lable at ScienceDirect

Biomaterials 30 (2009) 6825–6834

Contents lists avai

Biomaterials

journal homepage: www.elsevier .com/locate/biomateria ls

Influence of micro-well biomimetic topography on intestinal epithelialCaco-2 cell phenotype

Lin Wang a, Shashi K. Murthy a, William H. Fowle b, Gilda A. Barabino c, Rebecca L. Carrier a,*

a Chemical Engineering Department, Northeastern University, 360 Huntington Ave., 342 Snell Engineering, Boston, MA, 02115, USAb Biology Department, Northeastern University, Boston, MA, 02115, USAc Biomedical Engineering Department, Georgia Institute of Technology, Atlanta, GA, 30332, USA

a r t i c l e i n f o

Article history:Received 7 July 2009Accepted 27 August 2009Available online 19 September 2009

Keywords:Biomimetic topographySmall intestinal epithelial cell phenotypeScaffold for tissue engineering

* Corresponding author. Tel.: þ1 617 373 7126; fax:E-mail address: r.carrier@neu.edu (R.L. Carrier).

0142-9612/$ – see front matter � 2009 Elsevier Ltd.doi:10.1016/j.biomaterials.2009.08.046

a b s t r a c t

A microfabrication approach was utilized to create topographic analogs of intestinal crypts on a polymersubstrate. It was hypothesized that biomimetic crypt-like micro-architecture may induce changes insmall intestinal cell (i.e. Caco-2 cell) phenotype. A test pattern of micro-well features with similardimensions (50, 100, and 500 mm diameter, 50 mm spacing, 120 mm in depth) to the crypt structuresfound in native basal lamina was produced in the surface of a poly(dimethylsiloxane) (PDMS) substrate.PDMS surfaces were coated with fibronectin, seeded with intestinal-epithelial-cell-like Caco-2 cells, andcultured up to fourteen days. The cells were able to crawl along the steep side walls and migrated fromthe bottom to the top of the well structures, completely covering the surface by 4–5 days in culture. Thetopography of the PDMS substrates influenced cell spreading after seeding; cells spread faster and ina more uniform fashion on flat surfaces than on those with micro-well structures, where cell protrusionsextending to micro-well side walls was evident. Substrate topography also affected cell metabolic activityand differentiation; cells had higher mitochondrial activity but lower alkaline phosphatase activity atearly time points in culture (2–3 days post-seeding) when seeded on micro-well patterned PDMSsubstrates compared to flat substrates. These results emphasize the importance of topographical designproperties of a scaffolds used for tissue engineered intestine.

� 2009 Elsevier Ltd. All rights reserved.

1. Introduction

The inner surface of the intestine contains a monolayer ofenterocytes resting on a basement membrane; this epithelial layerhas a convoluted topography consisting of finger-like projections(villi) with deep well-like invaginations (crypts) between the villi.The dimensions of crypts and villi are on the order of hundreds ofmicrons (100–700 mm in height, and 50–250 mm in diameter) [1,2].The migration, proliferation, differentiation, and function of smallintestinal epithelial cells vary with position relative to these struc-tures. The differentiation of enterocytes into absorptive cells takesplace as they migrate upward from the crypt to the villus [3], sug-gesting that the crypt-villus micro-environment may play animportant role in cellular differentiation in the intestine. Thus far, thefunction of the crypt-villus micro-environment in regulating intes-tinal cell proliferation and differentiation has been poorly under-stood. The unique geometry of the crypt-villus micro-environmentmay play a role in controlling intestinal cell function. Specifically, the

þ1 617 373 2209.

All rights reserved.

crypt-villus topography may lead to different mechanical stressdistributions within the epithelial cell layer covering the extracel-lular matrix (ECM). Topography also may affect the distribution ofchemical factors, such as cell growth hormones and cell signalingmolecules along the crypt-villus axis on the apical side of theintestinal epithelium. The tight spacing within crypts and betweenvilli likely creates an environment of diffusion-limited transport withassociated concentration gradients of chemical factors, and prox-imity of apical cell surfaces may facilitate cell–cell signaling. There-fore, we hypothesized that crypt micro-architecture (micro-wells)may induce changes in intestinal epithelial cell phenotype in vitro.

It is generally accepted that cultured mammalian cells respondto topographical cues, and their response depends on cell type,feature size, shape, and physical and chemical properties ofsubstrate material [4–6]. A number of researchers have found thatcells (e.g. fibroblasts, macrophages, mesenchymal tissue cells,epithelial cells) will align with multiple grooves on a substrate[7–11]. The degree of cell alignment depends on the cell type, thesurface material, groove density, groove width, and groove depth. Insome cases, response to grooves has been quantified, for example, itwas found that cells responded to grooved surface topography as anautomatic controller, with cell migration on a microgroove surface

L. Wang et al. / Biomaterials 30 (2009) 6825–68346826

directly proportional to the square of the product of groove heightand spatial frequency, when groove height was in the range of 25–200 nm and spatial frequency was between 100 and 500 mm�1 [10].Pillar and well/pit structures have been much less studied comparedto groove/ridge structures [4,6,12]. Green et al. [13] found thatfibroblasts cultured on silicone substrates patterned with micro-pillar and micro-well arrays (2 and 5 mm in diameter 0.5 mm inheight) showed an increased rate of proliferation and cell density.Chehroudi et al. [14] studied bonelike tissue formation on subcu-taneous titanium-coated epoxy implants patterned with eithergroove or pit structures ranging from 30 to 120 mm in depth. It wasreported that both pit and groove patterns promoted bonelike tissueformation, however, pitted surfaces enhanced bone formation astheir depth increased, while bone formation on grooved surfacesfollowed the opposite trend, suggesting shape of surface patternmight significantly alter cell behavior.

Relative to nano-scale or single-cell-scale topography, there havebeen few attempts to study the effect of topography on the scale offunctional subunits of tissue, around 100 mm in size, on cellbehavior. As noted above, the intestine has particularly interestingtopography on this length scale. Reports on subunits scale topo-graphy are limited to groove/ridge structures, and tissues such asbone and skin, but not intestinal tissues [4,14,15]. The present studyaddresses the influence of micron-scale (w100 mm) crypt-likemicro-well topography on intestinal cell behavior to determinewhether the unique topography of intestinal crypt-villus unit playsan important role in regulating intestinal cells. Caco-2, a humancolon carcinoma cell line which exhibits features similar to humanintestinal epithelium, was cultured on PDMS substrates withbiomimetic topography (micro-well arrays). It was found thatintestinal crypt-like topography affect Caco-2 cell spreading,migration, metabolic activity, as well as cell differentiation. Thebroad significance of this work lies in understanding critical designelements for biomimetic scaffolds that can ultimately be used inintestinal tissue engineering.

2. Materials and methods

2.1. Fabrication of micro-patterned poly(dimethylsiloxane) (PDMS) substrates

To fabricate a patterned PDMS replica, a master SU-8 mold was created viaphotolithography. Briefly, 3 square test regions, each containing an array of circlescorresponding to the designed diameters, were drawn using AutoCAD software andprinted with high resolution on a transparency. The transparency was utilized asa photomask to create an SU-8 master mold with micro-pillar features using contactprint UV lithography at G.J. Kostas Micro and Nanofabrication Facility at North-eastern University [16]. The height of the columns on the SU-8 molds was measuredby a DEKTAK 3ST profilometer. A patterned PDMS replica with micro-well featureswas fabricated by pouring liquid PDMS (Sylgard 184�, Dow Corning) pre-polymer(10:1 base to curing agent weight ratio) onto the SU-8 master mold. After degassingfor 30 min and curing at 65 �C for 2 h, the solidified PDMS replica was peeled offfrom the SU-8 mold, producing the final patterned PDMS cell culture substrates.Micro-well depth was again measured by profilometry.

Patterned PDMS substrates were affixed to the bottom of wells of either 12 or 96well plates using uncured PDMS. Plates were then heated in oven at 65 �C for 2 h toallow scaffolds to completely adhere to the bottoms of the wells. For all the tests, flatPDMS substrate and cell culture treated polystyrene (PS) substrate were used ascontrols. All cell culture substrates were sterilized by immersing in 70% (v/v) over-night, followed by washing extensively with phosphate buffered saline (PBS, Sigma).PDMS has very poor cell adhesion properties due to its hydrophobicity [17]. Ourprevious study has found presorption of 50 mg/ml fibronectin (Fn) is sufficient topromote Caco-2 adhesion on PDMS to a level equivalent to that observed on cellculture treated PS. Therefore, before seeding cells, all substrates were coated with50 mg/ml fibronectin in PBS for 2 h at room temperature.

2.2. Cell culture

A human colon carcinoma cell line, Caco-2, was obtained from the American TypeCulture Collection (ATCC) and cultivated in Eagle’s minimum essential medium(ATCC) supplemented with 20% fetal bovine serum (FBS, ATCC) and 1% antibioticantimycotic solution (containing 10,000 units/ml penicillin G, 10 mg/ml

streptomycin sulfate and 25 mg/ml amphotericin B; Sigma). Confluent monolayerswere subcultured by incubating with 0.05% trypsin (Sigma) and 0.2% ethyl-enediaminetetraacetic acid (EDTA, Sigma) in Ca2þ- and Mg2þ-free PBS. Cultures wereincubated at 37 �C in a humidified atmosphere of 95% air, 5% CO2. For all experiments,cells were seeded at low density (2�104 cells/cm2) onto test surfaces and culturedfor specified time periods. Medium was aspirated and replaced after 2 days of seedingand every 2 days in culture.

2.3. F-actin staining (cytoskeletal observation)

Caco-2 cell morphology was assessed by staining with fluorescein phalloidin(Molecular Probes). Caco-2 cells were fixed in 3.7% formaldehyde (Sigma) in PBS for10 min at room temperature, washed twice with PBS, and permeabilized by 0.1%Triton X-100 in PBS for 5 min. After two washes with PBS, cells were incubated with0.16 mm phalloidin in 1% BSA/PBS for 20 min at room temperature. Cells werecounterstained with 5 mg/ml Hoechst 33 258 (Molecular Probes) in PBS for another20 min at room temperature. Cell cytoskeleton was observed using a fluorescencemicroscope (Olympus �51).

2.4. Image analysis

ImageJ software (http://rsb.info.nih.gov/ij/) was utilized to determine cellcoverage on flat and patterned PDMS substrates, and cell culture treated PS surfaces.Cell coverage was determined from F-actin stained images by measuring stainedarea and dividing by the entire area in a given field of view. It is acknowledged thatmethod does not consider the cell coverage on side walls of well structures.

2.5. Scanning Electron Microscopy (SEM)

Cells were washed twice with Hank’s Balanced Salt Solution (HBSS, Sigma) at37 �C, and then fixed with 3% glutaraldehyde (Sigma) in 0.1 M sodium cacodylate(Sigma), pH 7.4, containing 0.1 M sucrose (Sigma) for 45 min at room temperature.After fixation, the cells were washed in 0.1 M sodium cacodylate, pH 7.4, containing0.1 M sucrose and dehydrated through graded changes of ethanol solutions (35%,50%, 70%, 95%, and 100%). Cells were then dried using a Tousimis PVT-3 critical pointdryer. Dried samples were mounted on aluminum stubs and sputter coated withgold-palladium. The cells were examined using a Hitachi S4800 field emission SEM.

2.6. MTT assay (metabolic activity)

The MTT assay is used to estimate cellular metabolic activity; NADH- or NADPH-dependent mitochondrial dehydrogenases inside viable cells are able to reduce theyellow MTT tetrazolium salt to blue-purple formazan crystals. The amount of thecrystals formed reflects metabolic activity of cells at the end of essay period [18,19].For each sample, after the cells were incubated on patterned PDMS for 3 or 6 days,the medium was gently removed and replaced with 0.09 ml of phenol red-freeEagle’s minimum essential medium (Sigma). 0.01 ml of 5 mg/ml MTT (3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyl tetrazolium bromide) solution (Sigma) was thenadded to each sample located in each well of a 96 well plate. The cell cultures wereincubated at 37 �C for 3 h. The formazan crystals generated during the incubationperiod were dissolved by adding 0.1 ml of MTT solubilization solution (10% TritonX-100 plus 0.1 N HCl in anhydrous isopropanol, Sigma) and gently mixing thesolution by trituration. After the crystals were fully dissolved, the absorbances of thesolutions at 570 nm (OD570) were measured using a spectrophotometer. Cell culturemedium was used as a control. The MTT results are presented below as opticaldensity at 570 nm relative to total number of cells, which was measured bya CyQUANT� cell proliferation assay kit from Molecular Probes, as described below.

2.7. Cell counting assay

A CyQUANT� cell proliferation assay kit was utilized to count total cell numberin each sample. Briefly, cells were detached from each PDMS substrate by incubatingwith 0.05% trypsin and 0.2% EDTA in Ca2þ- and Mg2þ-free PBS for 20 min. An excessculture medium was added to stop the trypsin reaction, and PDMS substrates weregently removed from the bottom of cell culture plate, inverted, and immersed inmedium for 10 min, ensuring the cells trapped inside wells to settle into medium bygravity. The resulting cell suspension was then collected, washed twice with PBS,and concentrated by centrifuging for 5 min at 200� g. 1 mL of the CyQUANT� GRdye/cell-lysis buffer was added to the cell pellet and incubated for 5 min at roomtemperature. Fluorescence of the samples was measured using a fluorometer with480 nm excitation and 520 nm emission. A reference standard curve for convertingsample fluorescence values into cell numbers was prepared by counting cells usinga hemacytometer and measuring fluorescence intensity of known cell densitysamples.

2.8. Alkaline phosphatase activity assay

The alkaline phosphatase activity assay was performed on cell lysates. Justbefore conducting the assay, samples were washed twice with PBS and lysed by

L. Wang et al. / Biomaterials 30 (2009) 6825–6834 6827

CelLytic� M cell lysis/extraction reagent (Sigma) for 15 min at 37 �C. PDMSsubstrates were detached from the bottom of cell culture plates and inverted tofacilitate removal of cell contents inside micro-wells. Cell lysates were then collectedand subsequently centrifuged for 15 min at 16,000� g. Protein-containing super-natant was collected and divided into two parts. One part was used for the alkalinephosphatase activity assay and the other part was used to measure total protein bythe bicinchoninic acid (BCA, Pierce) assay. The alkaline phosphatase assay wasperformed utilizing an alkaline phosphatase detection kit (Sigma). For this assay,4-methylumbelliferyl phosphate is used as a fluorogenic substrate for alkalinephosphatase. The product, 4-methylumbelliferone, has intense fluorescence at pH10.3, with 360 nm excitation and 440 nm emission. Briefly, 1 ml of 10 mM 4-meth-ylumbelliferyl phosphate disodium (substrate), 20 ml of dilution buffer, and 160 ml offluorescent assay buffer were added to 20 ml cell lysate. After addition of substrate,samples were analyzed every 5 min for a total of 30 min using a fluorometewith360 nm excitation and 440 nm emission. The fluorescence intensity after 30 minincubation was used in data analysis. For total protein analysis, a BCA� protein assaykit (Pierce) was utilized. Briefly, 200 ml of BCA working reagent (BCA� reagent A andreagent B at 50:1 ratio) were added to 25 ml cell lysate, and samples were thenincubated at 37 �C for 30 min. After cooling to room temperature, the absorbance ofsamples at 562 nm was measured on a plate reader. Diluted bovine serum albumin(BSA) was used to prepare a standard curve. Results of alkaline phosphatase analysiswere expressed as fluorescence intensity per mg of protein.

2.9. Statistical analysis

Results are presented as mean� SEM. Statistical significance of differences wasanalyzed by oneway ANOVA followed by post hoc Tukey multiple comparison test.A value of p< 0.05 was regarded as statistically significant.

3. Results

3.1. Microfabrication of SU-8 master mold and PDMS substrateswith cypt-like topography

A microfabrication technique was used to re-create the micro-topography (crypt-like structures) of the native human smallintestinal epithelium on PDMS scaffolds. A test pattern of columns

Fig. 1. Design and microfabrication of a SU-8 master chip and a PDMS negative replicate. (A)scaffolds with micro-well structures, top-view (B) and cross-section (C).

(diameters: 50, 100, or 500 mm; wall to wall spacing: 50 mm;heights: 120 mm) was patterned into the surface of an SU-8 mastermold using photolithography (Fig. 1A). Profilometry analysis(DEKTAK 3ST profilometer) indicated that the height of thecolumns was around 120 mm. Negative replicates of the SU-8 moldwere produced using PDMS (Fig. 1B,C). The surfaces of PDMSsubstrates were comprised of micro-well arrays which mimic thetopography of small intestinal crypts.

Light microscopy and SEM were used to visually analyze topog-raphy of SU-8 molds and PDMS substrates. The images of the PDMSscaffolds demonstrated that the micro-wells had smooth side wallsand flat bottoms. However, an expansion of diameter was observedfrom the top to the bottom of each well, most likely as a result of theSU-8 fabrication procedure (Fig.1C). The dimensions and topographyof the patterns on the PDMS surfaces are shown in Fig. 1B, C. Thecircle patterns have a pitch distance (wall to wall) of 50 mm, whereasthe diameters of the circle patterns were 50, 100 and 500 mm,respectively. The depth of the patterns on PDMS substrates weremeasured by imaging the substrate cross-section by SEM. It wasfound that the depth of the PDMS micro-wells was approximately120 mm via SEM; this value was confirmed by profilometry.

3.2. Effect of crypt-like topography on Caco-2 cell morphology

Cell cytoskeletal staining of F-actin (Fig. 2 and Fig. 3) illustratesthe details of cell distribution and shape on patterned surfaces.When initially seeded, the majority of cells settled inside wells.After 1 day incubation, cells dwelling inside of 50 mm (Fig. 2B) and100 mm size wells were mostly still rounded. Meanwhile, cellsinside 500 mm wells were partially spread out (data not shown),and cells on flat PDMS were widely spread to a size of approxi-mately 100 mm (Fig. 2A). Quantitative analysis of cell coverage in 3

Light microscope images of SU-8 mold with micro-column arrays. SEM images of PDMS

Fig. 2. F-actin staining of Caco-2 cells cultured on flat PDMS substrates (A), and on PDMS substrates patterned with 50 mm well arrays for 24 h (B).

L. Wang et al. / Biomaterials 30 (2009) 6825–68346828

day Caco-2 cell cultures on patterned and flat surfaces indicatesthat the presence of micro-well topography delays cell spreadingand/or proliferation (Fig. 3A). The lag phase before cell spreading islonger for cells cultured on surfaces with smaller diameter wells.For example, after being incubated for 2 days, the coverages of cellson surfaces with 50 mm, 100 mm, 500 mm wells, flat PDMS, and PSwere 18.9%, 28.4%, 49.6%, 90%, and 77.5%, respectively. It was foundthat Caco-2 cells were able to conformally cover well structuresafter 3 days in culture (Fig. 3B,C), and cover the entire cell culturesubstrates after 5 days in culture (Fig. 3D,E).

In addition to the extent of cell spreading and substratecoverage, cell shape was also influenced by surface topography(Fig. 4). Cell spreading was more radially non-uniform when cellswere located inside wells. Cells tended to stretch out theirprotrusions or micro-spikes towards the side walls of the wellstructures. The formation of distinctive protrusions towards theside walls was observed on all patterned surfaces (i.e. 50 mm,100 mm, and 500 mm), while cells on flat surfaces spread evenly inall directions. In particular, as cells located inside of 50 mm wellsbegan to spread, they had much longer and more distinct protru-sions than cells cultured on surfaces with larger wells (i.e. 100 mmand 500 mm) and flat surfaces (Fig. 4).

SEM enabled observation of cell structure and migration relativeto micro-well structures. Depending on their initial seeding

Fig. 3. Caco-2 cell coverage on polystyrene (PS), flat PDMS (PDMS), and surfaces patterned w1, p< 0.0001; for day 2, p< 0.0001; for day 3, p¼ 0.006< 0.05. * represents statistical diffegroups which are statistically different from each other (A). Cells started to cover the wells asubstrate after 5 days in culture (D: 50 mm wells, E: 100 mm wells). The yellow arrow indicawall. Note the focal plane in (D) and (E) was at the top of the well; cells at the bottom of w

location relative to well structures, Caco-2 cells have differentmigration patterns. Cells located at the bottom corners of micro-wells 24 h after seeding tended to extend their protrusions to theside walls, bypassing sharp corners, suggesting a tendency ofmigrating up along side walls. This behavior is illustrated in Fig. 5Afor 500 mm micro-wells but was observed on all patterned surfacesfor approximately 100% of the cells settled at the bottom corners ofwells. Cells located on either the top flat surface or top corners ofwells 24 h after seeding appeared to be extended conformally alongthe substrate surface (Fig. 5B, C). For small diameter wells (i.e.50 mm and 100 mm) cells adjacent to the top boundaries of wellswere sometimes found to span over wells instead of extendingalong the side wall of the well structure and spreading into thewells (Fig. 5D).

The process of cells crossing over wells appeared to proceed viadefined stages (Fig. 6). Initially, cells close to the boundaries of wellsextended protrusions to sense well structure (Fig. 6A,D). Theextended protrusions aligned and extended along the brims of thewells (Fig. 6B, E), and eventually covered the wells (Fig. 6C,F).The nature of the cell cross over process depended upon the time ofculture and the size of the micro-wells (Fig. 7). The phenomenonwas more prominent after prolonged incubation and with increasedcell density (Fig. 7 L, M). Cells formed different cross-well structureswhen cultured for 2 days (Fig. 7A–D), or 3 days (Fig. 7E–K) on

ith 50, 100, and 500 mm well arrays after 1, 2, and 3 days in culture. ANOVA test for dayrence from other groups sampled on the same day; a, b, c, and d each represent twofter 3 days in culture (B: 50 mm wells, C: 100 mm wells), and covered entire cell culturetes a well covered by cells while the, white arrow represents cells attached to the side

ells were therefore out of focus.

Fig. 4. F-actin staining of Caco-2 cytoskeleton demonstrating interactions with flatpolystyrene (PS) surfaces, flat PDMS surfaces and side walls of 50, 100, and 500 mmwell structures 24 h after seeding. The scale bar corresponds to 30 mm.

Fig. 5. SEM images of Caco-2 spreading and migrating on micro-patterned PDMS 24 h afteron the top (flat) surface between wells (500 mm in diameter, B), and adjacent to the top bo

L. Wang et al. / Biomaterials 30 (2009) 6825–6834 6829

surfaces patterned with 50 mm wells (Fig. 7A, B, I) or 100 mm wells(Fig. 7C, D, E, F, G, H, J, K). Caco-2 cells were observed to cross over50 mm wells starting from day 1 in culture, while cells started cross100 mm wells from day 2 in culture. It was also found that 50 mmwells were covered either by a single cell (Fig. 7A) or multiple cells(Fig. 7B), 100 mm wells could only be covered by multiple cells(Fig. 7D, H, K). Single cells located on the top of 100 mm wellsappeared to preferentially spread and migrate between wellsinstead of across a well (Fig. 7C). After 3 days of incubation, once celldensity increased and substrate surfaces were almost completely(>90%) covered by cells, cell-cell contact spanning across 50 mm and100 mm wells was observed (Fig. 7E, F, G, I, J). Cells elongated andspanned across wells to establish contact with cells located at theopposite side of the well (Fig. 7E, I), or cells located on the oppositesides of wells extended long protrusions to build contact with eachother (Fig. 7G, J). It also appeared that cells spanning across wellsexpressed more actins (Fig. 7H), as more fine actin filamentsextending towards cell nuclei were observed. Once the substratesurfaces were completely covered by cells (about 5 days in culture),the fraction of wells bridged by cells also reached a plateau (Fig. 7 L,M). More than 30% of wells were found to be completely covered bycells in the case of 50 mm wells after 5 days in culture. About 30% ofwells were bridged by cells cultured on 100 mm wells, while lessthan 10% of wells were completely covered by cells after 6 days inculture. Cells were not found to span across 500 mm wells.

3.3. Effect of crypt-like topography on Caco-2 cell metabolic activity

Cells growing on patterned surfaces have significantly highermetabolic activity per cell than cells growing on flat surfaces earlyin culture (day 3 after seeding) (Fig. 8). The differences in metabolicactivity among different surfaces are no longer apparent by Day 6 inculture. Interestingly, at Day 3, cells cultured on surfaces with100 mm wells, which have dimensions closest to native humansmall intestinal crypts, had the highest metabolic activity. Theresults suggest that cells are more metabolically active on micro-well patterned surfaces when initially seeded.

3.4. Effect of crypt-like topography on Caco-2 cell differentiation

In the native small intestine, epithelial cell differentiation isdependent on the location of cells along the crypt-villus axis. Todetermine whether topological cues contribute to cell differentiationin vitro, expression of an enterocyte differentiation marker, alkalinephosphatase, was assessed. For all substrates, the alkaline phospha-tase activity increased with incubation time, and tended to level offbetween Day 8 and 14 in culture (Fig. 9). Compared to cells grown on

initial cell seeding: cells adjacent to a bottom corner of a well (500 mm in diameter, A),undary of a well (500 mm in diameter, C; 50 mm in diameter, D).

Fig. 6. F-actin staining (A,B,C) and SEM (D,E,F) of Caco-2 cells interacting with 50 mm well structures as they spread across the wells, 24 h after initial seeding. The scale barcorresponds to 25 mm.

L. Wang et al. / Biomaterials 30 (2009) 6825–68346830

patterned PDMS surfaces, cells grown on flat PDMS surfaces hadslightly higher alkaline phosphatase activity on Day 2 and Day 8, andreached maximal activity at an earlier time point (around 8 days)versus 14 days for cells grown on patterned PDMS. Among cellscultured on patterned surfaces, cells on surfaces with smaller sizewells had higher alkaline phosphatase activity than cells on largersize wells, as cells cultured on surfaces with 500 mm wells had thelowest alkaline phosphatase activity at day 8.

4. Discussion

It has become increasingly evident that cell behavior is influ-enced by topology of cell culture substrates or scaffolds [4,20,21].Benefitting from well-developed micro- and nano-fabricationmethods, various groups have probed important questions relatedto cellular response to topography, including: whether cellsrespond to topographic features using the same sensory systemused to sense changes in surface adhesiveness; whether the sizeand the shape of scaffolds may affect the nature of cell–cell inter-actions; whether the cellular response to topological cues varieswith the shape and dimension of the structure [20]; whether thebasement membrane topology plays a role in coordinating tissuefunction at a molecular level, rather than simply providing a phys-ical barrier or acting as a support or a reservoir for growth factorsand other molecules [21]; and whether microtopology affects localadhesion of cell binding proteins [22].

Small intestinal ECM consists predominantly of interwovenprotein fibers such as collagen or elastin that are 10–300 nm indiameter. These meshed fibers provide tensile strength as well asultra-structure to the basement membranes, while other glycopro-teins (such as fibronectin, and laminin), proteoglycans, and glycos-aminoglaycans form hydrated gels which function to resist

compressive forces [23]. Topographically, small intestinal ECM hasmany architectural scales. This hierarchical architecture accommo-dates a complex spectrum of property requirements for intestinalfunction [24]. The finest scale of topography consists of a heteroge-neous mixtures of pores, ridges and fibers which have sizes in thenanometer range [4]. The ECM sheets are folded and bent, creatinga micrometer range secondary structure. Most studies to date ontopographical effects of cell culture substrates are focused on surfacestructures less than a micrometer in size [25–28]; little has been doneto investigate the influence of micrometer range topography(secondary structure) on cell behavior. There are some studies,however, that have indicated that micrometer range surface topo-graphy may alter cell function [15,29], in accordance with resultspresented here.

We hypothesize that the crypt-villus micro-architecture mayinduce changes in cell phenotype. Micrometer scale structure canaffect cells by exerting different mechanical stresses, physicalbarriers, and chemical stimuli on them as they spread and migrate.For anchorage dependent cells, such as small intestinal epithelialcells, cell-matrix contacts link to the cytoskeleton and signalingpathways, which in turn regulate cell motility, morphology,proliferation, survival, and differentiation.

In the present work, it was observed that the majority (>90%) ofCaco-2 cells initially deposited on the micro-well patternedsubstrates fell inside the micro-wells. For 50 mm wells, most wellscontained a single cell, or possibly two cells. These results suggestthat the micro-well structure is able to trap, segregate, and confinecells. A similar passive cell trapping effect of micro-well structuresformed from poly(ethylene glycol) on glass (30 mm in diameter,approximately 20 mm deep) has also been reported; thesesubstrates were seeded with NIH 3T3 fibroblasts and rat hepato-cytes [30]. A PDMS substrate patterned with micro-wells

Fig. 7. F-actin cytoskeleton staining and SEM of Caco-2 cells during the process of crossing over the wells. In the cases of PDMS surfaces patterned with 50 mm well, wells can becovered by a single cell (A), or multiple cells (B) after 48 h of culture; on 100 mm well patterned surfaces, a single cell is not able to cross over well structures (C); the crossing over ofa well can be only achieved by multiple cell aggregates (D). The crossing over of wells is more prominent after 3 days of cell culture, when the cell density is high. The scale bar in(A)–(H) represents 50 mm, while that in I, J, and K represents 20 mm. Fractions of wells (i.e. 50 mm and 100 mm wells) which were bridged (L) or completely covered (M) by Caco-2cells were quantified based on manual counts of 3 substrates with w 1000 wells per substrate.

Fig. 8. Metabolic activity of Caco-2 cells incubated on polystyrene (PS), flat PDMS (PDMS), and surfaces patterned with 50, 100, and 500 mm well arrays for 3 days (A) and 6 days (B).ANOVA test for day 3, p< 0.0001, ‘‘a’’ represents statistical difference from PS, ‘‘b’’ stands for statistical different from PDMS, ‘‘c’’ stands for statistical different from 50 mm, ‘‘d’’ standsfor statistical different from 100 mm, and ‘‘e’’ stands for statistical different from 500 mm. ANOVA test for day 6, p¼ 0.276> 0.05.

L. Wang et al. / Biomaterials 30 (2009) 6825–68346832

(20–50 mm in diameter, approximately 20 mm deep) and seededwith rat basophilic leukemia (RBL-1) cells exhibited a similarphenomenon [30,31]. These studies focused mainly on cell distri-bution and initial cell spreading after 24 h in culture. They did not,however, investigate the influence of the topography on cellphenotype, and did not study the long-term effect of micro-wellstructure on cell behavior.

Newly-attaching cells undergo a sequential conversion fromforming spike-like protrusions to developing contractile focaladhesions to establishing contact with EMC. Initially, cells explorethe surrounding environment beyond the cell body using filopodiaand spikes, which allow cells to maintain a polarized leading edgeduring their migration. As cells continue spreading, lamellipodiaand associated ruffles start to form, which are followed by thedevelopment of contractile focal adhesions [32]. 24 h after seeding,cells located inside micro-well structures, especially small sizedwells (i.e. 50 mm and 100 mm), were still at the protrusion contactstage. Meanwhile, cells growing on flat surfaces were fully spread,indicating that the presence of micro-well structures hinders cellspreading. The delay of spreading was more prominent for cellsdwelling in small sized wells. The deep and narrow well structures

Fig. 9. Differentiation of Caco-2 cells (alkaline phosphatase activity) incubated onpolystyrene (PS), flat PDMS (PDMS), and surfaces patterned with 50, 100, and 500 mmwell arrays for 2, 8, and 14 days. ANOVA test among different surfaces: p¼ 0.039 forday 2, p¼ 0.038 for day 8, and p¼ 0.045 for day 14. a, b, and c each represent twogroups which are statistically different from each other.

might constrain cells by creating physical barriers (the walls of thewell structures) for cell spreading, or by limiting the diffusion ofregulatory molecules.

Micro-well structures also restrict cell-cell contact; as demon-strated in our study, 50 mm wells usually contain single cells whichare physically separated from cells dwelling in other wells. Theisolation of cells and consequent lack of cell-cell contact mightsuppress migration. It has been noted by other researchers thatisolated cells are poor migrators [33,34]. For example, Middletonet al. [34] observed that embryonic chicken pigmented retinalepithelial (PRE) cells barely migrated as single cells; however, afterthey contacted other cells and formed cell-cell adhesions, themulti-cell sheet formed broad lamellapodia and attempted tomove. The work of Nelson et al. [35] and Revin et al. [30] suggestedthat direct cell-cell contact promoted cell proliferation. It wasdemonstrated that endothelial or smooth muscle cells contactedwith one or more neighbors have a significantly higher growth ratethan single cells without contacts [35]. These observations agreewith our results, as the coverages of Caco-2 cells on surfaces with50 mm, 100 mm, 500 mm wells were significantly less than thecoverage on flat PDMS during the earlier days in culture.

Caco-2 cells cultured on micro-well patterned surfaces weremetabolically more active than those on flat surfaces in the earlierdays in culture. The high MTT reduction rate is associated with highactivity of NADH- or NADPH-dependent mitochondrial dehydro-genases [18]. NAD dependent dehydrogenases play an importantrole in citric acid cycle, which is a main source of ATP production[23]. Adequate ATP supply is crucial for active restructuring of theactin cytoskeleton during the processes of cell movement and thechange of cell shape [36]. Studies using fibroblast cells [36], heartendothelial cells [37], and erythrocytes [38] demonstrated thatcytoskeletal dynamics were energy dependent; in other words,they relied on global ATP supply. In this study, Caco-2 cells dwellinginside wells formed more distinctive filopodia and spike-likeprotrusions than the cells growing on flat surfaces. On the otherhand, it was observed that Caco-2 cells have a tendency to crossover 50 mm and 100 mm wells. In regions where cells span overwells, the cells do not have any solid support; thus, they may haveto either form stronger cell-cell adhesion (such as forming tighterjunctions) with neighboring cells or increase rigidity of cells byproducing more and longer actin fibers. Taken together, the resultssuggest that cells growing on patterned surfaces were more activein remodeling or constructing the actin cytoskeleton; these activ-ities requires more ATP supply and consequently could be related tothe higher metabolic activity observed. The higher metabolicactivity in early stages of cell culture might also indicate that cells

L. Wang et al. / Biomaterials 30 (2009) 6825–6834 6833

cultured on patterned surfaces have higher motility, or requiremore energy to conquer physical micro-barriers or to proliferateunder conditions lacking cell–cell interaction.

Caco-2 cells were found to adjust their migration path whengrowing on micro-well patterned surfaces, as they tended tochoose the shortest distance by avoiding the bottom corners ofwells or bridging wells when the well diameters (i.e. 50, 100 mm)were smaller than the well depth (120 mm). Cells are known to reactto the physical environment by a feed-back loop: they assert forceon their environment, sense the response force from the environ-ment, and then adjust their response [5]. It is well known thatmicron-scale groove/ridge patterns can influence cell migrationand morphology, and the extent of reaction is depending on width,and depth of groove/ridge topography. Sarkar et al. [39] developeda poly-caprolactone (PCL) scaffold with micron-scale groove/ridge(48 mm grooves; 5 mm deep; 12 mm spacing), which mimicked thesmall diameter blood vessel topography. The alignment of vascularsmooth muscle cells (VSMC) on micro-patterned PCL scaffolds wasobserved. Clark et al. [7,8] showed that BHK and chick embryonicneural cells’ alignment along a single ‘‘cliff’’ increased with depth ofcliff in the range of 1–25 mm. Cells have also been found to spanacross grooves, and this phenomenon is more prominent onsubstrates with deeper and narrower grooves [40,41]. Goldner et al.reported [40] that when the groove depth (50 mm) was more thangroove width (30 mm), rat dorsal root ganglion (DRG) neurons wereable to bridge across the grooves without underlying support,suggesting cells were smart enough to take the ‘‘shortcut’’. Thisresult is similar to that reported in our study. However, there iscontroversy regarding whether the alignment of cells on thetopography of a material is mediated through contact guidance, orwhether topographies actually creates chemical patterns guidingcell migration [12,42]. The presence of sharp corners of wellstructures might trigger morphological changes in Caco-2 cellswhich enable the observed bypassing of sharp corners. The sharpcorners could also create a zone of reduced diffusion (e.g., ofnutrient or growth hormones), thus chemically preventing cellsfrom approaching this area. The same reasoning might also apply tothe phenomenon of cell bridging across deeper and narrower wells.

The alkaline phosphatase activity suggested that Caco-2 cellsgrowing on patterned surfaces differentiate slower than cellsgrowing on flat surfaces. It has been reported by several researchersthat micron-scale biomimetic substrate topography affects celldifferentiation. For example, Pins et al. designed a skin basal laminaanalogue with ridges and channels, which had similar length scales(40–310 mm in width) to the invaginations found in a native skinbasal lamina [15]. They found that differentiation of keratinocyteslocated in the channels was enhanced. Chehroudi et al. [14] fixedtitanium-coated epoxy substrate patterned with pit structuresranging from 30 to 120 mm in depth to the parietal bone of rats.After 8 weeks, they found pit pattern promoted bonelike tissueformation, and bone formation enhanced as pit depth increased.The fact that in our study Caoc-2 cells grown on micro-well surfacesappeared to be less differentiated at least in terms of alkalinephosphatase activity indicates different mechanisms of cell differ-entiation depending on cell type, and could be related to a numberof experimental factors. For example, the high aspect ratio micro-wells may generate a concentration gradient of soluble chemicalfactors along the vertical direction. The concentrations of cytokinesand growth factors could be relatively high inside wells due todiffusion-limited transport. The cell-cell apical surface distancebetween cells locating inside wells is smaller than for cells locatedon tops of wells or on flat surfaces, which might also contribute tothe change of the cell phenotype. It is interesting to note that cellswithin crypts in native intestinal tissue also express a less differ-entiated cell phenotype than cells on villi.

5. Conclusions

A PDMS cell culture substrate with biomimetic small intestinalcrypt-like surface topography was fabricated. Qualitative andquantitative analyses suggested that crypt-like topography affectedCaco-2 cell morphology, migration, and differentiation. Spreadingand migration phenomena on micro-well patterned surfacesincluded the prominence of cellular protrusions extending to sidewalls, the defined progression of coverage over some smaller(50 mm and 100 mm) but not larger (500 mm) diameter wells, andthe bypassing of sharp well corners as cells migrated up side walls.The higher metabolic activity and lower level of differentiation, asindicated by alkaline phosphatase activity, of cells on the micro-well structures compared to flat-structures have interestingimplications regarding the significance of crypt-like structures ininfluencing cell phenotype, as cells in crypts in vivo multiply andare generally less differentiated than cells on villi. The significantinfluence of crypt-like micro-topography on intestinal cell pheno-type suggests that development of topographically biomimeticscaffolds may be a key enabling step for intestinal tissueengineering.

Acknowledgements

This work was financially supported by the National ScienceFoundation (Grant No. CBET 0700764). The authors gratefullyacknowledge Edward V. Kogan for his assistance of alkaline phos-phatase activity and metabolic activity assay.

Appendix

Figures with essential colour discrimination. Figs. 1–8 in thisarticle may be difficult to interpret in black and white. The fullcolour images can be found in the on-line version, at doi:10.1016/j.biomaterials.2009.08.046.

References

[1] Beaulieu J-F. Integrins and human intestinal cell functions. Front Biosci 1999;4:310–21.

[2] Weiss L. Histology, cell and tissue biology. 5th ed. New York: ElsevierBiomedical; 1983.

[3] Pageot L-P, Perreault N, Basora N, Francoeur C, Magny P, Beaulieu J-F. Humancell models to study small intestinal functions: recapitulation of the crypt-villus axis. Microsc Res Tech 2000;49(4):394–406.

[4] Flemming RG, Murphy CJ, Abrams GA, Goodman SL, Nealey PF. Effects ofsynthetic micro- and nano-structured surfaces on cell behavior. Biomaterials1999;20(6):573–88.

[5] Discher DE, Janmey P. Wang Y-l. Tissue cells feel and respond to the stiffness oftheir substrate. Science 2005;310(5751):1139–43.

[6] Bettinger CJ, Langer R, Borenstein JT. Engineering substrate topography at themicro- and nanoscale to control cell function. Angew Chem Int Ed Engl2009;48(30):5406–15.

[7] Clark P, Connolly P, Curtis AS, Dow JA, Wilkinson CD. Topographical control ofcell behaviour. I. Simple step cues. Development 1987;99(3):439–48.

[8] Clark P, Connolly P, Curtis AS, Dow JA, Wilkinson CD. Topographical control of cellbehaviour: II. Multiple grooved substrata. Development 1990;108(4):635–44.

[9] Wood A. Contact guidance on microfabricated substrata: the response ofteleost fin mesenchyme cells to repeating topographical patterns. J Cell Sci1988;90(4):667–81.

[10] Kemkemer R, Jungbauer S, Kaufmann D, Gruler H. Cell orientation bya microgrooved substrate can be predicted by automatic control theory. Bio-phys J 2006;90(12):4701–11.

[11] Meyle J, Gultig K, Nisch W. Variation in contact guidance by human cells ona microstructured surface. J Biomed Mater Res 1995;29(1):81–8.

[12] Curtis A, Wilkinson C. Topographical control of cells. Biomaterials 1997;18(24):1573–83.

[13] Green AM, Jansen JA, van der Waerden JPCM, von Recum AF. Fibroblastresponse to microtextured silicone surfaces: texture orientation into or out ofthe surface. J Biomed Mater Res 1994;28(5):647–53.

[14] Chehroudi B, McDonnell D, Brunette DM. The effects of micromachinedsurfaces on formation of bonelike tissue on subcutaneous implants as assessed

L. Wang et al. / Biomaterials 30 (2009) 6825–68346834

by radiography and computer image processing. J Biomed Mater Res1997;34(3):279–90.

[15] Pins GD, Toner M, Morgan JR. Microfabrication of an analog of the basallamina: biocompatible membranes with complex topographies. FASEB J2000;14(3):593–602.

[16] Sia SK, Whitesides GM. Microfluidic devices fabricated in poly(dimethylsiloxane)for biological studies. Electrophoresis 2003;24(21):3563–76.

[17] Ratner BD. Biomaterials science: an introduction to materials in medicine. 2nded. Amsterdam/Boston: Elsevier Academic Press; 2004.

[18] Blumenthal RD. Chemosensitivity. Totowa, NJ: Humana Press; 2005.[19] Schmidt JA, von Recum AF. Macrophage response to microtextured silicone.

Biomaterials 1992;13(15):1059–69.[20] Stevens MM, George JH. Exploring and engineering the cell surface interface.

Science 2005;310(5751):1135–8.[21] Nelson CM, Tien J. Microstructured extracellular matrices in tissue engi-

neering and development. Curr Opin Biotechnol 2006;17(5):518–23.[22] Sniadecki NJ, Desai RA, Ruiz SA, Chen CS. Nanotechnology for cell–substrate

interactions. Ann Biomed Eng 2006;34(1):59–74.[23] Alberts B. Molecular biology of the cell. 4th ed. New York: Garland Science;

2002.[24] Sarikaya M, Aksay IA. Biomimetics: design and processing of materials.

Woodbury, NY: AIP Press; 1995.[25] Karuri NW, Liliensiek S, Teixeira AI, Abrams G, Campbell S, Nealey PF, et al.

Biological length scale topography enhances cell-substratum adhesion ofhuman corneal epithelial cells. J Cell Sci 2004;117(15):3153–64.

[26] Kim M-H, Kino-oka M, Kawase M, Yagi K, Taya M. Response of humanepithelial cells to culture surfaces with varied roughnesses prepared byimmobilizing dendrimers with/without d-glucose display. J Biosci Bioeng2007;103(2):192–9.

[27] Teixeira AI, Abrams GA, Bertics PJ, Murphy CJ, Nealey PF. Epithelial contactguidance on well-defined micro- and nanostructured substrates. J Cell Sci2003;116(10):1881–92.

[28] Yim EKF, Reano RM, Pang SW, Yee AF, Chen CS, Leong KW. Nanopattern-induced changes in morphology and motility of smooth muscle cells.Biomaterials 2005;26(26):5405–13.

[29] Mata A, Boehm C, Fleischman AJ, Muschler G, Roy S. Analysis of connectivetissue progenitor cell behavior on polydimethylsiloxane smooth and channelmicro-textures. Biomed Microdevices 2002;4(4):267–75.

[30] Revzin A, Tompkins RG, Toner M. Surface engineering with poly(ethyleneglycol) photolithography to create high-density cell arrays on glass. Langmuir2003;19(23):9855–62.

[31] Rettig JR, Folch A. Large-scale single-cell trapping and imaging using micro-well arrays. Anal Chem 2005;77(17):5628–34.

[32] Adams JC. Regulation of protrusive and contractile cell-matrix contacts. J CellSci 2002;115(2):257–265.

[33] Thomas LA, Yamada KM. Contact stimulation of cell migration. J Cell Sci1992;103(4):1211–4.

[34] Middleton CA. The effects of cell-cell contact on the spreading of pigmentedretina epithelial cells in culture. Exp Cell Res 1977;109(2):349–59.

[35] Nelson CM, Chen CS. Cell-cell signaling by direct contact increases cellproliferation via a PI3K-dependent signal. FEBS Lett 2002;514(2–3):238–42.

[36] van Horssen R, Janssen E, Peters W, van de Pasch L, Lindert MMt, vanDommelen MMT, et al. Modulation of cell motility by spatial repositioning ofenzymatic ATP/ADP exchange capacity. J Biol Chem 2009;284(3):1620–7.

[37] Karl I, Bereiter-Hahn J. Tension modulates cell surface motility: a scanningacoustic microscopy study. Cell Motil Cytoskeleton 1999;43(4):349–59.

[38] Korenstein R, Tuvia S, Mittelman L, Levin S. Local bending fluctuations of thecell membrane. In biomechanics of active movement and division of cells. In:Akkas N, editor. Biomechanics of active movement and division of cells. Berlin:Springer-Verlag; 1994. p. 415–23.

[39] Sarkar S, Lee GY, Wong JY, Desai TA. Development and characterization ofa porous micro-patterned scaffold for vascular tissue engineering applications.Biomaterials 2006;27(27):4775–82.

[40] Goldner JS, Bruder JM, Li G, Gazzola D, Hoffman-Kim D. Neurite bridgingacross micropatterned grooves. Biomaterials 2006;27(3):460–72.

[41] den Braber ET, de Ruijter JE, Ginsel LA, von Recum AF, Jansen JA. Quantitativeanalysis of fibroblast morphology on microgrooved surfaces with variousgroove and ridge dimensions. Biomaterials 1996;17(21):2037–44.

[42] Tirrell M, Kokkoli E, Biesalski M. The role of surface science in bioengineeredmaterials. Surf Sci 2002;500:61–83.