ARTICLESPUBLISHED ONLINE: 20 DECEMBER 2009 | DOI: 10.1038/NMAT2606

Dynamic display of biomolecular patternsthrough an elastic creasing instability ofstimuli-responsive hydrogelsJungwook Kim, Jinhwan Yoon and Ryan C. Hayward*Surfaces with physicochemical properties that can be modulated using external stimuli offer great promise for designingresponsive or adaptive materials. Here, we describe biocompatible dynamic scaffolds based on thin hydrogel coatings thatreversibly hide and display surface chemical patterns in response to temperature changes. At room temperature, the gelabsorbs water, triggering an elastic creasing instability that sequesters functionalized regions within tight folds in the surface.Deswelling at ∼37 ◦C causes the gel surface to unfold, thereby regenerating the biomolecular patterns. Crease positionsare directed by topographic features on the underlying substrate, and are translated into two-dimensional micrometre-scalesurface chemical patterns through selective deposition of biochemically functionalized polyelectrolytes. We demonstratespecific applications of these dynamic scaffolds—selective capture, sequestration and release of micrometre-sized beads,tunable activity of surface-immobilized enzymes and reversible encapsulation of adherent cells—which offer promise forincorporation within lab-on-a-chip devices or as dynamic substrates for cellular biology.

Stimuli-responsive surfaces1–3 that show dynamicallycontrolledwettability4–11, adhesion12–14, optical properties15,16,cellular interactions17–20 anddisplay of proteinmicropatterns21

have been demonstrated using a number of creative approaches. Ofparticular interest aremethods bywhich a range of actuating stimuliincluding light10,22, temperature13,16, electric potentials17,18, solventenvironment9, ionic strength21 and mechanical loads20 can be usedto spatiotemporally generate or erase chemical patterns on thesurfaces. Although these pioneering methods to achieve dynamicchemical patterns have openedmany newpossibilities for the designof responsive and adaptive materials, they are typically limited in atleast one of three ways. The first is reversibility: when the patternedfunctionalities are deactivated by desorption from the substrate,reactivation requires the functional molecules to be re-patterned17or replenished18,21, constraining the ability of substrates to undergorepeated switching. The second is the choice of functionality: inmany cases, the motifs that give rise to stimuli-responsiveness,for example, azobenzene moieties or poly(N -isopropylacrylamide)(NIPAM) chains, are directly coupled to the resulting modulationin chemical functionality9,10,13, thereby restricting the types ofchemical contrast that can be obtained with a given platform. Thethird is pattern geometry: although micro-cracking of thin coatingson flexible substrates provides a powerful route to create chemicalpatterns actuated by mechanical strains, the patterns obtained arelimited to quasi-periodic structures of parallel or criss-crossinglines20. Here, we demonstrate biocompatible dynamic scaffoldsbased on an elastic creasing instability23–25 of surface-attachedhydrogels that overcome these limitations as follows: (1) patternedsurface functionalities are ‘deactivated’ by sequestration withinsurface folds, enabling repeated cycles of reactivation and deac-tivation by reversible unfolding and folding of creases; (2) thechoice of surface chemical functionality—applied by first graftingto poly(l-lysine) (PLL) and then adsorbing to the gel surface—isdecoupled from the responsive element (a NIPAM copolymer gel),

Department of Polymer Science and Engineering, University of Massachusetts, Amherst, Massachusetts 01003, USA.*e-mail: rhayward@mail.pse.umass.edu.

enabling great flexibility in the choice of functional (bio)moleculesdisplayed; and (3) the surface chemical patterns are spatiallyguided by topographic features on the underlying rigid substrates,providing access to nearly arbitrary two-dimensional patterns ofline segments on length scales down to at least several micrometres.Whereas the relatively slow timescale for swelling and deswelling ofgels is a well-known disadvantage of stimuli-responsive hydrogels,folding and unfolding of creases occurs fairly rapidly for the thingel layers considered here, for example in less than one second for∼10-µm-thick scaffolds.

The fabrication of dynamic scaffolds (Fig. 1a–h) begins bypolymerization of covalently crosslinked poly(NIPAM-co-sodiumacrylate) hydrogel films on rigid substrates. These copolymer net-works are well known to undergo a deswelling transition in aqueousmedia as the temperature is raised26. Polymerization is carried outbetween two parallel glass slides separated by a spacer to definethe gel thickness (8–160 µm); one of the slides is functionalizedwith methacrylate groups to anchor the gel to its surface. Onimmersion into aqueous media, swelling of the surface-attachedgel is constrained by the rigid glass substrate to occur only in thedirection normal to the surface, generating a biaxial compressivestress within the gel. Beyond a critical degree of swelling (when thethickness of the surface-attached gel grows by a factor of 2.0 fromits value at the time of network formation), the surface becomesunstable to an elastic creasing instability, wherein the surface formsa biaxial pattern of tight folds with a characteristic spacing similar tothe swelled gel thickness24,25. As previously demonstrated, swellingand deswelling of the surface-attached gel owing to changes intemperature can trigger reversible formation and disappearanceof creases27. Here we tune the composition of charged monomer(sodium acrylate, NaAc) and crosslinker (bisacrylamide) in thenetwork such that the transition between creased and flat statesoccurs under physiologically relevant conditions—that is, inphosphate buffered saline (PBS) at temperatures of 33–37 ◦C.

NATUREMATERIALS | VOL 9 | FEBRUARY 2010 | www.nature.com/naturematerials 159© 2010 Macmillan Publishers Limited. All rights reserved.

ARTICLES NATURE MATERIALS DOI: 10.1038/NMAT2606

CoverslipSpacers

NIPAMBisAAm

NaAc

APSTEMEDWater Hidden areas

within creases

23 οC

37 οCin PBS

100 μm

20 μm200 μm

23 οC 37 οC

AF

a c d

efgh

i j k

oo

SiCl3

b

Temperature

Deviation (μ

m)

3020100¬10¬20¬30

1.00

0.95

0.90

0.85

AF

LH H H H HL L L L L

Figure 1 | Fabrication and characterization of scaffolds with dynamic biomolecular patterns. a, A topographically patterned substrate is lithographicallyfabricated. b,c The substrate is modified with 3-(methacryloxy)propyltrichlorosilane to promote covalent anchoring of the hydrogel layer, and assembledwith a bare coverslip and spacers that define the hydrogel thickness into a channel (b) into which an aqueous pre-gel solution is loaded by capillary actionand left for 30 min to polymerize (c). BisAAm: bisacrylamide, APS: ammonium persulphate, TEMED: N,N,N′,N′-tetramethylethylenediamine. d, Afterdetaching the coverslip, the surface-attached hydrogel is swelled, causing creases to form above the centres of elevated features on the substrate, therebyhiding certain areas of the surface. e–g, Areas that remain exposed are coated with PLL-g-PEG (green) (e), then the hydrogel is deswelled to expose thepreviously hidden surface areas (f), into which a polyelectrolyte containing the desired biomolecular ligand (PLL-g-PEG-X, red) is selectively backfilled (g).h, Ligand-functionalized areas are then reversibly hidden and displayed through variations in temperature. i, In-plane projected images taken from LSCM at37 ◦C for hydrogels bound to substrates patterned with concentric circles (left) and parallel lines with edge-to-edge separations corresponding to thedisplayed numbers of microns (right). j, The actuation of dynamic patterns is revealed by LSCM images (in-plane projections above and cross-sectionsbelow) of the hydrogel bound to the substrate lithographically patterned with the letters ‘umass’ (inset: optical micrograph of the substrate pattern) at37 ◦C (top) and at 23 ◦C (bottom). k, Reproducibility of switching for 30-µm-wide TRITC stripes templated by 60-µm-wide raised parallel lines on thesubstrate is demonstrated by measuring the fraction of the exposed surface coated by FITC (AF, yellow lines) and deviations in crease locations from thefirst to subsequent cycles (blue lines) and from the centres of elevated substrate features (error bars). ‘L’ and ‘H’ represent temperatures of 23 ◦C and37 ◦C, respectively.

Control over crease locations is achieved by preparing hydrogelson substrates with lithographically patterned topographic features(Fig. 1a–c); creases are found to form selectively over the centresof elevated features on the substrate. The positions of folds arethen translated into surface chemical patterns by first depositingPLL tagged with a green fluorophore (fluorescein isothiocyanate,FITC) onto the creased gel surface (Fig. 1e), which saturates theaccessible surface area but does not modify the folded regions, thenunfolding the creases and selectively backfilling the formerly hiddenregions with functionalized PLL tagged with a red fluorophore(tetramethyl rhodamine isothiocyanate, TRITC; Fig. 1g). Thispatterning method—which is examined in greater detail below—allows for the preparation of patterns with feature sizes down to atleast a few micrometres (Fig. 1i).

We demonstrate the actuation of these dynamic chemicalpatterns by means of changes in temperature using a TRITC-modified surface pattern formed on a substrate with raisedtopographic features that spell out ‘umass’ (Fig. 1j, inset). At 37 ◦C(Fig. 1j, top), the gel is only slightly swelled relative to its volumeat the time of polymerization, leading to a nearly flat surfacewhere the TRITC-coated letters are fully exposed. On loweringthe temperature to 23 ◦C (Fig. 1j, bottom), laser scanning confocalfluorescence microscopy (LSCM) reveals that the TRITC-stainedregions become entirely hidden within the folds in the surface. Toquantify the reversibility of this process through repeated cyclesof swelling (23 ◦C) and deswelling (37 ◦C), we track the arealfraction of exposed surface coated by FITC (AF) and the averagedeviations of creases from the first to subsequent cycles for a

patterned hydrogel on a substrate with a series of 60-µm-wideraised parallel stripes. It can clearly be seen from Fig. 1k that AFvaries reproducibly between 1.00 (TRITC fully hidden) and ∼0.85(TRITC fully exposed) and that creases form in nearly identicallocations during each cycle, with average spatial deviations ofonly 0.45 µm, or less than 1% of the width of the templatingfeatures. By designing the substrates to fold slightly more than theamount required to completely hide the functionalized regions, itis straightforward to ensure that these small variations in creasepositions do not lower the efficiency of sequestration.

We use PLL as a platform for surface chemical modificationbecause, as illustrated in Fig. 2a, it provides essentially irreversiblephysical adsorption to the anionic gel surface28 as well asstraightforward attachment of any number of functionalities29,including fluorophores, grafted poly(ethylene glycol) (PEG) chainsto reduce nonspecific adsorption, and functional biomolecules30.To create dynamic biomolecular surface patterns, we first depositFITC-tagged PLL-g -PEG in the creased state. This saturates theaccessible surface, preventing subsequent adsorption of the secondpolymer in these regions, but diffusion of PLL-g -PEG into the gelis limited enough that the folded regions are effectively maskedoff by self-contact. In the second deposition step—conductedin the unfolded state—a similar TRITC-labelled graft copolymerbearing the desired biochemical ligand at the end of the PEG grafts(PLL-g -PEG-X) is used to selectively backfill the previously hiddensurface areas. To demonstrate the concept, we focus on two widelyused and broadly useful ligands, biotin and the integrin-bindingarginine–glycine–aspartate (RGD) tripeptide motif 31. However,

160 NATUREMATERIALS | VOL 9 | FEBRUARY 2010 | www.nature.com/naturematerials

© 2010 Macmillan Publishers Limited. All rights reserved.

NATURE MATERIALS DOI: 10.1038/NMAT2606 ARTICLES

O

NH N

H

O

NHOO O¬

Na+

xy

O

O

O

O4

4 4

nN

O

O

OO X

O

4

S C NFluorophore

+ +

HN

HN

HN

HN

HN

NH3

NH2 4

NH2

NH3 +NH3

18

20

1614121086420

0 10 20 0 10 20

0.8

0.6

0.4

0.2

0

a b

PLL-g-PEG-XPLL-g-PEG

Distance from surface (µm) Distance from surface (µm)

Lysi

ne c

once

ntra

tion

(mM

) Lysine concentration (mM

)

Figure 2 | Patterning functional polyelectrolytes on dynamic scaffolds. a, A schematic diagram showing the modification of PLL chains with fluorophoresand PEG grafts bearing terminal ligands (X), as well as adsorption of the graft copolymer onto a poly(NIPAM-co-NaAc) gel. b, The high degree of chemicalcontrast obtained between the green regions that are always exposed (left) and the red regions that form the crease interiors (right) is demonstrated bythe concentration of PLL-g-PEG-X (red) and PLL-g-PEG (green) determined by LSCM (concentrations expressed on the basis of lysine residues).

a wide variety of functional biomolecules can be introduced ina similar fashion by using biotin–avidin interactions to attachbiotinylated species to surfaces patterned with PLL-g -PEG-biotin,or commercially available heterobifunctional PEGs with an amine-reactive N -hydroxy succinimide ester at one end and variousfunctional groups such as an azide, maleimide or protectedhydrazide or thiol at the other. To quantify the chemical contrastobtained using this patterning technique, we compare the amountsof FITC-labelled PLL-g -PEG and TRITC-labelled PLL-g -PEG-Xdeposited in the two regions of the substrate (Fig. 2b). Within thedetection limits of the LSCM, the ‘green’ regions (that is, thosealways exposed on the surface) are coated exclusively with PLL-g -PEG, whereas the red regions (that is, those reversibly hidden anddisplayed by creasing) are coated exclusively with PLL-g -PEG-X .We note that appreciable diffusion of each polymer into the geldoes occur over several micrometres; for example, in Fig. 2b, ameasurable concentration of PLL-g -PEG persists to∼10 µm belowthe gel surface. This sets a lower limit on the resolution that canbe obtained, because it allows the PLL-g -PEG layer to penetratesome distance into the folded regions. In practice, feature sizesof a few micrometres are readily obtained (Fig. 1i, right), and weestimate the lateral resolution of the technique as roughly 2 µmon the basis of the width of the transition region at the edge ofpatterned features (see Supplementary Fig. S1). We anticipate thatfurther optimization of the conditions for surface-functionalizationusing higher-molecular-weight PLL or functionalized cationicnanoparticles should yield submicrometre resolution.

To understand how crease locations are directed by thetopographic patterns on the underlying substrate, we conductfinite-element analysis of the swelling of a hyperelastic gel32 attachedto a substrate with a step change in thickness (Fig. 3a). Sufficientlyfar from the step, the gel swells uniformly in the direction normalto the surface, whereas in the vicinity of the step a more complexdeformation field results. Consideration of the in-plane strain field(εx) reveals that the lateral expansion of the thicker portion of thegel induces a region of compressive strain in the thinner portion,with themaximum in-plane compression occurring a characteristicdistance d from the pattern edge (indicated by an arrow in Fig. 3a).This enhanced compression has two effects: it reduces the criticaldegree of vertical expansion necessary to trigger crease formation,from 2.0 for a flat substrate25, to 1.4 for the patterns shown in Fig. 1;and it templates the location at which the crease will form, that is,at the position of maximum in-plane compression. Numerically,we find that d scales linearly with, and is nearly equal to, the initialgel thickness in the thinner region, h0 (see Supplementary Fig. S2).Experimentally, when the width of the elevated features on thesubstrate (w) exceeds roughly 3d , two independent creases form,each a distance d from the corresponding edge of the pattern. As w

is reduced below 3d , these two creases merge into a single creasealong the centre of the pattern (see Supplementary Table S1); asboth d and the natural spacing between creases (λc) are similar inmagnitude to h0, it follows that the boundary between single anddouble crease formation occurs atw≈2d+λc≈3h0≈3d .

To demonstrate the utility of these dynamic patterned surfaces,we describe three example applications. In the first case, a scaffoldpatterned with regions of biotin is incubated with 2-µm-diameterstreptavidin-functionalized polystyrene beads. At room tempera-ture, the biotinylated regions are hidden within creases; thus, beadsnear the gel do not adhere to the surface, as demonstrated by theircontinued Brownian motion, and their ability to be completelydisplaced by a room-temperature PBS rinse. When the temperatureis raised to 37 ◦C during incubation however, the biotin-coatedregions become exposed, and beads stick to these regions overa timescale of ∼5min by means of strong biotin–streptavidinassociations (causing cessation of Brownian motion). Rinsingthe scaffold now leads to removal of any beads settled on thePEG-coated regions, whereas the beads adhered to the biotin-coatedregions are retained (Fig. 4a, top). As this scaffold is cooled to roomtemperature and creases form, the beads are either detached fromor encapsulated within the hydrogel, depending on the strengthof adhesion to the surface. Specifically, when the concentration ofbiotin in the near-surface region is ∼2.5 µM, beads are squeezedout of the creases as they form, completely detaching the beads andallowing them to resume Brownian motion (Fig. 4a, bottom right,and SupplementaryMovie S1). In this case, we roughly estimate theadhesive energy of a bead asUadh∼R2ΣbUb∼10−17 J, where R is thebead radius, Σb is the areal density of accessible biotins (estimatedas 75 µm−2 for a 50-nm-thick zone of interaction) and Ub is thebinding energy for a single association (31 kBT). This value is twoorders of magnitude lower than the estimated penalty for the gelto deform elastically around a bead, Uel ∼ R3G∼ 10−15 J, where Gis the gel modulus (∼1 kPa); thus, it follows that the elastic energyassociated with crease formation is sufficient to displace beads fromthe surface.However, when the density of biotin is increased 20-fold(to ∼50 µM or 1,500 µm−2), the beads remain fixed to the surfacethroughout, becoming encapsulated within the creases to depths of6–20 µmbelow the free surface, as dictated by the positions at whichthey initially adhered (Fig. 4a, bottom left, and SupplementaryFig. S3). By raising the temperature to 37 ◦C, the encapsulated beadsare once again displayed on the surface; although the adhesion ofeach individual bead is sufficient to force the gel to locally deform,the areal density of beads is sufficiently small that they do not‘crosslink’ the surface in the folded state. Surprisingly, we do notobserve beads swapping from one side of the fold to the other,suggesting that the initially formed bead/surface contact remainsstronger than the adhesion developed to the opposing surface

NATUREMATERIALS | VOL 9 | FEBRUARY 2010 | www.nature.com/naturematerials 161© 2010 Macmillan Publishers Limited. All rights reserved.

ARTICLES NATURE MATERIALS DOI: 10.1038/NMAT2606

Substrate

Ho ho h

d

x

y

xεa

b

0+ 0.173 ¬0.288

1.8

1.6

1.4

1.2

1.020 25 30 35 40

h/h o

Temperature (°C)

Figure 3 | The role of substrate topography in directing crease formation.a, Finite-element simulations of the swelling of a hyperelastic gel attachedto a substrate with a step change in thickness reveal a region of in-planecompressive strain (black arrow) that induces formation of creases acharacteristic distance, d, from the step edge. The level of swelling shown ish/h0∼ 1.4, corresponding to the critical value of h/h0 to trigger the creaseformation for the surface-attached hydrogel with the identical geometry.The white dashed line represents the initial position of the unswelled gelsurface. b, The vertical swelling of a hydrogel of identical composition tothose in Figs 1 and 2 attached to a planar surface plotted againsttemperature; for the step pattern geometry in a, crease formation occurs at33 ◦C (h/h0= 1.4, dashed lines), well below the critical swelling requiredfor creasing on a flat substrate (h/h0= 2.0). The error bars denote thestandard deviations for seven independent measurements at eachisothermal equilibrium (see Supplementary Information for details).

within a crease, probably reflecting the fact that in the swollen state,the surface of the gel is expanded and therefore presents a lowerareal density of biotin.

We next demonstrate spatiotemporal control over the catalyticactivity of dynamic hydrogel surfaces patterned with enzymes.As shown schematically in Fig. 4b (top), biotinylated lipase isconjugated to streptavidin-functionalized polystyrene beads, whichare then attached to the gel surface using avidin to couple biotinon the lipase-coated beads to the biotin-functionalized surfaceregions. This patterned scaffold is incubated at 26 and 38 ◦Cwith thechromogenic lipase substrates fluorescein diacetate (FDA) or 1,2-o-dilauryl-rac-glycero-3-glutaric acid-(6′-methylresorufin)ester

(DGGR; ref. 33), allowing spectrophotometric determinationof enzymatic hydrolysis. At 26 ◦C, the lipase-coated beads areentirely sequestered within the surface creases, thereby reducingthe effective catalytic activity of the dynamic scaffold. To isolatethe effects of sequestration from those of temperature alone, wecompare the activity of the dynamic scaffold to that of a staticscaffold that is otherwise identical but is coated with lipase overits entire surface and does not form creases owing to the lackof topographic substrate features. In both cases, we normalizethe measured activity by that of the corresponding scaffold at38 ◦C to compensate for differences in areal coverage of lipase.As shown in Fig. 4b (bottom), the normalized activity (AN) of thestatic scaffold (average of two trials) at 26 ◦C is only 40% towardsFDA and 83% towards DGGR, reflecting the inherent reductionin catalytic activity with decreasing temperature. For FDA as asubstrate, the dynamic scaffold shows a further reduction in AN to12%, indicating that sequestrationwithin creases is responsible for a3.3-fold reduction in the effective catalytic activity.Wenote that twoconsecutive trials were carried out with each scaffold to ensure thatthe observed reduction in catalytic activity was not simply the resultof irreversible denaturation of the enzyme on folding (see Supple-mentary Table S2). Given the small size of an FDAmolecule relativeto the typical pore size of the gel, it is perhaps surprising that anyreduction in activity occurs; however, we note that the resistance tomass transport provided by the gelmay be augmented both owing tothe barrier of PLL-g -PEG chains adsorbed to the gel surface, as wellas the increase in network density that occurs in the vicinity of thecrease relative to the rest of the gel (see Supplementary Fig. S4). ForDGGR as a substrate, sequestration of lipase is considerably moreeffective in reducing the catalytic activity of the dynamic scaffold,reducing AN 11-fold to 7.5%. As DGGR is water-insoluble, it isdispersed as submicrometre droplets that presumably face greatermass-transport barriers to reach the enzyme in the folded state.We anticipate that further tuning of substrate mass-transport rateswill allow much larger ‘on/off’ ratios in enzymatic activity than thesingle order of magnitude obtained so far.

Finally, we demonstrate the use of dynamic patterned substratesto provide reversible encapsulation of adherent cultured cells.Incubation of porcine epithelial cells on RGD-patterned scaffolds at37 ◦C in culture medium leads to selective adhesion and spreadingof cells in the RGD-functionalized areas (Fig. 4c, left). When thetemperature is lowered to 23 ◦C, the cells are pulled within creases,becoming effectively encapsulated within the gel (Fig. 4c, right).Repeated cycles of folding and unfolding are possible, with cellsmaintaining adhesion to the surface and showing no sign of reducedviability throughout (as determined by trypan blue staining; seeSupplementary Information for details). This represents an entirelynew approach to achieve reversible three-dimensional encapsula-tion of cells within a hydrogel, and may be used to temporarilysequester or protect adherent cells, or to study how cell behaviourschange on a transition from growth on a surface to that withina matrix. We note that in the folded state, the cells experience anearly uniaxial compressive stress of the order of the elasticmodulusof the gel (∼1 kPa); although at present this stress remains poorlycharacterized, these substrates provide potential opportunities forapplying forces to cells or characterizing cellmechanical properties.

We foresee this platform finding applications in a variety ofcontexts, for example: as components of microfluidic devices thatallow cells or target objects to be selectively captured, protected, re-leased and sorted; as dynamic substrates for cell culture that providespatiotemporal control of cell growth and exposure to insolublebiomolecules as well as possibilities for mechanically interrogatingcells; and as coatings for biomedical devices that provide triggeredrelease of therapeutic agents or changes in their interactions withsurrounding tissues. Efforts are underway to extend the flexibilityof this approach by generalizing to other actuating stimuli, for

162 NATUREMATERIALS | VOL 9 | FEBRUARY 2010 | www.nature.com/naturematerials

© 2010 Macmillan Publishers Limited. All rights reserved.

NATURE MATERIALS DOI: 10.1038/NMAT2606 ARTICLES

High surfacebiotin density

Low surfacebiotin density

Brownianmotion

10 µm

a b c

O

O

OH

O

OO

O

O

OO O

O O

O

O

O

O

O

O O N

N

O¬O

11

11

3

20 µm

CH3 CH3 CH3

CH3H3C

CH3

CH3

CH2

CH2HO

100

80

60

40

20

0A

N (

%)

Static scaffolds Dynamic scaffolds

FDADGGR

Biotin

Methylresorufin

DGGR

Streptavidin

Avidin

Hydrogel

PLL-g-PEG

PSbead

Fluorescein

Lipase

FDA

Figure 4 |Applications of dynamic hydrogel scaffolds. a, Streptavidin-coated beads selectively adhere to biotin-functionalized regions of the surface at37 ◦C (top), and are either encapsulated within creases (bottom left) or detached from the surface at 23 ◦C (bottom right), depending on the surfacedensity of biotin; beads and hydrogel surfaces are imaged by LSCM and optical microscopy, and the underlying substrate pattern is drawn to scale. (Thehourglass shape of beads in the inset at the bottom left is an artefact owing to their high fluorescence intensity.) b, A schematic representation oflipase-patterned dynamic scaffolds (top) and the normalized enzymatic activities of static and dynamic scaffolds towards substrates FDA (green) andDGGR (red) at 26 ◦C. PS: polystyrene. The error bars denote maximum and minimum values of AN from two trials. c, Optical micrographs of epithelial cellsselectively adhered to RGD-functionalized regions of the scaffold at 37 ◦C (left) and encapsulated within creases at 23 ◦C (right).

example, light34 or the presence of specific biomolecules35, andby achieving local actuation of creases to enable simultaneous,independent, dynamic patterns ofmultiple different biomolecules.

MethodsHydrogelation on patterned adhesive substrates. A master template of thedesired topographic substrate features was fabricated by photolithography usingSU-8 photoresist (MicroChem) spin-coated on a silicon wafer cleaned by Piranhasolution (H2O2 added slowly to concentrated H2SO4 in a volumetric ratioof 3:7). A polydimethylsiloxane negative stamp was made by curing Sylgard184 (Dow Corning, 10wt% of crosslinking agent) on the template for 2 h at70 ◦C. The topographic features of the master were then readily replicated byphoto-curing NOA81 (Norland Optical Adhesive 81, Norland Products) betweena coverslip and the patterned polydimethylsiloxane stamp, using either of twosoft lithographic techniques36—micromoulding in capillaries or microtransfermoulding—depending on the pattern geometry. The surfaces of crosslinkedNOA81replicas were activated with an oxygen plasma for 20min followed by silanizationwith [(3-methacryloxy)propyl]trichlorosilane (Gelest) in ethanol (0.4 vol%) for6 h. A 200 µl aqueous pre-gel solution with concentrations of 809mM NIPAM,6.3mM BisAAm and 80.6mM NaAc was degassed and backfilled with nitrogen,mixed with 0.3 µl of N ,N ,N ′,N ′-tetramethylethylenediamine and 1.0 µl of a10wt% aqueous ammonium persulphate solution, then rapidly loaded inside thecapillary channel formed by the silane-treated NOA81 replica and a bare coverslip,separated by spacers to define the gel thickness (8–160 µm, Kapton polyimide filmsfrom DuPont or no. 1 glass coverslips). Gelation was allowed to proceed for 30minbefore separating the bare coverslip from the gel surface.

Deposition of polylysine graft copolymers. To fabricate the scaffold described inFig. 1, the surface-attached hydrogel was immersed in the first deposition solutioncontaining 1.0mg of fluorescein-tagged PLL-g -PEG in 15ml of 20-fold diluted PBS(7mMNaCl) at 68 ◦C for 90min. This low-ionic-strength buffer was used to restrictpenetration of PLL-g -PEG beneath the gel surface by increasing the hydrodynamicsize of the polyelectrolye PLL-g -PEG chains, retarding their diffusion into thegel, and promoting electrostatic interaction between the polyelectrolytes and thegel. A high temperature (68 ◦C) was maintained during deposition to keep theswelling of the poly (NIPAM-co-NaAc) gel at low ionic strength slightly less than infull-strength PBS at room temperature. After the initial deposition in the creasedstate, the hydrogel was washed with warm PBS (40 ◦C) several times to removeexcess PLL-g -PEG on the surface. A 100 µl volume of the second depositionsolution, containing 1.0mg of PLL-g -PEG-X in 1ml of PBS, was next depositedon the scaffold at 40 ◦C for 5 s in a temperature-controlled convection oven and

the scaffold was washed with warm PBS (40 ◦C) several times. In the case ofPLL-g -PEG-biotin, the 20-fold reduced areal density of biotin was achieved bydepositing a 1:19mixture of PLL-g -PEG-biotin with PLL-g -PEG.

Lipase assay. The lipase activity assay using DGGR (Sigma-Aldrich) was conductedby incubating three different substrates—a static scaffold (S) which was coatedwith lipase-functionalized beads over its entire surface area, and which did notundergo creasing owing to the absence of topographic substrate features; a dynamicscaffold (D) patterned with lipase beads only in the regions that formed the interiorof the folds; and a control scaffold (C) not coated with lipase beads to determinebackground hydrolysis rates in the absence of enzymatic activity—in solutionsof DGGR at 26 and 38 ◦C for 10 h. The substrate solutions were freshly preparedbefore the incubation by adding 50 µl of DGGR solution (2mg DGGR in 1mldioxane, stored at −20 ◦C) in 2ml of 40mM Tris-HCl buffer (pH 8.0, NaCl0.14M) with vortexing. The incubated substrate solution was centrifuged to collectthe supernatant, and absorption of hydrolysed methylresorufin was measured at580 nm (Abs) by ultraviolet–visible spectrophotometry (U-3010, HITACHI). Thenormalized activity (AN) at 26 ◦Cwas calculated as follows:

AN(S or D;26 ◦C)=Abs(S or D;26 ◦C)−Abs(C;26 ◦C)Abs(S or D;38 ◦C)−Abs(C;38 ◦C)

The lipase assay using FDA (Sigma-Aldrich) was conducted using an identicalprocedure except that absorption was measured at 491 nm and the substratesolution was prepared by adding 20 µl of FDA solution (1mg FDA in 1ml acetone,stored at −20 ◦C) in 2ml of 40mM Tris-HCl buffer (pH 8.0, NaCl 0.14M)with vortexing. A super-saturated concentration of FDA was used with theaim of maintaining the maximum concentration of free FDA throughoutthe 10 h incubation.

See Supplementary Information for detailed methods related to synthesisof polylysine graft copolymers, characterization of dynamic scaffolds, lipaseimmobilization on scaffolds and cell culture on scaffolds.

Received 28 January 2009; accepted 16 November 2009;published online 20 December 2009

References1. Mather, P. T. Soft answers for hard problems. Nature Mater. 6, 93–94 (2007).2. Russell, T. P. Surface-responsive materials. Science 297, 964–967 (2002).3. Lahann, J. & Langer, R. Smart materials with dynamically controllable surfaces.

MRS Bull. 30, 185–188 (2005).

NATUREMATERIALS | VOL 9 | FEBRUARY 2010 | www.nature.com/naturematerials 163© 2010 Macmillan Publishers Limited. All rights reserved.

ARTICLES NATURE MATERIALS DOI: 10.1038/NMAT2606

4. Wang, R. et al. Light-induced amphiphilic surfaces. Nature 388,431–432 (1997).

5. Ichimura, K., Oh, S. K. & Nakagawa, M. Light-driven motion of liquids on aphotoresponsive surface. Science 288, 1624–1626 (2000).

6. Lahann, J. et al. A reversibly switching surface. Science 299, 371–374 (2003).7. Sun, T. L. et al. Reversible switching between superhydrophilicity and

superhydrophobicity. Angew. Chem. Int. Ed. 43, 357–360 (2004).8. Fu, Q. et al. Reversible control of free energy and topography of nanostructured

surfaces. J. Am. Chem. Soc. 126, 8904–8905 (2004).9. Ionov, L. et al. Reversible chemical patterning on stimuli-responsive

polymer film: Environment-responsive lithography. J. Am. Chem. Soc. 125,8302–8306 (2003).

10. Lim, H. S., Han, J. T., Kwak, D., Jin, M. H. & Cho, K. Photoreversiblyswitchable superhydrophobic surface with erasable and rewritable pattern.J. Am. Chem. Soc. 128, 14458–14459 (2006).

11. Chung, J. Y., Youngblood, J. P. & Stafford, C. M. Anisotropic wetting ontunable micro-wrinkled surfaces. Soft Matter 3, 1163–1169 (2007).

12. Crevoisier, G. B., Fabre, P., Corpart, J. M. & Leibler, L. Switchabletackiness and wettability of a liquid crystalline polymer. Science 285,1246–1249 (1999).

13. Jones, D. M., Smith, J. R., Huck, W. T. S. & Alexander, C. Variable adhesionof micropatterned thermoresponsive polymer brushes: AFM investigationsof poly (N-isopropylacrylamide) brushes prepared by surface-initiatedpolymerizations. Adv. Mater. 14, 1130–1134 (2002).

14. Lin, P. C., Vajpayee, S., Jagota, A., Hui, C. Y. & Yang, S. Mechanicallytunable dry adhesive from wrinkled elastomers. Soft Matter 4,1830–1835 (2008).

15. Kang, Y., Walish, J. J., Gorishnyy, T. & Thomas, E. L. Broad-wavelength-rangechemically tunable block-copolymer photonic gels. Nature Mater. 6,957–960 (2007).

16. Hu, Z. B., Chen, Y. Y., Wang, C. J., Zheng, Y. D. & Li, Y. Polymer gelswith engineered environmentally responsive surface patterns. Nature 393,149–152 (1998).

17. Jiang, X. Y., Ferrigno, R., Mrksich, M. & Whitesides, G. M. Electrochemicaldesorption of self-assembled monolayers noninvasively releases patterned cellsfrom geometrical confinements. J. Am. Chem. Soc. 125, 2366–2367 (2003).

18. Yeo, W. S., Yousaf, M. N. & Mrksich, M. Dynamic interfaces between cellsand surfaces: Electroactive substrates that sequentially release and attach cells.J. Am. Chem. Soc. 125, 14994–14995 (2003).

19. Akiyama, Y., Kikuchi, A., Yamato, M. & Okano, T. Ultrathinpoly(N-isopropylacrylamide) grafted layer on polystyrene surfaces forcell adhesion/detachment control. Langmuir 20, 5506–5511 (2004).

20. Zhu, X. Y. et al. Fabrication of reconfigurable protein matrices by cracking.Nature Mater. 4, 403–406 (2005).

21. Frey, W., Meyer, D. E. & Chilkoti, A. Dynamic addressing of a surface patternby a stimuli-responsive fusion protein. Adv. Mater. 15, 248–251 (2003).

22. Sikes, H. D. et al. Using polymeric materials to generate an amplified responseto molecular recognition events. Nature Mater. 7, 52–56 (2008).

23. Tojo, E., Nagao, K., Miura, T. & Nagamoto, S. in Photographic Gelatin(ed. Cox, R. J.) 49–61 (Academic, 1972).

24. Tanaka, T. et al. Mechanical instability of gels at the phase-transition. Nature325, 796–798 (1987).

25. Trujillo, V., Kim, J. & Hayward, R. C. Creasing instability of surface-attachedhydrogels. Soft Matter 4, 564–569 (2008).

26. Hirotsu, S., Hirokawa, Y. & Tanaka, T. Volume-phase transitions of ionizedN-isopropylacrylamide gels. J. Chem. Phys. 87, 1392–1395 (1987).

27. Li, C. F., Hu, Z. B. & Li, Y. Temperature and time dependencies of surfacepatterns in constrained ionic N-isopropylacrylamide gels. J. Chem. Phys. 100,4645–4652 (1994).

28. Bysell, H. & Malmsten, M. Visualizing the interaction between poly-l-lysineand poly(acrylic acid) microgels using microscopy techniques: Effect ofelectrostatics and peptide size. Langmuir 22, 5476–5484 (2006).

29. Hermanson, G. T. Bioconjugate Techniques 1st edn (Academic, 1996).30. Kenausis, G. L. et al. Poly(l-lysine)-g-poly(ethylene glycol) layers on metal

oxide surfaces: Attachment mechanism and effects of polymer architecture onresistance to protein adsorption. J. Phys. Chem. B 104, 3298–3309 (2000).

31. VandeVondele, S., Voros, J. & Hubbell, J. A. RGD-graftedpoly-l-lysine-graft-(polyethylene glycol) copolymers block non-specificprotein adsorption while promoting cell adhesion. Biotechnol. Bioeng. 82,784–790 (2003).

32. Hong, W., Liu, Z. & Suo, Z. Inhomogeneous swelling of a gel in equilibriumwith a solvent and mechanical load. Int. J. Solids Struct. 204, 3282–3289 (2009).

33. Panteghini, M., Bonora, R. & Pagani, F. Measurement of pancreatic lipaseactivity in serum by a kinetic colorimetric assay using a new chromogenicsubstrate. Ann. Clin. Biochem. 38, 365–370 (2001).

34. Kamenjicki, M., Lednev, I. K., Mikhonin, A., Kesavamoorthy, R. &Asher, S. A. Photochemically controlled photonic crystals. Adv. Funct. Mater.13, 774–780 (2003).

35. Miyata, T., Asami, N. & Uragami, T. A reversibly antigen-responsive hydrogel.Nature 399, 766–769 (1999).

36. Zhao, X. M., Xia, Y. N. & Whitesides, G. M. Soft lithographic methods fornano-fabrication. J. Mater. Chem. 7, 1069–1074 (1997).

AcknowledgementsWe are grateful to Z. Suo and X. Zhao for helpful discussions and for providing theuser-defined Abaqus subroutines used to model swelling of hyperelastic gels, and toP. Wadsworth and C. Fagerstrom for providing cells and assisting with cell cultureexperiments. This work was primarily financially supported by the National ScienceFoundation through grant DMR-0747756 with further support provided by a 3MNontenured Faculty Grant, and made use of facilities supported by the NSF MRSEC atUMass (DMR-0820506) and NSF grant BBS-8714235. J.Y. is partially supported by theKorean Research Foundation (KRF-2008-357-D00079).

Author contributionsR.C.H. and J.K. designed the research project and experiments. J.K. carried out the bulkof experiments and simulations; J.Y. characterized temperature-dependent swellingbehaviours of the gel networks. R.C.H. and J.K. primarily wrote the paper with inputand comments from J.Y.

Additional informationThe authors declare no competing financial interests. Supplementary informationaccompanies this paper on www.nature.com/naturematerials. Reprints and permissionsinformation is available online at http://npg.nature.com/reprintsandpermissions.Correspondence and requests formaterials should be addressed to R.C.H.

164 NATUREMATERIALS | VOL 9 | FEBRUARY 2010 | www.nature.com/naturematerials

© 2010 Macmillan Publishers Limited. All rights reserved.