titanium-coated micro-machined grooves of different dimensions affect epithelial andconnective tissue cells differently in vivo. J Biomed Mater Res1990;24:1203—19.

2230 J. Lincks et al. / Biomaterials 19 (1998) 2219—2232

[16] Cheroudi B, Gould TRL, Brunette DM. Effects of grooved tita-nium-coated implant surface on epithelial cell behavior in vitroand in vivo. J Biomed Mater Res 1987;23:1067—85.

[17] Brunette DM. The effects of implant surface topography on thebehavior of cells. Int J Oral Maxillofac Implants 1988;3:231—46.

[18] Van Noort R. Review titanium: the implant material of today.J Mater Sci 1987;22:3801—11.

[19] Grabowsky KS, Gossett CR, Young FA, Keller JC. Cell adhesionto ion implanted titanium. Mater Res Soc Symp Proc 1989;110:697—701.

[20] Rostoker W, Galante JO. Materials for human implantation.J Biomech Eng 1979;101:2—14.

[21] Lausmaa J, Mattson L, Rolander U, Kasemo B. Chemical com-position and morphology of titanium surface oxides. Mater ResSoc Symp Proc 1986;55:351—9.

[22] Golijanin L, Bernard G. Biocompatibility of implant metals inbone tissue culture. J Dent Res 1988;67:367.

[23] Golijanin G, Bernard G, Tuck M, Davlin L. Comparative studyof the canine bone/implant interface in vitro and in vivo. J DentRes 1989;68:307.

[24] Nowlin P, Carnes D, Windeler A. Biocompatibility of dentalimplant materials sputtered onto cell culture dishes. J Dent Res1989;68:275.

[25] Arai T, Pilliar M, Melcher AH. Growth of bone-like tissue ontitanium and titanium alloy in vitro (Abstract d3). Trans SocBiomet Ann Meeting 1989.

[26] Hambleton JC, Schwartz Z, Windeler SW, Luna MH, Brooks BP,Khare AG, Dean DD, Boyan BD. Culture surfaces coated withvarious implant materials affect chondrocyte growth and meta-bolism. J Orthop Res 1994;12:542—52.

[27] Evans EJ, Benjamin M. The effect of grinding conditions on thetoxicity of cobalt-chrome-molybdenum particles in vitro. Bio-materials 1987;8:377—84.

[28] Merritt K, Brown SA. Biological effects of corrosion productsfrom metals. In: Fraker AC, Griffer CD, editor. Corrosion anddegradation of implant materials: 2nd Symposium, ASTM STP859P. Philadelphia, PA: Am Soc Testing Materials, 1985:195.

[29] Franceschi RT, James WM, Zerlauth G. 1,a,25-dihydroxyvitaminD

3specific regulation of growth, morphology, and fibronectin in

a human osteosarcoma cell line. J Cell Physiol 1985;123:401—9.[30] Boyan BD, Schwartz Z, Bonewald LF, Swain LD. Localization of

1,25-(OH)2D

3responsive alkaline phosphatase in osteoblast-like

cells (ROS 17/2.8, MG63, and MC3T3) and growth cartilage cellsin culture. J Biol Chem 1989;264:11879—86.

[31] Schwartz Z, Schlader DL, Ramirez V, Kennedy MB, Boyan BD.Effects of vitamin D metabolites on collagen production and cellproliferation of growth zone and resting zone cartilage cells invitro. J Bone Miner Res 1989;4:199—207.

[32] Bretaudiere JP, Spillman T. Alkaline phosphatases. In: BergmeyerHU, editor. Methods of enzymatic analysis. Weinheim: VerlagChemica, 1984:75—92.

[33] Hale LV, Kemick ML, Wuthier RE. Effect of vitamin D meta-bolites on the expression of alkaline phosphatase activity byepiphyseal hypertrophic chondrocytes in primary cell culture.J Bone Miner Res 1986;1:489—95.

[34] Peterkofsky B, Diegelmann R. Use of a mixture of proteinase-freecollagenases for the specific assay of radioactive collagen in thepresence of other proteins. Biochemistry 1971;10:988—94.

[35] Lowry OH, Rosebrough NJ, Farr AL, Rano RI. Proteinmeasurement with the folin phenol reagent. J Biol Chem 1951;193:265—75.

[36] O’Keefe RJ, Puzas JE, Brand JS, Rosier RN. Effects of transform-ing growth factor-beta on matrix synthesis by chick growth platechondrocytes. Endocrinology 1988;122:2953—61.

[37] Groessner-Schreiber B, Tuan RS. Enhanced extracellular matrixproduction and mineralization by osteoblasts cultured on tita-nium surfaces in vitro. J Cell Sci 1992;101:209—17.

[38] Ong JL, Prince CW, Raikar GN, Lucas LC. Effect of surfacetopography of titanium on surface chemistry and cellular re-sponse. Implant Dentistry 1996;5:83—8.

[39] Windeler AS, Bonewald LF, Khare AG, Boyan BD, Mundy GR.The influence of sputtered bone substitutes on cell growth andphenotypic expression. In: Davies JE, editor. The bone-bi-omaterial interface. Toronto: Toronto Press, 1991:205—13.

[40] Vrouwenfelder WC, Groot CG, deGroot K. Better histology andbiochemistry for osteoblasts cultured on titanium-doped bioac-tive glass: bioglass 45S5 compared with iron-, titanium-, fluorine-,and boron-containing bioglasses. Biomaterials 1994;15:97—106.

[41] Braun G, Kohavi D, Amir D, Luna MH, Caloss R, Sela J, DeanDD, Boyan BD, Schwartz Z. Markers of primary mineralizationare correlated with bone-bonding ability of titanium or stainlesssteel in vivo. Clin Oral Implants Res 1995;6:1—13.

[42] Bordij K, Jouzeau JY, Mainard D, Payan E, Netter P, Rie KT,Stucky T, Hage-Ali M. Cytocompatibility of Ti—6A1—4V andTi—5A1—2.5Fe alloys according to three surface treatments,using human fibroblasts and osteoblasts. Biomaterials 1996;17:929—40.

[43] Rickard DJ, Gowen M, MacDonald BR. Proliferative responsesto estradiol, IL-1 alpha and TGF beta by cells expressing alkalinephosphatase in human osteoblast-like cell cultures. Calcif TissueInt 1993;52:227—33.

[44] Owen TA, Aronow M, Shalhoub V, Barone LM, Wilming L,Tassinari MS, Kennedy MB, Pockwinse S, Lian JB, Stein GS.Progressive development of the rat osteoblast phenotype invitro: reciprocal relationships in expression of genes asso-ciated with osteoblast proliferation and differentiation duringformation of the bone extracellular matrix. J Cell Physiol1990;143:420—30.

[45] Ali SY. Mechanisms of calcification. In: Owen R, Goodfellow J,Bollough P, editors. Scientific foundation of orthopaedics andtraumatology. London: Heinemann, 1984:175—95.

[46] Dean DD, Schwartz Z, Bonewald LF, Muniz OE, Morales SM,Gomez R, Brooks BP, Qiao M, Howell DS, Boyan BD. Matrixvesicles produced by osteoblast-like cells in culture become signif-icantly enriched in proteoglycan-degrading metalloproteinasesafter addition of b-glycerophosphate and ascorbic acid. CalcifTissue Int 1994;54:399—408.

[47] Anderson HC. Matrix vesicle calcification: review and update. In:Pick WA, editor. Bone and mineral research. Amsterdam: Ex-cerpta Medica, 1984:109—49.

[48] Boyan BD, Schwartz Z, Dean DD, Hambleton JC. Response ofbone and cartilage cells to biomaterials in vivo and in vitro. J OralImplantol 1993;19:116—22.

[49] Schwartz Z, Amir D, Boyan BD, Cochavy D, Muller-Mai C,Swain LD, Gross U, Sela J. Effect of glass ceramic and titaniumimplants on primary calcification during rat tibial bone healing.Calcif Tissue Int 1991;49:359—64.

[50] Choi JY, Lee BH, Song KB, Park RW, Kim IS, Sohn KY, Jo JS,Ryoo HM. Expression patterns of bone-related proteins duringosteoblastic differentiation in MC3T3-E1 cells. J Cell Biochem1996;61:609—18.

[51] Boyan BD, Schwartz Z, Park-Snyder S, Dean DD, Yang F,Twardzik D, Bonewald LF. Latent transforming growth factor-bis produced by chondrocytes and activated by extracellularmatrix vesicles upon exposure to 1,25-(OH)

2D

3. J Biol Chem

1994;269:28,374—81.[52] Dallas SL, Miyazono K, Skerry TM, Mundy GR, Bonewald LF.

Dual role for the latent transforming growth factor-b bindingprotein in storage of latent TGF-b in the extracellular matrixand as a structural matrix protein. J Cell Biol 1995;131:539—49.

[53] Norde W. Behavior of proteins at the interfaces, with specialattention to the role of the structure stability of the proteinmolecule. Clin Mater 1998;11:85—91.

J. Lincks et al. / Biomaterials 19 (1998) 2219—2232 2231

[54] Hench LL, Paschall HA. Direct chemical bond of bioactive glass-ceramic materials to bone and muscle. J Biomed Mater Res1973;7:25—42.

[55] Jarcho M, Kay JK, Gumaer RH, Doremus RH, Drobeck HP.Tissue, cellular, and subcellular events at a bone-ceramic hy-droxylapatite interface. J Bioeng 1977;1:79—92.

[56] Eisenbarth E, Meyle J, Nachtigall W, Breme J. Influence of thesurface structure of titanium materials on the adhesion of fibro-blasts. Biomaterials 1996;17:1399—403.

[57] Thompson GJ, Puleo DA. Ti-6A1-4V ion solution inhibition ofosteogenic cell phenotype as a function of differentiation timecourse in vitro. Biomaterials 1996;17:1949—54.

2232 J. Lincks et al. / Biomaterials 19 (1998) 2219—2232