Killian Paper Cell Adhesion surface shape in PNAS

8/11/2019 Killian Paper Cell Adhesion surface shape in PNAS

1/15


2/15

both (the mixedmedia ). Theculture wasstopped after oneweek and the cells were stained to reveal that MSCs differentiated tothe defined lineage and cells exposed to the mixed media showa mixture of both lineages ( Fig. S2). The remaining experiments were performed using patterned cells exposed to mixed media con-taining these competing soluble differentiation cues.

Directing MSC Differentiation with Geometric Cues. We first identi-fied the appropriate size of a patterned feature that would show

little bias in directing cell differentiation towards osteogenic andadipogenic fates. Cells that were adherent to mixed shapes with arange of geometric features and having areas of 1,000, 2,500 and,5 ; 000 m2 were treated with mixed media for one week and thenprobed with lineage specific markers. In all cases, approximately 50 60% of the cells had differentiated, but with different fates.For the small islands we observed primarily adipocyte character-istics (lipid vacuoles) while the large islands promoted an osteo-blast fate (alkaline phosphatase), consistent with the report by Chen and coworkers that cell size has a strong influence on dif-ferentiation (23) ( Figs. S3 and S4). For cells on patterns of inter-mediate area ( 2 ; 500 m2 ) the differentiation proceeded to give amixed population of adipocytes and osteoblasts. We next com-pared differentiation of MSCs on a series of shapes that main-tained this constant area but that displayed differences inaspect ratio and curvature.

Patterned cells that were cultured for one week in mixed mediaon rectangles having aspect ratios of 1 1 , 3 2 and 4 1 showed thatthe yield of osteogenesis increased with aspect ratio. Cells insquares (aspect ratio 1 1 ) showed 46% osteogenesis while cellson rectangles having aspect ratios of 3 2 and 4 1 displayed56% and 61% osteogenesis, respectively (Fig. 1 A ). We next eval-uated the influence of three shapes that each had pentagonal

symmetry but with different types of curvature (Fig. 1 B). The firstshape approximated a flower with large convex curves along eachedge. For this shape, 62% of MSCs differentiated to give adipo-cytes with the balance yielding osteoblasts. Cells patterned into apentagon shape with straight lines for the edges showed aneven distribution of adipocytes and osteoblasts. Cells having ashape approximating a star with concave edges and sharp pointsat thevertices hada 62%preference foran osteogenic fate. In allcases,a smallportion of differentiated cells ( 10% ) displayedboth

markers andwere notused in theanalysis.This experiment is strik-ing because the three shapes had a constant area and only subtlegeometric differences yet had a significant difference in directinglineage commitment. Significantly, the medium and soluble fac-tors wereunchanged in these experiments, showing that the shapecues imposed by theunderlying pattern were alone responsible forchanging the fate of the cells. This trend is consistent with ourear-lier study that showed an increasingly contractile cytoskeleton incells as they moved from a flower toa pentagonand finally to a starshape (18). In these experiments we used visualinspection to mark cells as adipocytes or osteoblasts, (or neither), as has been com-mon in other studies (23). To verify that this process reliably ca-tegorized the cells, we also repeated the analysis using anautomated image analysis algorithm. We acquired color imagesof individual cells and performed a color deconvolution to sepa-rate the red and purple channels for images of each dually stainedcell (see SI Text and Fig.S5 fordetails). Fig. 1 C showshistograms of color-specific intensities for the flower and star shapes. We as-signed fixed thresholds for designating lineage and report popula-tions of cells that show either of the stains, a combination of both,or neither. The trend in differentiation determined in this manneris comparable to that obtained by visual inspection and thereforeprovides strong evidence that shape alone is influencing thedifferentiation of adherent cells. For experiments that follow, we find that the fraction of cells that display either both or neitherstains is approximately constant and we therefore show only therelative populations of cells displaying one or the other marker.

Influence of Shape on Cytoskeleton in MSCs. We next characterizedthe cytoskeletal organization in MSCs on different shapes. Cells were cultured for 6 h under standard growth conditions and thenfixed and probed by immunofluorescence to observe the stressfibers and the focal adhesion complexes. Fig. 2 A C shows immu-nofluorescent images of cells cultured on 2 ; 500 m2 islandsstained for actin (green), vinculin (red), and merged images withnuclei in blue. Previous work using patterned substrates hasshown that cells assemble stress fibers along edges that overlapregions of substrate that are nonadhesive (15, 18). In the flowershape, stress fibers predominate at the acute corners between pe-tals where the cell spans nonadhesive regions. In sharp contrast tothe flower shape, the concave regions between points of the stargive rise to highly contractile regions where the cell spans acrossthe nonadhesive area. Similar results were obtained with increas-ing aspect ratio and contractility ( Fig. S6). To assess differences in

patterns of focal adhesion and stress fibers between cells, immu-nofluorescent heatmaps were generated for > 80 cells per shape(Fig. S7). On average, cells in star shapes show larger focal ad-hesions and stress fibers than cells in flower shapes. We note thatour finding is in contrast to a study by Chen and coworkers (17)that found that cell size, but not shape, governs the amount of focal adhesion. These results may stem from the use of differentcell lines, and perhaps a greater sensitivity of the MSCs to shape.

Immunofluorescent staining of myosin IIa, the primary motorprotein assembly that is responsible for cell contractility, was per-formed for MSCs adherent to the shapes (Fig. 2 D ). Average myo-sin IIa fluorescent heatmaps from a large number of cellsdemonstrate a higher degree of actomyosin contractility alongthe edges that overlap the nonadhesive regions of the monolayerfor cells on star shapes (Fig. 2 E ). These results suggest regions of

Adipocytes

Osteoblasts80

100AdipocytesOsteoblasts

A Badipogenesis

osteogenesis

% D

i f f e r e n

t i a t i o n

0

20

40

60

C 50osteogenesis

50 100 1500

10

20

30

40 adipogenesis

F r e q u e n c y

Intensity50 100 150

Intensity

0

20

40

60

80

c e l l f a

t e ( % )

100

Fig. 1. ( A) Percentage of cells captured on rectangles of varying aspect ratiodifferentiating to adipocyte or osteoblast lineage. ( B) Percentage of cells dif-ferentiating to either lineage when captured on fivefold symmetric shapes.(C ) Comparison of color intensity histograms generated by measuring the redand purple channels of cells in flower and star shapes ( n 393 ). The dotted line represents thresholds used to define lineage specification. The bar graphsummarizes lineage assignment from these cells using this method(R red channel, P purple channel). Error bars are standard deviationsfrom four separate experiments with over 200 cells per shape, p -value


3/15

local curvature on shapes that increase cytoskeletal tension of adherent cells promote osteogenesis relative to adipogenesis.The heatmaps show a slight variation (16% for the star and23% for the flower) in the intensities for symmetrical features.The heatmaps were constructed from cells that retained their re-lative alignment in the culture and therefore the asymmetry may reflect either a global polarization or intrinsic variability in cytos-keletal structure of patterned cells within a sample. We have not

performed experiments to address this question.Modulating Differentiation by Cytoskeletal Manipulation. We haveshown that the cytoskeleton in adherent cells is strongly influ-enced by subtle geometric shape cues, and suggest that the osteo-genic program in the cells is directly dependent on a contractilecytoskeleton. To further address this possibility, we evaluated theeffect of several pharmacological agents that are known to mod-ulate the cytoskeleton. Inhibitors were added to cell culture me-dia at concentrations that allowed complete spreading of theMSC to the underlying shape with no visually apparent changesto morphology. Control cultures without pharmacological agentsshow 72% of cells on the flower shape have an adipogenic fatecompared to 67% of cells on the star shape showing an osteogenicfate (Fig. 3 A ). We then compared the shape-dependent differen-tiation of cells that were treated with nocodazole, a potent micro-tubule depolymerizing agent that has been shown to increase cellcontractility (34). Indeed, the drug removed the influence of shape and in all cases gave a strong preference for osteogenesis(> 80% ) (Fig. 3 B ). Cytochalasin D is a small molecule that inhibitsF-actin polymerization and therefore reduces contractility of theactin cytoskeleton. This agent again removed the effect of shapeon differentiation but this time favored an adipogenic fate forcells on all shapes (Fig. 3 C ). Two other drugs that directly inhibitcontractility blebbistatin, which inhibits myosin II and Y-27632, which inhibits ROCK caused a decrease in osteogenesis with acorresponding increase in cells with adipocyte phenotypes andagain removed the influence of shape on differentiation(> 70% , Fig. 3 D , E ). Finally, we treated MSCs with an antibody

against the 51 integrin

which mediates coupling of the cytos-keleton to the extracellular matrix prior to adhesion, but using aconcentration of antibody that does not block spreading of thecell to fill the patterned island (35). We find that for bothshapes, approximately 85% of the cells choose an adipogenic fate(Fig. 3 F ). Treatment of cells on shapes with varying aspect ratio with these agents shows the same trend ( Fig. S8). Each of theseresults is consistent with the role of actomyosin contractility inpromoting an osteogenic fate in cells, and further serves to estab-lish the apparent requirement for actomyosin contractility inshape-dependent influence on differentiation.

RNA Expression Analysis of Cells on Flower and Star Shapes. Our ob-servations showing that osteogenesis is promoted by an increasedcytoskeletal contractility is consistent with previous reports de-

monstrating that stiff substrates promote osteogenesis (and whichis intuitively linked to the physical characteristics of bone) (23,27). To gain an insight on how actomyosin contractility regulatesthe osteogenesis program, we performed a gene expression analysis to establish the differentiated states of the MSCs and, as de-scribed further below, to identify gene families that were impor-tant in mediating the effect of shape on differentiation. RNA wasisolated from cells grown on 12 identical patterned substrates(maximum 2 ; 500 cells substrate) that were exposed for 1 week to mixed media on flower shapes, star shapes, and nonpatternedcontrols. When isolating RNA, we selected cultures whereingreater than 90% of the cells were attached to individual featuresand had spread to take on the shape of the feature. In this way,RNA derived from incompletely spread cells represented a minorfraction of the pool and was not expected to significantly alter theexpression profiles of the patterned cells. After one round of am-plification, we analyzed the RNA transcripts by hybridization to Affymetrix GeneChips and analyzed the relative expression levelsof known markers. A panel of common adipogenic transcripts

showed higher levels of expression in cells cultured on the flowershape relative to the star and unpatterned surfaces (normalizedto growth media control, Fig. 4 A . See Table S1 for fold change values). Similarly, a group of osteogenic transcripts displayedhigher expression for cells cultured in star shapes. The trendin expression level across lineage specific transcripts is consistent with the histological staining. To supplement the differential ex-pression of lineage specific transcripts we additionally performedimmunofluorescent staining and RT-PCR ( Fig. S9). Immuno-fluorescent staining of peroxisome proliferator-activated recep-tor (PPAR ) and Runt-related transcription factor 2(Runx2) markers for adipogenesis and osteogenesis that weredifferentially expressed in the microarray data reveal a compar-able trend to the histology stains in shape directed differentiation.The RT-PCR gels show increased expression of osteocalcin

n = 86

D ECAMax

Minn = 80

B

Fig. 2. ( A)-(C ) Immunofluorescent images of cells in flower and star shapesstained for F-actin ( green ), vinculin ( red ) and nuclei ( blue ). (D) Immunoflour-escent images of cells in flower and star shapes stained for myosin IIa. ( E )Fluorescent heatmaps of > 80 cells stained for myosin IIa as a quantitativemeasure of contractility. (Scale bar, 20 m).

% D

i f f e r e n

t i a t i o n

0

20

40

60

80

100AdipocytesOsteoblasts B

D

ANocodazolecontrol

% D

i f f e r e n

t i a t i o n

0

20

40

60

80

100Cytochalasin D Blebbistatin

Blocking antibodies 5, 1

C

F

% D i f f e r e n

t i a t i o n

0

20

40

60

80

100Y-27632

E

Fig. 3. ( A) Percentage of cells differentiating to adipocytes or osteoblastswhen cultured on patterns of the same shape, p -value < 0 .01 . (B)-(E ) Effecton shape-promoted differentiation in the presence of cytoskeleton disrup-tors and ( F ) integrin blocking antibodies.

4874 www.pnas.org/cgi/doi/10.1073/pnas.0903269107 Kilian et al.
http://www.pnas.org/cgi/data/0903269107/DCSupplemental/Supplemental_PDF#nameddest=SF8http://www.pnas.org/cgi/data/0903269107/DCSupplemental/Supplemental_PDF#nameddest=ST1http://www.pnas.org/cgi/data/0903269107/DCSupplemental/Supplemental_PDF#nameddest=SF9http://www.pnas.org/cgi/data/0903269107/DCSupplemental/Supplemental_PDF#nameddest=SF9http://www.pnas.org/cgi/data/0903269107/DCSupplemental/Supplemental_PDF#nameddest=ST1http://www.pnas.org/cgi/data/0903269107/DCSupplemental/Supplemental_PDF#nameddest=SF8http://www.pnas.org/cgi/data/0903269107/DCSupplemental/Supplemental_PDF#nameddest=SF8


4/15

(BGLAP) for cells in star shapes and lipoprotein lipase (LPL) forcells in flower shapes, consistent with the microarray results(Fig. S9 D ).

A comparison of the gene expression profiles in cells that wereadherent to star and flower shapes, but that were not yet treated with the differentiation cues shows significant differential expres-sion of genes associated with Wntsignalingandmitogen-activatedprotein kinasecascades (MAPK), both of these pathways arewell-studied in mechanotransduction and osteogenic regulation (36).

In the absence of soluble lineage guidance cues, both shapes ex-press transcripts associated with MAPK and Wnt signaling, butcells adherent to the star shape express higher levels of noncanonical (also known as Planar Cell Polarity (PCP) pathway) Wnt as-sociated transcripts including:Rac1, cell divisioncycle42 (Cdc42),RhoA, ROCK2, dishevelled 1, 2, and 3 (DVL1, 2, and 3) and di-shevelled associatedactivator of morphogenesis 1 (DAAM1). Thehigher expression of the GTPases involved in cell motility, Rac,and Cdc42, is consistent with recent work in our group that estab-lished greater polarity and protrusion with cells in a star geometry (18). Wnt11, the major player in the noncanonical PCP pathway,also shows > 10 -fold expression compared to cells adherent to theflower shape. Further, those cells adherent to the star shape havehigherexpression of Wntreceptors includingLRP5 and8 outof 10Frizzled receptors (FZD). (Fig. 4 B , left and Table S2 ). Previous

work has demonstrated the importance of Wnt signaling in regu-lating actomyosin contractility (37). Furthermore, high levels of the noncanonical Wnt downstream effector ROCK has beenshownto trigger osteogenesis through a myosin-generatedtensionfeedback loop (23, 38). It is therefore reasonable to suggest thatcontractile cells may be poised in a state of higher susceptibility tosoluble factors includingautocrine/paracrineWnt signals as wellasosteogenic media supplements.

Cells that were exposed to mixed media show significantly dif-

ferent expression profiles compared to patterned cells that hadnot been treated with soluble differentiation cues (Fig. 4 B , bluebox). Of particular interest, MSCs cultured on stars show elevated expression of genes involved in the MAP kinase pathwayslinked to both mechanotransduction and osteogenesis (39) in-cluding: extracellular signal-regulated kinases ERK1/2 (MAPK1,MAPK3), c-Jun N-terminal kinases JNK (MAPK8, MAPK9,MAPK10) and p38 kinases (MAPK11, MAPK12, MAPK13,MAPK14). We also find that star-shaped cells display higher ex-pression of 12 of the 17 Wnt transcripts and downstream effectorsof both the canonical and noncanonical pathways. In contrast,flower-shaped cells display increased expression of dikkopf-1(DKK1) and all of the secreted frizzled-related proteins (sFRP)involved in Wnt antagonism (see Table S2 for fold change values).

Inhibition of MAPK and Wnt Signaling in Patterned Cells. To deter-mine the extent to which MAP kinase cascades and Wnt signalingplay a role in shape-promoted osteogenesis, we tested the effectof inhibitors of these pathways on the differentiation. Addition of SB202190 a specific inhibitor of p38 phosphorylation (40) re-sulted in no discernible change in lineage commitment (Fig. 4 C ) whereas the ERK 1/2 inhibitor FR180204 and the JNK 1/2/3 in-hibitor SP600125 (41) caused a decrease in osteogenesis with aconcurrent increase in adipogenesis demonstrating the impor-tance of these cascades in osteoblast differentiation.

We also probed the role of Wnt signaling by supplementing themedia with the recombinant extracellular Wnt antagonists DKK1and sFRP3 (42). Treatment with DKK1 caused a slight decreasein osteogenesis and increase in adipogenesis relative to control(Fig. 4 D ). Incubation in the presence of sFRP3 caused a greaterdecrease in osteogenesis with an associated increase in adipogen-esis. Combining both DKK1 and sFRP3 led to a further increasein adipogenesis relative to control. Importantly, treatment withWnt inhibitors abrogated star-shape-promoted trends in osteo-genesis. From these results we speculate that cell contractility ele- vates ERK/JNK cascades and makes MSCs in star shapes moresusceptible to secreted molecules (Fig. 4 E ).

Limits of Geometric Control over MSC Fate. The studies describedabove used shapes that yielded differentiation of approximately 60 70% of one fate. We asked whether shapes that combined lo-cal curvature with aspect ratio could increase the preference forthe major outcome. Therefore, we compared the differentiationof cells on a circular shape and a holly leaf that incorporates

two osteogenesis-promoting cues: moderate aspect ratio ( 2 1 )and regions of subcellular curvature and concavity. Immunofluor-escence staining shows that cells confined to the holly shape havelarger focal adhesions and stress fibers spanning the nonadhesiveregions compared to cells in circular patterns ( Fig. S10). Further,heat maps showing the localization of myosin IIa show a strikingdifference in the contractility state of the cells (Fig. 5 A ). MSCs onboth shapes were exposed to mixed media for 1 week and stainedfor markers of osteogenesis andadipogenesis. Cells in the circularisland favored the adipogenic fate (74%) while the contractilecells in the holly shape favored the osteogenic fate (67%)(Fig. 5 B). The difference in differentiation between these shapesis approximately equal to that observed with the flower andstar shape. Thus, we infer that under the conditions employedin our work, we are realizing the maximum influence that shape

A B

ACSL4STAT5ASREBF1CEBPDNR1H3ACSL5LPLACSL1FABP4KLF15CEBPAPPARGPPARG1B

adipogenesis

MAPK12WNT7AMAPK1MAPK10WNT7BWNT4WNT6WNT5BGSK3BLEF1WNT2WNT1

Wnt/MAPK

-10

0

10

20

30JNKERK1/2p38

CNP NP

PPARG1BAACSLNCOR2EBF1ACSL6NCOR1KLF5EGR2ARNTL2CEBPGDDIT3KLF6EPAS1ACSL3PPARGC1ACEBPB

osteogenesis

WIF1WNT11

WNT8AFZD7DVL3WNT10AFZD2LRP5MAPK3BTRCCSNK1EAXIN1WNT3ROR2SOSTDVL2MAPK8AXIN2FZD8 0

10

20

30-30

-20

sFRP3DKK1D DKK1sFRP3

Osteo

OsteoAdipoAdipo

BMP15BMP3SMAD4SMAD9BMP5SMAD1SMAD3BMP8ABMP7RUNX2TGFB1BMP8BVDRBMP1BMP2BMP10

FZD8WNT9AFRAT1FZD5LRP6SFRP2SFRP5FZD4WNT8BCTNNB1WNT5AMAPK9SFRP1ROCK2SFRP4DAAM1MAPK13CDC42MAPK11

% D

i f f e r e n c e

% D

i f f e r e n c e

-30

-20

-10

0

ERK/JNK ROCKWnt

osteogenesis

E

NP

BMP10BGLAPTFIP11ALPLSMAD5SMAD2BMP6MGPSMAD6BMPR1ACOL1A2BMP4SMAD7SPP1SP7GDF2MSX2COL1A1IBSP

WNT9BFZD10FZD6WNT10BMAP2K1RHOARAC1APCWNT2BFZD9FZD1DVL1CTHRC1WNT16MAPK14DKK1FZD3

Fzd/LRP adhesion

RhoA/Rac/Cdc42Dsh

Daam1

-Max Max1

Fig. 4. Hierarchical clustering of ( A) osteogenic and adipogenic transcriptsexpressed in cells captured to both shapes and ( B) MAP kinase and Wnt sig-naling transcripts: Left cells in flower and star shapes cultured in standard

growth media for 1 week. Right (blue outline )

after 1 week exposure tomixed adipogenic and osteogenic media. ( C ) Change in differentiation fromcontrol when exposed to MAPK inhibitors. ( D) Change in differentiation ofcells in the presence of Wnt antagonists. ( E ) Speculative pathway for shapedirected differentiation of adherent MSCs.

Kilian et al. PNAS March 16, 2010 vol. 107 no. 11 4875
http://www.pnas.org/cgi/data/0903269107/DCSupplemental/Supplemental_PDF#nameddest=SF9http://www.pnas.org/cgi/data/0903269107/DCSupplemental/Supplemental_PDF#nameddest=SF9http://www.pnas.org/cgi/data/0903269107/DCSupplemental/Supplemental_PDF#nameddest=ST2http://www.pnas.org/cgi/data/0903269107/DCSupplemental/Supplemental_PDF#nameddest=ST2http://www.pnas.org/cgi/data/0903269107/DCSupplemental/Supplemental_PDF#nameddest=SF10http://www.pnas.org/cgi/data/0903269107/DCSupplemental/Supplemental_PDF#nameddest=SF10http://www.pnas.org/cgi/data/0903269107/DCSupplemental/Supplemental_PDF#nameddest=SF10http://www.pnas.org/cgi/data/0903269107/DCSupplemental/Supplemental_PDF#nameddest=ST2http://www.pnas.org/cgi/data/0903269107/DCSupplemental/Supplemental_PDF#nameddest=ST2http://www.pnas.org/cgi/data/0903269107/DCSupplemental/Supplemental_PDF#nameddest=SF9


5/15

can have in directing differentiation. We would expect, however,that synergistic combinations of growth factors and small mole-cules in conjunction with geometric cues can achieve a more spe-cific outcome.

DiscussionIn this paper we used patterned cells to investigate the relation-ship between shape and differentiation of multipotent MSCs. Thepatterned monolayers provide exquisite control over the shapesof individual cells, thereby enabling statistically relevant observa-tions over a large number of cells in a single well of a tissue cul-ture plate. In this study, we varied the aspect ratio and subcellularcurvature of individual cells and could include multiple shapes ona single substrate. MSCs survived and remained confined to pat-terns for a week allowing introduction of differentiation inducingchemical cues. We found that shapes promoting increased con-tractility led to preferential osteogenesis when cells were exposedto a mixture of lineage cues. In contrast, cells in shapes that pro-mote low contractility preferred to follow an adipocyte lineage.

The influence of geometry on differentiation that we observe isconsistent with a model wherein the local shape cues (i.e. pointedfeatures between concave regions) promote increased myosincontractility which enhances pathways associated with osteogen-esis. Immunofluorescence analysis of the cells, for example, con-firms the relationship between the shape cues and the enhancedstress filaments in the patterned cells (Fig. 2 and Fig. S7). Treat-ment of the cells with cytoskeletal-disrupting agents gave resultsthat are also consistent with this model. Nocodazole, for example,increases the contractility of the cytoskeleton and was shown todrive the majority of MSCs in shapes towards osteoblast fate.In contrast, cells exposed to molecules that inhibit contractility tended to differentiate into adipocytes. Further, by blocking aportion of cell surface integrin receptor with an antibody, wecould decrease the tension exerted by the cell on the substrate with a corresponding inhibition of osteogenesis. Taken togetherthese results demonstrate the importance of adhesion and ahighly contractile state for bone cell formation.

This interpretation is also consistent with the report by Chenand coworkers showing that myosin-generated cytoskeletal ten-

sion that follows spreading of MSCs leads to higher levels of RhoA, ROCK, and myosin light chain phosphorylation (23). Thistension-specific feedback loop was shown to act as a switch forosteogenesis. Further, a very recent report by Chien and cowor-kers studied the differentiation of MSCs on titanium oxide nano-tubes and found that larger tubes led to an increased spreading of cells and a corresponding increase in osteogenesis (31). Discherand coworkers varied the stiffness of the underlying matrix inMSC culture and showed elasticity to be a powerful mechanicalcue in directing MSC fate (27). Stiff matrices lead to enhancedcytoskeletal tension and osteogenesis while softer matrices direc-ted MSCs towards alternative lineages. Interestingly, the syn-thetic and natural matrices that had a comparable directingeffect in cell fate also shared a similar stiffness. We note the temp-tation to speculate that cells differentiate in response to shape

cues in a fashion that is consistent with the native geometry of cells of that lineage for example, round shapes with low stresspromote fat cells and contractile pointed shapes promote bone but we have no direct evidence to support this notion. In any event, this body of work demonstrates the significant roles thatphysical cues can play to increase cytoskeletal tension in theMSC microenvironment and the relevance of these effects to pro-moting osteogenesis.

To aid in understanding the role that shape plays in regulating

MSC differentiation, we performed gene expression analysisusing DNA microarrays. After exposing patterned cells to mixedmedia containing adipogenic/osteogenic soluble cues we see ahigher degree of osteogenic transcript expression for cells cul-tured in star shapes and higher expression levels of adipogenictranscripts in cells on flower shapes, consistent with the histo-logical studies. In both cases of normal media and mixed media,microarray analysis reveals differential gene expression for tran-scripts associated with MAP kinase pathways and Wnt signalingsuggesting a role for secreted factors. In particular, we find thatcells cultured on star shapes in growth media promote expressionof noncanonical Wnt/Fzd signaling molecules (including down-stream effectors RhoA and ROCK previously shown to be involved in osteogenesis (23, 38)). Moreover, when exposed tomixed media the gene expression of tension specific MAP kinases(p38, ERK1/2, JNK) and both canonical and noncanonical Wnttranscripts is increased for cells on star shapes relative to those onflower shapes and on unpatterned surfaces. In contrast, cells inflower shapes show elevated expression of Wnt inhibitory mole-cules (sFRPs), suggesting that this geometry may stimulate func-tional Wnt antagonism.

Wnt signaling is known to be important in osteoblast differen-tiation (36). Recent studies have also shown how Wnts are activated by mechanical stimuli during bone development (43, 44) with additional functions in regulating cell contractility during tis-sue morphogenesis (37, 45). p38, ERK, and JNK cascades areknown to be activated by mechanical stimuli but have also beenimplicated as targets of Wnt signaling through canonical (40) and noncanonical pathways (41, 46, 47) for regulation of osteogenesis.Taken together, these studies and our results support a picture where actomyosin contractility stimulates MAPK cascades andWnt signaling to regulate osteoblast differentiation.

To support this hypothesis, treatment of cells in both shapes with MAP kinase inhibitors shows that ERK1/2 and JNK are im-portant for osteogenesis and their inhibition causes increased adi-pogenesis when exposed to mixed media. This result is consistent with previous studies that point to the important role of ERK1/2and JNK in osteogenesis (41, 46, 48, 49) as well as an inhibitory role of JNK in adipogenesis (49). As further support, Wnt inhibi-tion by DKK1 and sFRP3 led to a decrease in osteogenesis with aconcurrent increase in adipogenesis. From these results, wepropose that cells in star shapes increase the levels of the noncanonical Wnt signaling molecules, their receptors, and their down-stream effectors. Subsequently, activationof ERK/JNK by soluble

cues andautocrine/paracrineWnt signalingis enhanced forcellsinthis geometry leading to elevated expression of master osteoblastregulators (see Fig. 4 E ). Likewise, it appears that cells in geome-tries associated with low contractility may not only make the cellsmore susceptible to soluble adipogenic cues but may also promotenegative regulation of Wnt signaling via the secretion of sFRPs.

This work has harnessed the utility of microcontact printingto control cell shape and allow systematic and highly parallelstudies of the factors that affect stem cell fate. This techniqueis also notable because it can be used to study large populationsof cells, providing good statistical data, and even providing suffi-cient RNA to permit gene expression profiling. By keeping thearea andshape of cells constant, the heterogeneities that normally attend cell cultures are reduced, giving a targeted investigation of the synergy between cell shape, biological, and chemical signals.

n = 85

n = 83

A B

% D

i f f e r e n

t i a t i o n

0

20

40

60

80


Fig. 5. ( A) Myosin IIa immunofluorescent heat maps of > 80 cells cultured incircles or holly shapes. ( B) Corresponding differentiation of cells in bothshapes after exposure to mixed media, p -value < 0 .005 .

4876 www.pnas.org/cgi/doi/10.1073/pnas.0903269107 Kilian et al.
http://www.pnas.org/cgi/data/0903269107/DCSupplemental/Supplemental_PDF#nameddest=SF7http://www.pnas.org/cgi/data/0903269107/DCSupplemental/Supplemental_PDF#nameddest=SF7http://www.pnas.org/cgi/data/0903269107/DCSupplemental/Supplemental_PDF#nameddest=SF7


6/15

Differentiation media supplements have been shown to operatethrough temporal regulation of different signaling cascades(48). Histology staining and RNA expression analysis shows highlevels of activation in oneweek compared to unpatterned cells un-der the same media conditions. Thus the use of shape cues can beemployed to enhance and rationally control differentiation speci-fic signaling, thereby accelerating biochemical assays and guideddifferentiation in engineered scaffolds. Therefore, this study pro- vides both unique tools and methodologies for investigative biol-ogy as well as design principles that will prove useful for designingmaterials for regenerative medicine.

Materials and MethodsFurther details of materials and methods are included in SI Text .

Cell Culture. Human MSCs were a gift from Dr. Andy Xiang Peng at Sun Yat-Sen University, and derived as described (33). Cells were cultured in DMEM-

low glucose supplemented with 10% FBS, 1% penicillin/streptomycin, pas-saged at 70 80% confluency and seeded at 5 ; 000 cells cm 2 . Differentiationmedia was added by itself or as a 1 1 (v/v) combination.

Surface Preparation. Microcontact printing of self-assembled monolayers onthin gold layers was performed as described previously (18).

Histology. Cells were fixed with 4% formaldehyde and stained for adipogen-esis (lipid droplets, Oil Red O), and osteogenesis (alkaline phosphatase, BCIP/ NBT) per manufacturer s instructions.

ACKNOWLEDGMENTS. We thank Dr. Vytas Bindokas for assistance with micro-scopy data analysis. This work was funded by the National Cancer Institute ofthe National Institutes of Health. K.A.K. is supported by a Ruth L. KirsteinNational Research Service Award Number F32GM087048 from the NationalInstitute of General Medical Sciences. Photolithography was performed atthe Chicago Materials Research Science and Engineering Centers microfluidicfacility (National Science Foundation).

1. Meirelles LdS, Chagastelles PC, Nardi NB (2006) Mesenchymal stem cells reside in vir-tually all post-natal organs and tissues. J Cell Sci 119:2204 2213.

2. Crisan M, et al. (2008) A perivascular origin for mesenchymal stem cells in multiplehuman organs. Cell Stem Cell 3:301 313.

3. Kassem M (2006) Stem cells: Potential therapy for age-related diseases. Ann N YAcad Sci 1067:436 442.

4. Ohlstein B, Kai T, Decotto E, Spradling A (2004) The stem cell niche: Theme and varia-tions. Curr Opin Cell Biol 16:693 699.

5. Pittenger MF, et al. (1999) Multilineage potential of adult human mesenchymal stemcells. Science 284:143 147.6. Caplan AI, Bruder SP (2001) Mesenchymal stem cells: Building blocks for molecular

medicine in the 21st century. Trends Mol Med 7:259 264.7. Derda R, et al. (2007) Defined substrates for human embryonic stem cell growth iden-

tified from surface arrays. ACS Chem Biol 2:347 355.8. Graziano A, et al. (2007) Concave pit-containing scaffold surfaces improve stem cell-

derived osteoblast performance and lead to significant bone tissue formation. PLoS One 2(6):e496 1 9.

9. McBride SH, Falls T, Knothe Tate ML (2008) Modulation of stem cell shape and fate b:Mechanical modulation of cell shape and gene expression. Tissue Eng Pt A14:1573 1580.

10. Wolff J (1986) The law of bone remodelling , ed Maquet P (Springer-Verlag and Hei-delberg GmbH and Co., Berlin), pp 1 126.

11. O Neill C, Jordan P, Ireland G (1986) Evidence for two distinct mechanisms of ancho-ragestimulation in freshly explantedand 3T3swiss mousefibroblasts. Cell 44:489 496.

12. Ingber DE (2006) Mechanical control of tissue morphogenesis during embryologicaldevelopment. Int J Dev Biol 50:255 266.

13. Spatz JP, Geiger B (2007) Molecular engineering of cellular environments: Cell adhe-sion to nano-digital surfaces. Methods Cell Biol 83:89 111.

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

15. Thery M, Pepin A, Dressaire E, Chen Y, Bornens M (2006) Cell distribution of stress fi-bres in response to the geometry of the adhesive environment. Cell Motil Cytoskele-ton 63:341 355.

16. Parker KK, Tan J, Chen CS, Tung L (2008) Myofibrillar architecture in engineered car-diac myocytes. Circ Res 103:340 342.

17. Chen CS, Alonso JL, Ostuni E, Whitesides GM, Ingber DE (2003) Cell shape providesglobal control of focal adhesion assembly. Biochem Biophys Res Commun307:355 361.

18. James J, Goluch ED, Hu H, Liu C, Mrksich M (2008) Subcellular curvature at the peri-meter of micropatterned cells influences lamellipodial distribution and cell polarity.Cell Motil Cytoskeleton 65:841 852.

19. Brock A, et al. (2003) Geometric determinants of directional cell motility revealedusing microcontact printing. Langmuir 19:1611 1617.

20. Parker KK, et al. (2002) Directional control of lamellipodia extension by constrainingcell shape and orienting cell tractional forces. FASEB J 16:1195 1204.

21. Jiang X, Bruzewicz DA, Wong AP, Piel M, Whitesides GM (2005) Directing cell migra-tion with asymmetric micropatterns. Proc Natl Acad Sci USA 102:975 978.

22. Kandere-Grzybowska K, et al. (2007) Cell motility on micropatterned treadmills andtracks. Soft Matter 3:672 679.

23. McBeath R, Pirone DM, Nelson CM, Bhadriraju K, Chen CS (2004) Cell shape, cytoske-letal tension, and RhoA regulate stem cell lineage commitment. Dev Cell 6:483 495.

24. Bhadriraju K, et al. (2007) Activation of ROCK by RhoA is regulated by cell adhesion,shape, and cytoskeletal tension. Exp Cell Res 313:3616 3623.

25. Murphy WL, Hsiong S, Richardson TP, Simmons CA, Mooney DJ (2004) Effects of abone-like mineral film on phenotype of adult human mesenchymal stem cells in vitro.Biomaterials 26:303 310.

26. Connelly JT, Garcia AJ, Levenston ME (2008) Interactions between integrin liganddensity and cytoskeletal integrity regulate BMSC chondrogenesis. J Cell Physiol 217(1):145 154.

27. Engler AJ, Sen S, Sweeney HL, Discher DE (2006) Matrix elasticity directs stem cell line-age specification. Cell 126:677 689.

28. Ward DF, Jr, et al. (2007) Mechanicalstrain enhancesextracellular matrix-induced genefocusing and promotes osteogenic differentiation of human mesenchymal stem cellsthrough an extracellular-related kinase-dependent pathway. Stem Cells Dev 16:467 480.

29. Yang Y, Relan NK, Przywara DA, Schuger L (1999) Embryonic mesenchymal cells sharethe potential for smooth muscle differentiation: Myogenesis is controlled by the cell sshape. Development 126:3027 3033.

30. Ni XF, et al. (2008) On-chip differentiation of human mesenchymal stem cells into

adipocytes. Microelectron Eng 85:1330 1333.31. Oh S, et al. (2009) Stem cell fate dictated solely by altered nanotube dimension. Proc

Natl Acad Sci USA , 106(7) pp:2130 2135.32. YuW, et al. (2008) Critical roleof phosphoinositide 3-kinase cascade in adipogenesis of

human mesenchymal stem cells. Mol Cell Biochem 310:11 18.33. Zhang A-X, et al. (2007) Proteomic identification of differently expressed proteins

responsible for osteoblast differentiation from human mesenchymal stem cells.Mol Cell Biochem 304:167 179.

34. Chang Y-C, Nalbant P, Birkenfeld J, Chang Z-F, Bokoch GM (2008) GEF-H1 couplesnocodazole-induced microtubule disassembly to cell contractility via RhoA. Mol Biol Cell 19:2147 2153.

35. Feng Y, Mrksich M (2004) The synergy peptide PHSRN and the adhesion peptide RGDmediate cell adhesion through a common mechanism. Biochemistry 43:15811 15821.

36. Hartmann C (2006) A Wnt canon orchestrating osteoblastogenesis. Trends Cell Biol 16:151 158.

37. Lee J-Y, et al. (2006) Wnt/frizzled signaling controls c. Elegans gastrulation by activat-ing actomyosin contractility. Curr Biol 16:1986 1997.

38. Arnsdorf EJ, Tummala P, Kwon Ronald Y, Jacobs CR (2009) Mechanically inducedosteogenic differentiation the role of RhoA, ROCKII and cytoskeleton dynamics. J Cell Sci 122:546 553.

39. Jaiswal RK, et al. (2000) Adult human mesenchymal stem cell differentiation to theosteogenic or adipogenic lineage is regulated by mitogen-activated protein kinase. J Biol Chem 275:9645 9652.

40. CaverzasioJ, Manen D (2007) Essential roleof Wnt3a-mediated activation of mitogen-activated protein kinase p38 for the stimulation of alkaline phosphatase activity andmatrix mineralization in C3H10T1/2mesenchymal cells. Endocrinology 148:5323 5330.

41. Fu L, et al. (2008) Stimulation of osteogenic differentiation and inhibition of adipo-genic differentiation in bone marrow stromal cells by alendronate via ERK and JNKactivation. Bone 43:40 47.

42. Alfaro MP, et al. (2008) The Wnt modulator sFRP2 enhances mesenchymal stem cellengraftment, granulation tissue formation and myocardial repair. Proc Natl Acad Sci USA 105:18366 18371.

43. Hens JR, et al. (2005) TOPGAL mice show that the canonical Wnt signaling pathway isactive duringbone development andgrowth and is activatedby mechanicalloading invitro. J Bone Miner Res 20:1103 1113.

44. Robinson JA, et al. (2006) Wnt/beta -catenin signaling is a normal physiologicalresponse to mechanical loading in bone. J Biol Chem 281:31720 31728.

45. Wozniak MA, Chen CS (2009) Mechanotransduction in development: A growing rolefor contractility. Nat Rev Mol Cell Biol 10:34 43.

46. Chang J, et al. (2007) Noncanonical Wnt-4 signaling enhances bone regeneration ofmesenchymal stem cells in craniofacial defects through activation of p38 MAPK. J Biol Chem 282:30938 30948.

47. Bikkavilli RK, Feigin ME, Malbon CC (2008) G o mediates WNT-JNK signaling throughDishevelled 1 and 3, RhoA family members, and MEKK 1 and 4 in mammalian cells. J Cell Sci 121:234 245.

48. Jaiswal RK, et al. (2000) Adult human mesenchymal stem cell differentiation to theosteogenic or adipogenic lineage is regulated by mitogen-activated protein kinase. J Biol Chem 275:9645 9652.

49. Tominaga S, Yamaguchi T, Takahashi S-I, Hirose F, Osumi T (2005) Negative regulationof adipogenesis from human mesenchymal stem cells by Jun N-terminal kinase.Biochem Biophys Res Commun 326:499 504.

Kilian et al. PNAS March 16, 2010 vol. 107 no. 11 4877
http://www.pnas.org/cgi/data/0903269107/DCSupplemental/Supplemental_PDF#nameddest=STXThttp://www.pnas.org/cgi/data/0903269107/DCSupplemental/Supplemental_PDF#nameddest=STXT


7/15

Supporting InformationKilian et al. 10.1073/pnas.0903269107SI TextSupplementary Methods. Materials. Laboratory chemicals and re-agents were purchased from Sigma-Aldrich unless otherwise spe-cified. Cell culture materials were purchased from FisherScientific. Media and reagents were purchased from Gibco. Bleb-bistatin, cytochalasin D, nocodazole, Y-27632, FR180204 (ERK inhibitor), SP600125 (JNK inhibitor), and SB202190 (p38 inhibi-tor) were purchased from Calbiochem. Recombinant humanDkk-1 and sFRP-3 were obtained from R&D systems. Mouse5 and 1 Integrin blocking antibodies were purchased from Che-micon. Rabbit anti-Myosin IIa and mouse anti-vinculin were pur-chased from Sigma. Mouse anti-Runx2 was purchased from Abcam (ab76956) and rabbit anti-PPAR (C26H12) was pur-chased from Cell Signalling Technologies. Alexa488-phalloidinand 4 ,6- diamidino-2-phenylindole (DAPI) were purchased fromInvitrogen. BCIP/NBTsolution for alkaline phosphatase was pur-chased from AMRESCO. Glass coverslips for surface prepara-tion were purchased from Fisher Scientific.

Surface preparation. Surfaces were fabricated by electron beamevaporation of a 40 thick layer of titanium followed by a280 thick gold layer onto glass coverslips. To create patternedsurfaces, polydimethylsiloxane (PDMS, Polysciences, Inc.)stamps were fabricated by polymerization upon a patternedmaster of photoresist (SU-8, MicroChem) created using UVphotolithography through a laser printed mask. The stamp wasinked with 5 mM ocatadecanethiol in ethanol, dried undernitrogen, and applied to the surface. 3 mM tri(ethylene glycol)undecanethiol in ethanol was applied for 16 h to discourageprotein and cell adhesion outside of patterns. Next, 25 g mL fibronectin in PBS was applied for 1 h and the surfaces storedunder sterile PBS until use. Equal deposition of matrix across shapes and surfaces was verified by antibody staining of Fibronectin.

Inhibition assays. Inhibitors were added to cell culture media at thefollowing concentrations that were determined empirically toallow complete spreading with no visually apparent changes tomorphology: Nocodazole (1 M), Cytochalasin D (0.2 M), Bleb-bistatin (1 M), and Y-27632 (2 M) (Calbiochem). Integrinblocking antibodies (5 and 1, Chemicon) were added to cellsin media prior to deposition at 1 g mL. MAP kinase inhibition was performed by adding supplemented media of the followingmolecules at 6 M after cell seeding and with each media change:FR180204 (ERK1/2), SP600125 (JNK), and SB202190 (p38)(Calbiochem). For Wnt inhibition, media was supplemented daily with Dkk-1 and sFRP-3 (5 nM) (R&D systems).

Histology and immunofluorescence. After incubation for 1 week,surfaces were fixed with 4% formaldehyde (Ted Pella, Inc.)and each sample stained with both a lipid staining solution foradipogenesis (Oil Red O, Sigma), and a alkaline phosphatasestain for osteogenesis (BCIP/NBT, Amresco) per manufacturersinstructions. For immunofluorescence, cells were permeablizedin 0.1% Triton X-100 inPBS, blockedwith 1% BSA. Primary anti-body labeling was performed in 5% goat serum containing 1%BSA in PBS for 20 min at 37 C with either mouse anti-vinculinor rabbit anti-MyosinIIa (Sigma). Primary antibody labeling formouse anti-Runx2 and rabbit anti-PPAR were performed in 1%BSA in PBS for 8 16 h at 4 C. Secondary antibody labeling wasperformed using the same procedure with Alexa-647 labeled goatanti-mouse or anti-rabbit IgG containing Alexa488-phalloidin

and DAPI (Invitrogen). Immunofluorescence microscopy wconducted using an Olympus IX81 spinning disk confocal miscope fitted with a Hamamatsu back-thinned EM-CCD cameImages were exported from Slidebook (Intelligent Imaging In vations) into ImageJ (NIH) for analysis.

RNA isolation for microarray experiments. Adherent cells were lyseddirectly in TRIZOL reagent (Invitrogen), purified according vendors protocol and used for microarray probe synthesis folloing standard Affymetrix protocols. RNA samples were ampliusing TargetAmp 1-Round aRNA Amplification Kit 103 (Epi-centre) according to vendor protocols. Hybridization, labeliand scanning were all performed by the University of ChicFunctional Genomics Facility. The labeled cRNA sample freach sample was hybridized to GeneChip Human GenomU133A 2.0 Array. Probe-level analyses of the images from sning of chips were performed using Affymetrix GeneChip Opating Software (GCOS). Threshold detection p-values were set toassign

present

( p < 0 .05 ),

marginal

(0 .05 p 0 .49 ), or absent ( p > 0 .49 ) decision calls for each gene assigned byMAS 5.0 criteria using GCOS. Data between samples were cpared by model-based expression index analysis using dC(http://biosun1.harvard.edu/complab/dchip).

Semiquantitative RT-PCR. Total RNA was reverse transcribed usingSuperscript III First Strand Synthesis System for RT-PC(Invitrogen). Primer sequences were as follows: BGLAPGGCGCTACCTGTATCAATGG and AGCAGAGCGA-CACCCTAGAC, LPL GAGATGGAGAGCAAAGCCCT and ATGAGGTGGCAAGTGTCCTC, GAPDH GCCACA-TCGCTCAGACAC and CTCGCTCCTGGAAGATGGT. Areactions were performed linearly by cycle number for easet of primers.

Microscopy data analysis. For the histological analysis, cells wereinspected by phase contrast at 40X to determine whether thexpressed lineage specific markers. For patterned MSCs only gle cells that adhered to the pattern were used for statistical an ysis. MSCs exposed to differentiation media cocktails eitstained strongly or notat all for the lineage specific markers. Cthat contained lipid vacuoles stained red by OilRedO wecounted as adipocyte specification. Cells that stained deep purfor alkaline phosphatase were counted as osteoblast committcells. A small subset of the differentiated cells (10% ) exposed tomixed osteo/adipo differentiation cues expressed both markand were ignored for the purpose of the analysis. For additioquantitation we performed a color deconvolution of all ceacross a surface stained with both markers. First, we usedpseudo-flat-field image correction algorithm (http:// ww w. macb iophot onic s. ca/ imagej /image_ in tensity_pro.htm#intensity_BG) to correct background and image intensitiesof 32-bit color images of cells taken at 10X. Next, color vec were assigned using a publically available color deconvoluplugin for ImageJ commonly used for separating similar histostains (http://www.dentistry.bham.ac.uk/landinig/softwarcdeconv/cdeconv.html). This takes a RGB image and using thecolor-specific vectors, separates out the red and purple channinto 8-bit binary images. Cells were then identified using a Gaian blur; the intensity was measured in each channel within a cified pixel radius and assigned a score based on user definedtensity thresholds. Thresholds were defined based on visuexamination of the binary images (see Fig. S5). For generat

Kilian et al. www.pnas.org/cgi/doi/10.1073/pnas.0903269107 1 of 9
http://biosun1.harvard.edu/complab/dchiphttp://www.macbiophotonics.ca/imagej/image_intensity_proce.htm#intensity_BGhttp://www.macbiophotonics.ca/imagej/image_intensity_proce.htm#intensity_BGhttp://www.macbiophotonics.ca/imagej/image_intensity_proce.htm#intensity_BGhttp://www.dentistry.bham.ac.uk/landinig/software/cdeconv/cdeconv.htmlhttp://www.dentistry.bham.ac.uk/landinig/software/cdeconv/cdeconv.htmlhttp://www.pnas.org/cgi/doi/10.1073/pnas.0903269107http://www.pnas.org/cgi/doi/10.1073/pnas.0903269107http://www.dentistry.bham.ac.uk/landinig/software/cdeconv/cdeconv.htmlhttp://www.dentistry.bham.ac.uk/landinig/software/cdeconv/cdeconv.htmlhttp://www.macbiophotonics.ca/imagej/image_intensity_proce.htm#intensity_BGhttp://www.macbiophotonics.ca/imagej/image_intensity_proce.htm#intensity_BGhttp://www.macbiophotonics.ca/imagej/image_intensity_proce.htm#intensity_BGhttp://www.macbiophotonics.ca/imagej/image_intensity_proce.htm#intensity_BGhttp://www.macbiophotonics.ca/imagej/image_intensity_proce.htm#intensity_BGhttp://biosun1.harvard.edu/complab/dchiphttp://biosun1.harvard.edu/complab/dchiphttp://biosun1.harvard.edu/complab/dchip


8/15

immunofluorescent heatmaps, MSCs cultured on both shapes were imaged on the same day using the same microscope andcamera settings. Raw fluorescent images were aligned in image

J with the same orientation as cultured across the surface, incporated into a Z stack and the average intensity calculated fheatmap generation.

A2) hydrophobic-CH 3 terminated

1) cell resistant-OH terminated

SAu

SAu

SAu

SAu

SAu

SAu

SAu

O

O

O

SAu

OH

O

O

O

SAu

OH

O

O

O

SAu

OH

O

O

O

SAu

OH

O

O

O

SAu

OH

O

O

O

SAu

OH

28 nm Au on Glass

B

Fig. S1. ( A) Schematic of self-assembled monolayers on gold. 1 protein and cell resistant tri(ethylene glycol) terminated monolayer and 2 printed hydro-phobic monolayer for fibronectin adsorption ( B) MSCs on fibronectin coated unpatterned surface ( Top ) and after patterning (in front of PDMS stamp photo-

graph). (Scale bar, 100 m).

B Osteogenic Media C Adipogenic Media D Osteo/Adipo MediaA Growth Media

Fig. S2. Dualstainingafter 1 weekfor osteogenesis (alkaline phosphotase

purple ) and adipogenesis (lipid vacuoles

red ) ( A) MSCsin growthmedia( B) MSCsin osteogenic promoting media containing dexamethasone, ascorbate and -glycerophosphate, ( C ) MSCs in adipogenic media containing dexamethasone, h-insulin, indomethacin and 3-isobutyl-1-methylxanthine and ( D) 1 1 combination of osteogenic and adipogenic supplements to expose cells to competingsoluble factors. (Scale bar, 100 m).

c e l l f a t e ( % )

0

20

40

60

80


1 0 0 0

m 2

2 5 0 0

m 2

5 0 0 0

m 2

A

B

C

Fig. S3. ( A) MSCs exposed for 1 week to mixed media after fixation and staining for alkaline phosphatase ( purple ) and lipid droplets ( red ). (B) 5 ; 000 m 2 starshowing preferential osteoblast commitment and 1 ; 000 m 2 star showing preferential adipocyte commitment. ( C ) % of cells across the same surface(1 ; 000 m 2 , 2 ; 500 m 2 and 5 ; 000 m 2 ) showing lineage specific markers demonstrating how spreading area influences lineage commitment. (Scale bar,50 m).

http://www.pnas.org/cgi/doi/10.1073/pnas.0903269107http://www.pnas.org/cgi/doi/10.1073/pnas.0903269107


9/15

E F G

H I J

DCBA

Fig. S4. Immunofluorescence of MSCs stained for actin (phalloidin, green ) and tubulin ( red ) for unpatterned surface ( A), 1 ; 000 m 2 (B), 2 ; 500 m 2 (C ), and5 ; 000 m 2 (D). Immunofluorescence reveals significant differences in the cytoskeletal architecture of MSCs that take on a more rounded morphology(1 ; 000 m 2 , (E )(G)) versus cells that are allowed to spread ( 5 ; 000 m 2 (H )( J )). The latter display increased organization of actin stress fibers and microtubulesacross the cell. (Scale bar, 20 m).

Fig. S5. Raw color images and 8-bit binary color-specific images after deconvolution for cells in flower shapes stained for both alkaline phosphatase ( purple )and lipid vacuoles ( red ).



10/15

C

B

A

Fig. S6. Immunofluorescence of MSCs in rectangles with increasing aspect ratio ( A) actin (phalloidin, green ) (B) focal adhesions (vinculin, red ) and ( C ) mergedimages with nuclei in blue (DAPI). Increasing the aspect ratio results in larger focal adhesions at the corners and short edge of the cells with an increase in stressfibers along the long edge. (Scale bar, 30 m).

n = 80 n = 86

C

A

D

B

Fig. S7. Immunofluorescent heatmaps of MSCs in flower and star shapes stained for actin ( A, B) and for vinculin ( C , D). (Scale bar, 20 m).



11/15

% D

i f f e r e n

t i a t i o n

0

20

40

60

80

100

% D

i f f e r e n

t i a t i o n

0

20

40

60

80

100

% D

i f f e r e n

t i a t i o n

0

20

40

60

80

100Y-27632

% D

i f f e r e n

t i a t i o n

0

20

40

60

80

100Blocking antibodies 5, 1

% D

i f f e r e n

t i a t i o n

0

20

40

60

80

100

A B

C D E

Nocodazole Cytochalasin D

Blebbistatin

Fig. S8. ( A)(D) Effect on shape promoted differentiation for rectangles with increasing aspect ratio in the presence of cytoskeletal disruptors and ( E ) integrinblocking antibodies.

0

1

2

3

4

5

6

7

8

I n u c l e u s

/ I c y

t o p l a s m

*p-value < 5E-9**p-value < 1E-7

BGLAP

LPL

GAPDH

B

C

A BM BM OAOA

PPAR

actinnuclei

**

**

*

*

BM BM OA OA

5

10

20

F l u o r e s c e n

t I n t e n s i

t y ( x 1 0 0 0 )

NP

Runx2 Nucleus Runx2 Cytoplasm PPAR Nucleus PPAR Cytoplasm

15

Runx2

D

Fig. S9. MSCs were cultured on patterned substrates presenting either flower or star shapes, treated with the mixed cues for differentiation to both osteo-blasts and adipocytes and then characterized to assess cell fate. ( A) Representative immunofluorescent images of MSCs on flower and star shapes stained foractin, PPAR , and Runx2 after one week exposure to basal media (BM) or mixed differentiation cues (OA). ( B) Quantitation of average PPAR and Runx2fluorescence for populations of adherent cells on nonpatterned surfaces (NP) and substrates presenting the flower and star shapes. Cells were fixed andstained with the marker-specific antibodies and imaged to quantitate both the nuclear and cytoplasmic fluorescence of the marker in each cell. ( C ) Ratiometricanalysis of nuclear to cytoplasmic fluorescence. The raw fluorescence data was first normalized by subtracting the average PPAR and Runx2 basal level ex-pression from control MSC cultures in the same shape followed by taking the ratio of nuclear to cytoplasmic fluorescence (I nucleus Icytoplasm ). Error bars arestandard deviations of 50 cells. The difference in means is statistically significant as determined by 2-population t-test (PPAR p -value < 5 E-9 , Runx2 p -value < 1 E-7 ). (D) Gel images of RT-PCR experiments that demonstrated higher expression of adipocyte specific genes (lipoprotein lipase, LPL) for cellsin flower shapes and osteoblast specific genes (osteocalcin, BGLAP) for cells in star shapes and unbiased expression of a housekeeping control (GAPDH).



12/15

A

B

Fig. S10. Designing shapes between contractile promoting extremes. ( A) circle showing low contractility by F-actin ( green ) and vinculin ( red ) with nuclei inblue compared to a holly leaf shape yielding large stress fibers along the concave long edge of the cell. ( B) Myosin IIa immunofluorescent staining.



13/15

Table S1. Fold change values for osteogenic and adipogenic transcripts relative to unpatterned MSCs (NP) in culture for flower and starshapes in basal media (BM) and mixed osteogenic/adipogenic media (OA).

Class Probe descriptionflower

BMstarBM

flowerOA

starOA

NPOA

Adipogenesis PPARGC1B:peroxisome proliferator-activatedreceptor gamma, coactivator 1 beta

1.72 1.1 4.45 4.19 1.6

PPARG: peroxisome proliferator-activated receptor gamma 1.22 1.4 1.88 1.62 1.03PPARGC1A: peroxisome proliferator-activated receptor gamma,

coactivator 1 alpha1.85 11.49 1.01 7.52 1.42

LPL: lipoprotein lipase 2.32 1.66 337.39 84.66 83.01ACSL1: acyl-CoA synthetase long-chain family member 1 4.68 2.55 12.54 5.35 4.24ACSL3: acyl-CoA synthetase long-chain family member 3 1.8 1.42 1.2 2.13 1.01ACSL4: acyl-CoA synthetase long-chain family member 4 2.05 1.28 1.76 2.88 1.8ACSL5: acyl-CoA synthetase long-chain family member 5 2.12 1.21 13.96 7.32 3.77ACSL6: acyl-CoA synthetase long-chain family member 6 1.48 1.02 1.08 1.07 1.15

AACSL: acetoacetyl-CoA synthetase-like 2.84 2.65 2.22 2.19 1.06FABP4: fatty acid binding protein4, adipocyte 1.03 1.14 800.77 610.76 527.87

CEBPA: CCAAT/enhancer-binding protein (C/EBP), alpha 4.11 9.04 322.21 454.1 1 171.34CEBPB: CCAAT/enhancer-binding protein (C/EBP), beta 1.44 1.08 1.12 3.23 1.26CEBPD: CCAAT/enhancer-binding protein (C/EBP), delta 1.33 4.18 1.42 1.03 2.85

CEBPG: CCAAT/enhancer-binding protein (C/EBP), gamma 7.28 4.29 2.71 28.62 2.72KLF5: Kruppel-like factor 5 (intestinal) 2.46 3.14 2.95 5.58 1.39

KLF6: Kruppel-like factor 6 2.51 2.28 1.49 11.7 1.27KLF15: Kruppel-like factor 15 2 1.76 51.06 21.92 38.12

EPAS1: endothelial PAS domain protein 1 1.63 1.32 1.12 2.24 1.03STAT5A: signal transducer and activator of transcription 5A 1.73 1 3.54 1.08 2.78

NCOR1: nuclear receptor corepressor 1 2.09 1.39 1.17 5.18 1.42NCOR2: nuclear receptor corepressor 2 1.78 2.39 1.15 1.68 1.91

DDIT3: DNA-damage-inducible transcript 3 2.42 1.49 2.3 6.25 2.18EGR2: early growth response 2 (Krox-20 homolog, Drosophila) 1.15 1.69 23.16 21.26 34.46

EBF1: early B-cell factor 1 1.87 2.21 1.78 1.62 4.71NR1H3: nuclear receptor subfamily 1, group H, member 3 1.93 1.12 8.36 4.84 4.16

SREBF1: sterol regulatory element binding transcription factor 1 2.11 1.07 1.97 1.48 1.51ARNTL2: aryl hydrocarbon receptor nuclear translocator-like 2 4.21 1.94 4.4 10.15 3.06

Osteogenesis SP7: Sp7 transcription factor 1.07 1.02 1.13 1.97 3.86ALPL: alkaline phosphatase, l iver/bone/kidney 1.63 1.11 3.44 5.81 5.38

RUNX2: runt-related transcription factor 2 5.56 6.74 2.43 10.67 5.82SPP1: secreted phosphoprotein 1 (osteopontin, bone sialoprotein I) 1.61 3.12 2.72 1.24 2.02

IBSP: integrin-binding sialoprotein (bone sialoprotein, bonesialoprotein II) 2.35 1.7 2.16 1.7 1.36

BGLAP: bone gamma-carboxyglutamate (gla) protein (osteocalcin) 2.34 1.55 1.05 1.3 1.65MGP: matrix Gla protein 2.55 2.38 1.42 1.99 3.48

TFIP11: tuftelin interacting protein 11 1.27 1.49 1.14 1.37 1.88MSX2: msh homeobox 2 1.77 2.88 2.09 1.74 1.24

VDR: vitamin D (1,25-dihydroxyvitamin D3) receptor 1.28 1.19 1.16 1.44 1.35COL1A2: collagen, type I, alpha 2 1.12 1.01 1.57 1.15 1.04COL1A1: collagen, type I, alpha 1 1.06 1.17 1.02 1.1 1.07

COL3A1: collagen, type III, alpha 1 (Ehlers-Danlos syndrome type IV,autosomal dominant)

1.25 1.13 1.8 1.43 1.68

TGFB1: transforming growth factor, beta 1 2.61 2.41 2.13 3.43 2.04SOX9: SRY (sex determining region Y)-box 9 (campomelic dysplasia,

autosomal sex-reversal)2.66 1.91 1.17 1.96 1.43

BMPR1A: bone morphogenetic protein receptor, type IA 1.75 1.01 1.13 2.71 1.15BMP1: bone morphogenetic protein 1 1.32 1.91 1.1 2.57 1.94

BMP2: bone morphogenetic protein 2 25.1 25.12 2.97 3.67 1.33BMP3: bone morphogenetic protein 3 (osteogenic) 1.88 3.49 5.37 2.93 2.16BMP4: bone morphogenetic protein 4 7.83 2.45 1.12 4.86 2.05BMP5: bone morphogenetic protein 5 2.42 1.47 1.02 3.42 2.59BMP6: bone morphogenetic protein 6 1.02 1.12 8.32 4.12 18.19

BMP7: bone morphogenetic protein 7 (osteogenic protein 1) 1.29 2.17 1.99 7.72 2.79BMP8B: bone morphogenetic protein 8b (osteogenic protein 2) 1.34 1.33 1.08 1.37 1.26

BMP8A: bone morphogenetic protein 8a 2.64 1.63 1.42 5.14 2.37GDF2: growth differentiation factor 2 1.29 4.35 1.99 1.37 2.88

BMP10: bone morphogenetic protein 10 2.13 2.32 1.26 1.04 1.67GDF11: growth differentiation factor 11 1.75 1.72 1.43 1.05 1BMP15: bone morphogenetic protein 15 1.68 1.16 1.1 1.23 1.02

SMAD1: SMAD family member 1 18.84 1.74 1.11 2.64 1.03SMAD2: SMAD family member 2 1.01 1.12 1.03 1.15 1.21SMAD3: SMAD family member 3 1.03 1.68 1.14 1.5 1.38



14/15


BMstarBM

flowerOA

starOA

NPOA

SMAD4: SMAD family member 4 2.2 1.16 1.51 1.03 1.12SMAD5: SMAD family member 5 1.5 1.08 1.24 1.82 2.99SMAD6: SMAD family member 6 1.28 1.46 2.92 4.58 1.9SMAD7: SMAD family member 7 1.17 1.13 1.06 1.39 1.41SMAD9: SMAD family member 9 1.92 1.14 1.08 2.25 1.21

SMURF1: SMAD specific E3 ubiquitin protein ligase 1 3.09 1.21 1.51 4.7 2.22SMURF2: SMAD specif ic E3 ubiquitin protein ligase 2 2.35 1.52 4.42 1.23 1.72



15/15

Table S2. Fold change values for Wnt signaling and MAP kinase transcripts relative to unpatterned MSCs (NP) in culture for flower and starshapes in basal media (BM) and mixed osteogenic/adipogenic media (OA).


BMStarBM

flowerOA

starOA

NPOA

Wnts/Fzds/Dvl WNT1: wingless-type MMTV integration site family, member 1 1.97 1.16 1.32 2.35 1.19WNT2: wingless-type MMTV integration site family member 2 3.71 1.11 1.05 3.57 1.2WNT3: wingless-type MMTV integration site family, member 3 2.3 1.41 1.29 1.66 1.41WNT4: wingless-type MMTV integration site family, member 4 1 1 1.11 16.25 1

WNT5A: wingless-type MMTV integration site family, member 5A 1.41 1.24 1.3 1.33 1.49WNT5B: Wingless-type MMTV integration site family, member 5B 4.38 1.87 8.49 4.39 1.15

WNT6: wingless-type MMTV integration site family, member 6 3.47 1.95 1.44 4.22 1.29WNT7A: wingless-type MMTV integration site family, member 7A 1.64 1.2 1.26 2.08 1.35WNT7B: wingless-type MMTV integration site family, member 7B 1.7 4.58 1.63 5.14 1.74WNT8A: wingless-type MMTV integration site family, member 8A 1.21 3.92 1.26 5.38 1.44WNT8B: wingless-type MMTV integration site family, member 8B 1 1 1.45 1.11 1WNT9A: wingless-type MMTV integration site family, member 9A 4.08 1.21 1.23 1.21 1.01WNT9B: wingless-type MMTV integration site family, member 9B 1.51 2.8 1.58 2.38 3.87

WNT10A: wingless-type MMTV integrat ion site family, member 10A 2.73 3.49 1.29 4.58 1.35WNT10B: wingless-type MMTV integration site family, member 10B 4.5 4.5 4.5 4.5 4.5

WNT11: wingless-type MMTV integration site family, member 11 1 13.59 1.11 13.09 2.56WNT16: wingless-type MMTV integration site family, member 16 1.07 1.04 1.24 7.53 1.38

FZD1: frizzled homolog 1(Drosophila) 1.68 1.11 1.19 2.35 1.3FZD2: frizzled homolog 2 (Drosophila) 1.98 2.52 1.56 2.95 1.46FZD3: frizzled homolog 3 (Drosophila) 1.3 1.32 1.18 1.33 1.03

FZD4: frizzled homolog 4 (Drosophila) 1.83 1.25 5.23 2.11 1.91FZD5: frizzled homolog 5 (Drosophila) 4.24 1.01 2.1 1.6 2.36FZD6: frizzled homolog 6 (Drosophila) 7.37 3.99 2.7 23.44 2.04FZD7: frizzled homolog 7 (Drosophila) 1.14 1.72 1.82 2.31 1.41FZD8: frizzled homolog 8 (Drosophila) 2.45 1.39 2.45 1.14 2.45FZD9: frizzled homolog 9 (Drosophila) 1.32 1.13 1.13 3.12 1.59

FZD10: frizzled homolog 10 (Drosophila) 1.43 1.14 1.74 2.77 1.22DVL1: dishevelled, dsh homolog 1 (Drosophila) 1.12 1.04 1.04 1.28 1.05DVL2: dishevelled, dsh homolog 2 (Drosophila) 1 1.04 1.29 1.27 1DVL3: dishevelled, dsh homolog 3 (Drosophila) 23.98 30.07 1.54 54.56 16.67

NoncanonicalWnt

DAAM1: dishevelled associated activator of morphogenesis 1 1.32 1.46 1.58 1.99 1.37

ROR2: receptor tyrosine kinase-like orphan receptor 2 4.46 2.1 2.65 2.81 1.85ROCK2: Rho-associated, coiled-coil containing protein kinase 2 1.06 1.42 1.47 1.26 1.32

CDC42: cell division cycle 42 (GTP binding protein, 25 kDa) 1.13 1.38 1.33 1.14 1.14RAC1: ras-related C3 botulinum toxin substrate 1 (rho family, small GTP

binding protein Rac1)1.7 1.43 1.17 2.76 1.24

RHOA: ras homolog gene family, member A 1.52 1.46 1.25 2.02 1.17Canonical

WntCTNNB1: catenin (cadherin-associated protein), beta 1, 88 kDa 1.17 1.19 1.35 2.28 2.87

LRP5: low density lipoprotein receptor-related protein 5 1.91 2.74 1.3 2.33 1.46LRP6: lowdensity lipoprotein receptor-related protein 6 1.01 3.03 1.54 1 1.49

AXIN2: axin 2 (conductin, axil) 1.51 2.06 1.15 1.77 1.64GSK3B: glycogen synthase kinase 3 beta 1.72 1.36 1.34 2.08 1.37

LEF1: lymphoid enhancer-binding factor 1 1.74 1.04 1.29 2.43 1.13CSNK1E: casein kinase1, epsilon 1.77 1.24 1.9 1.15 1.05

APC: adenomatous polyposis coli 6.65 3.56 1.15 19.01 1.87Wnt

InhibitoryWIF1: WNT inhibitory factor 1 2.99 1.7 1.91 2.97 1.04

SOST: sclerosteosis 1.32 1.12 2.87 1.72 2.12DKK1: dickkopf homolog 1 (Xenopus laevis) 1.85 1.2 1.1 2.05 1.06

SFRP1: Secreted frizzled-related protein 1 2.04 1.9 2.2 2.02 2.14SFRP2: secreted frizzled-related protein 2 4.59 1 10.09 2.38 6.4

FRZB: Frizzled-related protein 2.65 1.96 3.55 1.8 4.55SFRP4: secreted frizzled-related protein 4 1.11 1.51 8.4 20.78 2.2SFRP5: secreted frizzled-related protein 5 4.79 1.21 11.21 6.53 9.18

MAPK MAPK1: mitogen-activated protein kinase 1 1.05 1.06 1.09 1.51 1.28MAPK3: mitogen-activated protein kinase 3 1.22 1.37 1.05 1.4 1.61MAPK8: mitogen-activated protein kinase 8 1.26 4.48 2.09 2.12 1.31MAPK9: mitogen-activated protein kinase 9 1.23 1.19 1.32 1.43 1.32

MAPK10: mitogen-activated protein kinase 10 1.62 2.07 2.63 3.87 1.93MAPK11: mitogen-activated protein kinase 11 1.4 1.27 2.95 1.37 5.68MAPK12: mitogen-activated protein kinase 12 2.76 2.65 1.61 6.43 2.31MAPK13: mitogen-activated protein kinase 13 1.38 1.25 2.45 2.09 2.6MAPK14: mitogen-activated protein kinase 14 2.06 1.11 1.18 3.23 1.02

Kili t l g/ gi/d i/10 1073/ 0903269107 9 f 9

Date post:	03-Jun-2018
Category:	Documents
Upload:	heisenbergjones
View:	220 times
Download:	0 times

Killian Paper Cell Adhesion surface shape in PNAS

Documents