Invagination of Ectodermal Placodes Is Driven by Cell Intercalation-Mediated Contraction of the Suprabasal Tissue Canopy

Ectodermal organs such as teeth, hair follicles, and mammary glands begin their development as placodes. These are local epithelial thickenings that invaginate into mesenchymal space. There is currently little mechanistic understanding of the cellular processes driving the early morphogenesis of these organs and of why they lead to invagination rather than simple tissue thickening. Here, we show that placode invagination depends on horizontal contraction of superficial layers of cells that form a shrinking and thickening canopy over underlying epithelial cells. This contraction occurs by cell intercalation and is mechanically coupled to the basal layer by peripheral basal cells that extend apically and centripetally while remaining attached to the basal lamina. This process is topologically analogous to well-studied apical constriction mechanisms, but very different from them both in scale and molecular mechanism. Mechanical cell–cell coupling is propagated through the tissue via E-cadherin junctions, which in turn depend on tissue-wide tension. We further present evidence that this mechanism is conserved among different ectodermal organs and is, therefore, a novel and fundamental morphogenetic motif widespread in embryonic development.


Introduction
Understanding how tissues form physically (morphogenesis) is a major frontier, both in developmental biology and organ regeneration using stem cells [1,2]. Epithelial bending, especially invagination, is a recurrent morphogenetic event in development [3][4][5][6][7][8][9][10][11][12], but our knowledge of the underlying cellular mechanisms is quite limited. Ectodermal organs, such as teeth, hair follicles, and mammary glands all depend on epithelial bending at the start of their development: local epithelial thickenings (placodes) must invaginate into mesenchymal space for proper organ shaping (Fig 1A-1G) [13]. While molecular signaling involved in placode inductions is well described and well conserved [14,15], less attention has been given to the physical events required to execute this programme. The current view is that placodes thicken and invaginate either through vertically orientated cell divisions, in the case of teeth [16], and/or by centripetal cell migration, in the case of hair follicles [17]. However, neither of these processes as such can explain why the epithelium invaginates ( Fig 1D) rather than causing cells to pile up, merely thickening the epithelium (Fig 1B and 1C).
One plausible hypothesis is that cell shape changes in the basal cell layer (cells in contact with the basal lamina) drive invagination (Fig 1D). Such monolayer invagination occurs in other well-studied contexts either by apical constriction, in which actinomyosin contractile fibers narrow the apical ends of cells, making them wedge-or cone-shaped [5,7,[9][10][11], or "basal wedging" (observed in vertebrate neural tube formation), in which nuclei move basally in a pseudostratified epithelium to expand the basal area of highly columnar cells (S1A and S1B Fig) [3,10]. We set out to determine whether these processes or others might be the mechanism of placode invagination. We found that a third, entirely novel mechanism-suprabasal cell intercalation-is responsible.

Results
To determine whether apical constriction or basal wedging contributes to placodal invagination, we analysed cells in mouse molar tooth primordia (Fig 1H-1J). We labeled cells mosaically using tamoxifen-inducible membrane-GFP and quantified apical and basal widths of basal layer cells in invaginated versus flanking flat epithelium (S1C-S1I Fig). Cells with basal nuclei were not more abundant, nor did they have significantly wider bases, in the invaginating epithelium versus adjacent flat epithelium (S1D and S1E Fig), ruling out the basal wedging mechanism. These cells did exhibit apical narrowing (S1E and S1F Fig; note that ratio of means is not equal to the mean of ratios) but, surprisingly, had no apical enrichment of actin filaments or activated (phospho-) myosin (pMLC) (S1G-S1I Fig), the hallmarks of apical constriction [5,7,[9][10][11]. Instead, pMLC appeared elevated in horizontally elongated suprabasal cells (S1G Fig), as compared to underlying and neighbouring basal cells, suggesting that the superficial layers of the placode may be under contractile tension. Contractile suprabasal tissue drives bending of the tooth placode. A-G. Epithelial invagination is an early and critical step in the development of ectodermal organs like the tooth, hair follicle, and mammary gland. Locally increased proliferation or vertically orientated cell divisions (vertical arrows in B-D), which are textbook models of tooth invagination, do not explain why the resulting structure should invaginate (D) versus thicken (C) or evaginate (B). I.e., local stratification alone cannot explain the observed morphogenesis, specifically, why the basal profile bends to become concave. Active invagination of the basal cell layer Suprabasal tissue in the tooth placode undergoes gross narrowing and deepening during invagination (red dots in Fig 1I and 1J), bringing together "shoulders" on either side of the invagination (yellow arrowheads in Fig 1I and 1J) to make a tooth bud with a narrow neck ( Fig  1J). Tissue narrowing and lengthening, if active and autonomous, is termed convergent extension (CE), a highly conserved morphogenetic movement that simultaneously produces extending and contractile forces through coordinated cell rearrangements and drives processes ranging from early embryo axial extension to organ tubule extension [5,[18][19][20]. To test whether suprabasal CE drives placode invagination, we first determined whether the invagination of the epithelium is mechanically independent of underlying mesenchyme by explanting it. Explants rapidly bent themselves in the direction of normal invagination (S1 Movie), showing that mesenchyme is not a mechanical requirement for epithelial bending. A partial excision ("lateral cut") was sufficient to show that a 40°bend could take place in as little as 150 s (compared to 12 h in vivo) (Fig 1K and 1L; S2 Movie), indicating that bending is normally resisted by the attachment to flanking epithelium and underlying mesenchyme. The lateral cut, by releasing this resistance, allowed rapid elastic recoil from the incision site. Quantification showed the angle made by the epithelium after bending to be significantly less than the angle before cutting (pvalue < 0.001, t test; Fig 1Q). Similar explants of nasal epithelium (a nearby noninvaginating pseudostratified epithelium) at E12.5 showed no bending behaviour. Manual re-extension and release of the bent epithelium graphically demonstrated its elastic nature (S1 Movie).
To determine whether the contractile forces driving bending are generated in the suprabasal layers specifically, we made fine incisions through these layers, leaving basal cells intact ( Fig  1M and 1N; S2G-S2I Fig). We found that incisions immediately expanded by bidirectional recoil from 17 +/-2 μm wide at the start of recording to 27 +/-4 μm in less than one min ( Fig  1M, 1N and 1R; S3 Movie). Similar incisions in nasal epithelium from littermate embryos did not recoil upon cutting (Figs 1R, S2A and S2B; S3 Movie). This confirmed that tensile forces in the plane are not a general property of thickened epithelium. After incubation for a further 4 min, cut explants gradually adopted a "W" shape as the basal epithelium flopped open below the incision (S2C- S2F Fig), showing that the intact basal layer below the cut was unable to sustain the bend at this position. To test directly the insufficiency of the basal layer and requirement for suprabasal tissue, we released suprabasal tension with a suprabasal cut and then followed it by a lateral cut. If the suprabasal contractile force is required for invagination, then releasing it should result in loss of lateral cut-induced bending. This is what we observed, with no significant change in epithelial bend (p-value = 0.11; t test; Fig 1O-1Q; S2 Movie). Thus, intact, contractile suprabasal tissue is required for bending, providing force for bending-associated apical narrowing of basal layer cells extrinsically.
In convergently extending tissues in other systems, contractile forces are produced by the active intercalation of elongated neighbouring cells between one another [5,19,21]. To observe dynamic cell behaviours in the tooth placode, we made confocal movies of tooth primordia under conditions that faithfully recapitulate normal tooth morphogenesis [22]. This confirmed that suprabasal cells converge toward the middle of the placode as it invaginates (S4 and S5 Movies), where they undergo active intercalation, sliding over one another at many suprabasal locations (Fig 2 and S4 Movie). Often, intercalating cells display highly dynamic, lamelliform protrusions (S4 Movie). We examined multiple labelled cell pairs in different specimens and observed widespread cell intercalation as revealed by the significantly changing planar distance between cell centroids (p-value < 0.025; t test; Fig 2R). This demonstrates that the convergence of the suprabasal cell tissue of the tooth is associated with cell intercalation. However, in sharp contrast to CE in Xenopus mesoderm [5], and the recently described CE in mouse prenatal eyelid closure [23], both of which are fibronectin (FN)-enriched and FN-dependent, we did not find detectable levels of FN in suprabasal or basal placode tissue (consistent with a number of published papers and databases), indicating that placode invagination is mechanistically distinct from these known instances of CE.
We next asked how this cellular behaviour of dynamic cell intercalation is able to generate force at the tissue level. To address this, and to further understand the molecular agents that generate force, we carried out a series of experiments using chemical inhibitors against components of the actinomyosin contractility machinery. Both an intact F-actin cytoskeleton and myosin contractility are required to generate lamelliform protrusions and dynamic cell motility observed in intercalating cells [24][25][26]. Brief culture of tooth slices in the presence of cytochalasin and blebbistatin inhibited bending upon lateral cutting (Fig 3A-3F and 3M). Such loss of tension across the suprabasal cells was accompanied by a loss of elongated cell shapes (Fig 3G-3L), commonly associated with cells engaged in force-producing intercalation movements [24]. This demonstrated that, upon inhibition of the molecular agents known to generate tension and elongated cell shapes in a subcellular level, a concomitant loss in tissue-wide tension is observed, thereby establishing a link between cell intercalation behaviour and tissue tension.
We next asked if the cells at the tooth "shoulders" couple suprabasal contraction to the basal layer and analysed their shapes and behaviour. These were dramatically elongated cells that extended apical protrusions that leaned toward the midline of the invagination (outlined in yellow, Fig 4A). They also exhibited dynamic apical protrusions, which directly interacted with lamelliform protrusions of intercalating suprabasal cells (Fig 2N-2Q and S6 Movie) (unlike more central cells, which emerged apically in a rounder shape (Fig 2J-2M and S6 Movie). These apical protrusions appeared to contain E-cadherin (Fig 4B-4D), a normally suprabasal marker, despite the basal anchoring of the shoulder cells. These observations show that  peripheral basal cells participate at their apical ends in the adhesive and intercalatory processes of suprabasal tissue. The existence of these cells reconciles the appearance of cells actively leaving the basal layer and yet still exerting a force on it: they are delaminating despite the traction of their basally-anchored neighbours, presumably pulling themselves up by their apical/centripetally-pointing ends while being pulled on by their basal ends.
Cadherin junctions are well known to transmit forces within epithelia [9,10,12,20], but linkage to tension in highly motile, suprabasal placodal tissue has never been examined. In vitro, the size of E-cadherin adherens junctions increases when tensile force is applied [27,28], implying that releasing tension should shrink junctions. Accordingly, we found that upon releasing tissue-wide tension by lateral cutting, suprabasal E-cadherin puncta (spot adherens junctions) detected by an antibody specific to the E-cadherin extracellular domain diminished to background levels within 30 min (Fig 4E, 4F, 4I and 4J), suggesting junction disassembly. Phalloidin staining was also reduced around the cut, indicating some local disassembly of actin filaments, but remained detectable, showing generally intact tissue. The suprabasal tissue remained intact and E-cadherin puncta reappeared after 2-4 h (Fig 4G, 4H, 4K and 4L), indicating that lateral cuts do not result in tissue death or dissociation. Assuming only that E-cadherin focally concentrated at cell interfaces-detected as we did using an antibody against the extracellular domain-is involved in cell adhesion, these findings together show the presence of a feedback loop between cell-cell adhesion and tissue-wide contractility in vivo. To begin to test whether E-cadherin adhesion might be required, we treated explants briefly (10 min) with EGTA, a calcium chelator that dissociates cadherin adhesions, which are calcium-dependent. This treatment did not lead to tissue dissociation but inhibited tissue recoil following a lateral cut ( Fig  4M-4Q). This is consistent with an E-cadherin requirement, although it does not exclude other adhesive mechanisms.
Nuclear deformation is another widely useable in vivo force sensor, because relaxed nuclei are spherical and require external forces to become ellipsoid [29,30]. Consistent with this, we observed that, in intact tooth placodes, nuclei of superficial cells had high aspect ratios in explants ( Fig 5A) and in vivo in the direction of cell intercalation (but not in the orthogonal-zaxis; observe "lentil" cell shapes demonstrated in 3D in S7 Movie). Crucially, these shapes rapidly relaxed to a rounded shape upon tissue tension release following a lateral or suprabasal cut (Fig 5B). Consistent with this, cytochalasin and blebbistatin treatments that inhibited contractile tension also resulted in rounder nuclear shape (Fig 3G-3L). Interestingly, in 3D, the nuclei in vivo were lentoid (i.e., extended along the jaw as well as mediolaterally), consistent with centripetal tension around the whole placode (S6 Movie). Shoulder cells, too, showed tensiondependent elongation that was not only precisely correlated with invagination, but also relaxed upon lateral cutting, albeit more slowly than suprabasal cells (S3A-S3F Fig). At early stages of invagination (E11.5-E12.0) when any cells emerging from the basal layer to move toward the centre of the placode must "turn a corner," we also observed a few, but striking, "banana"shaped nuclei (Fig 5C), which relaxed within 30 min of a lateral cut (Fig 5D). This relaxation occurred on both sides of the invagination, indicating that tissue tension is propagated across the prospective tooth epithelium even before substantial placodal thickening has occurred. All the above nuclear deformation observations and functional relaxation experiments together demonstrate a direct link between cell-level tension and tissue-level contraction, thus mapping the forces involved.  The characteristic pattern of nuclear deformations allowed us to compare tooth primordia with forming hair follicles and mammary glands. Sectioning in the vertical plane revealed that all these invaginating structures contain the same patterns of nuclear elongation: arcs of cells with elongated nuclei linking suprabasal cells with inward-leaning basally attached shoulder cells (Fig 5E-5H). Cell tension mapping by nuclear deformation thus indicates that there is a tensile canopy of cells anchored at its periphery by a region of basal cells attached to the basal lamina in all these invaginating placodal structures. These findings suggest a new interpretation of the previously reported live imaging of centripetal cell migration in hair follicles [17]. Rather than migrating and converging to accumulate as a pile of cells, our analysis identifies these cells as likely being involved in active tension generation to converge placode shoulders, bending and buckling the underlying basal lamina (Fig 4E-4H).

Discussion
Taken together, our findings demonstrate that tooth placode suprabasal cells use dynamic rearrangements to form a contractile structure, which is capable of driving invagination concomitantly with stratification. At early stages, the suprabasal contractile tissue bends the noncontractile basal layer in a gentle curve (Fig 5I and 5J) while later, as nonintercalating cells emerge centrally, the contractile tissue forms a canopy over the invagination to draw the shoulders (rim) into close approximation (groove or pore) (Fig 5K and 5L). The contractile structure driving invagination in the ectodermal placodes is topologically somewhat analogous to the contractile actin cables that drive invagination by apical constriction in epithelial monolayers, but on a vastly different scale: intercalation among cells rather than actin and myosin filaments. Finally, the similarity, right down to the distribution of force-deformed nuclei, between tooth bud, hair follicle, and mammary gland primordia indicates that these ectodermal organs share the same mechanisms of physical morphogenesis. It seems highly likely that similarly detailed inspection in other organs that form by invagination of stratified placodes may reveal a similar morphogenetic signature at the cellular level.

Mouse Strains and Genetic Labelling
All animal experiments were conducted under licence from the United Kingdom Home Office and the institutional Ethical Review Board. Tissue-cutting experiments were performed on wild-type CD1 mice. All cell shape analysis and live imaging was performed using mT/mG mice [8] as follows: Mosaic labelling was achieved by injecting 4 mg per 30 g body weight tamoxifen into a pregnant female mT/mG mouse [Gt(ROSA)26Sortm4(ACTB-tdTomato,-EGFP)Luo/J (Jackson Laboratories strain 007576)] crossed with a tamoxifen-inducible Cre male B6.Cg-Tg(CAG-cre/Esr1 Ã )5Amc/J (Jackson Laboratories strain 004682).

Tissue Processing
For all tissue cultures (live imaging and cutting experiments), tissues were cultured in phenol red-free Dulbecco's Modified Eagle Medium: Nutrient Mixture F-12 (DMEM/F12). delaminate and migrate centripetally, they form a canopy of elongated suprabasal cells that intercalate when they meet in the centre of the invagination, while more cells extend centripetally from the shoulders (yellow cells in K). By E13.5 the arc of elongated suprabasal cells have stacked on top of each other into a narrower-longer array that form the "neck" of the tooth bud (L). Periderm, a superficial layer of squamous cells present throughout the epithelium in tooth and non-tooth forming regions, has been omitted for clarity. For staining, E12.5 mouse heads were fixed in MEMFA (3.7% formaldehyde in 0.1 MOPS, 1 mM MgSO4, 2 mM EGTA) and manually cut, using a disposable microtome blade, into approximately 250 μm frontal slices and stained with Alexa Fluor 633 Phalloidin, DAPI (Molecular Probes), anti-phosphomyosin light chain antibodies (Cell Signaling Technology 3674) or anti-DECMA-1, an antibody against the ectodomain of E-cadherin (Sigma U3254). For live imaging, slices were cultured under coverslip fragments anchored at the edges with Vaseline and imaged in Lumox culture dishes (Starstedt).

Imaging and Image Analysis
All confocal imaging was performed in a Leica SP5 inverted confocal laser scanning microscope using a 20X HCX-PLAPO dry objective lens, n.a.1.7. Z-stacks were acquired at half-optimal z. For cell shape analysis, z-stacks of individual cells were projected into a single plane in order to capture their maximum dimensions in xy. As the analysis was of bending as seen in the frontal plane, the posterior and anterior ends of the tooth primordium were excluded. In order to differentiate between short, intermediate, and tall cells, nuclear heights were measured (using the mid-height of the nucleus) and plotted as a proportion of each cell's height and the basal palisade height.
Cell widths were measured at the base, the apex, and at 1/4, 1/2, and 3/4 of the cell's height. As a measure of a cell's shape with respect to contributing to epithelial bending, we wished to capture deformations of the cell body rather than protrusions at basal and apical ends. Therefore, base-to-apex ratios shown in Fig 1E and 1F are calculated using widths at 1/4 and 3/4 height. To partition the tooth epithelium in flat and concave, the radius of curvature of the basal lamina was measured using the three-point curve plugin in ImageJ, at a resolution of three cell widths.
For nuclear aspect ratio quantification, tissues were imaged in 3D and the section with the maximum size in the frontal plane for each nucleus was used to measure the major and minor axis lengths. Sections were taken only from the middle of the tissue in which elongation is strictly in the frontal plane. Absence of elongation in the z-axis after cutting was confirmed for nuclei in all specimens to rule out rotation.

Cutting Experiments
For cutting experiments, we utilised the well-established slice culture system for studying tooth development ex vivo [31][32][33][34]. Briefly, E12.5 embryonic jaws (both maxillae and mandibles) were manually dissected from embryos in phenol-red-free Dulbecco's Modified Eagle Medium: Nutrient Mixture F-12 (DMEM/F12) and placed on a sterilised tissue chopper stage (McIlwain tissue chopper, Ted Pella Inc.) on which 250 μm slices were cut. This corresponds to six to eight cell layers, optimal to include enough cell layers to preserve histomorphology (compared to freshfixed material) but thin enough to allow air access for extended development in vitro (as published [31][32][33][34], and routinely in our department for several days, although we never went beyond 30 h). Frontal slices were then transferred in polyHEMA (poly-hydroxyl-ethyl-methacrylate; Sigma)-treated tissue culture plates in culture media (see above) and immobilised using a fragment of coverslip held down with Corning vacuum grease. Cuts were performed using BD Micro-Fine insulin syringes and timelapse was recorded with minimal delay (capturing every 5 s) on a Nikon SMZ645 dissecting microscope with a MicroPublisher 5.0 RTV camera (QImaging). For lateral cuts in the presence of inhibitors, slices were cultured in 0.1 μg/mL cytochalasin D (C2168, Sigma) or 3 μg/mL blebbistatin (B0560, Sigma) for 30 min before performing the cut. Slices were fixed within 10 min of the cutting experiment and processed as above. Visual inspection of slice cultures after labelling with DAPI and fluorescently labelled phallloidin revealed that there were no obvious signs of cell death caused by the cut or culture conditions (S2G- S2I Fig).

Cutting Experiment Quantification
The dimensions of the suprabasal cut width were measured from stills of the timelapse recording, immediately after the cut and at the end of recoil. The width of the cut at the oral surface was measured using the line tool and the command Analyse/Measure in ImageJ. To quantify tissue bending upon lateral, double cuts, and lateral cuts in the presence of cytochalasin and blebbistatin, the maximum angle of the basal lamina subtended at the centre of the invagination was measured using the ImageJ angle tool. Shoulder cell nuclei were more apicobasally elongated than their non-shoulder neighbours and aligned with the direction of suprabasal cell intercalation because of their lean (arrows in A-E). Increased aspect ratio following shoulder "lifting" (curvature) starts on the buccal (cheek side) first at E12.5 (A,D) but then becomes symmetrical as the lingual shoulder lifts (D,E). This elongation is lost when the tissue-wide tension is released upon a lateral cut (dashed line in C), as indicated by loss of peaks in aspect ratio (F). Note that the latter relaxation is slower than that in suprabasal cells (2-4 h versus 10 min). (TIF) S1 Movie. Explanted tooth epithelium bends autonomously (left), and manual re-extension and release of bent epithelium demonstrated tissue elasticity (right), related to Fig 1. The epithelium was microdissected from the mesenchyme using fine needles and cultured separately. Care was taken to explant the basal lamina intact. Timestamp is in mm:ss, 1 frame = 5 s.