Emergence of Patterned Stem Cell Differentiation WithinMulticellular Structures

SAMI ALOM RUIZ,a,b CHRISTOPHER S. CHENb

aDepartment of Biomedical Engineering, Johns Hopkins University School of Medicine, Baltimore, Maryland, USA;bDepartment of Bioengineering, University of Pennsylvania, Philadelphia, Pennsylvania, USA

Key Words. Mesenchymal stem cells • Differentiation • Three-dimensional • Patterning

ABSTRACT

The ability of stem cells to differentiate into specified lineagesin the appropriate locations is vital to morphogenesis and adulttissue regeneration. Although soluble signals are importantregulators of patterned differentiation, here we show that gra-dients of mechanical forces can also drive patterning of lin-eages. In the presence of soluble factors permitting osteogenicand adipogenic differentiation, human mesenchymal stem cellsat the edge of multicellular islands differentiate into the osteo-genic lineage, whereas those in the center became adipocytes.Interestingly, changing the shape of the multicellular sheetmodulated the locations of osteogenic versus adipogenic differ-entiation. Measuring traction forces revealed gradients of

stress that preceded and mirrored the patterns of differentia-tion, where regions of high stress resulted in osteogenesis,whereas stem cells in regions of low stress differentiated toadipocytes. Inhibiting cytoskeletal tension suppressed the rela-tive degree of osteogenesis versus adipogenesis, and this spatialpatterning of differentiation was also present in three-dimen-sional multicellular clusters. These findings demonstrate a rolefor mechanical forces in linking multicellular organization tospatial differentials of cell differentiation, and they representan important guiding principle in tissue patterning that couldbe exploited in stem cell-based therapies. STEM CELLS 2008;26:2921–2927

Disclosure of potential conflicts of interest is found at the end of this article.

INTRODUCTION

Morphogenesis involves a complex coordination of spatial re-arrangements that give rise to numerous complex structures, andthe differentiation of stem cells into the many cell types thatpopulate these structures [1, 2]. This orchestration of patterneddifferentiation within developmental structures is thought to beregulated primarily by spatial gradients of diffusible factors knownas morphogens [3–6]. It is generally thought that these morphogenscan simultaneously control morphologic changes and differentia-tion programs in cells, such that the appropriate cell types attaintheir corresponding locations within complex tissues.

In addition to such morphogenetic gradients, it is also clearthat gradients in mechanical forces, which could contribute tothe process of morphogenesis, also exist within developingembryos. During many of the massive rearrangements that occurin early development, such as convergence and extension, neu-ral tube formation, dorsal closure, and anterior gut formation,mechanical forces appear to be involved [7–10]. Furthermore,recent work suggests that such mechanical factors appear tocontribute to the specification of adult mesenchymal and em-bryonic stem cells to various lineage fates [11–15]. In thecontext of mesenchymal stem cells (MSCs), it has been shownthat progressively increasing RhoA-mediated cytoskeletal ten-sion can alter lineage specification from adipocytes or neuralprecursors to myocytes and osteoblasts [11, 12], raising the

possibility that spatial gradients of stress within a structurecould act to pattern differentiation. We recently introduced anapproach to generate such force gradients in micropatternedsheets of cells and demonstrated that such gradients can drivepatterns of cell proliferation in endothelial and epithelial cells[16]. It remains unclear, however, whether such gradients offorces can also determine patterns of lineage specification ofstem cells.

Here, we demonstrate that multicellular aggregates of MSCsundergo lineage specification as a function of aggregate geom-etry and of cell position within the aggregate. Cells on the outeredge of simple sheets committed to an osteogenic lineage,whereas interior MSCs differentiated to adipocytes. Three-di-mensional structures also exhibited partitioning of fat and bone.These patterns appear to result from nonuniform patterns inmechanical stress determined by multicellular shape and gener-ated by the actin-myosin cytoskeleton. Understanding thesemechanical patterning effects not only will improve our under-standing of in vivo differentiation but also may provide impor-tant insights for optimizing stem cell-based therapies.

MATERIALS AND METHODS

MaterialsThe following reagents were purchased from the following suppli-ers: polydimethylsiloxane (PDMS) (Sylgard 184; Dow Corning,

Author contributions: S.A.: conception and design, administrative support, provision of study materials, collection and assembly of data, dataanalysis and interpretation, manuscript writing; C.S.C.: conception and design, financial support, administrative support, provision of studymaterials, data analysis and interpretation, manuscript writing, final approval of manuscript.

Correspondence: Christopher S. Chen, M.D., Ph.D., 510 Skirkanich Hall, 210 South 33rd Street, Philadelphia, Pennsylvania 19104, USA. Telephone:215-746-1750; Fax: 215-746-1752; e-mail: chrischen@seas.upenn.edu Received May 1, 2008; accepted for publication August 5, 2008; first publishedonline in STEM CELLS EXPRESS August 14, 2008. ©AlphaMed Press 1066-5099/2008/$30.00/0 doi: 10.1634/stemcells.2008-0432

TISSUE-SPECIFIC STEM CELLS

STEM CELLS 2008;26:2921–2927 www.StemCells.com

at UN

IVE

RSIT

Y O

F PEN

N L

IBR

AR

Y on M

arch 5, 2009 w

ww

.StemC

ells.comD

ownloaded from

Midland, MI, http://www.dowcorning.com), Y-27632 (catolog no.1254; Tocris Bioscience, Bristol, U.K., http://www.tocris.com),blebbistatin (catalog no. 1760; Tocris Bioscience), ML-7 (catalogno. 475880; Calbiochem, San Diego, http://www.emdbiosciences.com), fibronectin (catalog no. 356008; BD Biosciences, San Diego,http://www.bdbiosciences.com), and low-melting-point agarose(catalog no. 15517-014; Gibco-BRL, Gaithersburg, MD, http://www.gibcobrl.com). Recombinant adenoviruses encoding greenfluorescent protein (GFP) and ROCK�3 were prepared using theAdEasy XL system (Stratagene, La Jolla, CA, http://www.stratagene.com) as previously described [17].

Cell CultureHuman mesenchymal stem cells (32-year-old male and 21-year-old male; Cambrex, Walkersville, MD, http://www.cambrex.com) were cultured in growth medium consisting of Dulbecco’smodified Eagle’s medium supplemented with 10% fetal bovineserum, 0.3 mg/ml L-glutamine, 100 units/ml penicillin, and 100�g/ml streptomycin. Adipogenic induction and osteogenic dif-ferentiation media were purchased from Cambrex. Medium waschanged every 3 days except when cells were treated withblebbistatin and Y-27632, in which case the medium was re-placed daily. When infected with adenoviruses containing GFPand ROCK�3, cells were incubated with the virus overnight andused the next day.

Pattern GenerationPatterns of fibronectin on PDMS-coated coverslips were generatedby microcontact printing [11]. The regions surrounding the patternswere rendered nonadhesive by incubating the coverslips in F127Pluronic (BASF, Ludwigshafen, Germany, http://www.basf.com)[18]. MSCs were seeded onto the coverslips at a density of 20,000cells per cm2. After an initial 24 hours in growth medium, the cellswere then cultured in either adipogenic or osteogenic medium or ina 1:1 mixture of both for the subsequent 14 days unless otherwiseindicated.

StainingAfter 14 days, the cells were fixed in 3.7% paraformaldehyde andstained for alkaline phosphatase using Fast Blue RR Salt/naphthol(FBS25-10CAP, 855-20ML; Sigma-Aldrich, St. Louis, http://www.sigmaaldrich.com), lipid droplets with 3 mg/ml oil red O (O-0625;Sigma-Aldrich) in 60% isopropanol, and nuclei with 4,6-diamidino-2-phenylindole (DAPI; D1306; Invitrogen, Carlsbad, CA, http://www.invitrogen.com). To assay for differentiation after 5 days inculture, the transcription factors peroxisome proliferator-activatedreceptor � (PPAR�) and cbfa-1 were stained for. After Triton-extracting the soluble component and fixing with 3.7% paraformal-dehyde, the samples were blocked overnight with 33% goat serum.The coverslips were then incubated in rabbit anti-PPAR� (sc-7196;Santa Cruz Biotechnology Inc., Santa Cruz, CA, http://www.scbt.com) and rat anti-cbfa-1 (MAB2006; R&D Systems Inc., Minne-apolis, http://www.rndsystems.com) antibodies for 1 hour. Afterthree phosphate-buffered saline (PBS) washes the coverslips wereincubated in 488-goat anti-rabbit and 647-goat anti-rat (A-11034,A-21247; Molecular Probes, Eugene, OR, http://probes.invitrogen.com) secondary antibodies, as well as DAPI. After three PBSwashes, the coverslips were mounted.

Microscopy and AnalysisAn Eclipse TE-200 microscope (Nikon, Tokyo, http://www.nikon.com) with a �10 objective was used to capture images of alkalinephosphatase and oil red O-stained patterns. An Axiovert 200Mmicroscope (Carl Zeiss, Jena, Germany, http://www.zeiss.com)with a �40 objective was used to capture fluorescence images oftranscription factor-stained patterns. Because the patterns werelarger than the �40 field of view, multiple images were capturedand stitched to form a single image of the entire pattern.

Four concentric circles were drawn around the circular pat-terns to divide them into four zones of equal area. Cells thatstained positive for oil red O or had PPAR� in their nuclei werescored as adipocytes, whereas cells that stained positive for

alkaline phosphatase or had cbfa-1 in their nuclei were countedas osteogenic cells. Cells with neither stain were scored asundifferentiated. The cells were then counted and used to calcu-late the percentage of each cell type in each zone. Means werecalculated from three experiments, with at least three circlesanalyzed from each. Bar graphs show mean � SEM. Statisticalsignificance between means of different lineages within a zoneor between means of the same cell type in two different zoneswas calculated using Student’s t test.

Measurement of Traction ForcesTraction forces were measured using microfabricated postarraydetectors as previously described [19]. Briefly, microfabricatedpostarray detectors (mPADs) were microcontact printed with fi-bronectin so that cells cultured on the devices spread and conformedto the printed patterns. The cells were cultured on mPADs in thepresence of growth medium for 24 hours before being fixed in 3.7%paraformaldehyde. Images of post tips and bases were capturedusing a Carl Zeiss Axiovert XX microscope with a �40 objective,stitched to form the pattern, and used to calculate the post deflec-tions and, from them, the forces exerted on the posts by the cells.Statistical significance between force distributions was calculatedby using the Wilcoxon rank-sum test.

Three-Dimensional Structure Generation andAnalysisMolds were produced by pushing PDMS stamps with relief features,cast from photolithographically fabricated Su-8 templates, ontoprepolymer PDMS in oxygen plasma-treated Petri dishes. Aftercuring for 2 hours at 60°C, the stamps were removed, leaving amold with wells. Prior to cell seeding, the molds were immersed inF127 Pluronic to render the PDMS nonadhesive. Prepolymer type Icollagen was added to the molds at 4°C. The molds were placed ina vacuum chamber to remove air bubbles trapped at the bottom ofthe well. MSCs were trypsinized, resuspended in gel precursor at adensity of 2 � 106 cells per milliliter, and then added to the molds.The molds were then centrifuged at 1,200 rpm for 2.5 minutes topool the cells into the wells. After the excess gel was aspirated, themold was exposed to a temperature of 37°C to polymerize thecollagen. One percent low-melting-point liquid agarose was addedto the molds to encase the microgel structures. The agarose wasgelled at 4°C for 1 minute, and fresh culture medium was added tothe dishes.

Microgels were stained for alkaline phosphatase and lipid drop-lets as described above, encased in Tissue-Tek OCT Compound(Electronic Microscopy Sciences, Hatfield, PA, http://www.emsdiasum.com), and snap frozen in liquid nitrogen. Cross-sectionswere taken using a Leica CM 1850 cryostat (Leica, Heerbrugg,Switzerland, http://www.leica.com) and imaged as described above.A MATLAB program was written to analyze the images. First, theprogram created two binarized images, one red, representing adi-pogenesis, and one blue, representing osteogenesis, from the orig-inal color image (supplemental online Fig. 1). Next, one-pixel“shells” starting from the perimeter of the structure and terminatingat its center were generated. Finally, each shell was comparedagainst the two binarized color images to compute the percentagesof red and blue pixels in the shell. The percentages were eitherplotted versus shell distance from the center (supplemental onlineFig. 2) or used to determine the percentages of fat and bone in fourconcentric zones of equal area. Means were calculated from threeexperiments, with cross-sections of at least three structures analyzedfrom each. Graphs show mean � SEM. Statistical significancebetween means of different lineages within a zone or betweenmeans of the same cell type in two different zones was calculatedusing Student’s t test. Statistical significance between curves insupplemental online Figure 2 was calculated by applying the signtest to the paired experiments.

2922 Multicellular Patterned Differentiation

at UN

IVE

RSIT

Y O

F PEN

N L

IBR

AR

Y on M

arch 5, 2009 w

ww

.StemC

ells.comD

ownloaded from

RESULTS

Patterning of Lineages Occurs Within MulticellularAggregatesTo produce monolayer sheets of human MSCs in a circularshape, we seeded cells onto substrates engineered with arrays ofadhesive islands generated by microcontact printing fibronectinonto coverslips (Fig. 1A, 1B). After cells were allowed to attachand form a sheet overnight, MSCs were exposed to a 1:1mixture of adipogenic and osteogenic induction media, and after14 days they were costained for lipid droplets to indicate adi-pogenesis and alkaline phosphatase to indicate osteogenesis.Cells within the interior of these circles contained lipid droplets(red), whereas those at the edge stained positively for alkalinephosphatase (blue) (Fig. 1C). This patterned segregation oflineages emerged in all sizes of circles examined (250–1,000�m diameter; approximately 20–250 cells) (Fig. 2A) but did notoccur in cells exposed to growth, adipogenic, or osteogenicmedium alone (Fig. 1D–1F). To quantify the distribution oflineages in these monolayers, we superimposed four concentriccircles on each monolayer to divide it into four zones of equalarea and counted the cells within each zone that were positivefor each marker (Fig. 1G). In the innermost zone, more than

85% of the cells had lipid droplets, and approximately 5% werealkaline phosphatase-positive, whereas in the outermost zone,less than 1% were adipocytes and 70% exhibited an osteo-genic phenotype (Fig. 1H). Interestingly, patterned differen-tiation appears to scale with size, where the relative radii atwhich cell fates segregate are proportional to the size of thecircle (Fig. 2B).

Although lipid accumulation is diagnostic for adipogenesis,alkaline phosphatase is not specific for osteogenesis. Althoughdetection of calcium deposition can be diagnostic for osteogenicdifferentiation, it was not possible in this setting because cellsescape the constraints of patterning before calcium deposition istypically seen in these cells (�21 days). As such, we confirmedour findings using two transcription factors for adipogenesis andosteogenesis: PPAR� and cbfa-1/runx2, respectively. PPAR� isexpressed within days of addition of adipogenic medium [20]and is a transcription factor that is required for adipogenesis[21], activating many of the genes necessary for fatty acidmetabolism, such as lipoprotein lipase and the fatty-acid-bind-ing protein aP2 [22]. cbfa-1/runx2 is similarly expressed at earlytime points during osteogenesis and is a transcription factoressential for osteoblast differentiation [23, 24], inducing expres-sion of bone matrix-generating genes, such as osteopontin,osteocalcin, type I collagen [25, 26], and alkaline phosphatase[26, 27]. Circular monolayers stained for these markers afteronly 5 days in culture demonstrated 47% of interior cells with

Figure 1. Patterned segregation of lineages occurs within multicellular aggregates. (A): Schematic showing substrate preparation by microcontactprinting. (B): Phase image of mesenchymal stem cells (MSCs) on 1-mm circular patterns at day 0. (C–F): One-mm circular MSC aggregates stainedfor fat droplets (red) and alkaline phosphatase (blue) after 14 days in mixed media (C), growth medium (D), adipogenic medium alone (E), orosteogenic medium alone (F). (G): Concentric circles were placed over circular aggregates to divide them into four zones of equal area. Nuclei werelabeled with 4,6-diamidino-2-phenylindole to determine cell location. (H): Bar graph showing fraction of each cell type within each zone after 14 daysin mixed media. (I): Bar graph showing fraction of each cell type within each zone determined by assaying for transcription factors after 5 days inmixed media. � indicates p � .05 for all comparisons of the same cell type across different zones or different cell types within the same zone, exceptfor those marked with #. Scale bars � 250 �m. Abbreviation: PDMS, polydimethylsiloxane.

Figure 2. Partitioning of differentiationscales with size. (A): Circular mesenchymalstem cell aggregates with diameters of 250,500, 750, and 1,000 �m stained for oil drop-lets (red) and alkaline phosphatase (blue)after 14 days in mixed media. (B): Bar graphshows fraction of each cell type within eachzone for circular aggregates of all four sizes.Scale bars � 250 �m. Abbreviations: adipo,adipogenic; osteo, osteogenic.

2923Alom Ruiz, Chen

www.StemCells.com

at UN

IVE

RSIT

Y O

F PEN

N L

IBR

AR

Y on M

arch 5, 2009 w

ww

.StemC

ells.comD

ownloaded from

PPAR� in the nucleus and 2% with nuclear cbfa-1/runx2 (Fig.1I). In contrast, 8% of edge cells exhibited nuclear PPAR�,whereas more than 73% were positive for cbfa-1/runx2. Theseresults confirmed the partitioning of differentiation within thesestructures.

Aggregate Geometry Affects Patterning ofDifferentiationTo assess the effect of monolayer geometry on the distributionof adipogenic and osteogenic lineages, MSCs were confined todifferent shapes. Squares, rectangles, ellipses, and half-ellipsesall resulted in adipogenesis in the interior and osteogenesis atthe edges, regardless of whether the edges were straight, cor-nered, or curved (Fig. 3A–3D). It was previously observed thatthe inner, concave edges of annuli exhibited low proliferationcompared with outer edges of solid patterns, suggesting thatthese inner edges may have different effects [16]. Indeed, MSCmonolayers cultured in such patterns exhibited osteogenesis atthe outer edge and adipogenesis at the inner edge (Fig. 3E, 3F).

It is unclear, however, whether the edges imposed this distribu-tion of differentiation because of their location (i.e., outer versusinner) or because they are concave. To address this ambiguity,we generated sinusoidal bands where the same edge oscillatesbetween convex and concave. MSCs in these undulating shapesunderwent osteogenesis at convex edges and adipogenesis atconcave edges, regardless of the amplitude and thickness of thebands (Fig. 3G, 3H). These results show that the curvature of theedge, as well as the overall shape of the monolayers, likelycontributes to spatial patterning of the differentiation.

Role of Tension in Patterning of DifferentiationMyosin-generated cytoskeletal tension has been implicated inregulation of spatial patterning of a variety of responses, such asproliferation and gastrulation [7, 16]. To determine whetherspatial differentials in cellular forces arose in our multicellularsheets and whether such forces correlated to the observed pat-terns of MSC differentiation, we plated MSCs in sinusoidalmonolayers onto mPAD substrates, which consist of an array ofelastomeric microneedles that report local traction forces [19](Fig. 4A). Indeed, the magnitudes of forces at the convex edgeof the undulations were nearly three times those at the concaveedge (66.3 vs. 23.6 nanonewtons [nN] per post) and mirrored thepattern of osteogenic versus adipogenic differentiation. Plottingthe distribution of forces at convex edges revealed a peak at 110nN, compared with 40 nN at concave edges (Fig. 4B). Thus, thegeometry of the multicellular structure can give rise to localgradients of mechanical forces.

To test whether the high tension generated at convex edgesis responsible for increased osteogenesis at those edges, wetreated MSCs in such patterns with the nonmuscle myosin IIinhibitor blebbistatin. Inhibiting myosin-generated tension ab-rogated osteogenesis almost completely, while preserving adi-pogenesis (Fig. 5A, 5B, 5E, 5F). Given possible nonspecificeffects of the inhibitor, we confirmed that inhibition of theupstream regulators of myosin, Rho-kinase (ROCK) withY27632 and myosin light chain kinase with ML-7, also resultedin specific suppression of osteogenesis (Fig. 5C, 5D, 5G, 5H).

Figure 3. Geometry determines spatial patterning of differentiation.(A–H): Mesenchymal stem cell aggregates in the shape of a square (A),rectangles (B), an ellipse (C), a half-ellipse (D), an offset annulus (E),an elliptical annulus (F), and sinusoidal bands (G, H) stained for oildroplets (red) and alkaline phosphatase (blue) after 14 days in mixedmedia. Red arrows indicate adipogenesis at concave edges, and bluearrows indicate osteogenesis at convex edges. Scale bars � 250 �m.

Figure 4. Geometry determines spatial distribution of cytoskeletaltension. (A): Phase images of mesenchymal stem cells on microfabri-cated postarray detectors in the shape of a sinusoidal band. Middle panelshows a traction force vector map superimposed on the image. Righttwo panels show close-ups of the convex and concave edges. Scalebars � 50 �m. (B): Frequency distribution of forces along the convexand concave edges. � indicates p � .05. Dashed lines show means.Abbreviation: nN, nanonewtons.

2924 Multicellular Patterned Differentiation

at UN

IVE

RSIT

Y O

F PEN

N L

IBR

AR

Y on M

arch 5, 2009 w

ww

.StemC

ells.comD

ownloaded from

These results suggest that localized patterns of cytoskeletaltension within these multicellular sheets trigger patterns ofdifferentiation to different lineages.

Patterning of Differentiation in Three-DimensionalStructuresThe patterning of differentiation that we observed may arise asa specific response to the micropatterned, flat substrates andmay not represent what occurs when cells are embedded withinthree-dimensional contexts. To determine whether patterneddifferentiation can occur in three-dimensional matrix, we devel-oped a method to construct three-dimensional, multicellularstructures of MSCs. PDMS molds produced from photolitho-graphically fabricated templates with wells of controlled dimen-sions were used to create MSC-laden collagen constructs ofdefined three-dimensional geometries (Fig. 6A). MSCs sus-pended in unpolymerized collagen type I were pooled into thewells by centrifugation, and excess collagen was removed be-fore polymerization. The resulting constructs were then releasedand encased in agarose to allow culture without diffusion bar-riers. These three-dimensional structures could be peeled out ofthe template while retaining their integrity (Fig. 6B). After 14days in mixed media, cells at the edge of the constructs differ-entiated to an osteogenic lineage (as indicated by alkaline phos-phatase staining), whereas those at the center underwent adipo-genesis (as indicated by lipid droplet staining) (Fig. 6C). Toconfirm that patterning of differentiation occurred throughoutthe structures, they were sectioned and imaged (Fig. 6D, 6E).To quantify the distribution of lineages in these structures,the spatial distribution of red and blue pixels within theircross-sections was determined (supplemental online Fig. 1).In the innermost zone, MSCs predominantly underwent adi-pogenesis, whereas cells switched to osteogenic fate in theoutermost zone (Fig. 7A, 7E). Although endogenous forcescould not be measured within these structures, we examinedwhether up- or downregulating cytoskeletal tension alteredthe patterns of differentiation. Constructs treated with bleb-bistatin decreased osteogenesis, mimicking our findings onflat substrates (Fig. 7B, 7F; supplemental online Fig. 2A).Expressing ROCK�3, a mutant form of ROCK that is con-stitutively active and results in high myosin activity, resultedin a thicker layer of osteogenesis at the edge of the structures(Fig. 7C, 7D, 7G, 7H; supplemental online Fig. 2B). These

Figure 6. Three-dimensional multicellular structures undergo parti-tioning of differentiation. (A): Schematic showing method to createthree-dimensional multicellular structures of mesenchymal stem cells(MSCs). (B): Phase image of MSCs in three-dimensional structures atday 0. (C): Phase image of a three-dimensional MSC structure stainedfor oil droplets (red) and alkaline phosphatase (blue) after 14 days inmixed media. (D, E): Longitudinal section and cross-section of MSCmulticellular structures. Scale bars � 250 �m. Abbreviation: PDMS,polydimethylsiloxane.

Figure 5. Cytoskeletal tension is necessary for patterned differentiation. (A–D): Circular mesenchymal stem cell aggregates stained for oil droplets(red) and alkaline phosphatase (blue) after 14 days in mixed media without inhibitor (A) or in the presence of 50 �M blebbistatin (B), 10 �M Y-27632(C) or 10 �M ML-7 (D). (E–H): Bar graphs showing fraction of each cell type within each zone. � indicates p � .05 compared with correspondingmean in the untreated condition (E). Scale bar � 250 �m.

2925Alom Ruiz, Chen

www.StemCells.com

at UN

IVE

RSIT

Y O

F PEN

N L

IBR

AR

Y on M

arch 5, 2009 w

ww

.StemC

ells.comD

ownloaded from

results demonstrate that patterned partitioning of lineagespecification also occurs in three-dimensional matrices in atension-dependent manner.

DISCUSSION

During morphogenesis, cells execute massive structuralchanges that lead to new tissue forms while simultaneouslydifferentiating to populate those structures. In the adult, cellsreside in specific locations within multicellular structures andare subjected to a wide variety of stresses, such as shear,stretch, tension, and compression. The findings presentedhere suggest a direct link connecting multicellular form,stress distributions, and lineage fate of differentiating stemcells. This link may restrict the differentiation potential ofstem cells to specific lineage fates as a function of positionwithin a structure. It is interesting that in our three-dimen-sional cylinders, MSCs form a hollow shaft of bone with acore of fat, not unlike the anatomy universal to all vertebratelong bones [28]. Such hardwiring between tissue form andfate may provide a key mechanism to prevent spurious fatedecisions and therefore coordinate tissue development in avariety of settings. For example, geometric factors present inthe bone microenvironment could regulate the balance ofadipogenic and osteogenic differentiation critical to bonehomeostasis. The loss of trabecular bone in osteoporosis maynot only be a primary defect in the disease but also could feedback to alter the balance of convex and concave surfaceswhere differential forces exist, thereby further exacerbatingany population imbalances. Whether such geometric distur-bances actually contribute to such diseases remains to beexplored. Furthermore, in the context of engineering tissuesfor regenerative medicine, such constraints may limit ourability to generate differentiated tissues with arbitrary shapes.

Our findings also indicate that mechanical forces gener-ated by contractile activity of cells provide the patterningcues that couples form to fate decisions. The existence of

such cytoskeletal forces has been implicated in numerousdevelopmental processes, such as germ band elongation, pri-mary invagination, and neural convergent extension [29 –31].Cytoskeletal tension had previously been shown to result inmechanical loading of three-dimensional matrices [32],which can feed back to modulate cell behavior, such asdifferentiation of fibroblasts into myofibroblasts [33]. Here,we demonstrated that gradients of stress are generated simplyby the existence of tissue geometry, even when massiverearrangements are not occurring, and therefore should beconsidered as a potential candidate for driving patterns in cellfunction. Importantly, although such force gradients appearto be omnipresent in the experimental system presented here,likely as a result of basal contractility of MSCs, our studiesalso suggest that form and fate could be decoupled or en-gaged to a different extent through genetic or environmentalregulation of cellular contractile activity and that such ma-nipulations can be exploited in tissue engineering. Althoughour studies implicate force as one factor that can be patternedby multicellular form, others have reported localized gradi-ents in soluble morphogens also patterned by multicellularstructures [34 –36]. Together, gradients in forces and mor-phogens, then, likely provide the key links that orchestratethe shape changes that occur during morphogenesis and thecoordinate spatial differentials in cell differentiation.

CONCLUSION

The findings presented here suggest a role for mechanical forcesin linking multicellular organization to spatial differentials ofcell differentiation. Understanding how fate decisions are tied totissue form will provide a better appreciation of how cellsorchestrate morphogenetic processes, as well as roadmap fordirecting stem cell fates in regenerative therapies. Understand-ing how fate decisions are tied to tissue form will provide abetter appreciation of how cells orchestrate morphogenetic pro-

Figure 7. Three-dimensional (3D) multicellular structures undergo partitioning of differentiation that can be modulated by cytoskeletal tension. (A,B): Cross-sections of 3D mesenchymal stem cell (MSC) structures stained for oil droplets (red) and alkaline phosphatase (blue) after 14 days in mixedmedia in the absence or presence of 50 �M blebbistatin. (C, D): Cross-sections of 3D MSC structures produced with either green fluorescent proteinor ROCK�3-infected cells stained for oil droplets (red) and alkaline phosphatase (blue) after 14 days in mixed media. (E–H): Bar graphs showingfraction of each cell type within each zone. In (E), � indicates p � .05 for all comparisons of the same cell type across different zones or differentcell types within the same zone except for those marked with #. In (F) and (H), � indicates p � .05 compared with the corresponding mean in theirrespective controls (E, G). Scale bar � 100 �m. Abbreviations: Ad-GFP, adenovirus containing green fluorescent protein; Ad-ROCK�3, adenoviruscontaining ROCK�3.

2926 Multicellular Patterned Differentiation

at UN

IVE

RSIT

Y O

F PEN

N L

IBR

AR

Y on M

arch 5, 2009 w

ww

.StemC

ells.comD

ownloaded from

cesses, as well as a roadmap for directing stem cell fates inregenerative therapies.

ACKNOWLEDGMENTS

We thank Y. Wang for technical assistance and S. Raghavan, R.Desai, Y. Wang, D. Cohen, and E. Issa for helpful discussions.This work was supported in part by grants from the NIH (Grants

EB00262, HL73305, and GM74048), the Army Research OfficeMultidisciplinary University Research Initiative, and the PennCenter for Musculoskeletal Disorders.

DISCLOSURE OF POTENTIAL CONFLICTS

OF INTEREST

The authors indicate no potential conflicts of interest.

REFERENCES

1 Salazar-Ciudad I, Jernvall J, Newman SA. Mechanisms of patternformation in development and evolution. Development 2003;130:2027–2037.

2 Keller R. Shaping the vertebrate body plan by polarized embryonic cellmovements. Science 2002;298:1950–1954.

3 Johnson RL, Tabin CJ. Molecular models for vertebrate limb develop-ment. Cell 1997;90:979–990.

4 Nilsson O, Parker EA, Hegde A et al. Gradients in bone morphogeneticprotein-related gene expression across the growth plate. J Endocrinol2007;193:75–84.

5 Stoykova A, Treichel D, Hallonet M et al. Pax6 modulates the dorso-ventral patterning of the mammalian telencephalon. J Neurosci 2000;20:8042–8050.

6 Nemer G, Nemer M. Regulation of heart development and functionthrough combinatorial interactions of transcription factors. Ann Med2001;33:604–610.

7 Keller R, Davidson LA, Shook DR. How we are shaped: The biome-chanics of gastrulation. Differentiation 2003;71:171–205.

8 Kiehart DP, Galbraith CG, Edwards KA et al. Multiple forces contributeto cell sheet morphogenesis for dorsal closure in Drosophila. J Cell Biol2000;149:471–490.

9 Brouzes E, Farge E. Interplay of mechanical deformation and patternedgene expression in developing embryos. Curr Opin Genet Dev 2004;14:367–374.

10 Farge E. Mechanical induction of twist in the Drosophila foregut/sto-modeal primordium. Curr Biol 2003;13:1365–1377.

11 McBeath R, Pirone DM, Nelson CM et al. Cell shape, cytoskeletaltension, and RhoA regulate stem cell lineage commitment. Dev Cell2004;6:483–495.

12 Engler AJ, Sen S, Sweeney HL et al. Matrix elasticity directs stem celllineage specification. Cell 2006;126:677–689.

13 Kurpinski K, Chu J, Hashi C et al. Anisotropic mechanosensing by mes-enchymal stem cells. Proc Natl Acad Sci U S A 2006;103:16095–16100.

14 Yamamoto K, Sokabe T, Watabe T et al. Fluid shear stress inducesdifferentiation of Flk-1-positive embryonic stem cells into vascular en-dothelial cells in vitro. Am J Physiol Heart Circ Physiol 2005;288:H1915–H1924.

15 Schmelter M, Ateghang B, Helmig S et al. Embryonic stem cells utilizereactive oxygen species as transducers of mechanical strain-inducedcardiovascular differentiation. FASEB J 2006;20:1182–1184.

16 Nelson CM, Jean RP, Tan JL et al. Emergent patterns of growth con-trolled by multicellular form and mechanics. Proc Natl Acad Sci U S A2005;102:11594–11599.

17 Pirone DM, Liu WF, Ruiz SA et al. An inhibitory role for FAK inregulating proliferation: A link between limited adhesion and RhoA-ROCK signaling. J Cell Biol 2006;174:277–288.

18 Tan JL, Liu W, Nelson CM et al. Simple approach to micropattern cells

on common culture substrates by tuning substrate wettability. Tissue Eng2004;10:865–872.

19 Tan JL, Tien J, Pirone DM et al. Cells lying on a bed of microneedles:An approach to isolate mechanical force. Proc Natl Acad Sci U S A2003;100:1484–1489.

20 Sekiya I, Larson BL, Vuoristo JT et al. Adipogenic differentiation ofhuman adult stem cells from bone marrow stroma (MSCs). J Bone MinerRes 2004;19:256–264.

21 Rosen ED, Sarraf P, Troy AE et al. PPAR gamma is required for thedifferentiation of adipose tissue in vivo and in vitro. Mol Cell 1999;4:611–617.

22 Spiegelman BM. PPAR-gamma: Adipogenic regulator and thiazo-lidinedione receptor. Diabetes 1998;47:507–514.

23 Komori T, Yagi H, Nomura S et al. Targeted disruption of Cbfa1 resultsin a complete lack of bone formation owing to maturational arrest ofosteoblasts. Cell 1997;89:755–764.

24 Komori T, Kishimoto T. Cbfa1 in bone development. Curr Opin GenetDev 1998;8:494–499.

25 Komori T. Requisite roles of Runx2 and Cbfb in skeletal development.J Bone Miner Metab 2003;21:193–197.

26 Byers BA, Garcia AJ. Exogenous Runx2 expression enhances in vitroosteoblastic differentiation and mineralization in primary bone marrowstromal cells. Tissue Eng 2004;10:1623–1632.

27 Kojima H, Uemura T. Strong and rapid induction of osteoblast differen-tiation by Cbfa1/Til-1 overexpression for bone regeneration. J Biol Chem2005;280:2944–2953.

28 Wilson FJ. Histology Image Review. New York: McGraw Hill, 1997.29 Bertet C, Sulak L, Lecuit T. Myosin-dependent junction remodelling

controls planar cell intercalation and axis elongation. Nature 2004;429:667–671.

30 Davidson LA, Koehl MAR, Keller R et al. How do sea-urchins invagi-nate? Using biomechanics to distinguish between mechanisms of primaryinvagination. Development 1995;121:2005–2018.

31 Elul T, Koehl MAR, Keller R. Cellular mechanism underlying neuralconvergent extension in Xenopus laevis embryos. Dev Biol 1997;191:243–258.

32 Grinnell F. Fibroblast biology in three-dimensional collagen matrices.Trends Cell Biol 2003;13:264–269.

33 Vaughan MB, Howard EW, Tomasek JJ. Transforming growth factor-beta 1 promotes the morphological and functional differentiation of themyofibroblast. Exp Cell Res 2000;257:180–189.

34 Ruhrberg C, Gerhardt H, Golding M et al. Spatially restricted patterningcues provided by heparin-binding VEGF-A control blood vessel branch-ing morphogenesis. Genes Dev 2002;16:2684–2698.

35 Metzger RJ, Krasnow MA. Development: Genetic control of branchingmorphogenesis. Science 1999;284:1635–1639.

36 Nelson CM, VanDuijn MM, Inman JL et al. Tissue geometry determinessites of mammary branching morphogenesis in organotypic cultures.Science 2006;314:298–300.

See www.StemCells.com for supplemental material available online.

2927Alom Ruiz, Chen

at UN

IVE

RSIT

Y O

F PEN

N L

IBR

AR

Y on M

arch 5, 2009 w

ww

.StemC

ells.comD

ownloaded from

DOI: 10.1634/stemcells.2008-0432 2008;26;2921-2927; originally published online Aug 14, 2008; Stem Cells

Sami Alom Ruiz and Christopher S. Chen Structures

Emergence of Patterned Stem Cell Differentiation Within Multicellular

This information is current as of March 5, 2009

& ServicesUpdated Information

http://www.StemCells.com/cgi/content/full/26/11/2921including high-resolution figures, can be found at:

Supplementary Material http://www.StemCells.com/cgi/content/full/2008-0432/DC1

Supplementary material can be found at:

at UN

IVE

RSIT

Y O

F PEN

N L

IBR

AR

Y on M

arch 5, 2009 w

ww

.StemC

ells.comD

ownloaded from