microengineering the ...“top-down” study of cells, beginning with phenotypic cellular behaviors,...

194 MRS BULLETIN • VOLUME 30 • MARCH 2005

IntroductionCells are the fundamental living units of

organisms: the response of an organism todisease, injury, or therapy is the response ofits cells. To be able to model life and predictthe response of organisms to therapeuticor pathological stimuli ultimately requiresthe ability to model the behavior of the cells(individually) and tissues (as collections ofcells) to these stimuli. Thus, characterizingthe full range of cellular behaviors, mappingthese behaviors to normal function and todisease, and reducing these behaviors tomolecular processes are three of the primarygoals of biomedical research.

The study of the molecular aspects ofcells, especially understanding the genomeand how the information it encodes is con-verted into proteins, has exploded in the last

40 years. Understanding how the pheno-type (the actualized cellular characteristics,processes, and behaviors, such as shape,size, growth rate, migratory behavior, andresponse to stimuli) arises from its genotype(molecular composition) is, however, a muchmore complex subject and is only just begin-ning to be explored. Both approaches—the“top-down” study of cells, beginning withphenotypic cellular behaviors, and the“bottom-up” study of nucleic acids, pro-teins, and networks of reactions—mustultimately combine if we are to fully un-derstand the cell.

Even without this full understanding,however, the ability to modulate the pheno-type of cells is potentially enormously use-ful. These behaviors are the result of the

operation of all the processes in living cellsand thus provide integrated, if very com-plex, information about the cellular re-sponse to stimuli (including the passage oftime). As it is not necessary to understandeverything about how a molecule interactswith the human organism in order to de-velop a useful drug, it is also not necessaryto understand the molecular basis of allcellular behaviors to be able to use cells in assays, as sensors, and in engineeredtissues.

These uses of cells do, however, requirethat their responses to stimuli be repro-ducible, especially if they are not entirelycharacterized. In science, it is understoodthat the reproducibility of an experimentis the sine qua non for transforming obser-vational science into reliable engineering.There is no such understanding among cellbiologists. One may well ask how it is pos-sible to do science without the repro-ducibility of experiments, but cells are, infact, very complicated systems and far fromcompletely understood; the fact that met-allurgy places much greater emphasis onreproducibility than cell biology stems fromthe fact that experiments in metallurgyusually are reproducible; those in cell biol-ogy often are not, and it is very difficult tounderstand all the reasons that might con-tribute the answer to the question, why not?

Characterizing cellular behavior requires,at minimum, two capabilities: (1) detailedcontrol over the features of the environmentthat influence the behavior of the cell; and(2) access to cells that are themselves suffi-ciently well defined so that they respondin reproducible ways to environmentalstimuli. Both of these subjects—the envi-ronment of the cell and the cell itself—areincreasingly important research topics astheir centrality to bioscience and biotech-nology becomes more obvious.

Organisms are, of course, composed ofcells. They are equally composed oftissues—functional assemblies of cells (usu-ally of several types) and structural ele-ments. In these tissues, cells are highlyorganized in their physical and chemicalinteractions with their neighbors and withother aspects of their environment. The architecture of the tissue—the spatial organization of cells and surroundingscaffolding—defines the local cues thatmay be experienced by an embedded cell.Such cues might include gradients (of con-centrations) in soluble and immobilized fac-tors (e.g., cell adhesion molecules andextracellular matrix proteins); networks of molecularly mediated adhesive and sig-naling interactions arrayed on adjacentcells and nonliving scaffolds; or static andtime-dependent forces acting on the cells.The complex structure of tissue—from

Microengineering theEnvironment ofMammalian Cellsin Culture

Christopher S. Chen, Xingyu Jiang, and George M.Whitesides

AbstractAssays based on observations of the biological responses of individual cells to their

environment have the potential to make enormous contributions to cell biology andbiomedicine.To carry out well-defined experiments using cells, both the environments inwhich the cells live and the cells themselves must be well defined. Cell-based assays arenow plagued by inconsistencies and irreproducibility, and a primary challenge in thedevelopment of informative assays is to understand the fundamental bases for theseinconsistencies and to limit them. It now seems that multiple factors may contribute tothe variability in the response of individual cells to stimuli; some of these factors may beextrinsic to the cells, some intrinsic. New techniques based on microengineering—especially using soft lithography to pattern surfaces at the molecular level and to fabricatemicrofluidic systems—have provided new capabilities to address the extrinsic factors.This review discusses recent advances in materials science that provide well-definedphysical environments that can be used to study cells, both individually and in groups, inattached culture. It also reviews the challenges that must be addressed in order to makecell-based assays reproducible.

Keywords: biosensors, cell-based assays, diagnostics, microfabrication, microfluidics,PDMS, poly(dimethylsiloxane), soft lithography.

www.mrs.org/publications/bulletin

Microengineering the Environment of Mammalian Cells in Culture

MRS BULLETIN • VOLUME 30 • MARCH 2005 195

molecular to organismic and from static totime-dependent—determines the worldin which the cell lives. Tissue structure iscritical to tissue function, and tissue func-tion reflects collective behaviors of cells.1,2

Although the spatial presentation of envi-ronmental signals at cellular and subcellularlength scales is essential in determiningmany cellular behaviors, understandingthe interactions of cells with their environ-ment has been much less studied than theinternal, molecular workings of the cell.The origin of this disparity lies in the toolsthat are available: molecular biology andbiochemistry have produced a flood of extraordinarily useful tools for exploringthe molecular aspects of the cell;3 there arefewer tools to define and manipulate itsmore macroscopic, external environment.Most of our current understanding of theinteraction between the environment andbehavior of cells is based on the presenta-tion of soluble and insoluble environmentalcues to large populations of a single type(ostensibly) of cell cultured as monolayersheets on the surface of a “substrate” (typi-cally, a lightly surface-oxidized polystyrenePetri dish). These limited conditions forpresenting stimuli to cells have limited ourunderstanding of how cells actually senseand respond to their environment. Newtools are now becoming available that makeit possible to (1) tailor many aspects of thesurface on which attached cells rest at themolecular level; (2) determine the shape andlocation of individual cells; (3) pattern cer-tain aspects of the liquid culture medium atsubcellular scales; and (4) release cells fromcertain environmental constraints at a speci-fied time during the experiment to studythe time-dependence of their behaviors (forexample, subsequent motion across thesubstrate).

The development of tools to build struc-tured environments for cells is making itpossible to study how cues presented withcellular and subcellular spatial granularityaffect cell behavior and function. Resultsfrom experiments carried out in these en-gineered environments establish clearly thatcells have a remarkable ability to senseand respond to their environment.4,5 Theyalso demonstrate that the “cell”—as oftenused in bioengineering and some areas ofcell biology—is ill defined as an experi-mental object and so heterogeneous andirreproducible in its response to stimuli thatthe development of better-defined systemsfor studying cells must be a central infra-structural objective for cell biology andbioengineering.

The understanding that cells in cultureare heterogeneous, but heterogeneous inways that we do not understand and cannoteasily characterize, emphasizes the impor-

tance of reductionist studies of single cells.It also suggests important limitations ofthese studies: a single cell can never speakfor a tissue, any more than an individual canspeak for a crowd.

New Tools to Control theEnvironment of Cells in Culture

The last 10 years have seen the rapid de-velopment of methods of presenting spa-tially and chemically well-defined, soluble,adhesive, and mechanical cues to cells; the spatial resolution of these techniquesextends easily to the 10 �m scale and, with increasing difficulty, to sub-100-nmscales.6,7 The majority of these techniqueshave arisen from two new capabilities:(1) the ability to pattern surfaces with bio-logically active (or inert) molecules and(2) the ability to build microfluidic sys-tems with dimensions relevant to cell biology and that take advantage of thecharacteristic fluid physics of liquids flow-ing in microchannels.

The tools used to pattern surfaces use,and are related to but quite distinct from, thephotolithographic tools used in the micro-electronics industry.6,8,9 Photolithographyis one of the most successful of all tech-nologies within its domain, but it is not wellsuited to the materials needed in cell biol-ogy, nor is it well suited to manipulating organic molecules at the interfaces with asophistication equivalent to that which ispossible and demanded in molecular biology and biochemistry. The emphasis inmicroelectronics is on defect elimination,small feature sizes, extreme accuracy in com-plex patterns, and long device lifetimes;the needs in cell biology are experimentalsimplicity, one-time use of devices, low cost,the ability to produce relatively simplepatterns, and compatibility with cells andbiomolecules. Feature sizes for microfabri-cated devices used in cell biology are usu-ally relaxed, as compared with traditionalmicroelectronic devices (1–100 �m featurescover the needed range of sizes), and de-fects are not important.

The patterning of substrates and the fab-rication of microfluidic systems rely on acommon technology—soft lithography.6,8,10

Many aspects of this technology have beendescribed in reviews; from the vantage ofthe discussions here, their most importantaspects are that they are relatively straight-forward to use experimentally, and theyprovide molecular-level control of surfacesusing the often structurally complex, deli-cate organic molecules required in cell biol-ogy. The extreme experimental care neededto manufacture microelectronic devices isnot required to make patterned (printed andmolded) microsystems for use in cell biol-ogy and bioengineering.

Adhesive cues can now be easily and re-producibly presented to cells in specificgeometries by patterning the adhesive extra-cellular matrix proteins onto cell culture sub-strates.11–13 Soluble factors can be deliveredto cells as gradients4,14 or in concentrationsthat vary with time, using microfluidic sys-tems (Figure 1).These patterning techniqueshave been used to demonstrate the impor-tance of the spatial presentation of solubleand adhesive cues to basic cell functionssuch as migration, proliferation, differentia-tion, and apoptosis (cellular “suicide”).5,11,15,16

Studies of Cell Function andBehavior Using Well-Defined,Patterned Environments

Here, we sketch representative experi-ences with these new tools for fabricatingmicropatterned environments to use in cellbiology and biotechnology. These tech-niques are bringing a new level of repro-ducibility to certain types of experiments incell biology and to enable certain studies—especially studies involving adhesion ofcells to patterned substrates and movementof cells on substrates—with levels of pre-cision that have not been possible in thepast. With this capability has come both theability to do new kinds of experiments thatexplore the relationship between the celland its environment and the suggestion thatcell populations, as they are currently stud-ied, are more heterogeneous in their re-sponse than has been fully appreciated inthe past. This realization provides importantchallenges for the future: high-resolutionstudies of phenotypic behaviors of cells isprobably going to require better-definedenvironments (to which one currently cansee a clear experimental pathway) andbetter-defined cells (to which the pathsnow appear more arduous).

It has long been known that cells are ableto sense spatially encoded soluble and ad-hesive cues.3,17,18 Immune cells, for example,can migrate to sites of infection and in-flammation by moving through gradientsin the concentration of soluble cytokines, inprocesses broadly called chemotaxis.19,20 Incell culture, chemotaxis can be observed byplacing the tip of a micropipette filledwith a chemotactic agent in a culture of re-sponsive cells; cells soon migrate and ac-cumulate at the pipette tip.20,21 Similarly,adhesion to an extracellular matrix also car-ries spatial information. Adhesion of cellsto a flat surface causes them to polarizenumerous subcellular organelles with re-spect to the surface plane. Adherent cellsalso sense and migrate toward a higherdensity of extracellular matrix ligands.22–24

Unfortunately, in many of the experimen-tal accounts of this type of study, the cuesused to stimulate the cells were spatially



ill defined; this poor definition made thecharacterization of the cell response am-biguous. For example, for cells migrating

in chemotactic gradients, it has been diffi-cult to define whether cells respond to ab-solute concentrations of soluble factors,gradients in concentration across the cellsurface, or some time-averaged sampling ofthe environment as cells move irregularlyacross the surface.

Here, soft lithography has made it pos-sible to define, with high spatial resolution,environmental cues—both those immobi-lized on surfaces and those dissolved in themedium—that influence cell behavior. Inso doing, they make it possible to study theways in which cells examine and respondto their surroundings. For example, by usingparallel laminar flows in a microfluidicchannel, we and others have been able to ex-pose cells to simple linear gradients to studychemotaxis (Figure 1).5,25 Microcontact print-ing or stencil printing with elastomericmembranes can also be used to generatewell-defined, micrometer-scale patterns ofproteins on surfaces.26,27

These approaches have demonstratedthat changes in cell shape that result as cellsattach, spread, and flatten on an adhesivesurface can regulate cell growth and death(Figures 2a and 2b).11,15 Earlier studies hadshown that increasing the density of extra-cellular matrix ligands on a substrate couldchange the shape and behavior of cells,but it left unresolved whether the changesin cell shape were causally linked to thechanges in cell behavior.28 Culturing cells onprogressively larger islands coated withextracellular matrix enabled cells to spreadto greater extents without the need for vary-ing the density of immobilized ligands. Thismore direct approach to modulate cellshape was shown to be the critical factorin triggering informative phenotypic be-haviors in cells.11,15

In some cases, the capability to modulatea particular environmental cue has pro-vided the means to discover previouslyunknown cellular responses. For example,soft lithography made it possible, for thefirst time, to pattern surfaces into adhesiveand non-adhesive regions and to define theshapes of these regions and, hence, theshapes of cells that adhered to and spreadon them (Figures 2a and 3). This new capa-bility has enabled studies of cells growinginto shapes defined by printing to be regulargeometrical figures (e.g., squares, triangles,and polygons) and demonstrated that boththe extent of cell spreading and flatteningand the geometric shape of the adhesivearea could regulate cell behavior.29,30 Cellsconforming to equilateral triangles orsquares polarize their migrating leadingedges (i.e., their lamellipodia) toward oneof the corners of the polygon rather thanalong a straight edge (see the bright areason the corners in Figure 3a).30 How cells de-

tect such geometric features, however, re-mains to be discovered.30,31

Do those microengineering technologiesthat have already been developed and

Figure 1. Photograph showing amicrofluidic device used for generatinggradients of soluble species.Theoperation of the device is illustrated withblue and red dyes. Solutions in channelssplit, mix by diffusion, split again, andultimately combine into a single widechannel (bottom); the stream in thischannel forms a gradient perpendicularto the direction of flow.

Figure 2.The influence of the footprintof a cell on its choice between growthand apoptosis. (a) (top) Schematicdiagram showing the pattern: squareislands of self-assembled monolayers(SAMs) to which proteins stick,surrounded by a different SAM to whichproteins do not adsorb; (bottom) Nomarskimicroscopic image of the shapes ofbovine adrenal capillary endothelialcells confined to the patterns. Scalelabels indicate lengths of the sides ofthe squares. (b) Changes in cell shapeas cells attach, spread, and flatten onan adhesive surface can regulate cellgrowth and death: apoptotic (cell death)index versus DNA synthesis (cellgrowth) index after 24 h, plotted as afunction of the area of the spread cell.



demonstrated in biomedical research rep-resent the majority of work that must bedone to enable hypotheses in cell biology tobe tested in reliable experimental systems?Hardly. We know only a few of the envi-ronmental signals that influence cells andare unable to investigate most of the onesthat we have identified. For example, cellshape, cell structure, cell–substrate adhe-sion, and cell mechanics are believed to be linked. Cells attached to soft substratesform fewer adhesions than do cells on stiffsubstrates.32 Cells also generate contractiletension; this tension generates mechanicalstresses at the cell–substrate boundary,and these stresses in turn seem to influencethe strength of adhesions. Because ad-hesions and mechanical stress are linked,it has been difficult to examine the in-fluence of adhesion, cell shape, and cellu-lar mechanics independently: the toolsnecessary to separate them have not beenavailable.

Approaches to these sorts of complexproblems—problems that span a hierarchyof dimensions, from the molecular to thecellular—are now beginning to emerge. Forexample, the ability of soft lithography toform beds of soft, elastomeric microneedlesby molding has made it possible simulta-neously and independently to measure cel-lular stresses and to provide a well-definedadhesive substrate (Figure 4).31 We haveused this combination to begin exploringthe relationships between cell shape, con-tractile stress, and cell–substrate adhesion.31

This prototype study demonstrates the

potential for microfabricated structures toprovide informative new approaches to thestudy of the mechanics of cells.

An Emerging Picture of CellularMechanics

Together, these and other studies of theinfluence of cell–substrate adhesion, cellshape, and cell mechanics, and the influ-ence of soluble factors on cells, combine tosuggest several hypotheses.

First, all of these environmental signalsseem to be linked: that is, these cues are non-orthogonal.2 This observation raises thequestion of how to design and execute awell-defined study of cellular responses. Forexample, at the simplest level, if the shapeof a cell affects its response to a growth fac-tor, then a complete understanding of theinfluence of the growth factor on cellularbehavior must include integrated studiesof the shape of the cell: practically, the in-fluence of the growth factor must be deter-mined by using it to stimulate cells havingmany, well-defined shapes. We do not cur-rently know what a complete set of cellularshapes would be. We are beginning to un-derstand how to control the shape of cellsattached to planar (two-dimensional) sub-strates, but have only a few tools to generatedefined three-dimensional scaffolds for cellgrowth; we have only weak control overthe nature of contacts between cells in cellculture and especially in cell cultures thatdefine cellular shapes.

Second, cells interact dynamically withtheir environment: they both respond to and

influence that environment. Most tools formicron-scale investigation of cells give onlylimited information about dynamics, al-though this situation is now beginning to change. Mrksich33,34 (see article in thisissue) and we35 have introduced tech-niques that enable cells to be patterned inshape and location using printed self-assembed monolayers (SAMs) and then re-leased from the constraints imposed by thispatterning using a short (and apparentlynon-damaging) pulse of electrical current(Figure 5). This procedure offers the ex-amination of a range of cellular behaviorsthat are reflected in motility. It provides,however, only the crudest glimpse into thetime domain. Providing better—and ide-ally, interactive—control of the environ-ment of cells in culture will be useful inextending the range of biologically rele-vant processes that can be studied usingcultured cells.

Third, using cells in well-defined en-vironments for experiments increases thesignal and reduces the noise and variabil-ity in the experimental system. Processesthat were difficult to study can becometractable with the use of microengineeredsubstrates. For example, cell-to-cell com-munication between neighbors has longbeen suspected to be an important regula-tor of cell behavior. Cell-to-cell contact,and presumably communication, increaseswhen cells are seeded at increasing surfacedensity. However, because cells are ran-domly distributed on the surface, any onecell may contact none, one, or even six

Figure 3. (a) A cell spread on a square 30 �m � 30 �m island; the cell was stimulated with platelet-derived growth factor (PDGF) and stainedfor F-actin with fluorophore-labeled phalloidin. (b) Cells confined to various shapes (all with areas of 900 �m2) were stimulated with PDGF andstained with phalloidin and 4 -6-diamidino-2-phenylindole to visualize F-actin (green) and nuclei (blue). Images obtained by fluorescence microscopy.�



neighbors. As a result, the effects of contactand number of contacts are often difficultto characterize. We have developed meth-ods that restrict cells to forming pairs (Fig-ure 6). This approach has allowed us toshow that cell proliferation increases witha small number of contacts and then sub-sequently decreases.36 Without these well-defined systems, this biphasic responsewas not previously seen. Exposing statisti-cally significant numbers (hundreds tothousands) of cells to indistinguishable con-ditions is necessary to distinguish variabilityto response caused by variations in the en-vironment from intrinsic variability in thecells themselves.

The Single-Cell Advantage (or Disadvantage)

Perhaps one of the most useful aspectsof soft lithographic technologies that enablepatterning of the shape and location of in-dividual cells is the ability to study the behaviors of individual cells in parallel ex-periments; these sorts of experiments havethe potential to bring a new level of preci-sion and reproducibility to cell-based ex-periments. Conventional cellular assays uselarge numbers of cells (thousands to tens ofmillions) that are pooled into a single assay.These sorts of assays accept the fact thatindividual cells may be entirely distinct intheir response to a stimulus and assume that

a response averaged over a population ofcells provides the most relevant measureof the stimulus–response characterizationof cells in realistic biological environments.These experiments also assume that anyvariations in the local environment (e.g.,the local density of cells on a surface) willaverage in the population response. In ad-dition, they ignore the obvious differencesamong cells—their position in the cell cycle,their lineage, their history of epigeneticmodification, and all the other characteris-tics that make individual cells individual.

Every field has its own justifications forusing averaged behaviors, but whatevertheir virtues, it is almost always useful tobe able to also look at individual behaviors.(Even in fields such as neuroscience andtissue engineering, where collective behav-iors are the real focus of research, charac-terizing the behaviors of individual cells isimportant.) Microengineered surfaces andmicrofluidic channels now make it possibleto provide indistinguishable—thus effec-tively identical—environments for multiple,individual cells and open the door to experi-ments that will deconvolve the variabilityin biological responses of cell populationsfrom those that reflect differences betweenindividual cells in the cell population.

In those instances in which statisticallysignificant numbers of individual cells havebeen studied in parallel experiments, whathave we learned about the robustness ofcellular responses? The answer is, onlyenough to begin to phrase intelligent ques-tions. It has been said that cells sense ananalog world, but make binary decisions.That is, cells cannot and do not partiallyreplicate, divide, differentiate, or die: theyeither do or they do not. They assess theirenvironment and, based on that assessment,commit irreversibly to a process. (If re-sponses to major choices were not binary,the opportunities for cellular dysfunctionthat could result, say, from having somecomponents preparing for mitosis whileothers prepared for apoptosis would bevery large).

Thus, when individual cells are allowedto spread to cover different “footprints,”the cells that are most spread have the great-est tendency to divide.28,37 Loosely inter-preted, single cells must occupy an areaabove some threshold value before they canswitch into the proliferative response. Ex-perimental observations of populations of in-dividual cells, uniformly patterned intoindistinguishable shapes, show that differentcells in these populations respond at differ-ent thresholds. That is, a sharp transition doesnot exist where, below a certain size offootprint, all cells do not divide, and abovethat size, all do; rather, within the popula-tion, an ever larger fraction of cells will

Figure 4. Elastomeric silicone microposts used to modulate the deformability of a substrate tocellular contractions. (a) Schematically, cells adhere to the tips of an array of closely spaced,vertically oriented elastomeric posts.When cells contact and probe the surface, the micropostsdeflect, depending on their mechanical stiffness, which can be easily manipulated by alteringtheir dimensions. (b) Scanning electron micrograph showing a cell attached to the entiremicropost substrate, including the shafts of the microposts and the underlying base.(c) Schematic illustration showing the process for microprinting adhesive proteins on themicroposts. (d) Differential interference contrast (top) and immunofluorescence (bottom)micrographs of the same region of posts where a 2 � 2 array of posts has been printed withfibronectin.When adhesive proteins are selectively microprinted on the tips of the microposts(e), cells only attach on the tips. Scale bars indicate 10 �m.



divide in response to incremental increasesin footprint size. We do not understand themechanistic basis of these observations. Dothey reflect the fact that individual cellshave different, preprogrammed thresholdsrequired to trigger proliferation, or is therea stochastic aspect to the response of indi-vidual cells even within collections of cellsthat are otherwise identical (if “identical”cells actually exist)? If one were to examinethe daughter cells of a cell that proliferatedfollowing an increase in its footprint, wouldthe daughters predictably respond at thesame threshold in footprint? In the case ofa non-dividing response (e.g., the produc-tion of lamellipodia at the corners of poly-gons), only a certain fraction of superficiallyindistinguishable (in shape and lineage)cells again exhibit the behavior.29 Are thereresponders and non-responders for everystimulus—a variability built into the popu-lation of cells—or do cells respond idio-syncratically at some times but not atothers? Why?

These studies with ensembles of (appar-ently) indistinguishable isolated cells raisefundamental questions about the designof biological experiments. If individual cellsare inherently different from one another—and different in ways that we cannot at

present characterize—then what fraction ofthe ensemble must respond to constitute astatistically significant response? Howmany cells must be included in a study toproduce a response that can be reproduced?How should statistics be treated in theseexperiments? What population of cells—prepared and characterized how—consti-tutes one cell type? These questions areuncomfortable ones for cell biologists, forwhom thoughtful and rigorous statisticalanalysis is not a natural act.

To begin to explore these fundamentalquestions of experimental design—charac-terization of the starting sample of cells,understanding the significance of the ex-perimental result—one experimental de-sign (and unfortunately not one that is oftenpractical) is to use the cell as its own control:that is, to repeatedly and non-invasivelystimulate a single cell and characterize the range of its responses. This approacheliminates cell-to-cell variability by focus-ing on a single cell, but it must still dealwith the ambiguities resulting from thehistory of that cell: its experience on multiple stimulations, its age, and its re-sponse to the rigors of being the object of investigation. Further, this approach isnot applicable to the major events of thelife and death of a cell: mitosis and apopto-sis. Still, the availability of “well-definedenvironments”—defined shapes, on defined

Figure 5. Bovine capillary endothelial cells were initially confined to rectangular patternsusing methyl-terminated self-assembled monolayers (SAMs) surrounded byethylene-glycol–terminated (EG-terminated) SAMs. Application of a cathodic voltage pulse(�1.2 V for 30 s) released the cells from the constraints of the microislands by desorbing theEG-terminated SAMs and enabling proteins to adsorb from the culture medium that allowedthe attachment and movement of cells.The numbers indicate the time elapsed (in minutes)after the voltage pulse.

Figure 6. Method to induce cells to culture in pairs with control over the contact betweenthem. (a) Agarose is wicked into channels formed by a poly(dimethylsiloxane) stamp sealedagainst a glass coverslip and allowed to gel before the stamp is peeled off. (b) When cellsare seeded onto these substrates containing bowtie-shaped wells, cells attach and cultureas either single cells or pairs. (c) The single cell-to-cell contact formed in these pairs can beblocked by fabricating substrates in which the agarose forms a thin wall, cutting thebowtie-shaped wells into separate, although closely spaced, wells.



substrates, in contact with defined culturemedia—constitutes the first step towardsuch experiments.

ConclusionsThe most elementary complete unit of

biology is the cell: it is the simplest unit thatis “alive.” To understand the cell is to un-derstand what it is to be alive. Among themany difficulties in the task of under-standing the cell is the variability of cells.Even when a population of cells is mono-clonal and shares an identical (or nearlyidentical) genome, individual cells are dif-ferent. The potential contributions to thesedifferences are beginning to be understood:the position in the cell cycle, a wide range ofepigenetic phenomena (from gene silencing,through splice variation, to post-translationalmodification of functional proteins), andpassage number (clearly marked in thestructure of telomers, but certainly recordedin many less easily read parts of the cell)are a few. In general, these differences mayreflect the fundamental truth that the historyof experience for each cell, and the conse-quent changes that result from that history,are simply different. The individual andcollective importance of these differences forthe conduct of research in cell biology isnot understood.

Cell biology has been a qualitative (or, atbest, semi-quantitative), observational sci-ence. It has—understandably, in light of thecomplexity of the cell—lagged far behindgenomics in moving toward a more quan-titative basis, one in which defining the re-producibility and statistical significance ofexperiments and their interpretations be-comes a routine part of interpreting the ex-periment. In bioanalysis, an area where thepotential for applications that are immedi-ately useful in the development of newpharmaceuticals has stimulated substantialeffort to quantitate cell-based assays, the re-putation for reproducibility of these assays is poor. There is an important opportunityto build the foundation for a “better-interpretable cell biology” by understand-ing the basis for the variability in cellularresponses to stimuli (including the simplepassage of time) and in improving the reproducibility of experiments involvingcells—both individually and in ensembles—and the ability to design experiments thatyield meaningful results.

One of the simplest parts of this problemis that of controlling the static physical en-vironment of the cell: its location, its shape,the mechanical and chemical properties ofthe interface to which it is attached, and thecomposition of the medium in which it isimmersed. The development of the tools to define these environmental factors isadvancing rapidly: soft lithography, repre-

sented in microcontact printing (to controlthe surface) and micromolding (to controlthe topography of the surface and to fabri-cate microfluidic systems), and materialsscience to define the mechanical propertiesof the surface are important parts of thisdevelopment. The toolset is not yet com-plete; it still is not possible to define the fourimportant elements of the environment—the composition/properties of the support,the nature of the interface between the sub-strate and the cell, the shape and location of the cell, and the medium surroundingthe cell—arbitrarily and independently, but the degree of control over these sys-tems has increased enormously in the lastdecade.

Controlling the dynamic aspects of theenvironment of the cell is just beginning tobe a focus of research. It is now possible topattern the composition of the fluidmedium surrounding the cell and to changethat composition (within some range) withtime constants of �1 s; it is possible to illuminate or irradiate the cell; it is alsopossible to release the constraints that de-termine the shape of a patterned cell. Thesetools are making it possible to conduct ex-periments on ensembles of individual cells,in statistically significant numbers (hun-dreds to thousands), and in indistinguish-able environments. These experiments willbe the first step toward understanding thevariation in response to a stimulus within apopulation of cells.

The next stages in the development of this technology have not really started. We know that cells attached to a cultureplate are very different from cells in tissue;how can one mimic the complex, time-dependent, multidimensional (in both to-pography and in chemistry) worldexperienced by a cell in tissue? Moving inthis direction will require integration ofthe mesoscopic, “cell-sized” tools devel-oped by the bioengineer with the molecu-lar approaches of the molecular biologistand geneticist.

Based on early experience, it appears thatcells do not treat soluble ligands, surface-bound ligands, and substrate mechanicsas orthogonal inputs to their function. Be-cause the cellular control systems are inter-connected, all factors in the environmentmust be simultaneously well defined. Bi-ology needs, inter alia, (1) temporal controlof cell–material interactions, (2) spatialcontrol of chemistry and mechanics at thenanometer scale, (3) spatiotemporal controlin three-dimensional settings, (4) presen-tation of complex (heterogeneous) surfaces,(5) understanding what properties of natu-ral materials should be mimicked, and(6) ways of mimicking cell–cell and cell–tissue interactions.

Microengineered systems now enable thestudy of single cells in well-defined environ-ments. An early lesson from these studiesis that even in the best-defined environ-ments now available, the response of cells to stimuli can be highly variable, and thereason for that variability is largely unde-fined. In particular, we do not understandhow the relative contributions of cell-to-cell,cell-to-environment, and environment-to-environment variability contribute to thiseffect. Nor do we know whether cell-to-cellvariability is inherent to the cells and de-terministic, or idiosyncratic because of un-characterized and idiosyncratic variabilityin response. Until we understand the fun-damental basis for population variance,developing reliable assays based on cells willuse population-averaged responses, and we will just have to live with the ambigui-ties that come with the difficulties of thesesystems.

Approaches to cell biology based onmicroengineering constitute necessary stepsin understanding the processes used byindividual cells as well as those used byensembles of cells to form tissues thatfunction in ways that individual cells donot. The materials science community willcontinue to play a critical role in generatingthe tools necessary to fuel this work in bio-medical research.

AcknowledgmentsWork at the University of Pennsylvania

was funded by the National Institutes of Health (NIH), grants EB00262 andHL073305. Work at Harvard University wasfunded by the NIH, grant GM065364.

References1. C. Arthur, M.D. Guyton, and J.E. Hall, Textbookof Medical Physiology, 10th Ed. (W.B. Saunders,Philadelphia, 2000).2. C.S. Chen, J. Tan, and J. Tien, Ann. Rev. Biomed.Eng. 6 (2004) p. 275.3. B. Alberts, A. Johnson, J. Lewis, M. Raff, R. Keith, and P. Walter, Molecular Biology of the Cell, 4th Ed. (Garland Science, New York, 2002).4. S. Takayama, E. Ostuni, P. LeDuc, K. Naruse,D.E. Ingber, and G.M. Whitesides, Nature 411(2001) p. 1016.5. S.K.W. Dertinger, X. Jiang, Z. Li, V.N. Murthy,and G.M. Whitesides, Proc. Natl. Acad. Sci. USA99 (2002) p. 12542.6. G.M. Whitesides, E. Ostuni, S. Takayama, X.Jiang, and D.E. Ingber, Annu. Rev. Biomed. Eng.3 (2001) p. 335.7. B.D. Gates, Q.B. Xu, J.C. Love, D.B. Wolfe,and G.M. Whitesides, Annu. Rev. Mater. Res. 34(2004) p. 339.8. X. Jiang and G.M. Whitesides, Eng. Life Sci. 3(2003) p. 475.9. P. Kenis, R. Ismagilov, S. Takayama, and G.Whitesides, Acc. Chem. Res. 33 (2000) p. 841.10. Y. Xia and G.M. Whitesides, Angew. Chem.Int. Ed. Engl. 37 (1998) p. 550.



11. C.S. Chen, M. Mrksich, S. Huang, G.M.Whitesides, and D.E. Ingber, Science 276 (1997)p. 1425.12. S. Takayama, J.C. McDonald, E. Ostuni,M.N. Liang, P.J.A. Kenis, R.F. Ismagilov, andG.M. Whitesides, Proc. Natl. Acad. Sci. USA 96(1999) p. 5545.13. D.T. Chiu, N.L. Jeon, S. Huang, R.S. Kane,C.J. Wargo, I.S. Choi, D.E. Ingber, and G.M.Whitesides, Proc. Natl. Acad. Sci. USA 97 (2000)p. 2408.14. S.K.W. Dertinger, D.T. Chiu, N.L. Jeon, andG.M. Whitesides, Anal. Chem. 73 (2001) p. 1240.15. S. Huang, C.S. Chen, and D.E. Ingber, Mol.Biol. Cell. 9 (1998) p. 3179.16. R. McBeath, D.M. Pirone, C.M. Nelson, K.Bhadriraju, and C.S. Chen, Dev. Cell 6 (2004)p. 483.17. M. Tessier-Lavigne and C.S. Goodman, Science 274 (1996) p. 1123.18. B.K. Brandley and R.L. Schnaar, Dev. Biol.135 (1989) p. 74.19. D. Bray, Cell Movements: from Molecules toMotility, 2nd Ed. (Garland Publishing, New York,2001).20. P.J.M. Van Haastert and P.N. Devreotes,Nat. Rev. Mol. Cell Biol. 5 (2004) p. 626.21. K. Hong, M. Nishiyama, J. Henley, M.Tessier-Lavigne, and M.M. Poo, Nature 403(2000) p. 93.22. S.L. Zackson and M.S. Steinberg, Dev. Biol.124 (1987) p. 418.23. J.B. McCarthy and L.T. Furcht, J. Cell Biol. 98(1984) p. 1474.24. V. Rigot, M. Lehmann, F. Andre, N. Daemi,J. Marvaldi, and J. Luis, J. Cell Sci. 111 (1998)p. 3119.25. N.L. Jeon, H. Baskaran, S. Dertinger, G.M.Whitesides, L. Van De Water, and M. Toner, Nat.Biotechnol. 20 (2002) p. 826.26. E. Ostuni, R. Kane, C.S. Chen, D.E. Ingber, and G.M. Whitesides, Langmuir 16 (2000) p. 7811.27. A. Folch, B.H. Jo, O. Hurtado, D.J. Beebe, andM. Toner, J. Biomed. Mater. Res. 52 (2000) p. 346.28. D.E. Ingber, Proc. Natl. Acad. Sci. USA 87(1990) p. 3579.29. K.K. Parker, A.L. Brock, C. Brangwynne,R.J. Mannix, N. Wang, E. Ostuni, N.A. Geisse,J.C. Adams, G.M. Whitesides, and D.E. Ingber,FASEB J. 16 (2002) p. 1195.30. A. Brock, E. Chang, C.-C. Ho, P. LeDuc, X.Jiang, G.M. Whitesides, and D.E. Ingber, Langmuir19 (2003) p. 1611.31. J.L. Tan, J. Tien, D.M. Pirone, D.S. Gray, K.Bhadriraju, and C.S. Chen, Proc. Natl. Acad. Sci.USA 100 (2003) p. 1484.32. C.M. Lo, H.B. Wang, M. Dembo, and Y.L.Wang, Biophys. J. 79 (2000) p. 144.33. M.N. Yousaf, B.T. Houseman, and M. Mrksich, Angew. Chem. Int. Ed. 40 (2001) p. 1093.34. W.-S. Yeo, M.N. Yousaf, and M. Mrksich,J. Am. Chem. Soc. 125 (2003) p. 14994.35. X. Jiang, R. Ferrigno, M. Mrksich, and G.M.Whitesides, J. Am. Chem. Soc. 125 (2003) p. 2366.36. C.M. Nelson and C.S. Chen, FEBS Lett. 514(2002) p. 238.37. R. Singhvi, A. Kumar, G.P. Lopez, G.N.Stephanopoulos, D.I.C. Wang, G.M. Whitesides,and D.E. Ingber, Science 264 (1994) p. 696. ■■

For more information, see http://advertisers.mrs.org

www.mrs.org/publications/bulletin

microengineering the ...“top-down” study of cells, beginning with phenotypic cellular behaviors,...

Documents