Scanning probe-enabled nanocombinatoricsdefine the relationship between fibronectinfeature size and stem cell fateLouise R. Giama,b,1, Matthew D. Massichb,c,1, Liangliang Haob,c, Lu Shin Wongb,c, Christopher C. Maderb,c, andChad A. Mirkina,b,c,2

aDepartment of Materials Science and Engineering; bInternational Institute for Nanotechnology; and cDepartment of Chemistry, Northwestern University,2145 Sheridan Road, Evanston, IL 60208

Contributed by Chad A. Mirkin, January 20, 2012 (sent for review December 28, 2011)

We report the development of a powerful analytical method thatutilizes a tilted elastomeric pyramidal pen array in the context of ascanning probe lithography experiment to rapidly prepare librarieshaving as many as 25million features over large areas with a rangeof feature sizes from the nano- to microscale. This technique canbe used to probe important chemical and biological processes,opening up the field of nanocombinatorics. In a proof-of-conceptinvestigation of mesenchymal stem cell (MSC) differentiation, com-binatorial patterns first enabled a rapid and systematic screeningof MSC adhesion, as a function of feature size, while uniformpatterns were used to study differentiationwith statistically signif-icant sample sizes. Without media containing osteogenic-inducingchemical cues, cells cultured on nanopatterned fibronectin sub-strates direct MSC differentiation towards osteogenic fates whencompared to nonpatterned fibronectin substrates. This powerfuland versatile approach enables studies of many systems spanningbiology, chemistry, and engineering areas.

nanopatterning ∣ polymer pen lithography ∣ focal adhesions ∣ osteogenesis ∣stem cell differentiation

It is important to note and distinguish that while the prevailingemphasis in lithography concerns the exact replication of struc-tures across large areas, methods to rapidly generate combinatoriallibraries having a tunable gradient of feature sizes with a single,maskless tool would be powerful for systematic investigations infields spanning biology, materials engineering, and nanotechnol-ogy (1). Herein we report the development of a powerful and un-ique analytical method that utilizes a tilted elastomeric pyramidalpen array in the context of a scanning probe lithography experi-ment to rapidly prepare libraries having as many as 25 million fea-tures over large areas with a range of feature sizes from the nano-to microscale that can be used to probe important chemical andbiological processes, opening up the field of nanocombinatorics.In a proof-of-concept investigation of mesenchymal stem cell(MSC) differentiation, combinatorial patterns first enabled a rapidand systematic screening ofMSC adhesion, as a function of featuresize, while uniform patterns were used to study differentiationwith statistically significant sample sizes. Without media containingosteogenic-inducing chemical cues, cells cultured on nanopat-terned fibronectin substrates direct MSC differentiation towardsosteogenic fates when compared to nonpatterned fibronectin sub-strates.

Polymer pen lithography (PPL) is a massively parallel direct-write technique that uses an array of elastomeric pyramidal pensto deliver materials (e.g., alkanethiols, proteins) to a surface andgenerate customizable patterns with feature sizes ranging fromthe nano- to microscale simply by changing (i) the amount offorce applied to the pen array or (ii) the tip-substrate contact time(2–5). Indeed, PPL has already been shown to be a high-through-put patterning technique capable of generating arbitrary designsuniformly on a surface. By using a tilted elastomeric pyramidalpen array in the context of a scanning probe lithography experi-

ment, however, one can rapidly prepare combinatorial librariesconsisting of millions of structures having a gradient of tunablefeature sizes that can be used to probe important chemical andbiological processes. This approach opens up the field of nano-combinatorics, as defined by the creation and use of customlibraries with precise control over feature size (nano- to micro-scale), spacing, pattern design, and composition. Importantly, wedemonstrate that such capabilities are enabling and useful forresearchers who wish to quickly validate or refute hypotheses ina nanocombinatorial manner and then systematically prototypestructures in a high-throughput, straightforward, and low-costway.

Though the broad field of combinatorics has primarily focusedon compositional changes (6–8), which can also be achieved usingmultiplexed PPL (4), altering biophysical parameters such as ex-tracellular matrix protein feature size, spacing, and pattern designin a rapid, low-cost, and robust manner remains challenging forresearchers. Given that microcontact printing (9) and PPL bothdepend on a photolithography step, the initial stamps and penarrays, respectively, require the same fabrication costs. Micro-contact printing stamps and polymer pen arrays, both primarilycomposed of elastomers [e.g., polydimethylsiloxane, (PDMS)],themselves are inexpensive and cost less per printed pattern withrepeated usage. Every time a pattern parameter (e.g., design, fea-ture size, feature spacing) is altered, however, new masks (>$500for a 4′′ wafer having simple features tens of microns in dimen-sions, and much more for submicron features) are needed formicrocontact printing; moreover, some pattern designs requiremechanical support for submicron features (10) or large featurespacings. PPL, on the other hand, takes advantage of high-preci-sion piezoelectric x − y − z positional control offered by scanningprobe instruments and the versatility of the elastomeric pen ar-rays to easily produce a diverse array of pattern designs, featuredimensions, and feature spacings without the need for newmasks.This advantage becomes especially important when screening foroptimal pattern parameters.

Rather than designing multiple masks for microcontact print-ing molds or being limited in throughput by serial conventionale-beam lithography, a single inexpensive pen array (square inchesin area) in a scanning probe experiment can be easily used in (i) atilted configuration to generate libraries having tunable combina-torial feature sizes between the nano- to microscale and (ii) alevel configuration to prepare homogeneous patterns over large

Author contributions: L.R.G., M.D.M., L.H., L.S.W., C.C.M., and C.A.M. designed research;L.R.G., M.D.M., L.H., and L.S.W. performed research; L.R.G., M.D.M., L.H., L.S.W., C.C.M.,and C.A.M. analyzed data; and L.R.G., M.D.M., and C.A.M. wrote the paper.

The authors declare no conflict of interest.1L.R.G. and M.D.M. contributed equally to this work.2To whom correspondence should be addressed. E-mail: chadnano@northwestern.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1201086109/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1201086109 PNAS ∣ March 20, 2012 ∣ vol. 109 ∣ no. 12 ∣ 4377–4382

ENGINEE

RING

APP

LIED

BIOLO

GICAL

SCIENCE

S

Dow

nloa

ded

by g

uest

on

Apr

il 6,

202

1

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1201086109/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1201086109/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1201086109/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1201086109/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1201086109/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1201086109/-/DCSupplemental

areas that allow quantification of mRNA and protein expressionin statistically significant populations of cells (Fig. 1 A and B).This approach not only enables qualitative screening of promisingstructures that affect differentiation, but also quantitative inves-tigation of cell-substrate interactions. Furthermore, it can be gen-erally applied to study myriad fundamental questions not limitedto biology.

Recently, it has been observed that environmental factors suchas substrate stiffness (11), topography (12), geometry (13, 14),and chemical cues (15) can guide stem cell fate. The exact struc-tures and factors that govern such responses, however, are notcompletely understood (16). To evaluate the importance and uti-lity of this nanocombinatorics approach in a proof-of-concept sys-tem, we have chosen to study human MSC differentiation (17),where literature has indicated that neurogenic, myogenic, andosteogenic lineages have focal adhesions of different character-istic sizes ranging between the nano- and microscale (11).

We therefore pose the as-yet unanswered question of whethersystematically programming extracellular matrix protein feature sizealone might dictate cell fate in the absence of chemical cues fromdifferentiation media. Specifically, we used a scanning probeapproach to investigate whether nanometer or micrometer-sizedfibronectin features can control MSC fate towards an osteogeniclineage. Fibronectin is an extracellular matrix protein that isknown to bind the α5β1 and αVβ3 integrin receptors of cellmembranes via an Arg-Gly-Asp (RGD) domain (18). Upon celladhesion and spreading, these integrin receptors aggregate andassemble into multimolecular complexes termed focal adhesions,which are involved in cell mechanosensing and mechanotrans-duction (19). The work reported herein departs from previousstudies (20, 21) which examine how cells differentiate whencultured on electrospun-nanofibers, carbon nanotubes (22), ortopographical surfaces (12), in particular, because immobilizedfibronectin on PPL-patterned substrates directly engage MSCsin biochemical and biomechanical interactions that recapitulatea more physiologically representative environment.

Results and DiscussionTo pattern substrates, a square-inch PDMS polymer pen arrayattached to a glass support made according to published methods(2) is coated with a 10 mM ethanolic solution of 16-mercaptohex-adecanoic acid (MHA) and blown dry with nitrogen. Afterwards,the pen array is mounted in a scanning probe instrument (XE-150, Park Systems) equipped with a tilt stage and a decoupledx-and y-piezo 100-μm closed loop scanner inside an environmen-tal chamber having humidity control. Commercial lithographysoftware provides customizable control over position, tip-sub-strate contact time (0.1 to 0.5 s), amount of z-piezo extension(range of approximately 25 μm), and z-piezo extension speed(100 μm∕s) for each feature in a desired pattern. The polymerpen array can be aligned by top-down optical observation oftip deformation using a microscope (2) or force maximizationstrategies (23).

The inked pen array is used to pattern monolayers of MHA onan electron-beam evaporated gold substrate (14 nm Au with a4 nm Ti adhesion layer on #1 cover slip glass). By intentionallytilting the polymer pen array (in this case, approximately 0.01°)relative to the plane of the substrate, it is possible to generatecombinatorial patterns having different feature sizes, but thesame feature spacing (Fig. 1A, Fig. S1). A small section of thepatterned substrates is used specifically to confirm pattern qualityand feature sizes while the remaining portion is kept for subse-quent cell biology experiments. To check pattern quality, a pat-terned area of the substrate is chemically etched to produceraised Au features representative of the remaining unetchedportion of the substrate and which can be visualized by optical orscanning electron microscopy (SEM). Uniform Au featuredimensions ranging from 300 nm to 1.2 μm are fabricated in thismanner and observed using SEM (Fig. 1A, Fig. S2). The remain-ing unetched patterned substrate is passivated using a 1 mMethanolic solution of hexa(ethylene glycol)-undecanethiol for1 h at room temperature to reduce nonspecific protein adsorption(Fig. 1C). The substrate is rinsed with ethanol and dried withnitrogen before immersion in a 10 mM ethanolic solution ofcobalt nitrate for 30 min at room temperature. In this step, cobalt

Fig. 1. Preparation of patterned substrates with fibronectin features. (A)Scheme of a tilted polymer pen array, which can be used to generate com-binatorial libraries. Below, there are SEM images of raised Au featurescreated after chemical etching MHA patterns from a single substrate. Withinthis one substrate, a range of feature sizes (widths of 475 nm, 1.2 μm) canbe generated by tilting the polymer pen array just 0.01° across the overallwidth of the patterned area of 1.8 cm. Because the pen array is attachedto a piezoelectric scanner in a scanning probe instrument, it is possible tocontrol the pattern design (e.g., square pattern with 2 μm pitch). (B) Schemeof a level polymer pen array, which can be used to generate homogeneousfeatures. Below, there are also SEM images of raised Au features after che-mical etching MHA patterns from a single substrate. (C) Scheme for immo-bilizing fibronectin in desired patterns beginning with PPL patterning ofMHA on Au substrates. Then, hexa(ethylene glycol)-undecanethiol is usedto passivate nonpatterned areas before fibronectin is adsorbed to MHAfeatures. MSCs are subsequently cultured on these fibronectin patterned sub-strates. (D) AFM topography image of patterned substrates after fibronectinimmobilization, showing that a monolayer of protein (height of ∼2.2 nm) isadsorbed. The corresponding AFM phase image shows that each MHA fea-ture is completely covered with adsorbed fibronectin.

4378 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1201086109 Giam et al.

Dow

nloa

ded

by g

uest

on

Apr

il 6,

202

1

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1201086109/-/DCSupplemental/pnas.1201086109_SI.pdf?targetid=SF1http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1201086109/-/DCSupplemental/pnas.1201086109_SI.pdf?targetid=SF1http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1201086109/-/DCSupplemental/pnas.1201086109_SI.pdf?targetid=SF2

cations are chelated by the carboxylic acid groups in MHA, whichenable selective orientation of fibronectin such that it does notaffect the cell-binding domain of the protein (24). A single anddense layer of human fibronectin is immobilized on the patternedsubstrates to provide defined regions of potential integrin engage-ment (Fig. 1D). Given that an “adhesive unit” of approximately60 nm has at least four clustered integrin heterodimers, fibronec-tin features having diameters of 300 nm and 1 μm could supportcontact of at least 100 to 1,400 integrins, respectively (25).

To begin, we tilted a single elastomeric pen array (squareinches in area) relative to the substrate for creating combinatorialpatterns having a range of nano- to microscale feature sizes(Fig. 1A, Fig. S1). The combinatorial patterns can be used to qua-litatively investigate cell adhesion and osteogenic marker expres-sion in individual MSCs by immunofluorescence (Fig. S3). Weidentified that even when the total projected cell area is heldconstant at 3;600 μm2 (square pattern of 60 × 60 μm), the totalfibronectin contact area per cell controls MSC adhesion. Whena MSC attaches to 182 μm2 of immobilized fibronectin, it is notfully spread on the patterned regions, whereas at 248 μm2 offibronectin, it covers the entire 60 × 60 μm pattern. These resultsare consistent with previous observations where capillary epithelialcell survival requires large projected cell areas and a minimumlevel of fibronectin contact area (26). Based on these results, weused 225 μm2 of total fibronectin contact area as sufficient fortesting if and how fibronectin feature size affects differentiation ofMSCs towards osteogenic fates.

Level polymer pen arrays were used to fabricate 10,000 equiva-lent patterns for two different feature sizes (1 μm and 300 nm)over square inch areas (Fig. 2A, Fig. S4). Importantly, total fibro-nectin contact area was held constant at 225 μm2 for both 1 μmand 300 nm feature sizes by controlling the feature density per60 × 60 μm square pattern; specifically, the 1 μm features weremade in a 15 × 15 array with a 4 μm spacing and the 300 nmfeatures were made in a 50 × 50 array with a 1.2 μm spacing.MSCs were seeded on these substrates (Fig. 2B) and culturedin normal growth media or osteogenic media (OM) containingchemical factors such as ascorbic acid, β-glycerophosphate,and dexamethasone, a critical component known to induce osteo-genesis (27). With this approach, large area homeogeneous pat-terns can be used to quantitatively measure the expression ofosteogenic markers at the mRNA and protein levels across entirecell populations.

In these experiments, a 1.8 × 1.8 cm2 pen array having 10,000pens (spacing between pens ¼ 180 μm) was used to preparepatterned substrates. The distance between pens within an arraycan be easily modified by using different pen molds (Fig. S5).Given that each pen only makes one pattern, 10,000 patternsare made on a substrate; the 180 μm pattern separation corre-sponds to a density of approximately 3;000 patterns∕cm2. If asingle cell occupies every pattern, the cell density remains about3;000 cells∕cm2; MSCs seeded at this density do not contribute tocell density-dependent observations of osteogenic differentiation(13, 27). When each pen in a 10,000 pen array prints a single pat-

Fig. 2. MSCs cultured on patterned substrates and their osteogenic marker expression levels after 7 d. (A) Large area fluorescence image of fibronectinpatterns (feature size ¼ 1 μm, pitch ¼ 4 μm) stained with mouse anti-fibronectin antibody and AlexaFluor488-conjugated goat anti-mouse IgG. Every penin the array produces a single pattern comprised of a square array of 15 × 15 individual features. (B) Phase contrast optical microscopy image of MSCs seededon fibronectin patterns (feature size ¼ 1 μm, pitch ¼ 4 μm). Quantitative RT-PCR results for (C) alkaline phosphatase (ALP), (D) osteocalcin (OCN), and (E) core-binding factor α (CBF-α) normalized to GAPDH levels. For each gene of interest, samples were also normalized to the negative control: no pattern, OM−. Resultsare presented as the mean� standard deviation, n ¼ 3, statistics were conducted using Student’s two-tailed t-test where asterisks indicate statistically sig-nificant differences relative to the negative control (no pattern, OM−): *p < 0.05, **p < 0.01. (F) WB results for ALP and CBF-α. GAPDH was used as a loadingcontrol.

Giam et al. PNAS ∣ March 20, 2012 ∣ vol. 109 ∣ no. 12 ∣ 4379

ENGINEE

RING

APP

LIED

BIOLO

GICAL

SCIENCE

S

Dow

nloa

ded

by g

uest

on

Apr

il 6,

202

1

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1201086109/-/DCSupplemental/pnas.1201086109_SI.pdf?targetid=SF1http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1201086109/-/DCSupplemental/pnas.1201086109_SI.pdf?targetid=SF1http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1201086109/-/DCSupplemental/pnas.1201086109_SI.pdf?targetid=SF3http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1201086109/-/DCSupplemental/pnas.1201086109_SI.pdf?targetid=SF4http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1201086109/-/DCSupplemental/pnas.1201086109_SI.pdf?targetid=SF4http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1201086109/-/DCSupplemental/pnas.1201086109_SI.pdf?targetid=SF5http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1201086109/-/DCSupplemental/pnas.1201086109_SI.pdf?targetid=SF5

tern spanning 60 × 60 μm (15 × 15 array of 1 μm MHA featuresspaced 4 μm apart or 50 × 50 array of 300 nm MHA featuresspaced 1.2 μm apart), 2.25 to as many as 25 million individualfeatures are generated on a substrate in about 5 and 25 min, re-spectively. Rather than only relying on time-dependent ink diffu-sion from the tip as with conventional dip-pen nanolithography(3, 28), PPL can rapidly access large features (>1 μm) so that sub-strate preparation is not prohibitively time-consuming for subse-quent experiments. Importantly, PPL itself allows one to prepareany custom pattern comprised of complex nanoscale to micro-scale features and is not limited to the relatively simple squarearrays used in this study.

After one week of growth on the patterned substrates, thelevels of osteogenic markers expressed by the MSCs were ana-lyzed using quantitative real-time reverse transcriptase polymer-ase chain reaction (qRT-PCR), Western blotting (WB), andimmunofluorescence (IF) confocal microscopy. Following qRT-PCR analysis, cells grown on fibronectin patterns with 300 nmfeatures in the absence of OM demonstrated a statistically signif-icant increase in the expression levels of the osteogenic markersALP, OCN, and CBF-α when compared to the negative control-cells grown on nonpatterned, uniformly coated fibronectin sub-strates (Fig. 2 C–E). Though the total surface area of fibronectinpresented to the cells was held constant, the 1 μm feature sizesexhibit modest increases in the expression levels of the testedosteogenic markers. Surprisingly, the cells grown on patterns con-sisting of 300 nm feature sizes expressed increased levels of OCN

and CBF-α relative to the positive control-cells grown on nonpat-terned fibronectin substrates in the presence of OM (Fig. 2). Thecombination of patterned substrates with OM, however, led to adecrease in expression of osteogenic markers for OCN, CBF-α,and ALP (Fig. 2), which may be explained by a previous study(27) that showed a biphasic dependency of osteogenic markerexpression on dexamethasone concentration. Below and above10 nM dexamethasone, ALP activity levels decreased (27). Toconfirm that trends in protein levels match mRNA results,osteogenic markers were measured by WB (Fig. 2F, Table S1)and IF (Fig. 3). Consistent with the qRT-PCR findings, proteinlevels of MSCs cultured in the absence of OM showed increasedosteogenic marker expression (ALP and CBF-α) especially forsubstrates having 300 nm features, but also those having 1 μmfeatures. It is worth noting that though the addition of OM de-creased mRNA levels of CBF-α in patterned substrates (1 μm and300 nm, Fig. 2E), protein amounts from these samples remaincomparable to cells cultured on patterned substrates withoutOM (Fig. 2F). Indeed, CBF-α mRNA and protein levels maynot directly reflect similar trends because of posttranscriptionalgene regulation that govern mRNA stability (29), posttransla-tional protein modifications that activate the transcription factor(30), and protein half-life (31).

Though the mechanism by which cells transduce environmen-tal signals is not well understood, it has been observed previouslythat focal adhesions assemble into a range of sizes dependingon the lineage of MSC differentiation (11). The work discussed

Fig. 3. Immunofluorescence images show the presence of osteogenic marker alkaline phosphatase in MSCs. Confocal microscopy images of immunofluor-escently stained samples where in the merged channel, actin is red, ALP is green, and the nucleus is blue.

4380 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1201086109 Giam et al.

Dow

nloa

ded

by g

uest

on

Apr

il 6,

202

1

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1201086109/-/DCSupplemental/pnas.1201086109_SI.pdf?targetid=ST1

herein not only provides a general tool for systematically studyingpopulations of MSCs, but also demonstrates that nanoscale fibro-nectin patterns increase the expression of osteogenic markers inMSCs over microscale features. In order to address the potentialmechanism by which osteogenesis may be occurring, we investi-gated focal adhesion formation and focal adhesion kinase (FAK)phosphorylation. Specifically, auto-phosphorylation of FAK attyrosine-397 occurs during cell adhesion and spreading inresponse to integrin engagement and clustering (32), which is im-portant for interactions that activate downstream pathways invol-ving Src-family nonreceptor tyrosine kinases and extracellularsignal-regulated kinase (ERK) proteins in somatic cells (Fig. 4A).In MSCs it has been observed that FAK phosphorylation, whichcan occur within hours and remain abundant for days, is necessaryfor and precedes osteogenic marker expression (e.g., CBF-α),which can occur after several days and increase in the followingweeks of cell differentiation (33). Consistent with qRT-PCR andWB analysis at the one week time point, it is possible to measurethe relative amount of phospho-FAK to FAK by enzyme-linkedimmunosorbent assay (ELISA). The ratio of phospho-FAK toFAK, however, is over 1.5 times greater for patterned samples(1 μm and 300 nm features) both in the absence and presence ofOM (Fig. 4B) as compared to the negative control (OM−NP).Furthermore, the relative amount of phospho-FAK to FAK forthe 300 nm-feature substrates without OM exceeded the positivecontrol (OMþNP), which likely contributes to the observed in-crease in the previously noted osteogenic markers. Consistentwith the observed increases in osteogenic marker expression, the300 nm features induce more FAK phosphorylation than the 1 μmpatterns. Importantly, the amount of FAK across all samples isstatistically equivalent (Fig. 4B).

In the absence of OM containing chemical factors, we observethat patterned fibronectin substrates direct MSC differentiationtowards osteogenic fates when compared to nonpatterned fibro-

nectin substrates. Furthermore, substrates having nanoscale fea-tures, optimized through the nanocombinatorics approach, aremore effective at inducing osteogenesis than microscale features,and for certain biomarkers, even more effective than the givenOM. Though observations of how programmed extracellular ma-trix feature size controls MSC fate are important from a biolo-gical standpoint, the impact of this unique nanocombinatoricstool will extend beyond the study of stem cell differentiation.The straightforward, massively parallel approach uses an elasto-meric pen array in a tilted configuration at the point-of-use torapidly generate custom libraries with a tunable gradient of fea-ture sizes to systematically screen nanostructures, study large po-pulations of cells, and address important fundamental scientificquestions. Indeed, one can imagine using this approach for a vari-ety of substrates to generate biochemical gradients (e.g., havingdifferent feature sizes and spacings across a substrate, changingbiomolecule concentration using multiplexed PPL) or probe mul-tivalent interactions in the context of many biological systems,including cancer cell metastasis, neuron signalling, embryonic de-velopment, or tissue engineering, among other potential struc-ture-function relationships in chemistry, materials engineering,and nanotechnology areas.

Materials and MethodsSubstrate Preparation. Glass coverslips (Fisher) were rinsed in pure ethanoland dried under a stream of N2. They were mounted in an electron-beamevaporator (Lesker) and when vacuum reached 2 × 10−7 mTorr, 4 nm of Tiand 14 nm of Au were evaporated. Polymer pen arrays were prepared usingconventional photolithography techniques according to published methods(2). Each pen array was oxygen plasma cleaned at 200 mTorr on mediumpower for 2 min. Arrays were then coated with a 10 mM MHA solution(in ethanol) and spuncoat for 30 sec at 1,000 rpm. After mounting the Ausubstrate and pen array on a scanning probe lithography instrument (ParkSystems), the chamber relative humidity was held at 45% for patterning.The scanning probe lithography instrument stage can be tilted and alignedwith the pen array according to established procedures (2, 23). Note that for300 nm features, deformation can be observed by slight contrast changes inthe pens. Patterns were programmed in the instrument with tip-substratecontact times ranging from 125 ms to 1 s for 300 nm and 1 μm features, re-spectively. Feature size and quality can be confirmed by sacrificing a portionof the substrate, etching Au in the unpatterned areas with a mixed aqueoussolution of 13.3 mM FeðNO3Þ3 · 9H2O and 20mM thiourea, and observing theresults under an optical or scanning electron microscope.

After patterning, the Au substrates were immersed in a 1mM 1-mercapto-11-undecyl hexa(ethylene glycol) solution in ethanol (Asemblon) for at least1 h to reduce nonspecific protein adsorption. The substrates were then im-mersed in a 10 mM CoðNO3Þ2 solution in ethanol (Sigma) for at least 30 min.After rinsing with pure ethanol and drying with N2, the substrates were ex-posed to 25 μg∕mL of human plasma fibronectin (Millipore) in phosphate-buffered saline (PBS) and shaken overnight at 4 °C.

Cell Culture. Human MSCs (Lonza) were cultured at 37 °C with 5% CO2 inMSC growth medium that was supplemented with MSC growth supplement(Lonza), L-glutamine (Lonza), and gentamycin sulfate amphotericin-1(Lonza). The cells were used before passage 2. For chemical induction ofdifferentiation, cells were cultured in osteogenic media (Lonza), which isMSC growth medium supplemented with dexamethasone, ascorbate, MSCgrowth supplement, L-glutamine, penicillin/streptomycin, and β-glyceropho-sphate (Lonza). Approximately 10,000 cells were seeded per substrate(1.8 × 1.8 cm2), which corresponds to about 3; 000 cells∕cm2.

Quantitative RT-PCR. Cells were harvested after one week of growth, andtotal RNA was extracted using TRIzol reagent (Invitrogen) following manu-facturer’s recommended protocol. Approximately 1 μg of RNA was then

Fig. 4. FAK phosphorylation levels for MSCs. (A) Proposed signaling path-way by which integrin-dependent MSC cell adhesion leads to osteogenesis.(B) ELISA results for FAK (white) and phosphorylated (Tyr397) FAK (gray) nor-malized to the negative control (no pattern, OM−). The relative amountsof FAK are statistically the same for all samples. Increased relative amountsof phosphorylated FAK are shown for the positive control (no pattern,OM+) and for MSCs cultured on patterned substrates (1 μm and 300 nm fea-tures in the presence and absence of OM). Results are presented as themean� standard deviation, n ¼ 3, statistics were conducted using Student’stwo-tailed t-test where asterisks indicate statistically significant differencesrelative to the negative control (no pattern, OM−): *p < 0.05.

Table 1. Primer Sequences 5 0 → 3 0

Gene name Forward Reverse

Osteocalcin GACTGTGACGAGTTGGCTGA GCAAGGGGAAGAGGAAAGAAALP GGAACTCCTGACCCTTGACC TCCTGTTCAGCTCGTACTGCCBF-α CCTCTGACTTCTGCCTCTGG TATGGAGTGCTGCTGGTCTGGAPDH ACAGTCAGCCGCATCTTCTT ACGACCAAATCCGTTGACTC

Giam et al. PNAS ∣ March 20, 2012 ∣ vol. 109 ∣ no. 12 ∣ 4381

ENGINEE

RING

APP

LIED

BIOLO

GICAL

SCIENCE

S

Dow

nloa

ded

by g

uest

on

Apr

il 6,

202

1

reverse transcribed using Superscript III (Invitrogen). PCR was performedon the cDNA using LightCycler 480 SYBR Green Master (Roche) on a RocheLight Cycler 480 II System following manufacturer’s recommended protocol.The relative abundance of the mRNA levels for the genes investigated wasnormalized to glyceraldehyde 3-phosphate dehydrogenase (GAPDH) expres-sion and compared to untreated cells to determine expression levels. Primersequences are listed in Table 1.

Protein Extraction and WB. For WB, cells were collected one week after seed-ing on the substrates and lysed in mammalian cell lysis buffer (ThermoScientific) containing protease and phosphatase inhibitor (Thermo Scientific).Proteins from the total cell lysates were run on a 4–15% precast polyacryla-mide gradient gel (Bio‐Rad), transferred to nitrocellulose membranes,blocked in 1% (wt∕vol) casein in TBS, and blotted with primary antibodiesfor CBF-α (1∶200) (Santa Cruz Biotechnology), alkaline phosphatase(1∶200) (Santa Cruz Biotechnology), GAPDH (1∶200) (Santa Cruz Biotechnol-ogy) and anti-mouse/rabbit IgG-AlexaFluor-680 (1∶5;000) secondary antibo-dies (Invitrogen). The blots were imaged using the Odyssey® InfraredImaging System (LI-COR Biosciences) and quantified using ImageJ.

Immunofluorescence and Confocal Microscopy. Cells cultured on substrateswere fixed in 3.7% paraformaldehyde/PBS for 15 min and then gentlywashed three times with PBS. Cells were permeabilized using 0.1% TritonX-100 in PBS solution with 2 mg∕mL of bovine serum albumin for 1 min be-fore immersing in the primary antibody solution (rabbit anti-ALP, Santa Cruz

Biotechnology) in PBS overnight at 4 °C on a shaker. Samples were gentlywashed three times with PBS before incubation with the secondary antibodysolutions (Invitrogen, AlexaFluor488-conjugated goat anti-rabbit) for 1 h atroom temperature. AlexaFluor568-labeled phalloidin (Invitrogen) was usedaccording to manufacturer’s instructions for actin staining. Samples weregently washed three times in PBS and then water before mounting onto glassslides using ProLong Gold antifade reagent with DAPI (Invitrogen). Cells wereimaged using a Nikon C-Si inverted laser confocal microscope.

ELISA. Commercial ELISA kits for FAK and phospho-FAK (Millipore) were usedaccording to manufacturer’s instructions.

ACKNOWLEDGMENTS. The authors thank A. Rademaker, the Electron ProbeInstrumentation Center, and the Northwestern University Cell Imaging Facil-ity for assistance. This work is based on research sponsored by the Air ForceOffice of Scientific Research (AFOSR) under award number FA9550-08-1-0124, SPAWAR Systems Center Pacific Code 22550 under award numberN66001-08-1-2044, the National Cancer Institute under award numberU54CA151880, the National Science Foundation (NSF) Nanoscale Scienceand Engineering Initiative under award number EEC-0647560, and theDefense Advanced Research Projects Agency (DARPA). L.R.G. acknowledgesan Achievement Rewards for College Scientists Scholarship and L.S.W. isgrateful to the Engineering and Physical Sciences Research Council (EPSRC)for a Life Science Interface Fellowship (EP/F042590/1).

1. Ratner BD, Bryant SJ (2004) Biomaterials: where we have been and where we aregoing. Annu Rev Biomed Eng 6:41–75.

2. Huo F, et al. (2008) Polymer pen lithography. Science 321:1658–1660.3. Giam LR, Mirkin CA (2011) Cantilever-free scanning probe molecular printing. Angew

Chem Int Ed 50:7482–7485.4. Zheng Z, et al. (2009) Multiplexed protein arrays enabled by polymer pen lithography:

addressing the inking challenge. Angew Chem Int Ed 48:7626–7629.5. Liao X, Braunschweig AB, Zheng ZJ, Mirkin CA (2010) Force- and time-dependent

feature size and shape control in molecular printing via polymer-pen lithography.Small 6:1082–1086.

6. Flaim C, Chien S, Bhatia S (2005) An extracellular matrix microarray for probing cellulardifferentiation. Nat Methods 2:119–125.

7. Mei Y, et al. (2010) Combinatorial development of biomaterials for clonal growth ofhuman pluripotent stem cells. Nat Mater 9:768–778.

8. Gobaa S, et al. (2011) Artificial niche microarrays for probing single stem cell fate inhigh throughput. Nat Methods 8:949–955.

9. Mrksich M, Dike L, Tien J, Ingber D, Whitesides G (1997) Using microcontact printingto pattern the attachment of mammalian cells to self-assembled monolayers of alka-nethiolates on transparent films of gold and silver. Exp Cell Res 235:305–313.

10. Csucs G, Kunzler T, Feldman K, Robin F, Spencer N (2003) Microcontact printingof macromolecules with submicrometer resolution by means of polyolefin stamps.Langmuir 19:6104–6109.

11. Engler AJ, Sen S, Sweeney HL, Discher DE (2006) Matrix elasticity directs stem celllineage specification. Cell 126:677–689.

12. Dalby MJ, et al. (2007) The control of human mesenchymal cell differentiation usingnanoscale symmetry and disorder. Nat Mater 6:997–1003.

13. McBeath R, Pirone D, Nelson C, Bhadriraju K, Chen C (2004) Cell shape, cytoskeletaltension, and RhoA regulate stem cell lineage commitment. Dev Cell 6:483–495.

14. Kilian KA, Bugarija B, Lahn BT, Mrksich M (2010) Geometric cues for directing thedifferentiation of mesenchymal stem cells. Proc Natl Acad Sci USA 107:4872–4877.

15. Li L, Klim JR, Derda R, Courtney AH, Kiessling LL (2011) Spatial control of cell fateusing synthetic surfaces to potentiate TGF-β signaling. Proc Natl Acad Sci USA108:11745–11750.

16. Lutolf MP, Gilbert PM, Blau HM (2009) Designing materials to direct stem-cell fate.Nature 462:433–441.

17. Pittenger M, et al. (1999) Multilineage potential of adult human mesenchymal stemcells. Science 284:143–147.

18. Geiger B, Bershadsky A, Pankov R, Yamada K (2001) Transmembrane extracellularmatrix-cytoskeleton crosstalk. Nat Rev Mol Cell Bio 2:793–805.

19. Vogel V, Sheetz M (2006) Local force and geometry sensing regulate cell functions.Nat Rev Mol Cell Bio 7:265–275.

20. Kim D-H, Lee H, Lee YK, Nam J-M, Levchenko A (2010) Biomimetic nanopatterns asenabling tools for analysis and control of live cells. Adv Mater 22:4551–4566.

21. Dvir T, Timko BP, Kohane DS, Langer R (2011) Nanotechnological strategies for engi-neering complex tissues. Nat Nanotechnol 6:13–22.

22. Hong S, Baik KY, Park SY, Heo K, Lee KB (2011) Carbon nanotube monolayer cues forosteogenesis of mesenchymal stem cells. Small 7:741–745.

23. Liao X, Braunschweig AB, Mirkin CA (2010) “Force-feedback” leveling of massivelyparallel arrays in polymer pen lithography. Nano Lett 10:1335–1340.

24. Gmeiner B, Leibl H, Zerlauth G, Seelos C (1995) Affinity binding of distinct functionalfibronectin domains to immobilized metal-chelates. Arch Biochem Biophys 321:40–42.

25. Schvartzman M, et al. (2011) Nanolithographic control of the spatial organization ofcellular adhesion receptors at the single-molecule level. Nano Lett 11:1306–1312.

26. Chen CS, Mrksich M, Huang S, Whitesides GM, Ingber DE (1997) Geometric control ofcell life and death. Science 276:1425–1428.

27. Jaiswal N, Haynesworth S, Caplan A, Bruder S (1997) Osteogenic differentiation ofpurified, culture-expanded human mesenchymal stem cells in vitro. J Cell Biochem64:295–312.

28. Piner R, Zhu J, Xu F, Hong S, Mirkin C (1999) “Dip-pen” nanolithography. Science283:661–663.

29. Gilbert L, et al. (2002) Expression of the osteoblast differentiation factor RUNX2(Cbfa1/AML3/Pebp2αA) is inhibited by tumor necrosis factor-α. J Biol Chem277:2695–2701.

30. Park OJ, Kim HJ, Woo KM, Baek JH, Ryoo HM (2010) FGF2-activated ERK mitogen-activated protein kinase enhances Runx2 acetylation and stabilization. J Biol Chem285:3568–3574.

31. Suh JH, Lee HW, Lee JW, Kim JB (2008) Hes1 stimulates transcriptional activity ofRunx2 by increasing protein stabilization during osteoblast differentiation. BiochemBioph Res Co 367:97–102.

32. Mitra S, Hanson D, Schlaepfer D (2005) Focal adhesion kinase: in command and controlof cell motility. Nat Rev Mol Cell Bio 6:56–68.

33. Salasznyk RM, Klees RF, Williams WA, Boskey A, Plopper GE (2007) Focal adhesionkinase signaling pathways regulate the osteogenic differentiation of humanmesench-ymal stem cells. Exp Cell Res 313:22–37.

4382 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1201086109 Giam et al.

Dow

nloa

ded

by g

uest

on

Apr

il 6,

202

1