Compartmentalized morphogenesis in epithelia: From cell to tissue shape


  • Fanny Pilot,

    1. Laboratoire de Génétique et de Physiologie du Développement (LGPD), UMR 6545 CNRS-Université de la Méditerranée, Institut de Biologie du Développement de Marseille (IBDM), Marseille, France
    Search for more papers by this author
  • Thomas Lecuit

    Corresponding author
    1. Laboratoire de Génétique et de Physiologie du Développement (LGPD), UMR 6545 CNRS-Université de la Méditerranée, Institut de Biologie du Développement de Marseille (IBDM), Marseille, France
    • Laboratoire de Génétique et de Physiologie du Développement (LGPD), UMR 6545 CNRS-Université de la Méditerranée, Institut de Biologie du Développement de Marseille (IBDM), Campus de Luminy case 907, 13288 Marseille cedex 09, Marseille, France
    Search for more papers by this author


During development, embryonic tissues are shaped in a species-specific manner. Yet, across species, general classes of tissue remodeling events occur, such as tissue infolding and tissue elongation. The spatiotemporal control of these morphogenetic processes is responsible for the organization of different body plans, as well as for organogenesis. Cell morphogenesis in a mesenchyme contributes to the shaping of embryonic tissues. Epithelial cells, despite that they need to maintain an apicobasal organization, play an equally important role during morphogenesis. Moving from apical to basal, we review compartmentalized cellular rearrangements underlying tissue remodeling in Drosophila and compare them with those found in other organisms. Contractile activity at the apical surface triggers tissue folding and invagination. The regulation of adhesion at adherens junctions controls polarized neighbor exchange during intercalation and tissue elongation. Basolateral protrusive activity underlies other cases of intercalation. These localized cell shape changes are spatially regulated by developmental signals. Some signals define a local change in cell behavior (e.g., apical constriction), others orient a dynamic process in the plane of the tissue (e.g., junction remodeling). Developmental Dynamics 232:685–694, 2005. © 2005 Wiley-Liss, Inc.


One of the most fascinating problems in developmental biology is understanding how tissues are shaped to produce highly organized embryos. Gastrulation, the stage during which the different germ layers are set aside, follows highly reproducible patterns of cell movements and rearrangements in each organism. Despite differences in the speed of the process (from 1 hr in Drosophila to ∼8 hr in Xenopus and more than 24 hr in the mouse), gastrulation shares similarities in different organisms. As the mesoderm and endoderm appear in deeper layers inside the embryo (through invagination or delamination), the whole embryonic axis elongates along the anterior/posterior (A/P) axis. An important realization has been that complex three-dimensional morphogenetic rearrangements stem from a very limited number of cell biological processes: loss of epithelial polarity during the epithelial to mesenchymal transition (EMT; see review in Shook and Keller, 2003), apical constriction of epithelial cells, cell intercalation (Keller, 2002; Myers et al., 2002; Wallingford et al., 2002), cell migration (Myers et al., 2002; Wallingford et al., 2002), and polarized cell division (Concha and Adams, 1998; Gong et al., 2004). The central problem, thus, becomes specifying the correct cell behavior in the appropriate group of cells. First, groups of cells are specified such that they follow a given morphogenetic pathway; second, the morphogenetic process is executed through changes in the cell architecture and/or polarity. In the course of executing a morphogenetic process, the cell often needs to orient its behavior in the plane of the whole tissue to which it belongs, be it cell division, cell migration, or cell intercalation. The orientation of such cell behaviors has a profound impact on the macroscopic rearrangements within the organism.

It is now an exciting time when cell biology, which tracks the intricacies of cell behavior at the molecular level in a whole organism, is uniting with developmental biology, which, more traditionally, identifies the signals that determine and orient specific classes of cell behaviors in groups of cells. Genetics, modern microscopy, and biochemistry join forces to explain how morphogenetic processes are executed and controlled in cells during development.

As already mentioned, cells in a developing organism may be epithelial, with a typical apicobasal polarized architecture (Knust and Bossinger, 2002; Nelson, 2003), or mesenchymal and, hence, without such polarity. Mesenchymal cells can migrate, thus contributing to morphogenesis, and in doing so, their behavior may be polarized: a good example is the polarized organization of mesenchymal cells undergoing directed cell migration (Myers et al., 2002) or cell intercalation (Keller, 2002). Epithelial cells need to maintain an apicobasal polarity. Despite this requirement, which might constrain their morphogenetic potential, they adopt a different repertoire of rearrangements, and also contribute to profound tissue reorganization events. Here, we review how the compartmentalization of subcellular rearrangements in epithelial cells accounts for the diversity of cell behaviors in epithelia. Apical constriction is required for tissue folding and invagination; remodeling of adherens junctions and basolateral protrusive activity controls cell intercalation and cell migration.


Together with the EMT, the invagination of epithelial tissues is a common process used to create multiple tissue layers. For example, invagination of epithelial tissues is important for the proper organization of the endoderm during gastrulation of Drosophila (Leptin and Grunewald, 1990; Sweeton et al., 1991), sea urchin (Anstrom, 1992; Davidson et al., 1995), and Xenopus embryos (Keller, 1981; Hardin and Keller, 1988). In Drosophila, the presumptive mesoderm invaginates along the ventral midline and separates from the ectoderm (Leptin and Grunewald, 1990; Sweeton et al., 1991; Fig. 1a). Neurulation in vertebrates also depends on the bending and invagination of the dorsal neuroepithelium (Schoenwolf and Franks, 1984; Schoenwolf and Alvarez, 1989; reviewed in Colas and Schoenwolf, 2001). The invagination of epithelial tissues relies on the polarized reorganization of the cells whose apical surface faces the outside of the embryo during gastrulation. Apical constriction leads to a stereotyped change in the cell shape from cuboidal to bottle-shape (Keller, 1981; Schoenwolf and Franks, 1984; Kam et al., 1991; Sweeton et al., 1991). The apical constriction of a group of cells creates a bending of the tissue and local infolding (Fig. 1a). In some cases, tissue folding leads to a complete invagination of the tissue (mesoderm, endoderm). In other cases, such as during the segmentation of the Drosophila embryo, the local folding of the tissue persists (Larsen et al., 2003).

Figure 1.

Apical constriction underlies Drosophila mesoderm invagination. a: Transverse section of a gastrulating embryo showing the ventral invaginating mesoderm (green). b: Infolding cells (red) undergo apical constriction, which correlates with the apical recruitment of Myosin II (black). Inset: ventral view of the apical surface of the invaginating cells showing the localization of a Myosin-II Light Chain green fluorescent protein fusion protein(black). HC, Heavy Chain; RLC, Regulatory Light Chain. c: Signals underlying the recruitment and activation of an apical actin–myosin contractile network in Drosophila and vertebrates. GPCR, G protein coupled receptor.

Apical constriction is linked to the specific recruitment of Myosin-II at the apical cell cortex together with F-actin as shown in Drosophila (Young et al., 1991; Fig. 1b), and chick embryos (Ferreira and Hilfer, 1993). It is believed that the contractile activity of an actin–myosin network reduces the apical cell perimeter. The apical surface either blebs as a result of constriction or is internally retrieved by endocytosis in parallel with constriction.

During early Drosophila gastrulation, the presumptive mesoderm and endoderm of the blastodermal epithelium invaginate, respectively, ventrally and posteriorly (a.k.a. mesoderm invagination and posterior midgut invagination, respectively). During Drosophila mesoderm invagination, it was shown that apical constriction requires activation of the small GTPase Rho by RhoGEF2 (Barrett et al., 1997). Rho may activate Rho-kinase Drok, a specific activator of Myosin-II Regulatory Light Chain (Winter et al., 2001) encoded in Drosophila by spaghetti-squash (sqh; Karess et al., 1991). Consistent with this explanation, the microinjection of the specific Rho-kinase inhibitor Y27632 in Drosophila embryos causes the down-regulation of Sqh recruitment at the apical cell surface in the presumptive mesoderm and posterior midgut anlage, and an inhibition of both mesoderm and posterior midgut invagination (T. Lecuit and C. Bertet, unpublished observations). This finding suggests that Rho signaling recruits and activates Myosin-II in the context of Drosophila apical constriction. In mammals too, apical constriction requires Rho signaling. Defects in mice mutant for p190RhoGAP show that Rho signaling is also required for neurulation (Brouns et al., 2000). Because p190RhoGAP is an inhibitor of Rho signaling and Rho-kinase a direct effector of Rho, this finding suggests that Rho signaling must be controlled precisely in the context of apical constriction. During Xenopus neurulation, however, Rho signaling may be dispensable (Haigo et al., 2003), suggesting that other signaling molecules might play a redundant function. Consistent with this view, another small GTPase, Rap1, was shown to be required for apical constriction during neurulation in Xenopus (Haigo et al., 2003). The use of parallel pathways to control apical constriction may in fact be more general, because Rap1 is also required for mesoderm invagination in Drosophila (Asha et al., 1999). Together, these results suggest that specific small GTPases may be required to a different extent in numerous examples of tissue folding, likely by ensuring the recruitment of an apical contractile network.

In vertebrates, the PDZ domain and F-actin binding protein Shroom may integrate signaling inputs involved in apical cytoskeletal organization and constriction. Shroom is essential for apical constriction during mouse (Hildebrand and Soriano, 1999) and Xenopus (Haigo et al., 2003) neurulation. Shroom expression in relatively naive Xenopus epithelial cells can induce apical actin accumulation and constriction in a cell-autonomous manner. Shroom may be part of a complex involved in anchoring Myosin-II apically during constriction. In Drosophila, in which no Shroom orthologue is found, it remains unclear which molecule(s) recruit(s) Myosin-II at the apical surface (Fig. 1c).

What developmental signals define which cells undergo apical constriction? This question has received the most complete answer in the context of Drosophila mesoderm and posterior midgut invaginations. The secreted molecule encoded by folded-gastrulation (fog) is both necessary and sufficient to trigger apical constriction in epithelial cells (Costa et al., 1994). fog expression is induced zygotically in the presumptive mesoderm as well as in the posterior midgut primordium. The localized expression of fog activates the G-alpha like protein Concertina (Parks and Wieschaus, 1991) by means of an as yet unknown transmembrane receptor and leads to the activation of RhoGEF (Fig. 1c). Experiments in which fog is ubiquitously expressed, however, show that apical constriction must be localized to a group of cells to induce furrowing and invagination (Morize et al., 1998). Apical constriction may indeed create a local instability in the epithelium that initiates the process of furrowing. In vertebrates, it is still unclear what triggers the localized expression of shroom, which in turn triggers local apical constriction and tissue bending. Studies on mouse neural tube closure suggest that the combination of an inducing signal from the adjacent ectoderm and of an inhibitory signal mediated by Shh expressed in the underlying notochord defines dorsolateral bending points in the neuroepithelium (Ybot-Gonzalez et al., 2002).


Intercalation in a Mesenchyme and in an Epithelium

Tissue morphogenesis can occur in the absence of cell shape changes and oriented cell division and can rely exclusively on cells exchanging places with their neighbors (Keller, 2002). When this process is spatially oriented, it is called intercalation and is responsible for the elongation of many embryonic tissues both in vertebrates and invertebrates. Cell intercalation along the mediolateral axis of an embryo causes the convergence of cells from either side of the embryo toward the midline and is responsible for the extension of the embryonic tissue along the anteroposterior axis. Developmental biologists have called this universal process convergence–extension (CE; Keller et al., 1992, 2000). In some cases, such as during gastrulation of the zebrafish embryo, CE also depends on polarized cell migration (Myers et al., 2002).

Intercalation in mesenchymal tissues has been well documented in vertebrate embryos such as zebrafish, Xenopus, and chick (Keller et al., 1992; Shih and Keller, 1992; Davidson and Keller, 1999; Heisenberg et al., 2000; Lawson and Schoenwolf, 2001; Topczewski et al., 2001; Jessen et al., 2002; Marlow et al., 2002). In this case, intercalation relies on monopolar or bipolar protrusive activities of the cells, which support their migration and the formation of adhesive contacts with new neighboring cells (Brieher and Gumbiner, 1994; Keller, 2002; Wallingford et al., 2002). Polarization of the protrusive activity ensures directionality of CE and relies on the Planar Cell Polarity (PCP) pathway (Wallingford et al., 2000).

Intercalation also occurs in epithelial cells and accounts for many morphogenetic processes during embryonic development. During gastrulation, intercalation has been described in CE of the primary epithelium of the Drosophila embryo during germ-band elongation (Irvine and Wieschaus, 1994; Bertet et al., 2004; Zallen and Wieschaus, 2004); of the dorsal ectoderm of Caenorhabditis elegans (Williams-Masson et al., 1998); and of the notochord of ascidians (Munro and Odell, 2002b), zebrafish (Glickman et al., 2003), and chick (Schoenwolf and Alvarez, 1989). At later stages of development, medial edge epithelial cells also undergo intercalation during mouse palate fusion (Tudela et al., 2002). During organogenesis, cell intercalation is also important in shaping tissues: gut morphogenesis in C. elegans (Leung et al., 1999) and Drosophila (Iwaki et al., 2001; Lengyel and Iwaki, 2002; Johansen et al., 2003), posterior spiracle morphogenesis (Hu and Castelli-Gair, 1999; Brown and Castelli-Gair Hombria, 2000), and the formation of the dorsal branches of the respiratory system known as the tracheal system, in Drosophila (Ribeiro et al., 2002; Affolter et al., 2003; Jazwinska et al., 2003). In these latter cases, cell intercalation occurs in epithelial tubes and causes a reduction in the tube circumference and a corresponding lengthening along its main axis.

At first sight, intercalation in epithelial cells poses a paradox: how can cell movements occur within a cohesive tissue formed of strongly adhesive cells? At least three options can be envisioned to account for cell intercalation of epithelial cells. Cells could transiently down-regulate adhesion with their neighbors and delaminate from the epithelium, migrate, and restore their epithelial properties at another location. Alternatively, cells could maintain their epithelial behavior and adopt a transient migratory behavior in between other epithelial cells. Finally, the epithelial cells could simply remodel specific contacts in an ordered directional pattern so that they progressively exchange places with their neighboring cells.

Planar Polarized Junction Remodeling and Intercalation in an Epithelial Sheet

This latter scenario was shown recently to occur during anteroposterior elongation of the Drosophila embryo, called germ-band elongation, hereafter GBE (Bertet et al., 2004; Zallen and Wieschaus, 2004). In this case as well as in the majority of epithelia, epithelial cells are arranged in a honeycomb pattern in which cells minimize cell surface contacts and optimize packing. GBE results in an approximate doubling in length of the A/P axis of the Drosophila embryo. The epithelial sheet folds back dorsally at the posterior end of the embryo (Fig. 2a, arrow). This process relies exclusively on cell intercalation and not on cell shape changes or cell division. Time-lapse recording of the cell adherens junctions (AJ) marked with an E-cadherin–green fluorescent protein fusion protein, reveals that cells do not delaminate and do not migrate across the tissue but, instead, remodel specific contacts after an ordered spatial–temporal pattern of junction remodeling (Fig. 2b; Bertet et al., 2004).

Figure 2.

Intercalation during Drosophila germ-band elongation involves polarized junction remodeling in the plane of the tissue. a,b: Early gastrulating embryo undergoing gem-band elongation. The intercalating epithelial sheet (red) elongates along the anteroposterior axis and folds back dorsally (white arrow). The arrowhead marks the posterior tip of the germ-band, whereas the cephalic furrow (b, CF) marks the anterior limit. c: Time-lapse recording of embryos expressing an E-cadherin–green fluorescent protein fusion protein reveals a polarized, irreversible remodeling of junctions from a type-1 to a type-3 geometrical conformation, leading to intercalation as cells exchange neighbors. Notice the enrichment of Myosin II in type-1 junctions (orange), which is required for this process. d: The segmental expression of the pair rule gene product Even-skipped (Eve) is disrupted in a Krüppel mutant. In this mutant, the polarized remodeling of cell junctions during intercalation is strongly reduced and the polarized enrichment of Myosin-II in type-1 junctions no longer happens. Cell identity, as monitored by the expression of Eve (red), Odd-skipped (Odd, green), and Hairy (blue) is lost in this mutant (insets). (Confocal images are generous courtesy of David Kosman and John Reintiz.) e: Models accounting for the role of Myosin II in junction remodeling. Top: Clustering of E-cadherin complexes. Middle: Enhancement of E-cadherin endocytosis. Bottom: Down-regulation of E-cadherin stabilizing proteins.

In each hexagonal cell, the pair of contacts with adjacent anterior and posterior neighboring cells shrink so the cell contacts progress from a so-called type-1 to type-2 configuration (Fig. 2b). In a second step, new junctions form between contacting dorsal and ventral neighbors, leading to a so-called type-3 configuration (Fig. 2b). The cell perimeter does not change during this process. Thus, intercalation and the associated elongation of the embryonic tissue, boil down to a simple, irreversible pattern of junction remodeling, whereby each cell undergoes the shrinkage of a set of junctions and the subsequent expansion of another (Bertet et al., 2004). The whole process is relatively rapid, between 20–25 min. This polarized behavior reveals that each intercalating cell “knows” where the head and the tail of the embryo are, because only the contacts with anterior and posterior neighbors are rearranged. The process of junction remodeling, thus, is polarized in the plane of the epithelium. The findings show that spatially oriented cell movements can result from subcellular rather than global changes in cell adhesion.

Morphogenetic processes at the boundaries of the elongating germ-band might pull the tissue and orient cell intercalation. However, in fog mutants in which both the mesoderm and the posterior midgut fail to invaginate, intercalation occurs, suggesting that cells can remodel cell contacts in the absence of morphogenetic processes at the tissue boundaries. This explanation suggests that external forces do not play an important role and that, instead, intercalation is controlled by local mechanisms operating at cell boundaries (Bertet et al., 2004).

Strikingly, Myosin-II was shown to localize at cell junctions with a planar polarized pattern (Bertet et al., 2004; Zallen and Wieschaus, 2004). Myosin-II localizes first uniformly at AJs but is subsequently enriched in the type-1 junctions at a time when junction remodeling and intercalation begin. Myosin-II is still enriched at the vertex of type-2 junctions but localizes at a lower level in expanding type-3 junctions (Fig. 2c). This polarized localization is essential for junction remodeling. In embryos mutant for zipper, which codes for Myosin-II heavy chain (Kiehart, 1990; Young et al., 1993), junction remodeling is as if frozen and intercalation fails to occur (Bertet et al., 2004). This finding results in a defect in embryonic axis elongation. Similar observations are made in embryos injected with Y-27632, a pharmacological inhibitor of Rho-kinase (Uehata et al., 1997; Narumiya et al., 2000), a specific regulator of Myosin-II regulatory light chain (Winter et al., 2001). Myosin-II, thus, is required for the type-1 to type-2 junction remodeling event. The irreversibility of the evolution from the type-2 to type-3 configuration might rely on the persistent destabilizing action of Myosin-II toward the type-1 configuration and on the intrinsic properties of cells to maintain or expand adhesion contacts where possible (Bertet et al., 2004).

How does Myosin-II control the polarized down-regulation of cell adhesion in type-1 junctions? Three mechanisms can be envisioned (Fig. 2e). Myosin-II could force the clustering of E-cadherin in the plasma membrane by the local contraction of the actin cytoskeleton at AJs. Alternatively, Myosin-II could trigger the polarized endocytosis of E-cadherin by changing the recruitment of actin filaments involved in endocytosis. Both mechanisms rely on the notion that E-cadherin is present at the plasma membrane in a stable form that needs to move or to be removed to rearrange adhesion. A third model refers to a more realistic scenario in which E-cadherin is present in a very dynamic fraction, stabilized by specific regulators, such as the component of the Par3-Bazooka/Par6/aPKC complex (Kuchinke et al., 1998; Wodarz et al., 2000; Petronczki and Knoblich, 2001; Knust and Bossinger, 2002; Nance and Priess, 2002). In the absence of Myosin-II or before Myosin-II is enriched in type-1 junctions, all cell contacts are equally stabilized leading to an equilibrium with six junctions of approximately equal length. Myosin-II might locally destabilize adhesion and place the type-1 junction at a disadvantage with the other one, thus causing it to shrink toward the type-2 configuration. The observation that the PDZ domain protein Par3/Bazooka localizes in a complementary domain to Myosin-II (Zallen and Wieschaus, 2004) lends support to this model. The local down-regulation of Par3/Baz indeed may cause the local destabilization of E-cadherin–based adhesion. However, it should be noted that the mechanism whereby Par3/Baz controls the stable localization of E-cadherin at AJs is unclear. Future work is expected to shed light on the mechanism underlying Myosin-dependent destabilization of adhesion.

This pattern of junction remodeling is simple enough to suggest that it could well underlie intercalation of other hexagonal epithelial tissues. The precise mechanism might be different, but the pattern of junction remodeling is likely to be conserved.

Coupling Embryonic Polarity and Planar Cell Intercalation

Which signals orient adhesion remodeling and intercalation in whole organisms? For instance, during Drosophila GBE, junction remodeling is polarized along the A/P axis of the embryo, and A/P patterning is required for A/P axis elongation (Irvine and Wieschaus, 1994; Bertet et al., 2004; Zallen and Wieschaus, 2004) as it is in the Xenopus mesoderm (Ninomiya et al., 2004). In a Drosophila embryo mutant for Krüppel, a gap gene that is required, along with other gap genes, to restrict the segmental expression of many pair-rule genes (including even-skipped [Clyde et al., 2003], odd-skipped, and hairy), cell intercalation is very strongly reduced during GBE and is occasionally reversible (Fig. 2d). In both Krüppel and even-skipped (eve) mutants, Myosin-II localization is no longer enriched in type-1 junctions (Bertet et al., 2004; Zallen and Wieschaus, 2004). These data show that the restricted expression of pair-rule genes such as eve is essential to initiate the polarized enrichment of Myosin-II at cell borders. It is still unclear, however, how polarizing signals downstream of the transcription factor Eve control this process.

In other cases, specific polarity signals have been suggested to spatially bias the activity of the PCP pathway. The PCP pathway orients a variety of planar polarized morphogenetic processes, such as mesenchymal cell intercalation, the polarized location of hairs in Drosophila, the chiral organization of photoreceptor cells in the Drosophila ommatidia, and the stereociliate cells in the vertebrate cochlea (Tree et al., 2002a; Fanto and McNeill, 2004). The core PCP pathway consists of the seven-pass transmembrane protein family of G-protein coupled receptors Frizzled (Fz); the adhesion protein Flamingo/Starry-night (Fmi/Stan); and the cytoplasmic proteins Dishevelled (Dsh), Prickle (Pck), and Strabismus (Stbm). During vertebrate mesenchymal cell intercalation, Wnt secreted molecules such as Wnt11 and Wnt5 have been implicated upstream of Fz (Heisenberg et al., 2000; Topczewski et al., 2001; Kilian et al., 2003). In Drosophila imaginal discs, an epithelial tissue, a feedback loop mechanism that causes the asymmetric recruitment of Fz, Dsh, Pck, Fmi, and Stbm at opposite poles of the cells (Strutt et al., 2002; Tree et al., 2002b) is spatially biased by the protocadherin molecules Dachsous and Fat as well as Four-jointed, all of which are expressed in shallow gradients across different Drosophila epithelial tissues (Casal et al., 2002; Lawrence et al., 2002; Yang et al., 2002; Fanto et al., 2003; Matakatsu and Blair, 2004). Recent experiments in which Fat is expressed uniformly across the wing epithelium, suggest that additional mechanisms can polarize epithelial cells in the plane of the tissue (Matakatsu and Blair, 2004). Significantly, the PCP pathway does not appear essential during Drosophila GBE, as Fz, Fz2, and Dsh are dispensable for the process (Zallen and Wieschaus, 2004). This finding suggests the existence of different or additional signals. This explanation does not rule out the possibility that, as in the PCP pathway, the source of the polarity consists of an upstream graded activity linked to the A/P patterning of the embryo and of a symmetry breaking process at the cell junctions. However, such polarity is presumably established by means of a distinct molecular mechanism.

Junction Remodeling and Intercalation in an Epithelial Tube

Intercalation occurs in quite different cellular contexts, where both cell shape and tissue topology differ significantly from the simple hexagonal organization of epithelial sheets reported above. A remarkable example of this is the intercalation underlying tubular elongation in the Drosophila respiratory system, the tracheal system (Fig. 3a; Jazwinska et al., 2003; Ribeiro et al., 2005). The dorsal branches initially consist of adjacent pairs of epithelial cells, their apical surfaces forming the lumen of the tube. The dorsal elongation of the branches depends on intercalation and results in the formation of chains of cells, each cell forming a ring with the apical surface covering the lumen around the full circumference of the tube. Although the process is more complicated than in the germ-band, it is striking to see that, here again, intercalation relies on the remodeling of cell junctions following a stereotyped pattern of junction shrinkage followed by junction expansion. Some of the cell contacts undergo an irreversible change from a type-1 toward a type-2 and type-3 configuration (Fig. 3b). The shrinking type-1 junction is inter-cellular, whereas the expanding type-3 junction is auto-cellular.

Figure 3.

Intercalation in the Drosophila tracheal system. a: Dorsal branches of the embryonic tracheal system elongate dorsally, through intercalation of the epithelial tube cells and pulling of the leading tip cells (asterisks, top, adapted with permission from Jazwinska et al., 2003). b: Observation of the apical surface of the intercalating cells reveals that a stereotyped junction remodeling event with intercellular junctions (type-1) shrinkage and subsequent autocellular (type-3) expansion transforms a bicellular into a unicellular epithelial tube.

In the case of the tracheal system, the irreversibility of the process depends, to a large extent, on the pulling force exerted by the migratory behavior of the leading tip cells at the dorsal most region of the elongating dorsal branch (Fig. 3a, asterisk; Ribeiro et al., 2002, 2005). The remarkable property of this process is that the tube architecture is preserved during intercalation and the lumen remains intact. Maintaining the lumen requires precise coordination of junctional remodeling and changes in cell position.

Elongation of leg imaginal discs of Drosophila during metamorphosis is another case of dramatic epithelial remodeling. Major cell shape changes have been described that could account for elongation (Condic et al., 1991). Cell intercalation may also drive this process as suggested by reports of changes in the geometrical organization of cell contacts in imaginal disc cells (Fristrom, 1988). In addition to differences in the precise source of polarity underlying intercalation, hexagonal intercalation and the more original pathway in the tracheal epithelial tube boil down to an irreversible and polarized remodeling of cell adhesion consisting of a first phase where a subset of junctions shrink, followed by a phase of junction expansion.


Another common case of epithelial intercalation, to date not reported in Drosophila but in other organisms, is associated with a specific cell behavior, namely basolateral cell protrusions. This process concerns, for instance, intercalation in the notochord of ascidians, amphibians, and fish during the A/P elongation of the embryo during gastrulation. In these organisms, the notochord is an epithelium (Munro and Odell, 2002a, b; Crawford et al., 2003). Such intercalation also occurs in C. elegans, during dorsal intercalation of the epidermis (Heid et al., 2001) as well as during gut organogenesis. It is very likely that, with the emergence of new visualization methods in whole embryos, other examples representative of this intercalation pathway will be found.

As shown on Figure 4 in the case of dorsal epidermal intercalation in the worm, intercalation occurs between two adjacent rows of cells. Pairs of trapezoidal cells change their shape to become wedges and expand contacts between themselves both at the AJs area and along the basolateral surface. Cells slide between each other until the cells now straddle the dorsoventral axis.

Figure 4.

Basolateral surface dynamics during Caenorhabditis elegans dorsal epidermal intercalation. a: Dorsal view of a C. elegans embryo epidermis undergoing intercalation along the dorsal midline. Insets: note the cell shape changes. (Data from Jeff Hardin, adapted with permission from Heid et al. [2001].) b: Schematic representation of cell shape changes accounting for intercalation. The intercellular junction along the dorsal midline dramatically expands and finally becomes an anteroposterior junction. c: A distinct protrusive activity is apparent along the basolateral surface, which could account for the oriented new intercellular contacts.

How might adhesion be remodeled to account for this process? The intercellular junction along the dorsal midline (Fig. 4a,b, interface between the green and purple cells) expands significantly during this process without major change in cell perimeter, suggesting that adhesion may be reinforced here, at the expense of other cell contacts. However, this possibility is not supported by any observation. Studies in different organisms in which this type of intercalation occurs have suggested alternative mechanisms of rearrangements (Williams-Masson et al., 1998; Heid et al., 2001; Munro and Odell, 2002a, b) as the basolateral cell surface contains several membrane extensions supported by actin filaments, projecting toward the adjacent cells. This protrusive activity may allow adhesive contacts with neighboring cells and could reflect a form of migratory behavior that would orient cell movement (Fig. 4b,c). A polarized remodeling of cell adhesion contacts could act in parallel to a protrusive pulling activity to account for cell movement. Alternatively, apical adhesion contacts could passively follow cell cytoskeleton deformation induced by the protrusive activity. A mutant situation showed that the protrusive activity along the basolateral surface of the cells is not sufficient to drive cell movement but that this process must be coupled to the translocation of the cell body to cause the net movement of the cell (Heid et al., 2001). The analysis of the die-1 mutant, which codes for a zinc-finger transcriptional regulator and which is required for C. elegans dorsal intercalation reveals that cell intercalation can be blocked, despite normal protrusive activity in epithelial cells (Heid et al., 2001). In the die-1 mutant, the nuclear translocation of intercalating cells is affected, suggesting that a global reorganization of the whole cell may be also required for intercalation to occur.


Mesenchymal cells play an important role in development. Their unpolarized nature endows them with extensive mobility. And indeed, cell migration is essential during several developmental processes such as neural crest cell development and mesoderm morphogenesis. Mesenchymal cells can intercalate by means of a process akin to cell migration. In contrast, epithelial cells display a typical apicobasal polarity, which is essential to ensure the proper architecture of the tissues. Through the formation and consolidation of strong adhesive interactions at cell junctions, epithelial cells are constrained in their ability to move, thereby limiting the possibility of tissues to change shape appropriately during development. However, we have shown in this review that, in fact, far from being constrained in their morphogenetic potential, epithelial cells engage in a large number of morphogenetic rearrangements. This property stems, we argue, from the compartmentalized organization of epithelial cells, allowing differential remodeling of the apical surface, the junctional area, or the basolateral surface. If we consider that different signals could polarize the remodeling of each of these membrane domains, it becomes apparent that epithelial cells can progress through complex and numerous morphogenetic pathways, while keeping intact the integrity of epithelial tissues.

Observing cells in their native environment, that of a whole, living organism, illustrates vividly this notion and shows how important it is to study cell behavior and morphogenesis during development. An accurate account of how, at the molecular level, a given membrane domain can be remodeled, together with a knowledge of how this can be oriented in a tissue will provide the essential building blocks to understand how tissues are shaped at a macroscopic level.


We thank David Kosman and John Reinitz for providing us with the confocal images of embryos shown in Figure 2d. We acknowledge the very precious suggestions of an anonymous reviewer to improve an earlier version of this manuscript. We thank members of our lab for discussions and Steve Kerridge for comments on the manuscript. Research in our lab is supported by an ATIP grant from the CNRS. F.P. is supported by the Académie de Médecine.