Micropatterned mammalian cells exhibitphenotype-specific left-right asymmetryLeo Q. Wana, Kacey Ronaldsona, Miri Parka, Grace Taylora, Yue Zhanga,Jeffrey M. Gimbleb, and Gordana Vunjak-Novakovica,1

aDepartment of Biomedical Engineering, Columbia University, New York, NY 10027; and bPennington Biomedical Research Center, Louisiana StateUniversity System, Baton Rouge, LA 70808

Edited by Michael Levin, Tufts Center for Regenerative and Developmental Biology, Tufts University, Medford, MA, and accepted by the Editorial Board June 2,2011 (received for review March 10, 2011)

Left-right (LR) asymmetry (handedness, chirality) is a well-con-served biological property of critical importance to normal devel-opment. Changes in orientation of the LR axis due to genetic orenvironmental factors can lead to malformations and disease.While the LR asymmetry of organs and whole organisms has beenextensively studied, little is known about the LR asymmetry atcellular and multicellular levels. Here we show that the cultivationof cell populations on micropatterns with defined boundariesreveals intrinsic cell chirality that can be readily determined byimage analysis of cell alignment and directional motion. By pat-terning 11 different types of cells on ring-shaped micropatternsof various sizes, we found that each cell type exhibited definiteLR asymmetry (p value down to 10−185) that was different betweennormal and cancer cells of the same type, and not dependent onsurface chemistry, protein coating, or the orientation of the grav-itational field. Interestingly, drugs interfering with actin but notmicrotubule function reversed the LR asymmetry in some cell types.Our results show that micropatterned cell populations exhibit phe-notype-specific LR asymmetry that is dependent on the function-ality of the actin cytoskeleton. We propose that micropatterningcould potentially be used as an effective in vitro tool to studythe initiation of LR asymmetry in cell populations, to diagnose dis-ease, and to study factors involved with birth defects in laterality.

cell patterning ∣ tissue morphogenesis ∣ cell polarity ∣ cell migration

Left-right (LR) asymmetry is seen in the development ofnumerous living organisms, including climbing plants (1),

helices of snail shells (2), and the human body (3, 4). Geneticdiseases and prenatal exposure to teratogens can cause birth de-fects in laterality (5, 6). Using embryos of Drosophila, zebrafish,frog Xenopus, mouse, and chicken, several models have beenproposed for the establishment of LR asymmetry, such as direc-tional nodal flow driven by primary cilia (7–9), voltage gradientsresulting from asymmetric expression of ion channels (10, 11),and asymmetric vesicular transport via myosin 1D along actincable networks (12–14). In vertebrates, while some of themechanisms appear to be conserved, important differences areobserved between species such as differential regulation of asym-metric ion flux in frog and chicken embryos (10), and a distinctrole of FGF-8 in mouse and rabbit embryos as a sidedness deter-minant (15). Two main models describe the establishment ofLR assymetry. The primary cilium model holds that the LR asym-metry is a phenomenon initiated by the formation of a ciliatedembryonic structure known as a node, which generates direc-tional fluid flow being transduced into asymmetric gene expres-sion and giving rise to LR asymmetry (7–9, 16). The second modelholds that LR asymmetry is a fundamental property of the cell,based on the observations that LR asymmetry can be establishedwithout primary cilia, node, or fluid flow (10, 17). Recapitulatingchiral morphogenesis using in vitro models with human cellswould thus help further interrogate the existing competingmodels, toward better understanding of tissue morphogenesisin embryo.

The initiation of LR asymmetry in development is often firstobserved in populations of cells of the same type, such as snailembryonic cells at four-cell and eight-cell stages and mouse cellsat embryonic nodes. The establishment of LR asymmetry or chir-ality within such cell clusters may rely on some intracellular struc-ture, such as the hypothetical F-molecule or actin/microtubulecytoskeleton (18, 19) that can distinguish left from right by orient-ing the third axis with respect to predetermined dorsal–ventraland anterior–posterior axes. In addition, during development, thespecification and self-organization of migrating cells are mediatedby physical boundaries imposed by the extracellular matrix andthe surrounding cells and tissues. We, therefore, hypothesizedthat the populations of cells of the same type, if cultured withinpatterns with well-defined boundaries, would express directionalalignment and motion associated with the establishment of LRasymmetry.

We find that the cells form an invariant chiral alignment,depending on the cell phenotype and actin function. We infer thatall cells are intrinsically chiral and that the function of actin mayplay a significant role in switching the intrinsic chirality of cellsand their LR asymmetry in developmental biology. Due to itssimplicity, our 2D microscale system provides an effective meth-od to study the LR symmetry of tissue development in vitro andmay provide a useful tool for identifying biological and environ-mental factors involved in bilateral birth defects.

Results and DiscussionCells Exhibit Chiral Morphogenesis on Micropatterned Geometries.We cultured the C2C12 murine myoblasts on ring-shaped micro-patterns (Fig. S1 and Table S1 in SI Appendix) (20, 21). Phasecontrast images, in which cell contour appears brighter thanthe inner cell region, were used to measure cell alignment (22)(Fig. 1A), as indicated by green lines in Fig. 1B. Each green linewas assigned a biased angle between −90° and 90°, based on itsdeviation from the circumferential direction (blue dash line); apositive value represented a counter clockwise (CCW) alignment,while a negative value represented a clockwise (CW) alignment(Fig. 1C). An angular histogram (Fig. 1D) and the radial distribu-tion (Fig. 1E) of the measured angles revealed preference forpositive angles, corresponding to the CCW bias. In >30 indepen-dent series of experiments using >1;000 individual rings (Fig. 1Fand Table S2 in SI Appendix), the C2C12 cells showed CCWalignment with a biased angle of 8.47°� 0.20° (mean� SEM),with very strong statistical significance (p ¼ 9.3 × 10−186).

Author contributions: L.Q.W., J.M.G., and G.V.-N. designed research; L.Q.W., K.R., M.P.,G.T., and Y.Z. performed research; L.Q.W., K.R., J.M.G., and G.V.-N. analyzed data; andL.Q.W., J.M.G., and G.V.-N. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.M.L. is a guest editor invited by the Editorial Board.

See Commentary on page 12191.1To whom correspondence should be addressed: E-mail: gv2131@columbia.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1103834108/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1103834108 PNAS ∣ July 26, 2011 ∣ vol. 108 ∣ no. 30 ∣ 12295–12300

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

SEECO

MMEN

TARY

Dow

nloa

ded

by g

uest

on

Aug

ust 1

1, 2

020

Other geometries such as circles, squares, and linear stripswere also tested. Biased cell alignment was observed on linearstrips and rings but not on circles and squares (Fig. S2 A–F inSI Appendix), suggesting the importance of appositional bound-aries for the expression of chirality. Because equivalent bias incell alignment was observed on linear strips and rings of differentsizes (Fig. S2G in SI Appendix), rings with an inner diameter of250 μm and a distance of 200 μm between the inner and outerboundary were used in subsequent studies.

Cell Chirality on Patterned Surfaces Depends on Cell Phenotype. Wethen tested a panel of mouse, rat, and human cells derived fromskeletal muscle, bone, adipose tissue, skin, heart, and blood ves-sels (Fig. 2 and Table S1 in SI Appendix). Cell phenotypes in-cluded myoblasts, osteoblasts, endothelial cells, fibroblasts, andmesenchymal stromal/stem cells. All cells exhibited distinct chir-ality after they reached a confluency of approximately 75% andwere cultured overnight. Interestingly, mouse (C2C12) and hu-man skeletal muscle cells (hSkMCs) showed a CCW alignment,while all other cell types exhibited a CW alignment (Fig. 2 andFig. S3 and Tables S2 and S3 in SI Appendix). Cancer skin cellsalso exhibited a CCW alignment, different from the CW align-ment of skin fibroblasts from the same individual. These data sug-gest that adherent mammalian cells exhibit an invariant chiralitydetermined by the cell phenotype and disease condition.

LR Asymmetry on Micropatterns Is Established by Mechanisms Invol-ving Cellular Directional Migration on Boundaries. For cells on a pat-terned geometry to display their chirality and to distinguishbetween left and right (y axis), the polarity of the z axis (up-down)and x axis (front-back) must be established (Fig. 3A). Notably, thez axis was established independent of gravity direction, as thesame biased alignment (relative to the cellular apical-basal axis)was observed experimentally in regular and vertically inverted cellcultures (Table S4 in SI Appendix). The geometric boundaries de-termined organelle positioning (i.e., x axis), with centrosomes(bright green, Fig. 3B and Fig. S4 in SI Appendix) and Golgi ap-paratus (red, Fig. 3C) being positioned closer to the boundariesthan the cell nucleus (blue) (23), independent of the gravitationaldirection. In addition, cell chirality was maintained on patternswith different surface chemistry (24), after the disruption of cad-herin function by reducing calcium levels (25), and the inhibitionof cell proliferation (Fig. S1 and Tables S5–S7 in SI Appendix).

For further insight into the biased cell alignment, we analyzedC2C12 cell motion. When the cells were seeded sparsely, no clearbias in motion or alignment was observed over more than 20 hrsof culture (Movie S1). For a higher cell density seeding, cell align-ment did not show a clear bias before confluency (after 15 hrs;Fig. 3 D and E and Movies S2 and S3). These data suggest thatsignificant inhibition of the random walk of the cells through phy-sical cell–cell contact was necessary for the cells to express subtledirectional migration and biased alignment. The cells werelabeled and tracked along the inner and outer ring pattern(Movie S4), and the cell migration velocity was estimated bydigital image correlation (26). The speed of migration was higherat the inner and outer ring boundaries than within the interiorregion (25 μm∕hr vs. 10 μm∕hr; p < 0.05). The average velocity(Fig. 3F) and the velocity changes in radial and circumferentialdirections (Fig. 3G) demonstrated that the cells migrated in theCW direction (at 8 μm∕hr) at the inner ring boundary (p < 0.05),and in the CCW direction (at 4 μm∕hr) at the outer ring bound-ary (p < 0.05).

At the level of an individual cell, this seemingly opposite cir-cular motion of cells on the inner and outer boundary is in factconsistent with biased migration. Based on cell polarization atboundaries, the x axis can be defined as the direction from thenucleus to centrosome (27), as shown in Fig. 3A. In other words,the cells “face” outward on the outer ring, and inward to the cen-ter on the inner ring. Thus, for the C2C12 cells, the biased mi-gration can thus be considered as “leftward bias” along both theinner and outer ring boundary. Also, observed biased alignmentof the cells on micropatterns is related to the directional migra-tion at the boundaries, as seen for a C2C12 cell migrating towarda boundary and adopting the leftward biased migration. Becausecell polarization and biased migration occur at the boundaries ofmicropatterns, cell proximity to a boundary is necessary for theexpression of chirality. This finding was further supported by thebiased cell alignment being most clearly seen in the regions closeto the boundaries, especially for less elongated cells such as car-diac fibroblasts (Fig. 2A).

The mean biased angle of human umbilical cord endothelialcells (hUVEC) on rings was similar in magnitude to that ofC2C12 cells but was negative (−8.47°� 0.33°, n ¼ 388), indicat-ing a CWalignment (Table S3 in SI Appendix). By the time the cellchirality was established (15–20 h after seeding; Movies S5–S7),cells along the boundaries had a significantly higher migrationspeed than those in the interior regions (35 μm∕hr vs. 20 μm∕hr;

Fig. 1. Mouse myoblasts (C2C12) show distinct chirality on micropatterned surfaces. Scale bars: 100 μm. (A) Asymmetric cell alignment on ring patterns (phasecontrast image). (B) Cell alignment directions (green lines). (C) The biased angle of cell alignment (green lines) was defined as either CW or CCW, based onthe deviation from the circumferential direction (blue dash line). (D) The circular histogram of biased angles shows CCW chirality. (E) Circumferential averagesof the subregional biased angles at different radial positions on the ring (mean� SEM). (F) The histogram of the mean biased angles of C2C12 cells(33 experiments, >1;000 ring patterns).

12296 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1103834108 Wan et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 1

1, 2

020

p < 0.05). In contrast to C2C12 cells, hUVEC migration was inthe CCW direction (15 μm∕hr) at the inner ring boundary(p < 0.05) and in the CW direction (20 μm∕hr) at the outer ringboundary (p < 0.05) (Fig. 3H and I). Based on the x axis directedfrom the nucleus to the centrosome/Golgi apparatus, the migra-tion of hUVECs exhibited a “rightward” bias.

Chirality of Muscle Cells Requires Functional Actin but not Tubulin. Toinvestigate the roles of actin and tubulin, cyoskeletal proteins pu-tatively linked to cell chirality (1, 2, 12, 14, 27), we used drugs toalter the dynamics of their polymerization and depolymerization(Fig. 4 A and B). For C2C12 and hUVEC cells, cell alignmenton micropatterned rings followed a dose-dependent response(Table S8 in SI Appendix). Low concentrations of the actin tread-milling inhibitors (Latrunculin A, cytochalsin D, Jasplakinolide),which did not completely inhibit actin polymerization/depolymer-ization, reversed the chirality of C2C12 cells from CCW into CW.In contrast, inhibitors of tubulin dynamics (Nocodazole, Taxol) atconcentrations below those resulting in cell apoptosis or inhibi-

tion of cell migration did not change cell chirality. Similar resultswere obtained for human skeletal muscle cells (Table S8 in SIAppendix). The CCW bias of the cells depended on the functionof actin but not nonmuscle myosin II (Fig. S5 in SI Appendix). Incontrast to C2C12 cells, the drugs tested could not reverse thechirality of hUVEC cells or rat cardiac fibroblasts (Table S8 inSI Appendix). Collectively, these data suggest that functional actinis required for the muscle cells exhibiting CCW bias but not thecells exhibiting CW bias.

Interestingly, C2C12 cells treated with Latrunculin A polarizedin the same fashion as the untreated cells (Fig. 4C), as evidencedby organelle positioning relative to boundaries, suggesting thatthe drug did not alter cell polarization. Analysis of cell migration(Movies S8 and S9) showed reversal at the boundaries (Fig. 4 Dand E), with the cells migrating CW along the outer ring andCCW along the inner ring at 15 μm∕hr (p < 0.05). Similar ana-lyses showed that C2C12 cells treated with 200 nM Nocodazolepolarized and migrated in the same fashion as the untreatedcontrols (Fig. S6 in SI Appendix). Thus, inhibition of actin

Fig. 2. Cell chirality on patterned substrates depends on cell phenotype and disease condition. (A) Phase contrast images of various cell types. (B) Cell sourceand their chirality on micropatterned surfaces. Here, “−” indicates that cells were isolated in the lab; CW: clockwise alignment; CCW: counterclockwise align-ment; N/S: not significantly biased to CW or CCW. The C2C12 cells and human skeletal muscle cells exhibit a counterclockwise alignment while other cells thatwere studied show a clockwise alignment. The human skin cancer fibroblast cell line show a counterclockwise alignment, opposite from healthy human skinfibroblasts. Scale bars: 100 μm.

Wan et al. PNAS ∣ July 26, 2011 ∣ vol. 108 ∣ no. 30 ∣ 12297

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

SEECO

MMEN

TARY

Dow

nloa

ded

by g

uest

on

Aug

ust 1

1, 2

020

function reversed the CCW chirality and the biased migrationof C2C12 cells.

The establishment of cell asymmetric alignment on micropat-terns required definite polarization and directional migration ofthe cells on appositional boundaries of geometries such as ringsand long strips. Compared to the cells cultured on large surfaces,and microscale circles or squares, rings and strips provided nar-row space for the cells to align and to sense boundaries. In addi-tion, appositional boundaries offered a primary direction for thecells to elongate and migrate, such that any bias in cellular LRdecision would be amplified, leading to directional motion. Sucha mechanism required at least two-cell width between opposingboundaries (>20 μm). Otherwise, a single cell would receive con-fusing signals by simultaneously touching both boundaries, andnot express definite polarization or directional migration.

During native development, boundaries are established by cellpopulations compartmentalized into distinct functional units(e.g., in Drosophila wing), in response to gradients of morpho-gens (28), and differential adhesion and cortical tension (29).Interestingly, even at the four-cell and eight-cell stages, the snail

embryos exhibit a biased alignment, which has a similar responseto the drugs (i.e., Latrunculin A and Nocodazole) as mammalianmuscle cells (2).

We propose that cell chiral alignment is associated with cellmigration based on the following observations: (i) The cellsare randomly oriented right after seeding, and the establishmentof cell alignment required cell motion; (ii) biased alignment wasinitiated at the same time as the directional motion of cells onboundaries; and (iii) the chirality of cell alignment was consis-tently paired with the leftward or rightward bias of cell motionat the boundaries. Following Latrunculin A treatment, the chir-ality of C2C12 cells was reversed, as was the biased motion ofcells on boundaries.

From the chirality data of muscle cells treated with LatrunculinA, we infer that it is possible that two competing mechanisms co-exist within cells to determine their LR decisions or chirality. Onemechanism would require actin function and lead to the intrinsicleftward bias at the boundaries and subsequent CCW alignmentin chiral morphogenesis. The second mechanism would induceintrinsic rightward bias at the boundaries and a CW alignment

Fig. 3. Cellular LR asymmetry on micropatterned rings is established by mechanisms involving boundary effects. (A) Cells on a ring “sense” the z axis throughattachment to the substrate and the x axis through the ring boundaries. The cell alignment bias of the y axis (dash red lines) creates the observed cellular chiralbehavior or LR asymmetry. (B) Centrosomes (bright green) are positioned closer to each boundary than nuclei (blue) in C2C12 cells (actin: red, tubulin: green).Scale bars: 50 μm. (C) Golgi apparatus (red) is positioned closer to ring boundaries than nuclei (blue) in hUVEC cells. Scale bars: 50 μm. (D) Phase contrast imagesof the cells (top) at 5, 10, 20, and 30 hr after cell seeding and the corresponding histograms (bottom) of biased angles from the subregions for each image. Scalebars: 100 μm. (E) The time history of the mean biased angle of C2C12 cells on a ring, with the insert for cell number increasing exponentially with time. (F)Average velocity and direction of C2C12 cells are indicated by arrow length and direction, respectively. (G) Average velocity of C2C12 cell migration in thecircumferential (Vθ) and the radial direction (Vr ) as a function of radial position. (H) Average velocity and (I) the circumferential migration velocity of hUVECcells at the inner and outer ring boundary as a function of time.

12298 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1103834108 Wan et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 1

1, 2

020

in chiral morphogenesis. Our drug treatment studies do not ex-clude the role of microtubule/centrosome in determining cellchirality, especially for the rightward bias, which is worthy offurther investigation. The differences in chirality may be dueto the higher expression of actin in muscle cells than other cells(30, 31), necessitating identification of cell-type-related determi-nants of chirality in tissue development. Interestingly, actin-de-pendent mechanisms were reported to account for chiralproperties of the Xenopus egg cortex (32), early developmentof asymmetry in Xenopus embryos (33), and cardiac loopingin vertebrate embryos (34, 35). In addition, alternations in actinpolymerization were reported to regulate phonotypical events inmalignant cells (36). Further studies are necessary to evaluate therole of actin expression levels in the establishment and reversal ofcell chirality, possibly mediated through the noncanonical Wntsignaling pathway, which plays a critical role in pattern determi-nation during embryonic development (37).

In summary, cells cultured on micropatterns with well-definedappositional boundaries exhibit chiral morphogenesis that can bereadily determined by analysis of cell alignment and directionalmotion. For cell populations to express chirality on micropat-terns, it was necessary to provide: (i) close appositional bound-aries for the cells to sense and polarize; (ii) certain confluence(75% or more) to inhibit cell random walk; and (iii) sufficienttime (≥15 hours) for the cells to migrate. In studies of 11 differ-ent cell types cultured on thousands of ring-shaped patterns, weobserved that the cell chirality was defined by the cell phenotype,and that the loss of actin but not microtubule function couldreverse the CCW cell chirality. The simple and highly accuratein vitro platform developed in these studies could potentially

be used to study the initiation of chiral morphogenesis and iden-tify genetic, biochemical, and environmental factors leading tomalformations.

MethodsMicrocontact Printing. Cell patterning was done by using polydimethylsilox-ane (PDMS) elastomeric stamps and self-assembly monolayers (SAMs) (20, 21,24). A master mold was first fabricated with SU-8 2050 photoresist (Micro-Chem Corp.) and chromium masks with desired geometric features. The mix-ture (10∶1) of PDMS prepolymer and curing agent (Dow Corning) was pouredinto the mold and cured at 70 °C for 4 h.

In most studies (and unless indicated otherwise), an adhesive SAM octa-decanethiol (Sigma) was transferred onto the gold-coated (150 Å in thick-ness) glass slide with the PDMS stamp (Fig. S1 in SI Appendix) (20). Theslide was then immersed in a nonadhesive ethylene glycol-terminatedSAM (HS-ðCH2Þ11-EG3, Prochimia) for 3 h. Finally, patterned surfaces werewashed with ethanol and coated with 10 μg∕mL fibronectin (Sigma)for 30 min.

Alternatively (Fig. S1 in SI Appendix), the PDMS stamp was coated with50 μg∕mL fibronectin for 30 min, aspirated, and dried in the air for 1 min(24). The stamp was then gently placed onto tissue culture-treated dishesfor 2 min and incubated in a nonadhesion SAM, 100 μg∕mL poly-L-lysine-polyethylene glycol (PLL-g-PEG; Susos AG), for 1 h. Finally the surface waswashed with phosphate buffered saline (PBS).

Cell Culture. Cells were maintained in tissue flasks with culture media speci-fied in Table S1 in SI Appendix. After reaching 70% confluency, cells weretrypsinized and seeded onto protein-coated patterned surfaces. Once thecells attached, extra cells were washed off with fresh medium. At this timedrugs were added into the culture medium if necessary. Phase contrastimages were taken after overnight incubation at 37 °C and 5% CO2 whencells reached confluency on the ring patterns.

Drug Treatment. To examine the role of actin in LR asymmetry, cells weretested with 20–500 nM Latrunculin A, 0.05–1.0 μg∕mL Cytochalasin D, and1–100 nM Jasplakinolide. Latrunculin A inhibits actin polymerization by form-ing a 1∶1 molar complex with G-actin, thereby inhibiting its ability to poly-merize into F-actin (38). Cytochalasin D inhibits actin polymerization bybinding to the growing ends of F-actin chains and thus preventing the attach-ment and addition of G-actin monomers (39). Jasplakinolide is known to bindto and stabilize actin filaments in vitro (40).

To examine the role of microtubules in LR asymmetry, 0.2–2 μM Nocoda-zole, and 0.3–30 nM Taxol were used. Nocodazole suppresses microtubuledynamics by destabilizing and disassembling microtubules (39). Taxol inhibitsthe microtubule depolymerization and stabilizes microtubules (41).

To examine the role of the actomyosin motor, cells were tested with 0.2–10 mM Y-27632, 1–20 mM ML-7, and 0.5–10 mM Blebbistatin. Y-27632 worksas a selective inhibitor to prevent the phosphorylation of the myosin regu-latory light chain (42). ML-7 acts as a selective inhibitor of the myosin lightchain kinase (43). Blebbistatin forms a low actin affinity complex throughbinding to myosin heads, causing the inhibition of nonmuscle myosin IIATPase activity (44).

Immunofluorescence Staining.After imaging, cells were fixed with 4% formal-dehyde in cytoskeletal buffer (10 mM MES, 138 mM KCl, 3 mM MgCl2, 2 mMEGTA, and 0.32 M sucrose) for 25 min. For actin/tubulin double staining, thecells were incubated with phalloidin-TRITC (1∶400; Invitrogen) and anti-Tubu-lin-FITC (1∶50; Sigma) for 1 h. For the Golgi apparatus positioning inside pat-terned rings, the cells were incubated in 1 μg∕ml antihuman golgin-97(Invitrogen) for 1 h. After secondary antibodies, cell nuclei were stained with200 ng∕mL 40, 6-diamidino-2-phenylindole (DAPI; Sigma) for 10 min. Finally,the cells were mounted with Fluoromount-G medium (SouthernBiotech).

Analysis of Cell Alignment. High-resolution phase contrast images of live pat-terned cells were taken at a resolution of approximately 0.645 μm∕pixel, andanalyzed using a custom-written code in MatLab (MathWorks), based on theautomated detection of intensity gradient and circular statistics (22). In thisalgorithm, the intensity gradient was determined pixel by pixel with a Gaus-sian differential filter. In each subregion of the image, the dominant localdirection was determined using an accumulator scheme, in which the orien-tation of each pixel follows a von Mises distribution, a circular analogue ofthe linear normal distribution. Subsequently, the orientation in each subre-gion was converted into an angle bias based on its deviation from the circum-ferential direction (see Fig. 1C). Mean angle and standard deviation of LRasymmetry were calculated for all subregions, using circular statistics (45).

Fig. 4. Chirality of muscle cells requires functional actin but not tubulin.(A) Phase contrast images and (B) chirality of C2C12 cells on micro-patternedrings in the presence of 50 nM Latrunculin A, 0.2 μg∕mL Cytochalasin D,30 nM Jasplakinolide, 200 nM Nocodazole, and 10 nM Taxol. Scale bars:100 μm. (C) Latrunculin A does not change the polarity of C2C12 cells, asthe cells positioned their centrosome (bright green), rather than the nucleus(blue), closer to ring boundaries. Scale bars: 50 μm. (D) Migration of C2C12cells in the presence of Latrunculin A, with the direction and magnitude ofvelocity indicated by arrow direction and length, respectively. (E) Averagevelocity of the cells along the circumferential direction (Vθ) and the radialdirection (Vr ) as a function of radial position.

Wan et al. PNAS ∣ July 26, 2011 ∣ vol. 108 ∣ no. 30 ∣ 12299

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

SEECO

MMEN

TARY

Dow

nloa

ded

by g

uest

on

Aug

ust 1

1, 2

020

We verified that the variation of subregion size from 10 by 10 pixels to 60 by60 pixels would not significantly alter the judgment of cell chirality on rings.The subregion size was therefore set to 20 by 20 pixels (i.e., 13 by 13 μm).

Analysis of Cell Migration. For time-lapse videos, cells were patterned on agold-coated glass-bottom Petri dish. After cells attached, the dish was trans-ferred into an environmental chamber (37 °C and 5% CO2) and image timeseries were recorded every 5 min at a resolution of 1.56 μm∕pixel for a totaltime of 20–40 h. As the image capture rate is much higher than the charac-teristic time for cell migration, digital image correlation, together with sub-pixel displacement estimation, were used to determine the displacements ofcell migration. A fast Fourier transform (FFT)-based method was first utilizedto match regions of two sequential phase contrast images. The calculateddisplacement values were then used as inputs for a more accurate estimationof displacement fields at a subpixel level, using a second-order image corre-lation algorithm described previously (26). To evaluate the bias of cell migra-

tion, the obtained velocity field was further projected in the circumferential(Vθ) and radial (Vr ) direction.

Statistical Analysis. Cell chirality on ring patterns (i.e., clockwise or counter-clockwise alignment) was determined from calculated biased angles in localregions with one sample test for the mean angle, analogous to the one sam-ple t test in linear statistics (45). The overall biased behavior of the cells wastested based on the number of rings exhibiting either clockwise or counterclockwise alignment in the rank test. The directionality of cell migration onboundaries was determined with the one-tailed Student t test. The confi-dence level was set to 0.05 for all statistical tests.

ACKNOWLEDGMENTS. We thank Dr. Oliver Hobert, Dr. Donald Freytes, andGeorge Eng for their insightful comments, and Dr. David Madigan for hisadvice on statistical analysis. We gratefully acknowledge funding receivedfrom the National Institutes of Health (Grant EB002520 to G.V.-N.) andPennington Biomedical Research Foundation (J.M.G.).

1. EdwardsW,Moles AT, Franks P (2007) The global trend in plant twining direction.GlobEcol Biogeogr 16:795–800.

2. Shibazaki Y, Shimizu M, Kuroda R (2004) Body handedness is directed by geneticallydetermined cytoskeletal dynamics in the early embryo. Curr Biol 14:1462–1467.

3. Afzelius BA (1976) A human syndrome caused by immotile cilia. Science 193:317–319.4. Li R, Bowerman B (2010) Symmetry breaking in biology. Cold Spring Harb Perspect Biol

2:a003475.5. Mercola M, Levin M (2001) Left-right asymmetry determination in vertebrates. Annu

Rev Cell Dev Biol 17:779–805.6. Aylsworth AS (2001) Clinical aspects of defects in the determination of laterality. Am J

Med Genet 101:345–355.7. Okada Y, Takeda S, Tanaka Y, Belmonte JC, Hirokawa N (2005) Mechanism of nodal

flow: A conserved symmetry breaking event in left-right axis determination. Cell121:633–644.

8. Hirokawa N, Tanaka Y, Okada Y, Takeda S (2006) Nodal flow and the generation ofleft-right asymmetry. Cell 125:33–45.

9. Tabin CJ (2006) The key to left-right asymmetry. Cell 127:27–32.10. Levin M, Thorlin T, Robinson KR, Nogi T, Mercola M (2002) Asymmetries in H+/K+-AT-

Pase and cell membrane potentials comprise a very early step in left-right patterning.Cell 111:77–89.

11. Adams DS, et al. (2006) Early, H+-V-ATPase-dependent proton flux is necessary for con-sistent left-right patterning of non-mammalian vertebrates. Development133:1657–1671.

12. Spéder P, Adám G, Noselli S (2006) Type ID unconventional myosin controls left-rightasymmetry in Drosophila. Nature 440:803–807.

13. Spéder P, Noselli S (2007) Left-right asymmetry: Class I myosins show the direction. CurrOpin Cell Biol 19:82–87.

14. Hozumi S, et al. (2006) An unconventional myosin in Drosophila reverses the defaulthandedness in visceral organs. Nature 440:798–802.

15. Fischer A, Viebahn C, BlumM (2002) FGF8 acts as a right determinant during establish-ment of the left-right axis in the rabbit. Curr Biol 12:1807–1816.

16. McGrath J, Somlo S, Makova S, Tian X, Brueckner M (2003) Two populations of nodemonocilia initiate left-right asymmetry in the mouse. Cell 114:61–73.

17. Levin M (2005) Left-right asymmetry in embryonic development: A comprehensivereview. Mech Dev 122:3–25.

18. Brown NA, Wolpert L (1990) The development of handedness in left/right asymmetry.Development 109:1–9.

19. Levin M, Mercola M (1998) The compulsion of chirality: Toward an understanding ofleft-right asymmetry. Genes Dev 12:763–769.

20. Wan LQ, et al. (2010) Geometric control of human stem cell morphology and differ-entiation. Integr Biol (Camb) 2:346–353.

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

22. KarlonWJ, et al. (1999) Measurement of orientation and distribution of cellular align-ment and cytoskeletal organization. Ann Biomed Eng 27:712–720.

23. Desai RA, Gao L, Raghavan S, Liu WF, Chen CS (2009) Cell polarity triggered by cell-celladhesion via E-cadherin. J Cell Sci 122:905–911.

24. Thery M, Piel M (2009) Adhesive micropatterns for cells: A microcontact printingprotocol. Cold Spring Harb Protoc 2009 pdb prot5255.

25. Collares-Buzato CB, McEwan GT, Jepson MA, Simmons NL, Hirst BH (1994) Paracellularbarrier and junctional protein distribution depend on basolateral extracellular Ca2þin cultured epithelia. Biochim Biophys Acta 1222:147–158.

26. Zhou P, Goodson KE (2001) Subpixel displacement and deformation gradient measure-ment using digital image/speckle correlation (DISC). Opt Eng 40:1613–1620.

27. Xu J, et al. (2007) Polarity reveals intrinsic cell chirality. Proc Natl Acad Sci USA104:9296–9300.

28. Irvine KD, Rauskolb C (2001) Boundaries in development: Formation and function.Annu Rev Cell Dev Biol 17:189–214.

29. Lecuit T, Lenne PF (2007) Cell surface mechanics and the control of cell shape, tissuepatterns and morphogenesis. Nat Rev Mol Cell Biol 8:633–644.

30. Shimizu N, Obinata T (1986) Actin concentration and monomer-polymer ratio indeveloping chicken skeletal muscle. J Biochem 99:751–759.

31. Herman IM (1993) Actin isoforms. Curr Opin Cell Biol 5:48–55.32. DanilchikMV, Brown EE, Riegert K (2006) Intrinsic chiral properties of the Xenopus egg

cortex: An early indicator of left-right asymmetry? Development 133:4517–4526.33. Qiu D, et al. (2005) Localization and loss-of-function implicates ciliary proteins in early,

cytoplasmic roles in left-right asymmetry. Dev Dyn 234:176–189.34. Taber LA (2006) Biophysical mechanisms of cardiac looping. Int J Dev Biol 50:323–332.35. Itasaki N, Nakamura H, Sumida H, Yasuda M (1991) Actin bundles on the right side in

the caudal part of the heart tube play a role in dextro-looping in the embryonic chickheart. Anat Embryol (Berl) 183:29–39.

36. Rao J, Li N (2004) Microfilament actin remodeling as a potential target for cancer drugdevelopment. Curr Cancer Drug Targets 4:345–354.

37. Pohl C, Bao Z (2010) Chiral forces organize left-right patterning in C. elegans byuncoupling midline and anteroposterior axis. Dev Cell 19:402–412.

38. Coue M, Brenner SL, Spector I, Korn ED (1987) Inhibition of actin polymerization bylatrunculin A. FEBS Lett 213:316–318.

39. Cooper JA (1987) Effects of cytochalasin and phalloidin on actin. J Cell Biol105:1473–1478.

40. BubbMR, Spector I, Beyer BB, Fosen KM (2000) Effects of jasplakinolide on the kineticsof actin polymerization. An explanation for certain in vivo observations. J Biol Chem275:5163–5170.

41. Derry WB, Wilson L, Jordan MA (1995) Substoichiometric binding of taxol suppressesmicrotubule dynamics. Biochemistry 34:2203–2211.

42. Ishizaki T, et al. (2000) Pharmacological properties of Y-27632, a specific inhibitor ofrho-associated kinases. Mol Pharmacol 57:976–983.

43. Isemura M, Mita T, Satoh K, Narumi K, Motomiya M (1991) Myosin light chain kinaseinhibitors ML-7 and ML-9 inhibit mouse lung carcinoma cell attachment to the fibro-nectin substratum. Cell Biol Int Rep 15:965–972.

44. Kovacs M, Toth J, Hetenyi C, Malnasi-Csizmadia A, Sellers JR (2004) Mechanism ofblebbistatin inhibition of myosin II. J Biol Chem 279:35557–35563.

45. Fish NI (1993) Statistical Analysis of Circular Data (Cambridge Univ Press,Cambridge, UK).

12300 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1103834108 Wan et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 1

1, 2

020