Developmental Cell 11, 459–470, October, 2006 ª2006 Elsevier Inc.DOI 10.1016/j.devcel.2006.09.007
Multicellular Rosette Formation Links
Planar Cell Polarity to Tissue Morphogenesis
J. Todd Blankenship,1Stephanie T. Backovic,1
Justina S.P. Sanny,1Ori Weitz,1
and Jennifer A. Zallen1,*
1Developmental Biology Program
New York, New York 10021
Elongation of the body axis is accompanied by the as-
sembly of a polarized cytoarchitecture that provides
the basis for directional cell behavior. We find that pla-
nar polarity in the Drosophila embryo is established
through a sequential enrichment of actin-myosin
cables and adherens junction proteins in complemen-
tary surface domains. F-actin accumulation at AP
interfaces represents the first break in planar symme-
try and occurs independently of proper junctional
protein distribution at DV interfaces. Polarized cells
engage in a novel program of locally coordinated
behavior to generate multicellular rosette structures
that form and resolve in a directional fashion. Actin-
myosin structures align across multiple cells during
rosette formation, and adherens junction proteins
assemble in a stepwise fashion during rosette resolu-
tion. Patterning genes essential for axis elongation
selectively affect the frequency and directionality of
rosette formation. We propose that the generation of
higher-order rosette structures links local cell interac-
tions to global tissue reorganization during morpho-
Convergent extension is a conserved morphogenetic
event that generates one of the striking properties of
embryonic form—the elongated body axis. A common
mechanism for elongation of epithelial and mesenchy-
mal tissues is cell intercalation, in which oriented cell
movements cause a tissue to narrow in one dimension
and lengthen in a perpendicular dimension (Keller
et al., 2000; Wallingford et al., 2002; Nikolaidou and
Barrett, 2005; Solnica-Krezel, 2005). Cell intercalation
involves a diverse repertoire of behaviors, including
polarized protrusive activity (Hardin, 1989; Shih and Kel-
ler, 1992; Elul and Keller, 2000; Munro and Odell, 2002),
differential adhesion (Wieschaus et al., 1991; Irvine and
Wieschaus, 1994), and cell-shape changes (Fristrom,
1988; Bertet et al., 2004). These behaviors are associ-
ated with the formation of a polarized cytoarchitecture
that concentrates the Bazooka/PAR-3 scaffolding pro-
tein and the Myosin II actin motor in distinct cortical
domains in Drosophila (Bertet et al., 2004; Zallen and
Wieschaus, 2004), as well as Par-6 and atypical protein
kinase C at sites of motile activity in vertebrates
(Hyodo-Miura et al., 2006). However, it is not known
how these polarized proteins engage the cellular
machinery that provides the basis for directional cell
In an emerging theme from studies in vertebrates and
invertebrates, the spatial information that guides cell
intercalation is provided locally by contact between dif-
ferent cell types (Irvine and Wieschaus, 1994; Ninomiya
et al., 2004). In Drosophila, the eve and runt transcrip-
oping anterior-posterior (AP) axis, and axis elongation
is disrupted when either gene is absent or expressed
uniformly (Irvine and Wieschaus, 1994; Zallen and Wie-
schaus, 2004). Moreover, ectopic eve or runt expression
is sufficient to reorient the polarity of adjacent cells
(Zallen and Wieschaus, 2004), demonstrating that polar-
ity is actively modulated by local interactions. A related
mechanism operates in vertebrates, where AP differ-
ences promote intercalary behavior when Xenopus me-
sodermal cells from different axial positions are juxta-
posed in culture (Ninomiya et al., 2004). These studies
indicate a deep homology in the spatial mechanisms
that organize cell rearrangement during tissue morpho-
Diverse cellular mechanisms translate global spatial
cues into directional cell behavior in different tissues.
In vertebrates, intercalating cells characteristically elon-
gate in the direction of cell movement (Concha and
Adams, 1998; Keller et al., 2000; Topczewski et al.,
2001), while no significant cell-shape anisotropy is
observed in Drosophila (Irvine and Wieschaus, 1994).
Here, we show through live imaging studies that interca-
lating cells in the Drosophila germband locally organize
resolve in a directional fashion. Germband cells become
polarized prior to intercalation through a sequence of
events that leads to the asymmetric distribution of F-
actin and adherens junction proteins. These polarities
are dynamically remodeled during the processes of
rosette formation and resolution. Rosette frequency is
reduced in eve mutants that partially elongate the
body axis, and rosette frequency and directionality are
both disrupted in bicoid nanos torso-like mutants that
lack AP patterning and fail to elongate. These results
suggest that multicellular rosette structures, and not in-
dividual cells or cell interfaces, represent the functional
units of cell behavior during tissue elongation.
Planar Polarity Is Established Prior to Intercalation
During axis elongation in Drosophila, polarized cell
movements cause the embryonic germband to narrow
along the dorsal-ventral (DV) axis and lengthen by 2.5-
fold along the anterior-posterior (AP) axis (Hartenstein
These directional behaviors require the Bazooka/PAR-3
scaffolding protein and the Myosin II actin motor pro-
tein, which localize to complementary surface domains
along the planar axis (Bertet et al., 2004; Zallen and
Wieschaus, 2004). To ask when planar polarity is first
generated, we developed a quantitative immunofluores-
cence assay for protein localization (Experimental
Procedures). In early stage-6 embryos, Bazooka and
Myosin II were localized uniformly at the apical cell sur-
face (Figures 1A and 1E). By contrast, late stage-6 em-
bryos displayed an enrichment of Bazooka at the bor-
ders between dorsal and ventral cells (DV interfaces,
Figure 1B) and Myosin II at the borders between anterior
Figure 1. Polarized Localization of Bazooka/PAR-3 and Myosin II Prior to Intercalation
(A–L0) (A–D) Bazooka (Baz), (E–H) Myosin II (Myo), (I–L) Neurotactin (Nrt). Anterior is oriented toward the left, and dorsal is up. (A–H) Baz (A) and
Myo (E) localized uniformly at the apical surface in early stage-6 embryos.Late stage-6 embryos displayed an enrichment of Baz (B) at horizontal
DV interfaces and Myo (F) at vertical AP interfaces. Baz and Myo polarities were also present in stage 7 (C and G) and stage 8 (D and H). (I–L) Nrt
was not polarized in (I) early stage 6, (J) late stage 6, (K) stage 7, or (L) stage 8. (A0–L0) Relative edge intensities plotted over the full angular range
(2–7 embryos/histogram). The same embryos were analyzed for all three proteins. Bars represent the average relative intensity of edges within
a 15?angular range. Far-left (0?–14?) and far-right (165?–179?) bars are nearly parallel to the AP axis (DV interfaces); central bars (75?–89?and
90?–104?) are nearly normal to the AP axis (AP interfaces). The scale bar is 10 mm.
and posterior cells (AP interfaces, Figure 1F). These re-
sults indicate that Bazooka and Myosin II initially coloc-
alize at the apical surface and segregate into comple-
mentary subdomains during stage 6, prior to the
initiation of cell movement. Bazooka and Myosin II re-
mained polarized at the onset of intercalation in stage
7 (Figures 1C and 1G) and throughout intercalation in
stage 8 (Figures 1D and 1H). By contrast, the lateral
membrane protein Neurotactin was uniformly localized
at these stages (Figures 1I–1L; Figure S1, see the Sup-
plemental Data available with this article online). The es-
tablishment of a planar polarized cytoarchitecture prior
to cell movement is consistent with a role for Bazooka
and Myosin II in promoting intercalary behavior.
To ask if AP and DV domains form independently or
sequentially, we examined whether Bazooka localiza-
tion is required for the distribution of other polarized
proteins. The DmPar-6 PDZ-domain protein and DaPKC
associate biochemically with Bazooka (Wodarz et al.,
2000; Petronczki and Knoblich, 2001; Hutterer et al.,
et al., 2004; Harris and Peifer, 2005). Bazooka and
adherens junction proteins failed to localize apically in
DmPar-6 and DaPKC maternal and zygotic mutants,
and Bazooka was present in ectopic basolateral puncta
thatretained anaffinity for DV interfaces (Figures 2B and
2D and data not shown). Consistent with these results,
germband extension was disrupted in embryos mutant
for DmPar-6 (49% of progeny of females bearing
DmPar-6 germline clones, n = 137) and DaPKC (46%
of progeny of females bearing DaPKC germline clones
crossed to DaPKC heterozygous males, n = 167).
Despite the mislocalization of Bazooka and junctional
proteins in DmPar-6 and DaPKC mutants, Myosin II
localized correctly to AP borders at the apical cell sur-
face (Figures 2D and 2E). These results demonstrate
that distinct planar domains of Bazooka and Myosin II
can form independently of apical-basal polarity and
apical adherens junctions.
F-Actin Polarity Represents the First Break
in Planar Symmetry
The asymmetric distribution of Myosin II suggests a
eton in intercalating cells. Consistent with this possibil-
ity, we found that filamentous actin (F-actin) accumu-
lated at AP cell borders in a polarized fashion (Figures
2G and 3A–3C). F-actin was enriched at AP interfaces
in early stage-6 embryos in which Bazooka and Myosin
II were uniformly distributed (Figure 2G), indicating that
Myosin II assembles onto an already polarized F-actin
lished despite Bazooka mislocalization in DmPar-6 and
DaPKC mutants(Figure 2F anddatanot shown), indicat-
ing that AP polarity forms upstream or independently of
DV polarity. By contrast, F-actin polarity was abolished
Figure 2. AP Polarity Occurs Upstream or Independently of DV Polarity
(A–G00) (A–F) Bazooka (Baz, green), (A and B) Armadillo (Arm, red), (C–E) Myosin II (Myo, red), (C and D) Neurotactin (Nrt, blue), (F) F-actin (red).
Anterior is oriented toward the left, and dorsal is up in (A), (B), (E), and (F); apical is up, and basal is down in (C) and (D). (A,B) Arm (red) and Baz
(green) were apical in wild type (A stage 7) and basolateral in DmPar-6 (B stage 7, shown 6 mm basal to the plane in [A]). Mislocalized Baz puncta
retained an affinity for DV interfaces. (C,D) Myo (red) and Baz (green) were apical in wild type (C stage 8), but Baz was basolaterally displaced in
DmPar-6 (D stage 8). (E) Myo polarity (red) was maintained despite Baz mislocalization (green) in DmPar-6 (stage 8 in [E], stages 7–8 in [E0]). (F) F-
actin polarity (red) was maintained despite Baz mislocalization (green) in DmPar-6 (stage 7 in [F], stages 7–8 in [F0]). (G) F-actin was polarized in
early stage-6 embryos in which Myosin (G0) and Bazooka (G00) were uniform (Myosin localization assessed by the functional Sqh:GFP regulatory
light-chain fusion). (E0, F0, and G–G00) Relative edge intensities plotted over the full angular range (2–4 embryos/histogram). Scale bars are 10 mm.
Rosette Formation and Axis Elongation
in the absence of AP patterning in the progeny of bicoid
nanos torso-like females (bcd nos tsl mutants, Fig-
ure 3D). Therefore, asymmetric F-actin distribution in re-
sponse to the AP-patterning system is the first evidence
of planar polarity in the Drosophila embryo.
Junctional Proteins Are Enriched at DV Interfaces
at the Onset of Intercalation
Bazooka and Myosin II accumulate in the vicinity of
adherens junctions inintercalating cells (Zallen andWie-
schaus, 2004). To ask if adhesion is spatially regulated
Figure 3. Polarized Localization of F-Actin and Adherens Junction Proteins in Intercalating Cells
(A–L0) (A–D) F-actin (visualized with phalloidin), (E–H) Armadillo (Arm), (I–L) DE-cadherin (Ecad). Anterior is oriented toward the left, and dorsal is
up.(A–D) F-actin accumulated atAPinterfacesinwild-type embryosprior tointercalation instage6(A),atthe onset of intercalation instage7(B),
and during intercalation in stage 8 (C). F-actin polarity was abolished in bcd nos tsl mutants (stage 6 in [D] and [D0]). (E–L) Wild-type embryos
displayed uniform localization of Arm (E) and Ecad (I) prior to intercalation at stage 6. DV enrichment of Arm (F and G) and Ecad (J and K)
was observed at the onset of intercalation in stage 7 (F and J) and during intercalation in stage 8 (G and K). Arm (H) and Ecad (L) were uniformly
localized in bcd nos tsl mutants (stage 7 in [H] and [L], stages 7 and 8 in [H0] and [L0]). (A0–L0) Relative edge intensities plotted over the full angular
range (3–6 embryos/histogram). The scale bar is 10 mm.
during intercalation, we examined the distribution of the
core adherens junction proteins DE-cadherin/Shotgun
and its cytoplasmic binding partner, Armadillo/b-cate-
nin. While adherens junctions are present at intercellular
contacts throughout the germband (Tepass and Harten-
stein, 1994), we found that junctional proteins displayed
a subtle and reproducible enrichment at DV borders of
intercalating cells in stages 7 and 8 (Figures 3F, 3G,
3J, and 3K), but not prior to intercalation at stage 6 (Fig-
ures 3E and 3I). The polarized distribution of these junc-
tional proteins was eliminated in bcd nos tsl mutants
(Figures 3H and 3L). Together, these experiments indi-
cate that F-actin asymmetry is the primary planar polar-
ity in the Drosophila germband, followed by the segre-
gation of Myosin II and Bazooka into complementary
surface domains. Junctional proteins subsequently be-
come enriched at DV interfaces at the time of intercala-
tion. The asymmetric distribution of cytoskeletal and
junctional proteins could contribute directly to polarized
cell behavior during tissue elongation.
Cellular Mechanisms of Axis Elongation: Formation
of Multicellular Rosette Structures
To understand how this polarized cytoarchitecture is
translated into directional cell behavior, we performed
time-lapse confocal imaging of embryos expressing
GFP-tagged membrane markers (Experimental Proce-
dures; Movie S1). Germband extension occurs over a
2 hr period (Campos-Ortega and Hartenstein, 1985).
We focused on the first 40 min, during which 80% of
stein and Campos-Ortega, 1985). Germband cells devi-
ate from a uniform hexagonal organization and display a
wide range of neighbor relationships (Zallen and Zallen,
2004). This property is also observed in other epithelial
tissues (Fristrom, 1988; Classen et al., 2005). Prior to
intercalation, many germband cells were hexagonal in
topology, having interfaces with six neighboring cells
(58% of cells in 4 stage-6 embryos, >250 cells/embryo).
heptagonal (9%). The topological range increased at the
onset of intercalation in stage 7, and nonhexagonal
cells dominated by stage 8 (73% of cells in 6 stage-8
embryos, >200 cells/embryo). Germband cells therefore
become highly disordered during intercalation (Fig-
ure 4A) (Zallen and Zallen, 2004).
In a honeycomb, as well as in disordered foams, three
cells meet at every vertex (Weaire and Rivier, 1984).
Cells that change neighbors in a two-dimensional sheet
undergo an obligatory intermediate in which four cells
meet; this neighbor-exchange event is referred to as
an elementary T1 process (Weaire and Rivier, 1984).
Accordingly, the number of 4-cell vertices (vertices at
which four cells and four edges meet) increases during
intercalation (Figure 4B) (Bertet et al., 2004; Zallen and
Zallen, 2004). Our live imaging studies revealed that
germband populations form unexpected vertices at
which 5 or more cells meet (Figure 4C; Movie S1), with
up to 11 cells converging at a single point (Figures
occur through T1 processes and indicate a higher-level
organization of cell behavior.
To investigate the behaviors that lead to rosette
formation, we developed a computational approach to
identify rosette configurations in germband populations
(Experimental Procedures). This algorithm highlights
clusters of five or more cells that circumscribe short
edges (less than one-half the median length) and/or
high-order vertices (at which four or more cells meet).
In wild-type embryos, up to 61% of germband cells
formed rosettes of five or more cells at a single time
point (Figure 4D), and up to 27% formed rosettes of six
or more cells (Figures 4E and 5A–5F). Similar morphol-
ogies were observed in fixed embryos (Figure S2). To
detect cells that participate transiently in rosette struc-
tures, we tracked individual cells for 20–25 min from the
onset of intercalation and found that 87% of germband
cells were incorporated into rosettes (n = 514 cells in 3
embryos) and that 56% of cells engaged in multiple
rounds of rosette formation (up to 5 rosettes/cell that
average 1.7 rosettes/cell). These results demonstrate
that multicellular rosette formation is a prevalent behav-
ior in intercalating populations.
Figure 4. Dynamic Cell Patterns in Intercalating Populations
Wild type (nine embryos, blue), eve mutant (seven embryos, red), and bcd nos tsl (bnt) mutant (five embryos, green). Bars indicate the mean, and
error bars indicate standard error of the mean.
(A) In wild type, the variance of the cell topology (number of sides) distribution,PpðnÞðn2? nÞ2, displays a sharp increase in slope at the begin-
intercalation during stage 7 (0–10 min) and stage 8 (10–35 min). The increase in the topological variance was reduced in eve and bcd nos tsl
(B) The fraction of 4-cell vertices increased at the onset of intercalation in stage 7 and plateaued in stage 8 in wild type; eve and bcd nos tsl
mutants displayed a slower rise in 4-cell vertices, and bcd nos tsl mutants reached a lower peak value.
(C) Vertices at which five or more cells meet were present in wild-type embryos and were reduced in eve and bcd nos tsl mutants.
(D) A majority of wild-type cells formed rosettes of five or more cells. Fewer cells formed rosettes in eve and bcd nos tsl mutants.
(E) Rosettes of six or more cells were present in up to 27% of wild-type cells at a single time point and were less frequent in eve and bcd nos tsl
mutants. Cell patterns in a field of 111–424 cells were analyzed at 2.5 or 5 min intervals.
ning of stage 7 that provides an objective reference for the onset of intercalation and defines t = 0. The variance continued to rise throughout
Rosette Formation and Axis Elongation
Rosette formation and resolution proceed in a strictly
directional fashion that parallels structural changes at
the tissue level. In every case, a cellular array that was
elongated along the DV axis rearranged to form an array
that was elongated along the AP axis (100%, n = 90
rosettes). Rosette progression consists of three charac-
teristic phases: (1) formation, the constriction of linked
interfaces between adjacent columns of cells (Fig-
ure 5H, w10 min); (2) a transient high-order vertex inter-
mediate in which multiple cells meet at a single point
Figure 5. Multicellular Rosette Formation in Intercalating Cells
(A–J0) (A–J) Images from a time-lapse movie of wild-type cells expressing GFP:Resille. Time is given in minutes relative to the onset of interca-
lation as defined in Figure 4A. Anterior is oriented toward the left, and dorsal is up. (A–F) Time-lapse images of the germband delimited by the
necessarily indicate the same cells in sequential images. (G–J) Time-lapse images of a single rosette. (G) During rosette formation, two columns
of cells initially align (edges that will contract are indicated by yellow arrows in [G0]). (H) Adjacent pairs of cells constrict their shared interfaces in
at a high-order vertex (the 11-cell rosette is indicated by a blue circle in [I0]). (J) This apparently symmetric structure resolves in a strictly direc-
tional fashion during rosette resolution (the direction of extension is indicated by green arrows in [J0]). Scale bars are 10 mm.
(Figure 5I, <1 min); and (3) resolution, the establishment
of contact between cells that were previously 2–5 cell
diameters apart along the DV axis and the separation
w10 min). Rosette behaviors produced a change in the
AP/DV aspect ratio (the length of a multi-cell array along
the AP axis divided by its length along the DV axis) by
a factor of 2.1 for rosettes of 6–7 cells (n = 74) and by
a factor of 3.0 for rosettes of 8–10 cells (n = 9). The
change in the aspect ratio for the germband sheet dur-
ing the same time period was 2.7 6 0.2 (for a field en-
compassing one-third of the germband, n = 3 embryos).
The directionality of rosette formation arises from the
concerted contraction of AP interfaces shared by neigh-
boring pairs of cells. Five-cell rosettes formed by loss of
two linked AP interfaces (98%, n = 190), 6-cell rosettes
formed by loss of 3 linked AP interfaces (100%, n = 65,
Figures6Aand6F), andupto8linked APinterfaces con-
tracted in tandem to produce rosettes of 7–11 cells
(100%, n = 25, Figures 5G–5I). Contraction of isolated
AP interfaces leads to local interactions among four
cells that can also result in neighbor exchange (T1
processes [Weaire and Rivier, 1984]) and have been im-
plicated in germband elongation (Bertet et al., 2004).
Consistent with this, 99% of 4-cell vertices formed
through loss of an AP interface (n = 376, Figures 6D–
6F), and 72%of4-cellvertices resolvedby T1processes
(n = 715, Figures 6D and 6E). These events created con-
tact between cells that were separated by less than one
cell diameter along the DV axis (92%, n = 391) and
increased the AP/DV aspect ratio by a factor of 1.79 (n =
tet et al., 2004) therefore contributes to axis elongation
and represents one manifestation of the general cell be-
thedynamic behaviors of linked interfaces informing ro-
sette structures were distinct from isolated interfaces.
AP interfaces rapidly contracted to form 4-cell vertices
during T1 neighbor exchange (average time to disap-
pearance 7.5 6 0.3 min, n = 56 interfaces in 3 embryos).
By contrast, AP interfaces in forming rosettes persisted
significantly longer before collapsing to produce a high-
order vertex (10.8 6 0.5 min, n = 49 interfaces in 3
embryos) and were more likely to undergo alternating
rounds of contraction and extension (Movie S1). These
results suggest that interface dynamics are actively
modulated by input from neighboring pairs of cells in a
mechanism that promotes rosette formation.
Dynamic Protein Localization during Rosette
Formation and Resolution
To understand the molecular basis of rosette behavior,
we analyzed the localization of cytoskeletal and junc-
tional proteins in fixed cells that display morphological
hallmarks of rosette formation and resolution. F-actin
and Myosin II were enriched in high-order vertices at
Figure 6. Rosette Behaviors Are Directional in eve, but Not bcd nos tsl, Mutants
(A–I00) Images from time-lapse movies of intercalating cells in wild type (A and D–F), eve mutants (B, G, and H), and bcd nos tsl mutants (C and I)
visualized with GFP:Resille (A, C–F, and I) or GFP:Spider (B, G, and H). Time is given in minutes. Anterior is oriented toward the left, and dorsal is
up. (A) Wild-type rosette progression causes a multicellular array to narrow along the DV axis during rosette formation (A0) and to elongate along
the AP axis during rosette resolution (A00). (B) In an eve mutant, 6-cell rosettes formed (B0) and resolved (B00) in a directional fashion. (C) In a bcd
nos tsl mutant, 6-cell rosettes formed (C0) and resolved (C00) along random axes. (D–F) Wild-type 4-cell vertices (D0–F0) formed through loss of an
AP interface and resolved to create contact between dorsal and ventral cells (D00and E00) or to join rosettes (F00). (G and H) In eve mutants, 4-cell
vertices resolved (G00) or joined rosettes (H00) with wild-type directionality. (I) In bcd nos tsl mutants, 4-cell vertices often formed through loss of
a DV interface (I0), and new edges often restored old contacts (I00). Scale bars are 10 mm.
Rosette Formation and Axis Elongation
the rosette center (Figures 7B and 7E) and in linear
chains of short AP interfaces in configurations that re-
semble forming rosettes (compare Figures 7A and 7D
with Figures 5H and 6A). Resolving rosettes were identi-
fied by linear chains of short DV interfaces and the char-
acteristic hourglass morphology of associated cells
(compare Figures 7C and 7F with Figures 5J and 6A00).
In contrast to its general association with AP interfaces,
F-actin was also enriched at DV interfaces in resolving
rosettes (Figure 7F). A majority of these interfaces dis-
played gaps in Bazooka localization (Figures 7C0, 7G0,
and 7H0; 43/49 rosettes), but they were usually positive
for DE-cadherin (Figure 7F0; 14/15 rosettes) or Arma-
dillo/b-catenin (Figure 7H; 21/25 rosettes). Apparent
gaps between cells stained positively for the Neurotac-
tin membrane marker (data not shown), demonstrating
richment and Bazooka depletion atnewly forming DV in-
terfaces do not correlate with their global distribution
and suggest a role in rosette resolution.
cadherin and Armadillo/b-catenin are recruited to sites
of new cell contacts prior to Bazooka association. To
test this directly, we analyzed protein localization in
Figure 7. Dynamic Localization of Cytoskeletal and Junctional Proteins during Rosette Progression
(A–L00) (A–C) Myosin II (Myo, red), (A0–C0and G0–I0) Bazooka (Baz, green), (D–F) F-actin (red), (D0–F0) DE-cadherin (Ecad, green), (G–I) Armadillo
(Arm, red). Anterior is oriented toward the left, and dorsal is up. (A–C) Myo was enriched in shrinking AP interfaces in forming rosettes (arrows
in [A00]) and at high-order vertices (arrowhead in [B00]). Baz was depleted from growing DV interfaces in resolving rosettes (arrows in [C00]). (D–F) F-
actin was enriched in shrinking AP interfaces in forming rosettes (arrows in [D00]), in high-order vertices (arrowhead in [E00]), and in growing DV
interfaces in resolving rosettes (arrows in [F00]). (G–I) Short DV interfaces either lacked Arm (G) and Baz (G0arrows point to missing edges, 4/
25 rosettes), displayed high levels of Arm (H) and low levels of Baz (arrows in [H0], 19/25 rosettes), or were positive for Arm (I) and Baz (arrowhead
in [H0] and arrows in [I0], 2/25 rosettes). Similarly, a majority of short DV interfaces displayed high levels of Ecad and low levels of Baz (11/15 ro-
settes), while asubset was negative for both (1/15rosettes) orpositive forboth (3/15rosettes).(J–L)Images from atime-lapse movie of wild-type
cells expressing (J) Bazooka:GFP, (K) Ecad:GFP, or (L) GFP:Moesin. Vertices (yellow arrows), growing edges (green arrowheads), and shrinking
edges (red arrowheads) are indicated. Bazooka:GFP was recruited to new interfaces after a delay ([J0], t = 5 min; [J00], t = 11 min), while Ecad:GFP
was present at new interfaces from the earliest stages of interface formation ([K0], t = 0.5 min; [K00], t = 2.5 min). (L) GFP:Moesin accumulated
transiently at new interfaces ([L0], t = 2 min) and quickly returned to steady-state levels ([L00], t = 3 min, GFP remains enriched at the interface
indicated by the central arrowhead). Scale bars are 10 mm.
living embryos that express functional GFP fusions
to DE-cadherin (Oda and Tsukita, 2001) or Bazooka
(Benton and St. Johnston, 2003). We found that DE-
cadherin:GFP accumulated at new interfaces from the
earliest time points examined, within 30 s of the appear-
first detected at new interfaces after a delay of several
minutes (3.8 6 0.6 min, n = 10 rosettes; Figure 7J). The
F-actin markerGFP:Moesin (Kiehart etal., 2000) was en-
n = 13 rosettes; Figure 7L). These results demonstrate
that cell contacts created by intercalation first accumu-
late DE-cadherin and F-actin and subsequently recruit
Bazooka during their eventual stabilization (Figure 8).
Rosette Frequency and Directionality Require
We have shown that a majority of germband cells partic-
ipate in multicellular rosette structures that undergo
directional convergence and extension. If rosette be-
haviors are important for elongation, then germband ex-
tension should be disrupted in mutants that form fewer
rosettesor rosettesthatlack directionality. Totest these
predictions, we performed time-lapse imaging of eve
mutants that display reduced germband elongation
(Irvine and Wieschaus, 1994) and bcd nos tsl mutants
that fail to elongate (Nusslein-Volhard et al., 1987).
These experiments revealed that fewer rosettes formed
in eve and bcd nos tsl mutants (Movies S2 and S3). The
creased from 61% in wild-type embryos to 40% in eve
and 31% in bcd nos tsl (Figure 4D), and rosettes of six
or more cells decreased from 27% in wild-type to 10%
in eve and 3% in bcd nos tsl (Figure 4E). Consistent
with these results, fewer 4-cell vertices joined rosettes
ineve(14%,n=448)andbcdnos tsl (11%,n=265)com-
pared to wild-type (27%, n = 715). In addition, less than
2% of wild-type 4-cell vertices failed to either resolve or
join a rosette within 15 min, while 20% of 4-cell vertices
in eve mutants and 18% of 4-cell vertices in bcd nos tsl
mutants persisted for 15 min or longer. The failure to
form higher-order rosettes in AP-patterning mutants is
thus accompanied by an aberrant stabilization of 4-cell
Surprisingly, some rosettes still formed despite the
absence of AP-patterning information in bcd nos tsl mu-
tants, but these behaviors occurred along random axes
and did not promote elongation. In eve mutants, 4-cell
vertices formed through contraction of an AP interface
as in wild-type (93%, n = 328, Figures 6G and 6H), while
a DV interface (64%, n = 237, Figure 6I). Similarly, 6-cell
rosettes in eve mutants formed and resolved with wild-
type directionality (12/13, Figure 6B), while 6-cell ro-
settes in bcd nos tsl mutants formed through loss of
DV interfaces (6/11), AP interfaces (3/11), or both (2/11,
Figure 6C) and resolved to promote elongation (4/9),
sion (1/9). A subset of 6-cell rosettes in bcd nos tsl mu-
an interface that had recently been lost (5/9 rosettes).
These results demonstrate that rosette resolution does
not always occur perpendicular to the direction of ro-
sette formation and suggest that both processes are
regulated downstream of the AP-patterning system.
A major challenge in developmental biology is to under-
stand how changes in tissue structure are generated on
cytoskeletal and junctional proteins in intercalating cells
are polarized with respect to the axes of the embryo and
aredynamically remodeledduring rosetteformationand
resolution. Three lines of evidence indicate that rosette
structures are closely associated with the essential cell
behaviors that drive axis elongation. First, rosettes
form and resolve in a strictly directional fashion that re-
Second, a majority of germband cells participate in mul-
directionality are selectively disrupted in AP-patterning
mutants that are defective for axis elongation. Rosette
behaviors account for morphological properties of
germband populations that cannot be explained by sim-
ple neighbor exchange, including the diverse cell geom-
etries present in intercalating populations, the meeting
of more than four cells at a single point, and the rapid
juxtaposition of widely separated cells.
Creating Asymmetry along Planar Axes
The first evidence of planar polarity in intercalating cells
is an asymmetric enrichment of F-actin in response to
the AP-patterning system. Cytoskeletal reorganization
is also an early event in polarization of the C. elegans
Figure 8. Multicellular Rosette Formation
and Axis Elongation
Sequence of events from left to right. F-actin
enrichment (purple) at AP interfaces is the
earliest evidence of planar polarity (left).
Myosin II (pink) subsequently accumulates
at AP interfaces and may coordinate the
contraction of linked edges to drive rosette
formation. F-actin and Myosin II colocalize
at the vertex where multiple cells meet (cen-
ter). Rosettes resolve in a directional fashion
to create contact between cells that were
previously separated along the DV axis.
DE-cadherin association (green) is an early step in new contact formation and coincides with formation of a transient F-actin structure (right).
Bazooka (blue) is recruited to new interfaces after a delay, which may create local differences in adhesion that are important for rosette
Rosette Formation and Axis Elongation
embryo in response to sperm entry (Nance, 2005).
F-actin polarity in the Drosophila germband is followed
by segregation of Myosin II and Bazooka into distinct
cortical domains and enrichment of adherens junction
proteins at DV interfaces. Polarized F-actin and Myosin
II distribution does not require proper Bazooka localiza-
tion to DV interfaces, indicating that AP polarity forms
upstream or independently of DV polarity. These results
argue that Bazooka does not delimit the boundary of
Myosin II expression, since reduction of the Bazooka
DV domain in DmPar-6 and DaPKC mutants is not ac-
companied by an expansion of Myosin II. Moreover, api-
cal Myosin II does not simply exclude Bazooka, since
Bazooka aggregates display an affinity for DV interfaces
at basolateral locations. These results indicate that pla-
nar polarized Bazooka and Myosin II domains can form
in the absence of direct contact upon disruption of
Actin-Myosin Cables Align across Multiple Cells
during Rosette Formation
It has been proposed that axis elongation in Drosophila
occurs through stereotyped cell-shape changes across
a uniform field (Bertet et al., 2004). However, germband
cellsdisplay awiderange ofgeometries,andonly asub-
set of cells is present in the prescribed orientation prior
to intercalation (Zallen and Zallen, 2004). Moreover, this
multiple rounds are required for full elongation. Here, we
find that germband cells display a range of dynamic be-
allow for both simple neighbor exchange and higher-
order rosette formation. Actin-myosin networks could
coordinate interface contraction in adjacent pairs of
cells to create multicellular rosette structures that
amplify the elongation produced by local cell rearrange-
ment. Supracellular actin-myosin cables are implicated
in morphogenetic processes such as wound healing
(Martin and Lewis, 1992; Wood et al., 2002), epithelial
closure (Young et al., 1993; Jacinto et al., 2002; Franke
et al., 2005), and cell sorting (Wei et al., 2005). Of note,
germband cells in bcd nos tsl mutants still undergo ro-
sette formation in the absence of AP patterning, but it
is nondirectional. The capacity for locally coordinated
cell-shape changes may represent an intrinsic property
of epithelial tissues that can be mobilized in response to
distinct spatial inputs during morphogenesis.
Dynamic Regulation of Adherens Junction Proteins
stages of germband extension: first, through the general
enrichment of junctional proteins at DV interfaces, and,
later, through transient differences in junctional compo-
sition at new cell contacts. The asymmetric distribution
of Bazooka prefigures the formation of a planar polar-
ized junctional network, consistent with known roles of
Bazooka in promoting the assembly or stabilization of
adherens junctions (Muller and Wieschaus, 1996; Abde-
lilah-Seyfried et al., 2003; Harris and Peifer, 2004; Chen
and Macara, 2005). Adhesive differences could perform
multiple functions during intercalation, such as facilitat-
ing the alignment of cells prior to rosette initiation, pro-
viding structural support for the extreme morphologies
produced byrosette formation, orpromoting membrane
growth during rosette resolution.
We demonstrate that high-order vertices at which five
or more cells meet are resolved through a mechanism
that involves AP patterning and sequential recruitment
of cytoskeletal and junctional proteins at new mem-
branes. Nascent cell contacts display a transient F-actin
ute to dynamic junctional behaviors required for mem-
brane growth. In cultured cells, F-actin promotes the
expansion of junctional domains in a polymerization-
dependent process mediated by the Arp2/3 complex
and formin-1 (Kovacs et al., 2002; Kobielak et al., 2004;
Verma et al., 2004). To our knowledge, our data provide
the first evidence that F-actin is recruited to newly form-
ing junctions in vivo, where it may influence the forma-
tion or stabilization of cell contacts.
While the sequential association of cytoskeletal and
junctional proteins may contribute to the formation of
new cell interfaces, this does not explain why rosettes
resolve in a directional fashion. The simultaneous asso-
ciation of multiple cells provides a potential choice point
at which cells select new neighbors in a reproducible
manner. It has been proposed that 4-cell vertices allow
a cell to choose among three neighbors (Baum, 2004).
Here, we find that rosette formation can bring up to 11
cells into simultaneous contact, providing an opportu-
tions. The guidance information that directs this choice
is provided by the AP-patterning system, since rosette
resolution in bcd nos tsl mutants does not correlate
with the axis of rosette formation or the global axes of
the embryo. Moreover, 4-cell interactions are aberrantly
stabilized in eve and bcd nos tsl mutants, suggesting
that these structures are not intrinsically unstable, but
are actively resolved by specific cell affinities. The AP-
patterning system may allow a cell to precisely discrim-
inate among its neighbors during rosette resolution. Al-
ternatively, disassembly of the contractile actin-myosin
network involved in wild-type rosette formation could
render a subset of cells refractory to building the actin
structures required to establish new contacts.
Emergent Patterns of Tissue Organization Arise
from Single-Cell Behaviors
The study of morphogenesis lies at the interface be-
tween cell and tissue biology and requires new ways
of thinking about cell behavior. An outstanding question
is whether the organization of multicellular populations
can be described as the sum of single-cell events, or if
unique phenomena emerge atthe tissue level. Individual
cells often coordinate their behavior within a cohesive
multicellular structure. Perhaps the functional unit of
tissue morphogenesis is in fact a group of cells, each
of which senses its position within the group and carries
out distinct behaviors accordingly. For example, the
monolithic advance of a cell sheet mediates wound
healing and epithelial closure (Schock and Perrimon,
2002; Redd et al., 2004), and entire tissues can migrate
as a group to new locations in the body (Lecaudey and
Gilmour, 2006). In these cases, it may be appropriate to
consider tissue morphogenesis in terms of collective
actions by multi-cell assemblies, rather than as the
reiteration of single-cell behaviors. We show here that
multicellular rosette structures promote rapid and
efficient tissue elongation in the Drosophila germband.
An understanding of the principles that govern such
higher-order cell behaviors will provide insight into how
erties of tissue organization.
Fly Stocks and Genetics
Flies were maintained at 25?C by standard procedures. Oregon R
was the wild-type stock. Germline clones for the DaPKCk06403
(Wodarz et al., 2000) and DmPar-6D226(Petronczki and Knoblich,
2001; Hutterer et al., 2004) alleles were generated by the FLP-DFS
system (Chou and Perrimon, 1992). Time-lapse imaging was con-
stock were imaged starting at stage 6b/7 and were genotyped by
cephalic furrow defects and partial germband extension. Maternal
bcd nos tsl mutant embryos were progeny of GFP:Resille; bcdE1
nosL7tsl146females. Bazooka:GFP embryos were the F2 progeny
of UASp:Bazooka:GFP (Benton and St. Johnston, 2003) and
mataTub-Gal4VP16 67C;15 (gift of D. St. Johnston). Ecad:GFP (Oda
and Tsukita, 2001), GFP:Moesin (Kiehart et al., 2000), and Sqh:GFP
(Royou et al., 2004) were expressed from endogenous promoters.
Embryos were fixed by heat-methanol fixation (Muller and Wie-
schaus, 1996) for staining with rabbit anti-Arm (1:200 [Riggleman
by Wodarz et al., ), rabbit anti-Myo II (Zipper heavy chain,
1:1250, gift of C. Field [Foe et al., 2000]), and mouse anti-Nrt (1:200,
Developmental Studies Hybridoma Bank, DSHB). Embryos were
fixed for 1 hr at the interface of heptane and 3.7% formaldehyde in
0.1 M sodium phosphate buffer (pH 7.4) and were manually devitelli-
nized for staining with Alexa-488 phalloidin (Molecular Probes),
guinea pig anti-Baz (1:500), rat anti-DE-cadherin (1:100 [Oda et al.,
1994], DSHB), and rabbit anti-GFP (1:100, Torrey Pines). Secondary
antibodies (1:500) were conjugated to Alexa-488, Alexa-568, or
Alexa-647 (Molecular Probes). Embryos were mounted in Prolong
Gold (Molecular Probes). Images were acquired on a Zeiss LSM510
META confocal microscope with a PlanNeo 403/1.3NA objective.
Quantitation of Polarized Protein Distribution
Embryos were staged as described by Campos-Ortega and Harten-
stein (1985); stage-6 embryos displayed no pole cell movement and
donot include embryos in6a or6b. Images weremaximum-intensity
projections of up to eight 1 mm apical slices acquired at 0.5 mm steps
to accommodate stage-specific differences in apical-basal distribu-
tion. F-actin and Myosin II were analyzed in projections of the same
and mean fluorescence intensity were measured in Object Image for
all edges in a 50 mm2region (73–276 interfaces/embryo). Absolute
fluorescence intensities were ranked and normalized to a scale
from 20.5 to 0.5 to generate relative intensity values that were neg-
ative for edges with below-average intensity and positive for edges
with above-average intensity. Distributions were scored as polar-
ized if all positive and negative bars were consecutive.
In Figures 4 and 5, Z stacks of 1 mm steps were acquired at 15 s in-
tervals on a Perkin Elmer RS5 spinning disc confocal microscope
with a Zeiss PlanNeo 403/1.3NA objective. Single optical slices
(GFP:Resille) or projections of 1–3 slices (GFP:Spider) within 3 mm
of the apical surface were analyzed. In Figures 7J–7L, Z stacks of
1 mm slices at 0.5 mm steps were acquired at 30 s intervals on a Zeiss
LSM510 META confocal microscope with a PlanNeo 403/1.3NA ob-
jective. For Bazooka:GFP and Ecad:GFP, maximum-intensity pro-
and mean interface intensity was measured in ImageJ. Interfaces
were scored as positive if they reached >50% of their final intensity
(average of ten steady-state measurements) after subtracting back-
ground intensity (average of ten cytoplasmic measurements).
Confocal images were analyzed in a semiautomated fashion by us-
ing custom software written in Matlab. A binary image representing
cell boundaries was obtained by repeated use of filters and image
operations and was used to construct a polygonal cellular lattice
whose geometric properties were calculated. Four-cell vertices
were manually tracked in 6–10 embryos for each genotype. To iden-
tify cells involved in rosette formation, all cells in a 403 field at the
onset of intercalation in stage 7 were tracked for 20–25 min. To an-
alyze interface dynamics, AP interfaces that were 3–7 mm long and
oriented at 72?–108?relative to the AP axis (20% of the angular
range) were measured in ImageJ at 1 min intervals starting 3–4
min before the increase in topological variance at stage 7.
Supplemental Data include color-coded pixel-intensity plots for
Figures 1 and 3, images of cell patterns in fixed embryos, and time-
lapse movies of wild-type and mutant embryos. These files are
available at http://www.developmentalcell.com/cgi/content/full/
We are grateful to Richard Zallen (Department of Physics, Virginia
Tech) for many valuable discussions and for advice on the statistical
analysis of cell patterns. We thank Keith Amonlirdviman for gener-
ously sharing an algorithm that we used as the basis for developing
our computer program and Chris Field for providing Myosin II anti-
bodies. We also thank Kathryn Anderson, Mary Baylies, Natalie De-
nef, Mimi Shirasu-Hiza, Eric Wieschaus, and anonymous reviewers
roughs Wellcome Fund Career Award in the Biomedical Sciences,
a March of Dimes Basil O’Connor Starter Scholar Research Award,
and a Searle Scholar Award to J.A.Z.
Received: June 13, 2006
Revised: August 9, 2006
Accepted: September 12, 2006
Published: October 2, 2006
Abdelilah-Seyfried, S., Cox, D.N., and Jan, Y.N. (2003). Bazooka is
a permissive factor for the invasive behavior of discs large tumor
cells in Drosophila ovarian follicular epithelia. Development 130,
Baum, B. (2004). Animal development: crowd control. Curr. Biol. 14,
Benton, R., and St Johnston, D. (2003). A conserved oligomerization
domain in Drosophila Bazooka/PAR-3 is important for apical locali-
zation and epithelial polarity. Curr. Biol. 13, 1330–1334.
Bertet, C., Sulak, L., and Lecuit, T. (2004). Myosin-dependent
junction remodelling controls planar cell intercalation and axis elon-
gation. Nature 429, 667–671.
Campos-Ortega, J.A., and Hartenstein, V. (1985). The Embryonic
Development of Drosophila melanogaster (Berlin: Springer-Verlag).
Chen, X., and Macara, I.G. (2005). Par-3 controls tight junction
assembly through the Rac exchange factor Tiam1. Nat. Cell Biol.
Chou, T.B., and Perrimon, N. (1992). Use of a yeast site-specific
recombinase to produce female germline chimeras in Drosophila.
Genetics 131, 643–653.
Classen, A.-K., Anderson, K.I., Marois, E., and Eaton, S. (2005). Hex-
agonal packing of Drosophila wing epithelial cells by the planar cell
polarity pathway. Dev. Cell 9, 805–817.
Concha, M.L., and Adams, R.J. (1998). Oriented cell divisions and
cellular morphogenesis in the zebrafish gastrula and neurula:
a time-lapse analysis. Development 125, 983–994.
Elul, T., and Keller, R. (2000). Monopolar protrusive activity: a new
morphogenic cell behavior in the neural plate dependent on vertical
interactions with the mesoderm in Xenopus. Dev. Biol. 224, 3–19.
Rosette Formation and Axis Elongation
Foe, V.E., Field, C.M., and Odell, G.M. (2000). Microtubules and
mitotic cycle phase modulate spatiotemporal distributions of
F-actin and myosin II in Drosophila syncytial blastoderm embryos.
Development 127, 1767–1787.
Franke, J.D., Montague, R.A., and Kiehart, D.P. (2005). Nonmuscle
in multiple tissues during dorsal closure. Curr. Biol. 15, 2208–2221.
Fristrom, D. (1988). The cellular basis of epithelial morphogenesis.
A review. Tissue Cell 20, 645–690.
Hardin, J. (1989). Local shifts in position and polarized motility drive
cell rearrangement during sea urchin gastrulation. Dev. Biol. 136,
Harris, T.J., and Peifer, M. (2004). Adherens junction-dependent
in Drosophila. J. Cell Biol. 167, 135–147.
Harris, T.J., and Peifer, M. (2005). The positioning and segregation
of apical cues during epithelial polarity establishment in Drosophila.
J. Cell Biol. 170, 813–823.
Hartenstein, V., and Campos-Ortega, J.A. (1985). The spatio-tempo-
ral pattern of embryonic cell divisions. Roux’s Arch. Dev. Biol. 194,
Hutterer, A., Betschinger, J., Petronczki, M., and Knoblich, J.A.
(2004). Sequential roles of Cdc42, Par-6, aPKC, and Lgl in the estab-
lishment of epithelial polarity during Drosophila embryogenesis.
Dev. Cell 6, 845–854.
Hyodo-Miura, J., Yamamoto, T.S., Hyodo, A.C., Iemura, S., Kusa-
kabe, M., Nishida, E., Natsume, T., and Ueno, N. (2006). XGAP, an
ArfGAP, is required for polarized localization of PAR proteins and
cell polarity in Xenopus gastrulation. Dev. Cell 11, 69–79.
Irvine, K.D., and Wieschaus, E. (1994). Cell intercalation during
Drosophila germband extension and its regulation by pair-rule
segmentation genes. Development 120, 827–841.
Jacinto, A., Wood, W., Woolner, S., Hiley, C., Turner, L., Wilson, C.,
Martinez-Arias, A., and Martin, P. (2002). Dynamic analysis of actin
cable function during Drosophila dorsal closure. Curr. Biol. 12,
Keller, R., Davidson, L., Edlund, A., Elul, T., Ezin, M., Shook, D., and
Skoglund, P. (2000). Mechanisms of convergence and extension
by cell intercalation. Philos. Trans. R. Soc. Lond. B Biol. Sci. 355,
Kiehart, D.P., Galbraith, C.G., Edwards, K.A., Rickoll, W.L., and
Montague, R.A. (2000). Multiple forces contribute to cell sheet
morphogenesis for dorsal closure in Drosophila. J. Cell Biol. 149,
participates in adherens junctions and polymerization of linear actin
cables. Nat. Cell Biol. 6, 21–30.
Kovacs, E.M., Goodwin, M., Ali, R.G., Paterson, A.D., and Yap, A.S.
(2002). Cadherin-directed actin assembly: E-cadherin physically
associates with the Arp2/3 complex to direct actin assembly in
nascent adhesive contacts. Curr. Biol. 12, 379–382.
Lecaudey, V., and Gilmour, D. (2006). Organizing moving groups
during morphogenesis. Curr. Opin. Cell Biol. 18, 102–107.
Martin, P., and Lewis, J. (1992). Actin cables and epidermal move-
ment in embryonic wound healing. Nature 360, 179–183.
Muller, H.A., and Wieschaus, E. (1996).armadillo,bazooka,and star-
dust are critical for early stages in formation of the zonula adherens
and maintenance of the polarized blastoderm epithelium in
Drosophila. J. Cell Biol. 134, 149–163.
Munro, E.M., and Odell,G.M. (2002). Polarizedbasolateral cell motil-
ity underlies invagination and convergent extension of the ascidian
notochord. Development 129, 13–24.
Nance, J. (2005). PAR proteins and the establishment of polarity
during C. elegans development. Bioessays 27, 126–135.
Nikolaidou, K.K., and Barrett, K. (2005). Getting to know your neigh-
Ninomiya, H., Elinson, R.P., and Winklbauer, R. (2004). Antero-
posterior tissue polarity links mesoderm convergent extension to
axial patterning. Nature 430, 364–367.
Nusslein-Volhard, C., Frohnhofer, H.G., and Lehmann, R. (1987).
Determination of anteroposterior polarity in Drosophila. Science
Oda, H., and Tsukita, S. (2001). Real-time imaging of cell-cell adhe-
rens junctions reveals that Drosophila mesoderm invagination
begins with two phases of apical constriction of cells. J. Cell Sci.
Oda, H., Uemura, T., Harada, Y., Iwai, Y., and Takeichi, M. (1994). A
Drosophila homolog of cadherin associated with armadillo and
essential for embryonic cell-cell adhesion. Dev. Biol. 165, 716–726.
Petronczki, M., and Knoblich, J.A. (2001). DmPAR-6 directs epithe-
lial polarity and asymmetric cell division of neuroblasts in Drosoph-
ila. Nat. Cell Biol. 3, 43–49.
Redd, M.J., Cooper, L., Wood, W., Stramer, B., and Martin, P. (2004).
Wound healing and inflammation: embryos reveal the way to perfect
repair. Philos. Trans. R. Soc. Lond. B Biol. Sci. 359, 777–784.
Riggleman, B., Schedl, P., and Wieschaus, E. (1990). Spatial expres-
sion of the Drosophila segment polarity gene armadillo is post-
transcriptionally regulated by wingless. Cell 63, 549–560.
Royou, A., Field, C., Sisson, J.C., Sullivan, W., and Karess, R. (2004).
Reassessing the role and dynamics of nonmuscle Myosin II during
furrow formation in early Drosophila Embryos. Mol. Biol. Cell 15,
Schock, F., and Perrimon, N. (2002). Molecular mechanisms of
epithelial morphogenesis. Annu. Rev. Cell Dev. Biol. 18, 463–493.
Shih, J., and Keller, R. (1992). Cell motility driving mediolateral inter-
calation in explants of Xenopus laevis. Development 116, 901–914.
Solnica-Krezel, L. (2005). Conserved patterns of cell movements
during vertebrate gastrulation. Curr. Biol. 15, R213–R228.
Tepass, U., and Hartenstein, V. (1994). The development of cellular
junctions in the Drosophila embryo. Dev. Biol. 161, 563–596.
Topczewski, J., Sepich, D.S., Myers, D.C., Walker, C., Amores, A.,
Lele, Z., Hammerschmidt, M., Postlethwait, J., and Solnica-Krezel,
L. (2001). The zebrafish glypican knypek controls cell polarity during
gastrulation movements of convergent extension. Dev. Cell 1, 251–
Verma, S., Shewan, A.M., Scott, J.A., Helwan, F.M., den Elzen, N.R.,
Miki, H.,Takenawa,T.,and Yap, A.S.(2004).Arp2/3activity is neces-
sary for efficient formation of E-cadherin adhesive contacts. J. Biol.
Chem. 279, 34062–34070.
Wallingford, J.B., Fraser, S.E., and Harland, R.M. (2002). Convergent
extension: the molecular control of polarized cell movement during
embryonic development. Dev. Cell 2, 695–706.
Weaire, D., and Rivier,N. (1984). Soap, cells,and statistics—random
patterns in two dimensions. Contemp. Phys. 25, 59–99.
Wei, S.Y., Escudero, L.M., Yu, F., Chang, L.H., Chen, L.Y., Ho, Y.H.,
Lin, C.M., Chou, C.S., Chia, W., Modolell, J., and Hsu, J.C. (2005).
Echinoid is a component of adherens junctions that cooperates
with DE-Cadherin to mediate cell adhesion. Dev. Cell 8, 493–504.
Wieschaus, E., Sweeton, D., and Costa, M. (1991). Convergence and
extension during germband elongation in Drosophila embryos. In
Gastrulation, R. Keller, ed. (New York: Plenum Press), pp. 213–223.
Wodarz, A., Ramrath, A., Kuchinke, U., and Knust, E. (1999).
Bazooka provides an apical cue for Inscuteable localization in
Drosophila neuroblasts. Nature 402, 544–547.
Wodarz, A., Ramrath,A., Grimm, A., and Knust, E.(2000).Drosophila
atypical protein kinase C associates with Bazooka and controls
polarity of epithelia and neuroblasts. J. Cell Biol. 150, 1361–1374.
Wood, W., Jacinto, A., Grose, R., Woolner, S., Gale, J., Wilson, C.,
and Martin, P. (2002). Wound healing recapitulates morphogenesis
in Drosophila embryos. Nat. Cell Biol. 4, 907–912.
Young, P.E.,Richman,A.M.,Ketchum,A.S., andKiehart,D.P. (1993).
Morphogenesis in Drosophila requires nonmuscle myosin heavy
chain function. Genes Dev. 7, 29–41.
Zallen, J.A., and Wieschaus, E. (2004). Patterned gene expression
directs bipolar planar polarity in Drosophila. Dev. Cell 6, 343–355.
Zallen, J.A., and Zallen, R. (2004). Cell-pattern disordering during
convergent extension in Drosophila. J. Phys.: Condensed Matter