Fibroblast adhesion to micro- and nano-heterogeneoustopography using diblock copolymers and homopolymers

Irene Y. Tsai,1 Masahiro Kimura,2 Rebecca Stockton,3 J. Angelo Green,4 Ricardo Puig,1 Bruce Jacobson,4

Thomas P. Russell1

1Polymer Science and Engineering Department, University of Massachusetts, Amherst, Massachusetts 010032Films & Film Products Research Laboratories, Toray Industries, Inc., 1-1, Sonoyama 1-chome, Otsu, Shiga 520-8558,Japan3Cardiovascular Research Center, University of Virginia at Charlottesville, Charlottesville, Virginia4Program in Molecular and Cellular Biology and Department of Biochemistry and Molecular Biology, University ofMassachusetts at Amherst, Amherst, Massachusetts

Received 27 May 2004; accepted 8 July 2004Published online 13 October 2004 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/jbm.a.30183

Abstract: Polymeric substrates of different surface chemis-try and length scales were found to have profound influenceon cell adhesion. The adhesion of fibroblasts on surfaces ofoxidized polystyrene (PS), on surfaces modified with ran-dom copolymers of PS and poly(methyl methacrylate)[P(S-r-MMA)] with topographic features, and chemicallypatterned surfaces that varied in lateral length scales fromnanometers to microns were studied. Surfaces with hetero-geneous topographies were generated from thin film mix-tures of a block copolymer, PS-b-MMA, with homopolymersof PS and PMMA. The two homopolymers macroscopicallyphase separated and, with the addition of diblock copoly-mer, the size scales of the phases decreased to nanometerdimensions. Cell spreading area analysis showed that a thinfilm of oxidized PS surface promoted adhesion whereas athin film of P(S-r-MMA) surface did not. Fibroblast adhesion

was examined on surfaces in which the lateral length scalevaried from 60 nm to 6 �m. It was found that, as the laterallength scale between the oxidized PS surfaces decreased, cellspreading area and degree of actin stress fiber formationincreased. In addition, scanning electron microscopy wasused to evaluate the location of filopodia and lamellipodia.It was found that most of the filopodia and lamellipodiainteracted with the oxidized PS surfaces. This can be attrib-uted to both chemical and topographic surface interactionsthat prevent cells from interacting with the P(S-r-MMA) atthe base of the topographic features. © 2004 Wiley Periodi-cals, Inc. J Biomed Mater Res 71A: 462–469, 2004

Key words: heterogeneous surfaces; topography; fibroblasts;cell adhesion

INTRODUCTION

Cell adhesion, a key element in understanding cell–biomaterial interactions, underpins cell growth, func-tion, and survival. The parameters influencing celladhesion are critical to applications ranging from bio-sensors to implants to tissue engineering and bioreac-tors. Cell–substrate interactions are central in under-standing the fundamentals of cell adhesion andmigration. Recent studies have shown that topogra-

phy has an important role in cell adhesion and cellsrespond to topographic features on the nanometer andmicron length scales.1–7 In vivo, cells adhere to thebasement membrane substrata that are on the nano-meter length scale.8,9 Therefore, it is necessary to iden-tify the effects of length scale and surface chemistry oncell adhesion so as to optimize scaffolds for tissueengineering or biosensor surfaces.

With a uniform surface chemistry, cells will alignalong topographic features. On a grooved surface, forexample, cells will respond to grooves with the actincytoskeleton fibers oriented parallel to the grooves.This phenomenon is known as contact guidance. Cellsalso react differentially to chemical signals when at-tached to adhesive, as opposed to nonadhesive, pat-terns.6 Consequently, both chemistry and topographyhave important consequences on cell adhesion.2–5 Pat-terns generated by homopolymer blends and diblockcopolymers are well suited to study the role of topog-

Correspondence to: T. P. Russell; e-mail: russell@mail.pse.umass.edu

Contract grant sponsor: National Institutes of Health; con-tract grant numbers: GM-29127, T32-GM08515, 5 T32HL007284-27

Contract grant sponsor: U.S. Department of EnergyContract grant sponsor: National Science Foundation

© 2004 Wiley Periodicals, Inc.

raphy on cell adhesion in a very simple manner. Thinfilms of polymer mixtures macroscopically phase sep-arate with domain sizes typically on the micron lengthscale, whereas diblock copolymers microphase sepa-rate into well-defined arrays of domains tens of nano-meters in size.10–13 Thus, with mixtures of homopoly-mers and diblock copolymers, the size scale of themorphology can be systematically controlled from thenanoscopic to macroscopic length scales by varyingthe relative concentration of the components. In thisstudy, cell spreading area and actin cytoskeleton mor-phology were investigated on substrates in which thelateral length scale of the surface topography andchemical functionality were varied. A marked depen-dence on the length scale of the surface topographywas found.

MATERIALS AND METHODS

Surface preparation

Surfaces in which the lateral length scales of heterogene-ities varied from nanometers to micrometers were preparedusing thin film mixtures of homopolymers with diblockcopolymers. The relative concentrations of homopolymersand block copolymers were varied to control domain size.Asymmetric diblock copolymers of polystyrene (PS) andpoly(methyl methacrylate) (PMMA), denoted P(S-b-MMA),with a molecular weight of 73,000 and a polydispersity of1.04, were prepared by standard anionic polymerizationmethods. The volume fraction of PS in the PS-b-MMA is 0.7and, therefore, the equilibrium morphology of the copoly-mer is one with 20-nm cylindrical domains of PMMA withan average separation distance of 30 nm. PS and PMMAwith narrow molecular weight distribution and molecularweight of 52,000 and 29,000, respectively, were purchasedfrom Polymer Laboratories and used without further puri-fication. Five separate surfaces having different concentra-tions of PS-b-MMA mixed with PS and PMMA were usedand are listed in Table I. One percent blend solutions intoluene were spin coated onto a surface in which interac-tions were balanced. Specifically, a hydroxy-terminated ran-dom copolymer of styrene and methyl methacrylate[P(S-r-MMA)], having a styrene fraction of 0.58, was an-chored to the native oxide layer of a silicon substrate as

described previously.12,13 With the P(S-r-MMA), the interfa-cial interactions of the substrate with PS-b-PMMA, PS, andPMMA are balanced.

The thin film mixtures were annealed at 170°C, reactiveion etched with oxygen to remove the top �6 nm of the film,and then exposed to ultraviolet (UV) radiation for 35 min tocrosslink the PS and degrade the PMMA. The films weresubsequently washed with acetic acid to remove PMMA,rinsed with water, and dried. The PS remaining on thesubstrate is oxidized under these conditions. A schematicdiagram of the surface preparation is shown in Figure 1. Thescanning force microscopy images of the heterogeneous sur-faces were used to determine the autocorrelation,14 which isdefined as

C�x� � ���

�

h�x��h�x� � x�dx�

where h(x) is the height of the surface at a position x. Theposition of the first peak in C(x) was used as a measureof the lateral length scale of the heterogeneities.

To determine whether cells preferentially adhere tooxygen plasma-treated and UV-exposed PS orP(S-r-MMA), two surfaces of PS and PMMA were alsoprepared following the same procedure as the pat-terned surfaces, to produce a flat surface of uniformoxygen plasma-treated and UV-exposed PS andPS-r-MMA.

Tissue culture and cell spreading area assays

NIH 3T3 murine fibroblasts (American Type Culture Col-lection, Rockville, MD) were maintained as subconfluentmonolayers in Dulbecco’s modified Eagle medium (DMEM)(Sigma-Aldrich, St. Louis, MO) with 10% (v/v) calf serum(Atlanta Biologicals, Norcross, GA), 100 �g/mL streptomy-cin, and 60 U/mL penicillin (Fisher Scientific, Atlanta, GA)in a humidified 5% CO2 incubator at 37°C.

For cell spreading area assays, adherent cells were detachedwith 0.01% trypsin–ethylenediaminetetraacetic acid, washedwith 0.01% soybean trypsin inhibitor in serum-free medium,and resuspended at a density of 1–2 � 105 cells/mL DMEMwith 10% calf serum. Equal numbers of suspended cells wereplated onto prepared surfaces and glass coverslips coated uni-formly with fibronectin as control surfaces (Becton-Dickinson,

TABLE IAverage Length Scales of the Patterned Surfaces With the Corresponding Volume Fraction

of Homopolymer Blend and Diblock Copolymers

Surface NameVolume Fraction of

PS (MW 52K) PMMA (MW 29K)Volume Fraction of

PS-b-MMA (MW 71K)Autocorrelation

Period

6-�m surface 1 0 6 �m730-nm surface 0.83 0.17 730 nm580-nm surface 0.67 0.33 580 nm110-nm surface 0.33 0.67 110 nm60-nm surface 0.16 0.84 60 nm

FIBROBLAST ADHESION TO VARIED SURFACES 463

Franklin Lakes, NJ), placed in covered 35 mm2 PS tissue cultureplates. Plated cells were kept in a humidified 37°C incubator.Cells were allowed to spread on the surfaces for 30, 60, 90, and120 min, and the samples were fixed in 0.25% glutaraldehydefor 1 min, then permeabilized in Karsenti’s Buffer (0.5% TritonX-100, 80 mM PIPES, 1 mM MgSO4, 5 mM EGTA, pH 6.9) for1 min. Fixed cells on surface chips were washed and thenstored in phosphate-buffered saline containing 0.1% Tween-20and 0.02% Na azide. To analyze for cell spreading area, cellswere incubated for 15 min in rhodamine-labeled phalloidin(Molecular Probes, Eugene, OR) following the manufacturer’sinstructions. Samples were washed extensively with phos-phate-buffered saline and placed between a glass slide andglass coverslips after addition of ImmunoFluor mounting me-dium (ICN Biomedicals Inc., Irvine, CA) and sealed with nailpolish. Cells were photographed and evaluated for cell spread-ing area at indicated intervals by fluorescence microscopy,using an Olympus BX51 reflection fluorescence microscope.The data shown represent means and standard error of themean (SEM) of three experiments and five fields in each ex-periment. At least 150 cells were analyzed for each data point.Cell spreading area was analyzed using “Image J,” download-able from the National Institutes of Health (NIH).15

Actin cytoskeleton staining and scoring degree ofactin stress fiber formation

Cells were plated on patterned and flat surfaces and al-lowed to spread for either 2 or 18 h in DMEM with 10%serum, fixed, and permeabilized as above in Karsenti’sBuffer for 1 min followed by an additional fixation in 0.5%

glutaraldehyde for 10 min, then washed and stored in phos-phate-buffered saline containing 0.1% Tween-20 and 0.02%Na azide. To visualize F-actin, cells were incubated for 15min in rhodamine-labeled phalloidin (Molecular Probes) fol-lowing the manufacturer’s instructions. Samples werewashed extensively with phosphate-buffered saline placedbetween a glass slide and glass coverslips after the additionof ImmunoFluor mounting medium (ICN Biomedicals) andsealed with nail polish. F-actin fluorescence was photo-graphed using an Olympus BX51 reflection fluorescencemicroscope equipped with a digital camera. Images wereacquired at 400� magnification as TIFF files and processedusing Paintshop Pro software. The images shown are repre-sentative of results seen in three separate experiments.

A five-point scale measuring the degree of actin stressfiber formation was used to semiquantify the extent of theF-actin stress fiber formation in fibroblast NIH 3T3 cells onadhering to different patterned surfaces. The criteria forblind scoring was as described previously16 where

1. Few or no resolved F-actin stress fiber formation andmostly cortical actin

2. Thin, short F-actin filament generally occupying atleast 25% of the cell volume

3. Moderate stress fiber formation of F-actin wherestress fibers are thicker and occupy at least 50% of thecell volume

4. Extensive stress fiber formation where stress fibers arethick and well-defined; many traverse the full widthof the cell

5. The entire cell is densely packed with thick stressfibers; most traverse the width of the cell.

Figure 1. Schematic diagram of the fabrication of heterogeneous topography.

464 TSAI ET AL.

A minimum of 50 cells were counted for each patternedsubstrate. The average degree of stress fiber formation andstandard error of the mean (SEM) are shown in the bargraphs. The technique was developed for scoring F-actinbundling in HeLa cells that make much fewer stress fibers.However, the approach can be applied to fibroblasts.

Field emission scanning electron microscopy (SEM)

NIH 3T3 fibroblasts, plated on surfaces and incubated in10% serum medium, were rinsed twice with serum-freeDMEM maintained at 37°C. The cells were then fixed with0.5% glutaraldehyde in serum-free DMEM also maintainedat 37°C for 30 min. The cells were then rinsed again twice for5 min with DMEM and postfixed with 1% osmium tetroxidefor 30 min. The cells were then rinsed three times for 10 mineach in serum-free DMEM. After postfixation, the cells weredehydrated by immersing in the following concentrations ofethanol for 2 min each: 20, 30, 40, 50, 60, 80, 90, 96, and 100%.The cells were then dried by critical point drying using aBalzers CPD030 Critical Point Dryer. The samples weremounted and coated with gold-palladium by sputter-coat-ing using a Polaron E5100 SEM coating unit. Finally, thesamples were imaged using SEM JSM-6320FXV with anaccelerating voltage of 5 kV.

RESULTS AND DISCUSSION

Cell spreading area

In general, the cells did not align with any preferredorientation, because the patterns generated by ho-mopolymers/diblock copolymer mixtures were ran-domly arranged on the surface. In Figure 2(A), cellspreading area was plotted as a function of time. Thecells were highly polarized and spread more on ahomogeneous surface of oxidized PS than on a homo-geneous surface of P(S-r-MMA) at all time periods[Fig. 2(A)]. This suggests that the oxidized PS is morefavorable for cell adhesion than P(S-r-MMA). This isfurther evidenced by the preferential attachment ofthe cells’ filopodia on the oxidized PS, as will bediscussed later. Overall, spreading area shows a linearincrease with time and begins to plateau after 90 min.Spreading area for the nanometer patterned surface isclosely aligned with the homogeneous oxidized PSsurface, as is the surface with the micron-sized pat-terns with the homogeneous P(S-r-MMA) surface. Inprevious work by Stockton and Jacobson,17 it wasshown that, after maximal cell spreading is achieved,there is a pronounced change in cell shape where thecells move from a more rounded phenotype (associ-ated with spreading) to a more elongated phenotype(associated with migration). In this study, we saw thesame trend of change in cell shape after 2 h.

There is a general increase in spreading area as thelateral correlation length decreases, as seen in Figure2(B). As the lateral spacing decreases, the oxidized PSsurface area increases. This translates into increased

Figure 2. (A) Average cell spreading area of cells culturedon patterned substrate of average length scales of 6 �m and60 nm as a function of time. Each data point is the averageof 150 cells of three experiments and five fields in eachexperiment. Error bars indicate SEM. (B) Average cellspreading area of cells cultured on patterned substrate ofaverage length scales of 6 �m to 60 nm at 30 min (squares),60 min (circles), 90 min (triangles), and 120 min (diamonds).Each data point is the average of 150 cells of three experi-ments and five fields in each experiment. Error bars indicateSEM. *Statistically significant deviations (p 0.05 usinganalysis of variance) from 6-�m pattern. **Statistically sig-nificant deviations (p 0.05 using analysis of variance) from60-nm pattern.

FIBROBLAST ADHESION TO VARIED SURFACES 465

favorable surface interactions with the cells and, con-sequently, the cells spread more on surfaces withsmaller, rather than larger, correlation lengths. Theincreased nonfavorable interactions, as is the case withsurfaces characterized by larger correlation lengths,result in rounder cells with reduced spreading area asshown in Figure 2(B). Interestingly, this trend is stron-ger for cells analyzed at 60 and 90 min where there isa strong correlation between cell spreading area andlateral correlation length scale. The same general trendcan be seen at 30 and 120 min. At 30 min, apparentdeviations can be attributed to cells not completelyattached and just beginning to spread. Cells on 60-nmpatterned surfaces have begun to spread, whereas thecells on the other patterns are still relatively similar toeach other, as evidenced by the distinct turning pointbetween 60- to 110-nm patterned surfaces. This sug-gests that the initial event for threshold cluster spacingis between 60 and 110 nm, which is close to the re-ported value of 140 nm of Massia and Hubbell18 and�60 nm of Maheshwari et al.19 Furthermore, thechanges in the cell shape can affect the measurementsof spreading area, particularly after 120 min, wherecells usually take on a migratory phenotype andchange from the rounded phenotype, characterizingactively spreading cells, to more elongated migratorycells.

Another parameter that may also contribute to thegeneral trend of increasing spreading area with de-creasing lateral length scale is topographic confine-ment. This appears to prohibit cell spreading into the“valleys” or depressions between raised surface areas.Consequently, the cell has less contact area with thesurface of P(S-r-MMA). The decreasing exposure tothe P(S-r-MMA) and increasing exposure to the oxi-dized PS increases preferential cell interaction withthe surface, translating to an increased cell spreadingarea. This hypothesis is consistent with Figure 2(A)where it is seen that cell spreading on the P(S-r-MMA)surface is comparable to the 6-�m patterned surfaces.This is also true for cells on the oxidized PS surfaceand the 60-nm patterned surfaces.

Degree of actin stress fiber formation

The formation of stress fibers in fibroblasts is oftenassociated with cellular immobilization on the extra-cellular matrix.16 Actin stress fiber formation was alsoexamined on heterogeneous surfaces characterized bydifferent lateral length scales. In general, the semi-quantitative analysis of the actin stress fiber formationat 2 h showed an increase in actin stress fibers as thelateral length scale of heterogeneities decreased, asshown in Figure 3. Additionally, there appeared to betwo main transitions, namely, one for surfaces with

lateral correlation length from 6-�m to 730-nm sur-faces and a second with surfaces with 110- to 60-nmfeatures.

The majority of cells either have no resolved stressfibers or thin and short actin filaments occupying atleast 25% of the cell volume on the 6-�m patternedsurface at 2 h. They also have a small number of axialstress fibers, as seen in Figure 4. Most of the F-actinfibers are distributed in the cell periphery as shortcortical structures, randomly oriented at the leadingedge. This could be the consequence of a low adhesionarea of the cells that bridge across the topography, asevidenced by the SEM images in Figure 5. With pat-terns �6 �m in size, there are fewer cells with little orno resolved stress fibers. For the cells on surfaces withpatterns having 730-nm down to 110-nm features,they have thin, short actin stress fibers, taking up atleast 25% of the cell volume and slightly more axialstress fibers than cells on the 6-�m patterned surface.Some of the F-actin fibers are short cortical polygonalfibers ending in filopodia. The second noticeablechange is that the cells on the 60-nm patterned sur-faces have a larger number of thick stress fibers, oc-cupying at least 50% of the cell volume. Some of theF-actin fibers are not elongated into axial stress fiber-supported filopodia. Many large lamellipodia haveshort stress fibers of cortical actins. The actin stressfibers of cells on 60-nm features are closer to the actinstress fibers of cells on a homogeneous oxidized, PSsurface, as shown in Figure 3.

By 18 h, cytoskeletons on all patterned surfaces have

Figure 3. Degree of actin stress fiber formation was used tosemiquantify the extent of the actin stress fiber formation infibroblast NIH 3T3 for different patterned surfaces. Thenumbers shown are average degree of actin stress fiberformation. A minimum of 50 cells were counted for eachpatterned substrate of three experiments. Error bars indicateSEM. * and **Statistically significant deviations (p 0.05using analysis of variance) between 0.11 �m and 0.58 �m.

466 TSAI ET AL.

Figure 4. Fluorescent images of fibroblast, NIH 3T3, cytoskeletons on various lateral length scale patterned surfaces at 2 hand 18 h. The images shown are representative of results seen in three separate experiments.

FIBROBLAST ADHESION TO VARIED SURFACES 467

matured with more actin stress fibers seen throughoutthe cells (Fig. 4). This increase in the number of actinstress fibers after 18 h of incubation suggests that thecells have deposited extracellular matrix proteins dur-ing this period. By this simple semiquantitative anal-ysis of the actin stress fibers, an increase in the numberof actin stress fibers is seen with decreasing laterallength, even after 18 h. Cells on the 6-�m patternedsurface at 18 h have elongated and axially orientedstress fibers. Some cells still have little or thin stressfibers on the 6-�m patterned surface. These cells havedisorganized filopodia. For cells on surfaces with 730-to 110-nm features at 18 h, the actin fibers have thick-ened and elongated, but with fewer axial stress fiberslacking conformity or alignment, and some periph-eral, angular stress fibers. Finally, cells on the 60-nmpatterned surface have extensive stress fibers that arewell-defined and traverse the full width of the cell.Few cells on this surface have densely packed thickstress fibers, traversing the entire cell. Again, the actinstress fibers for cells on the homogeneous oxidized PSare comparable to cells on the 60-nm patterned sur-face.

Additional studies focusing on cell migration onsurfaces with heterogeneity from the nanometer tomicron lateral length scale, show that cells tend to

migrate more readily on surfaces with micron-sized,rather than nanometer-sized heterogeneities.20 Thisobservation is consistent with previous work,17,21

where it was demonstrated that cells migrate faster,form fewer stress fibers, and have less spreading area.

Location of filopodia and lamellipodia

SEM showed numerous microvilli on the cell sur-face. Both filopodia and lamellipodia were observedon all patterned substrates. Most of the filopodialanded on the oxidized PS, if the feature sizes werelarger than the size of the filopodia [Fig. 5(C)]. Other-wise, the filopodia covered more than one surfacefeature with no apparent protrusions into the groovesof P(S-r-MMA) [Fig. 5(A,B)]. In addition, the edges ofthe lamellipodia also tended to conform or end at theoxidized PS posts, as shown in Figure 5(D).

CONCLUSIONS

Cell adhesion of NIH 3T3 fibroblasts on texturedsurfaces of oxidized PS and P(S-r-MMA), character-

Figure 5. SEM images of cells cultured on patterned substrates with average length scales of (A) 60 nm at 60,000 originalmagnification, (B) 110 nm at 80,000 original magnification, (C) 730 nm at 45,000 original magnification, and (D) 6 �m at 10,000original magnification.

468 TSAI ET AL.

ized by correlation lengths from the nanometer tomicron size scale, was examined. Fibroblasts adheremore favorably on the oxidized PS than P(S-r-MMA).Decreasing the lateral length scales between the oxi-dized PS features increased cell spreading area andactin stress fiber formation. Both chemistry and geom-etry can limit cell interaction with the P(S-r-MMA).This is evidenced by the cell spreading area and actinstress fiber formation. These findings are relevant tothe fundamental understanding of cell–substrate in-teractions and development for the surfaces of medi-cal devices.

The authors thank Louis Raboin and Dale A. Callaham forassistance with SEM and the sample preparation for SEM.

References

1. Assender H, Bliznyuk V, Porfyrakis K. How surface topogra-phy relates to materials properties. Science 2002;297:973–976.

2. Desai TA. Micro- and nanoscale structures for tissue engineer-ing constructs. Med Eng Phys 2000;22:595–606.

3. Curtis A, Wilkinson C. Topographical control of cells. Bioma-terials 1997;18:1573–1583.

4. Tirrell M, Kokkoli E, Biesalski M. The role of surface science inbioengineered materials. Surf Sci 2002;500(1–3):61–83.

5. Flemming RG, Murphy CJ, Abrams GA, Goodman SL, NealeyPF. Effects of synthetic micro- and nano-structured surfaces oncell behavior. Biomaterials 1999;20:573–588.

6. Fairman K, Jacobson BS. Unique morphology of HeLa cellattachment, spreading and detachment from microcarrierbeads covalently coated with a specific and non-specific sub-stratum. Tissue Cell 1983;15(2):167–180.

7. Dalby MJ, Riehle MO, Johnstone H, Affrossman S, Curtis SG.In vitro reaction of endothelial cells to polymer demixed nano-topography. Biomaterials 2002;23:2945–2954.

8. Shirato I, Tomino Y, Koide H, Sakai T. Fine structure of theglomerular basement membrane of the rat kidney visualizedby high resolution scanning electron microscopy. Cell TissueRes 1991;266:1–10.

9. Yamasaki Y, Makino H, Ota Z. Meshwork structures in bovineglomerular and tubular basement membrane as revealed byultra-high resolution scanning electron microscope. Nephron1994;66:189–199.

10. Bruder F, Brenn R. Spinodal decomposition in thin films of apolymer blend. Phys Rev Lett 1992;69:624–627.

11. Ermi B, Karim A, Douglas JF. Formation and dissolution ofphase-separated structures in ultrathin blend films. J Polym SciPart B: Polym Phys 1998;36:191–200.

12. Mansky P, Russell TP, Hawker CJ, Pitsikalis M, Mays J. Or-dered diblock copolymer films on random copolymer brushes.Macromolecules 1997;30:6810–6813.

13. Mansky P, Huang E, Liu Y, Russell TP, Hawker CJ. Controllingpolymer-surface interactions with random copolymer brushes.Science 1997;275:1458–1460.

14. Autocorrelation analysis performed on a PC computer usingthe public domain NIH Image program (developed at the U.S.National Institutes of Health and available on the Internet athttp://rsb.info.nih.gov/nih-image/about.html); Scion soft-ware.

15. Downloadable Image J software from http://rsb.info.nih.gov/ij/index.html.

16. Glenn HL, Jacobson BS. Arachidonic acid signaling to thecytoskeleton: the role of cyclooxygenase and cyclic AMP de-pendent protein kinase in actin bundling. Cell Motil Cytoskel-eton 2003;55:265–277.

17. Stockton RA, Jacobson BS. 2001. Modulation of cell-substrateadhesion by arachidonic acid: lipoxygenase regulates cellspreading and ERK1/2-inducible cyclooxygenase regulatescell migration in NIH-3T3 fibroblasts. Mol Biol Cell 2001;12:1937–1956.

18. Massia SP, Hubbell JA. An RGD spacing of 440 nm is sufficientfor integrin alpha V beta 3-mediated fibroblast spreading and140 nm for focal contact and stress fiber formation. J Cell Biol1991;114:1089–1100.

19. Maheshwari G, Brown G, Lauffenburger DA, Wells A, GriffithLG. Cell adhesion and motility depend on nanoscale RGDclustering. J Cell Sci 2000;113:1677–1686.

20. Kimura M, Tsai IY, Green JA, Jacobson BS, Russell TP. Fibro-blast migration on heterogeneous topography using diblockcopolymers and homopolymers. Unpublished.

21. Stockton RA, Jacobson BS. 5-Lipoxygenase and cyclooxygen-ase-2 regulate fibroblast–matrix adhesion through differentialactivation of RAC and RHO. In preparation.

FIBROBLAST ADHESION TO VARIED SURFACES 469