Radially-patterned morphogenesis of murine hair follicle placodes ensures robust epithelial budding

Summary

The bending of simple cellular sheets into complex three-dimensional forms requires developmental patterning cues to specify where deformations occur, but how positional information directs morphological change is poorly understood. Here we investigate how morphogen signaling and cell fate diversification contribute to morphogenesis of murine hair placodes, in which collective cell movements transform radially-symmetric primordia into bilaterally-symmetric tubes. Through live imaging and 3D volumetric reconstructions we demonstrate that Wnt and Shh establish radial patterns of cell fate, cell morphology and movement within developing placodes. Cell fate diversity at different radial positions provides unique and essential contributions to placode morphogenesis. Further, we show that downstream of radial patterning, gradients of classical cadherin expression are required for efficient epithelial rearrangements. Given that the transformation of epithelial discs into three-dimensional tubes is a common morphological motif used to shape diverse organ primordia, mechanisms of radially-patterned morphogenesis are likely highly conserved across evolution.

Graphical Abstract

eTOC blurb

Leybova et al. utilize live imaging and 3D reconstructions to uncover a radial pattern of cell fate, morphology, adhesion and motility in developing hair follicle primordia. Radial patterning is established by graded Wnt signaling and Shh, and is necessary for the complex morphogenetic changes that drive placode budding and polarization.

Introduction

Organ shape arises in embryonic development through the bending, folding and elongation of simple cellular sheets into elaborate three-dimensional geometries^1–3. Out-of-plane deformations, such as invaginations and buds, give rise to tubes and branches, whereas in-plane deformations expand, thin and elongate developing tissues¹. The positions at which these deformations occur are dictated by inductive patterning cues, often in the form of morphogen gradients, which specify different cell fates at precise positions within a tissue^3,4. The direction/orientation of tissue deformation, by contrast, is determined by polarity cues that align with the body axes^2,5,6. How the positional information imparted by morphogen signals integrates with polarity cues to drive morphogenetic transformations is a central question in developmental biology.

Mammalian hair placodes are the precursors to hair follicles and are attractive model organ primordia to investigate how inductive patterning cues instruct morphogenetic change. Self-organizing gradients of diffusible activators and inhibitors specify hundreds of hair placodes across the skin surface in a spatially-ordered pattern^7–9. Each placode begins as a disc of ~100 epithelial cells that is patterned radially with at least two distinct cell types arranged in concentric rings^10–12. As placodes invaginate and grow into the underlying dermis, cells positioned in the center become the tip of the elongating hair follicle while the halo of outer cells trails behind to give rise to the upper follicle and its resident stem cell compartment^10,11. Thus, radial patterning in the placode directly relates to cell position and fate in the mature follicle, but how this positional information contributes to tissue mechanics that shape the developing follicle is unknown.

During placode morphogenesis, out-of-plane deformations bend the primordium into an epithelial bud, which grows and elongates to form a closed end tube^7,8,13,14. Concurrently, in-plane deformations polarize the bud, which bends asymmetrically with an anterior-facing tilt^10,15. Placode polarization is driven by stereotyped cell rearrangements that reposition centrally-located cells anteriorly and sweep peripheral cells posteriorly¹⁰, transforming the radially-symmetric placode into a planar-polarized bud with anterior-posterior (A-P) asymmetry. This coordinated cell motion is directed by planar cell polarity (PCP), an intrinsic feature of the epithelium that is aligned with the global A-P tissue axis and orients each developing hair follicle to point in a common direction^10,15,16. Although PCP is active in every cell of the epidermal progenitor layer, only cells of the placode undergo planar-polarized rearrangements. PCP must therefore act together with inductive cues that establish placode identity to generate this complex morphogenetic behavior.

Because a cell’s radial position determines its identiy¹¹, we hypothesized that radial patterning may also determine its morphogenetic behavior. Inner and outer cells display differential patterns of gene expression^11,12, suggesting they possess distinct physical and/or molecular properties that provide unique contributions to placode morphogenesis. Here, we investigate the mechanisms controlling radial patterning of the hair placode and determine how positional information contributes to morphogenesis. We show that Wnt and Shh signal sequentially to establish the radial pattern, where Wnt-dependent inner cells induce surrounding outer cells through Shh. Using live imaging and 3D volumetric reconstructions of hair placode morphogenesis, we demonstrate that cell morphology and movement are radially patterned; cells display distinct in-plane and out-of-plane deformations at different radial positions. Further, radially-patterned cell behaviors depend on proper cell type diversification at central and peripheral positions, and inner and outer cells provide unique contributions to placode morphogenesis. Inner cells initiate the movements that drive placode polarization, whereas outer cells form a dome-like compartment within which cell rearrangements occur. Finally, we show that gradients of cadherin expression, downstream of radially-patterned Wnt signaling, are required for efficient in-plane rearrangements. Given how the transformation of a flat, circular field of epithelial cells into a 3D tubular shape is a common morphological motif used in diverse organ primordia, mechanisms of radially patterned morphogenesis are likely to be highly conserved across evolution.

Results

Spatiotemporal establishment of radial patterning during placode formation

To define how the radial pattern of cell fates emerges during placode morphogenesis, we monitored the spatial and temporal expression of inner (Shh-GFP or Lhx2), outer (Sox9), and dermal condensate (DC; Sox2) markers in confocal reconstructions of placodes at different morphological stages present at E15.5 (Figure 1A–A’, Figure S1)^{10,11,17–22}. Inner cells and the DC were the first to be specified, where circular clusters of approximately 30 basal cells expressing Shh-GFP and Lhx2 (Figure 1A–B and S1A) were positioned directly above aggregates of dermal fibroblasts expressing Sox2 (Figure S1B)^13,23. In later-staged placodes that had just begun to bud downward into the dermis, Sox9 expression appeared in approximately 100 basal cells that formed a halo surrounding the inner, Shh-GFP cluster (Figure 1A–B)^10,20. A second population of Sox9 expressing cells also appeared suprabasally (Figure 1A’), which corresponds to a previously described population that arises from asymmetrically dividing Shh+ progenitors^11,20. In this study, we focus on the outer ring of Sox9-expressing basal cells that gives rise to the upper hair follicle and its resident stem cell population¹¹.

Figure 1. Temporal and spatial characterization of radial patterning during placode formation, specified by Wnt and Shh signaling.

(A-A’) Planar (A) and sagittal (A’) views of of hair placodes at the indicated stages. Flat-mounted, dorsal skin explants from Shh-GFP embryos at E15.5 were labeled with phalloidin (blue), GFP (green) and Sox9 (magenta) antibodies. Arrow in A’ denotes plane shown in A. Dotted line depicts the epidermal-dermal boundary. Anterior is left.

(B) Quantification of inner and outer cells during the budding stage. n=6 placodes from 2 embryos. Student’s paired t-test, p<0.0001.

(C) K14-Cre; β-catenin fl/fl; mTmG embryonic skin at E15.5 where K14-Cre activity is not mosaic. mGFP (Cre-positive, green) is shown.

(D) Placode from K14-Cre; β-catenin fl/fl; mTmG embryo with mosaic K14-Cre activity labeled with Lhx2 (right, magenta). Left panel shows mTomato cells (mT, Cre-negative and wild type for β-catenin; blue) are surrounded by mGFP cells (Cre-positive, β-cat mutant, green). Quantification shows percent of Lhx2-expressing cells that are mGFP or mT. n= 25 placodes from 3 embryos. Binomial distribution test, p<0.0001.

(E) Placode from K14-Cre; β-catenin fl/fl; mTmG embryo with mosaic K14-Cre activity labeled with Sox9 (right, magenta). Quantification shows percent of Sox9-expressing cells that are mGFP (β-cat mut) or mT (WT). n=26 placodes from 3 embryos. Binomial distribution test, p=0.1602, n.s.

(F) Placode from K14-Cre; Smoothened fl/fl; mTmG (Smo cKO) E15.5 skin labeled with Lhx2 antibodies. Quantification shows number of Lhx2-positive cells at budding. n=10 placodes from 4 embryos (control) and n=16 placodes from 5 embryos (Smo cKO). Student’s unpaired t-test, p=0.6260, n.s.

(G) Smo cKO placode labeled with Sox9 antibodies. Quantification shows Sox9 mean intensity ratio of outer:inner cells. Student’s unpaired t-test, p<0.0001. n=7 placodes from 5 embryos (control) and n=11 placodes from 5 embryos (Smo cKO). Scale bars, 10μm.

Following specification and downward budding, placode progenitors rearrange within the epithelial plane, transforming placode morphology from a radially-symmetric to planar-polarized organization^10,15. In these later-staged, polarizing placodes, Shh-GFP expressing inner cells were repositioned anteriorly, while Sox9-expressing outer cells were displaced posteriorly (Figure 1A–A’). Sox2+ dermal fibroblasts remained closely associated with the Shh-GFP cluster, becoming co-displaced toward the anterior (Figure S1B). As developing follicles elongated into cylinders at the hair germ stage, the anterior-posterior segregation of all three cell types became increasingly pronounced (Figure 1A–A’). Thus, radial patterning of the placode is established from the center-outward, and specification of inner and outer cell identities precedes in-plane cell rearrangements.

Wnt and Shh signal sequentially to establish the radial pattern

We next sought to define how morphogen signaling pathways known to be required for hair follicle formation contribute to radial patterning. Wnt/β-catenin signaling is one of the earliest events in hair placode specification, and is both necessary and sufficient for hair follicle induction^24–29. To determine the specific requirements for Wnt signaling in specifying the two major epithelial cell types within hair placodes, we generated mouse embryos with β-catenin deleted mosaically in the skin epithelium (K14-Cre; β-catenin fl/fl; mTmG). We used the mT/mG reporter system to distinguish wild-type from knockout cells, where mGFP expression reports Cre activity and thus marks β-catenin deficient cells while mTomato expression marks cells lacking Cre (referred to as wild type, WT)³⁰. To confirm mGFP expression from the mTmG cassette was an appropriate proxy for β-catenin deficiency, we labeled β-catenin mosaic skins with antibodies against the Wnt target gene LEF1, and found that LEF1 was present only in mTomato+ cells, and was lacking in all mGFP+ cells (Figure S2). In addition, and consistent with a loss of β-catenin activity, areas of the epithelium consisting entirely of mGFP+ cells were completely devoid of hair placodes (Figure 1C)^24,26. In contrast, placodes did form in mosaic regions of the epidermis where wild-type and mutant cells were distributed in a strikingly consistent pattern. Wild-type cells clustered at the center of mosaic placodes, expressed inner cell markers, and were surrounded by a ring of Sox9-expressing cells that could consist of wild-type cells, β-catenin mutant cells, or a mix of both (Figure 1D–E)²⁰. To quantify how Wnt signaling ability correlates with the acquisition of inner and outer fates, we categorized cells as wild type or β-catenin-deficient based on their mGFP versus mTomato intensity and measured Lhx2 and Sox9 levels. Of the cells that expressed Lhx2, virtually all were wild type and positioned in the placode center. Mutant cells, by contrast, were spatially excluded from the central cluster and did not express the inner cell marker (Figure 1D). Mutant cells encircling these wild type clusters did, however, express Sox9 with a similar frequency to wild type cells (Figure 1E). We conclude that Wnt-dependent inner cells are pre-requisite for outer cell fate, but that Wnt/β-catenin signaling is only needed directly for inner and not outer fate.

Inner cells produce Shh, which is presumably secreted and signals in a paracrine manner (Figure 1A)^{18,31,32 20}. In hair placodes of Shh-deficient embryos, Sox9 expression is largely absent suggesting that inner cells signal to their neighbors via Shh to induce Sox9 expression and outer cell identity^10,20. To test this hypothesis, we conditionally ablated the Smoothened (Smo) receptor required for Shh signal transduction in the skin epithelium using K14-Cre. As expected, elimination of epithelial Smo caused a marked reduction in Sox9 expression (Figure 1G, Figure S3B)²⁰, whereas the number of Lhx2 positive inner cells, as well as the DC, were unaffected (Figure 1F, Figure S3A and S3C). While additional signals likely contribute to the radial pattern, these data are consistent with a model where Wnt specifies inner cell identity and activates Shh expression. In turn, Shh signals outward to induce Sox9 expression thereby establishing a radial pattern of cell identities in the placode.

Placode epithelial morphology and movement are radially patterned

Patterning events that specify cell fates at specific positions are thought to drive the downstream morphogenetic changes that shape developing organs. To define the morphogenetic changes that accompany radial patterning, we performed automated segmentation of epithelial cell edges, measured two-dimensional morphological features and correlated these with cell fate and radial position (Figure 2A–E)³³. The earliest morphological sign of placode formation, which coincided with specification of inner cells, was the shrinkage of epithelial cross-sectional area (Figure 2A–B, D). Later, and coinciding with peripheral Sox9 expression, outer cell shapes became increasingly eccentric, elongating circumferentially and compressing radially around the inner cell cluster (Figure 2A,C, E). During polarization, cross-sectional area was further reduced, and outer cells maintained their eccentric shapes (Figure 2B–E).

Figure 2. Epithelial morphology is radially patterned in developing hair placodes.

(A) Planar views of the IFE and placodes at the indicated stages labeled with membrane-GFP. Inner and outer cells, identified by Lhx2 or Sox9 labeling (not shown), are shaded in green and magenta, respectively. IFE cells included in quantitation in B-E are shaded black. IFE cells not included in quantitation are shaded gray.

(B-C) Heat maps of cross-sectional area (B, scale in μm²) and cell eccentricity (C, scale 0–1) of the regions shown in A.

(D-E) Mean cross-sectional area (D) and mean eccentricity (E) of inner and outer cells at the indicated stages. n=389 cells from 2 skins (IFE); n=106 inner, 197 outer cells from 3 placodes (specification, spec); n=106 inner, 402 outer cells from 3 placodes (budding, bud); n=72 inner, n=227 outer cells from 3 placodes. (polarizing, pol). Error bars are SD, and samples were compared using two-way ANOVA. p=0.0002, row factor and p=0.3092 n.s. column factor (D, area). p=0.3086 n.s., row factor and p=0.0003, column factor (E, eccentricity).

(F-G) Mean cross-sectional area (F) and mean eccentricity (G) of placode cells tracked during live imaging. T=0hrs represents start of cell rearrangements.

(H) Still frames from live imaging of placode morphogenesis. E15.5 embryonic skin explants expressing mTomato were imaged for 20 hours (10 hours before and after initiation of cell rearrangements t=0hr). Inner and outer cells are inferred based on morphometric data in A-E and false colored green and magenta, respectively, to show their relative positions through time. Anterior is left. Scale bars, 10μm.

Live imaging of embryonic skin explants allowed us to monitor these morphological transitions through time, which occurred over approximately 20 hours. Using morphological criteria we had obtained from correlating fate markers with cell shape at fixed time points, we could identify and track inner and outer cells in a label-free manner (Figure 2F–H and Video S1). Over the first 10 hours, we observed the gradual shrinkage of placode cell areas and increase in shape eccentricity (Figure 2F–G). Few rearrangements occurred and inner and outer cells retained their radial positions. Over the next 10 hours, planar cell rearrangements repositioned inner and outer cells to the anterior and posterior, respectively (Figure 2H and Video S1). Plotting cross sectional area and cell shape eccentricity through time, we found that changes in cell morphology, both the reduction in area and increase in eccentricity, precede the initiation of cell rearrangements (Figure 2F–G).

Taken together, these results show that placode epithelial cells display distinct morphologies and movements at different radial positions that coincide in time and space with cell fate acquisition. First inner cells are specified and the epithelium condenses through a reduction in cross-sectional area. Next, outer cells are specified, which elongate circumferentially coincident with epithelial budding. Following specification of both cell types, polarized cell rearrangements initiate and gradually reposition inner and outer cells into new, planar-polarized locations.

3D reconstructions of placode morphology reveal radially patterned, out-of-plane shape changes

To extend our morphometric analysis to capture out-of-plane deformations, we developed an image analysis pipeline to volumetrically reconstruct the 3D architecture of developing hair placodes. Focusing on basal cells, we reconstructed placodes at the specification and early polarization stages, as well as non-placode, interfollicular basal cells as controls (Figure 3A–C, Video S2). Measurements of cell height, volume, and apical-to-basal ratio revealed several radially-patterned morphological features (Figure 3D–F). First, during placode specification, epithelial cell volume is roughly conserved; as cells shrink in cross-sectional area (Figure 2B, D), they increase in height (Figure 3B, D, E)¹⁴. The polarization stage, by contrast, is associated with a marked decrease in cell volume (Figure 3E). Because cell number is also increasing over this period, the reduction in volume suggests that placode cells divide without first doubling in size. Second, although epithelial budding in many organs is driven by apical constriction^34,35, inner cells of budding hair placodes narrowed roughly uniformly along the apical-basal axis (Figure 3B, green cells, and 3F)¹⁴. By contrast, outer cells adopted highly eccentric shapes along their apical-basal axes, bending their apical surfaces inward toward the placode center (Figure 3B–C, magenta cells). In sagittal cross-sectional views, outer cells reach over and above their inner cell neighbors to attach apically to a group of 3–5 suprabasal cells (Figure 3C, blue cells). Thus, through radially-patterned, out-of-plane shape changes, the placode forms a dome-like architecture where outer cells arch over inner cells and connect to a suprabasal roof. During polarization, the dome-like arrangement becomes asymmetric as outer cells on the anterior thin and extend more prominently inward than cells on the posterior (Figure 3B–C).

Figure 3. 3D morphometric analysis reveals distinct cell behaviors of inner and outer cells during placode morphogenesis.

(A) 3D reconstructions of the IFE (top), specifying (middle) and polarizing (bottom) placodes. Cells are color coded in a rainbow pattern from the center out. Scale bars, 10μm (x,y).

(B) Reconstructed volumes of selected cells from the IFE and placodes (white arrowheads in A).

(C) Sagittal views of placodes at the indicated stages. Outer cells (magenta) bend inward and connect apically to a cluster of suprabasal cells (blue).

(D) Quantification of cell height. n=110 cells (IFE); n= 33 inner, 72 outer (specifying); n=29 inner, 144 outer (polarizing). Same n-values are used in D-F. Student’s unpaired t-test, p<0.001 for all groups relative to IFE.

(E) Quantification of cell volume. Student’s unpaired t-test, IFE-inner specification, n.s., p=0.2462; IFE-outer specification, n.s., 0.3941; IFE-inner polarization, p<0.0001; IFE-outer polarization, p<0.0001.

(F) Quantification of apical:basal ratio. Student’s unpaired t-test, IFE-inner specification, n.ss, p=0.1871; IFE-outer specification, p=0.0078; IFE-inner polarization, p=0.0021; IFE-outer polarization, p<0.0001.

(G) Representative example of 5-cell cluster from specifying-stage placode. Single confocal slices at apical and basal positions are shown. Cells are false colored to highlight their connectivity. Scale bars, 10μm.

(H) 3D volume of 5-cell cluster (same as G) in two orientations.

(I) Distribution of the number of cell neighbors in a single 2D plane.

(J) Distribution of the cumulative number of cell neighbors along the apical-basal axis. Inner cells (top, green) and outer cells (bottom, magenta) are plotted separately.

An unexpected feature revealed by 3D volumes was the evidence of cell rearrangements even in fixed samples. Cells within the placode were often connected to different neighbors at their apical and basal surfaces, suggesting they had been captured in an intermediate state of intercalation (Figure 3G). In 3D reconstructions of 5-cell clusters, cells appeared to twist around their neighbors (Figure 3H). To describe quantitatively the unusual cell connectivity of placode cells in 3D, we calculated the total, cumulative number of neighbors per cell in 3D and compared this value to the number of neighbors in a single plane. A perfectly columnar epithelial cell will have the same cumulative number of neighbors in 3D as it has in a single plane. We found in a given 2D plane, the majority of cells are surrounded by 4–7 neighbors (Figure 3I). In 3D, however, most placode cells connected to ≥8 cumulative neighbors, and could be connected to >11 different neighbors cumulatively along their apical-basal axes. Further, during the polarization stage when cells extensively rearrange, we observed an increase in the proportion of cells with 11+ neighbors (Figure 3J). The majority of interfollicular epidermal cells, by comparison, had 4–7 neighbors in both 2D and 3D (Figure 3I–J). These volumetric data indicate that the cell intercalations associated with placode polarization occur asynchronously along the apical-basal axis. Notably, inner cells displayed features of cell rearrangements (>cumulative neighbors) prior to outer cells, suggesting epithelial movements initiate with inner cells (Figure 3J).

Radially-patterned cell morphologies require inner and outer cell fate diversification

To understand how radial patterning contributes to morphological change in the placode, we analyzed 2D and 3D placode morphology in the absence of epithelial Smo, which selectively impairs outer cell fate (Figure 1F–G, Figure S3). Two-dimensional analysis of epithelial morphology revealed that Smo cKO placodes condensed relatively normally, however, the reduction in cross-sectional area per cell was less uniform compared to controls (Figure 4A,C, compare to Figure 2B, D). Eccentric cell shapes were found at both peripheral and central positions of Smo cKO placodes and, unlike controls, there were no significant differences in cell eccentricity between inner cells and those at peripheral positions (Figure 4B,D, compare to Figure 2C, E). In volumetric reconstructions of Smo cKO placodes (Figure 4E–F, Video S3), peripheral cells did not consistently bend their apical surfaces inward and above inner cells (Figure 4F) and thus, when viewed in sagittal optical slices, Smo cKO placodes did not organize into a dome-like architecture (Figure 4G). Further, the geometric packing of placode progenitors was highly disorganized (Figure 4H). Cells twisted and intercalated along their apical-basal interfaces to a greater extent than controls (Figure 4I), which was reflected in a higher percentage of cells having >8 cumulative neighbors (Figure 4J–L). Thus, although the inner cells of Smo cKO placodes appear to be specified correctly (Figure S3A), the loss of outer cell fate non-autonomously impairs their morphology. These defects in epithelial geometry correlated with reduced downward budding and a failure of Smo cKO placodes to polarize (Figure S3C’ and S3D)³⁶. We conclude that radially-patterned morphology depends on radial fate acquisition, and that one important function of Shh-dependent outer cells is to form a radial and apical compartment within which inner cell rearrangements can efficiently occur.

Figure 4. 3D placode architecture depends on radial patterning.

(A-B) Heat maps of cross-sectional area (A, scale in μm²) and eccentricity (B, scale 0–1) at the indicated placode stages in Smo cKO E15.5 embryos (compare to WT in Fig. 2B–C).

(C-D) Quantification of mean cross-sectional area and cell eccentricity in Smo cKO placodes (compare to WT in Fig. 2D–E). Inner and outer cells are determined based on inner fate markers (not shown) and cell position. n=502 cells from 2 skins (IFE); n=106 inner, 167 outer cells from 3 placodes (specification, spec); n=102 inner, 277 outer cells from 3 placodes (budding, bud); n=65 inner, n=150 outer cells from 3 placodes (polarizing, pol). Error bars are SD, two-way ANOVA. p=0.0002, row factor and p=0.7460 n.s. column factor (C, area). p=0.3436 n.s., row factor and p=0.9727 n.s., column factor (D, eccentricity).

(E) 3D reconstruction of Smo cKO placodes at the indicated stages. Cells are color coded in a rainbow pattern from the center out. Scale bars, 10μm (x,y).

(F) Reconstructed volumes of selected placode cells (white arrowheads in E). Compare to WT in Fig. 3B.

(G) Sagittal views of Smo cKO placodes at the indicated stages. Outer cells (magenta shading) fail to bend inward and connect to the apical cluster of suprabasal cells (blue shading). Compare to WT in Fig. 3C.

(H) Representative example of a 4-cell cluster from Smo cKO placode. Single imaging planes at apical and basal positions are shown. Scale bar, 10μm.

(I) 3D volumes of a 4-cell cluster (same as in H) shown in two orientations.

(J) Distribution of the number of cell neighbors in a single 2D plane during placode specification.

(K-L) Distribution of the cumulative number of cell neighbors along the apical-basal axis during placode specification (K) or polarization (L). Inner cells (top) and outer cells (bottom) are plotted separately.

The contributions of radial patterning to collective cell motion

To investigate how radial patterning contributes to the cellular rearrangements that drive placode polarization, we combined live imaging and automated cell tracking with genetic and pharmacological perturbations that inhibit inner or outer fate specification. In control placodes, cells move in a stereotyped, counter-rotational pattern in which inner cells are displaced toward the anterior, while outer cells slide posteriorly past their more centrally positioned neighbors (Figure 5A–B, Video S4)¹⁰. Tracking cell clusters located at different positions highlights the stereotyped behaviors that contribute to this collection motion. First, inner cells converge toward the placode midline and extend anteriorly via mediolateral intercalations (Figure 5D, top; cells 1–4). Second, posteriorly moving outer cells intercalate circumferentially toward the rear of the placode gaining new neighbors along the circumferential axis (Figure 5D, middle; cells 5–8). Third, at the boundary between inner and outer cells, the two cell types slide past one another in opposing directions (Figure 5D, bottom; cells a-f). This converts an initial vertical stripe of cells at the center of the placode into a V-shape whose apex points toward the anterior (Figure 5C).

Figure 5. Differential contributions of inner and outer cells to cell rearrangements.

(A) Live imaging of placode morphogenesis. E15.5 embryonic skin explants expressing mTomato imaged for >10 hours. Segmented cells are false colored in a rainbow pattern at the start of cell rearrangements and tracked through time (10h). Dotted line denotes inner/outer cell boundary. Scale bar, 10μm.

(B) Smoothed trajectories of centroid positions of all tracked cells shown in (A) over 10 hours.

(C) Select inner and outer cells (false-colored green and magenta, respectively) tracked through time.

(D) Select cell clusters showing distinct cell rearrangements at inner (top), outer (middle) and the inner/outer boundary (bottom, dotted line) regions of the placode. Arrows indicate direction of movement.

(E) Live imaging of K14-Cre; Smo fl/fl ; mTmG (Smo cKO) placode at E15.5. Cells are false colored in a rainbow pattern at the start of cell rearrangements. Scale bar, 10μm.

(F) Smoothed trajectories of centroid positions over 10 hours in Smo cKO placode shown in (E).

(G) Select cell clusters from Smo cKO placode showing cell intercalations orthogonal to the plane. Top row shows a cell intercalate and remain in the plane. Bottom row shows a cell intercalate only temporarily.

(H) Quantification of new cells appearing in the imaging plane over 10-hour time course. Cumulative number includes cells appearing due to orthogonal intercalations and divisions.

(I) Quantification of cells intercalating orthogonally over 10-hour time course. Cumulative number includes only cells that intercalate transiently (remain in imaging plane for <3 hours).

(J) Live imaging of placode morphogenesis in E15.5 skin explant treated with LGK974, a Porcupine inhibitor. Cells are false colored in a rainbow pattern at the start of cell rearrangements. Scale bar, 10μm.

(K) Smoothed trajectories of centroid positions over 10 hours in LGK974-treated placode shown in (J).

In placodes consisting only of inner cells (Smo cKO), progenitors did undergo neighbor exchanges, but they were highly erratic and did not result in productive, anterior and posterior-directed movements (Figure 5E–F; Video S5). Centrally positioned cells did not converge toward the placode midline or extend anteriorly (Figure 5E; cells 1–4), and peripheral cells did not intercalate along the placode circumference (Figure 5E; cells 5–8). Thus, when cells were color coded into vertical stripes, the pattern was retained through time (Figure 5E–F). Additionally, we observed many intercalations oriented orthogonal to the epithelial plane in Smo cKO placodes (Figure 5G), which was reflected in a greater number of new cells appearing in the imaging plane compared to controls, both in total and transiently, over the 10-hour time course (Figure 5H–I). Further, because downward budding was severely compromised in Smo cKO placodes, cell movements took place within a single plane level with the interfollicular epidermis (Figure 5E). These defects in epithelial rearrangements likely relate directly to the twisted intermingling of cells observed in Smo cKO placodes reconstructed in 3D.

We next attempted to generate placodes consisting mainly of outer cells. Based on prior data indicating that Wnt/β-catenin signaling suppresses Sox9 expression in later-staged hair follicles^20,37,38, we hypothesized that inhibiting Wnt in early placodes would lead to de-repression of Sox9, and convert inner cells to an outer-like fate. E15.5 Shh-Cre; mTmG skin explants were grown in the presence of LGK974, a pharmacological inhibitor of Porcupine, which is broadly required for secretion of Wnt family ligands^38,39. In this experiment, mGFP expression driven by Shh-Cre serves as a lineage reporter for inner cells specified prior to inhibitor treatment (Figure S4A). Compared to DMSO-treated control placodes, in which Sox9 expression is suppressed in mGFP-positive cells, Shh-Cre>mGFP cells in LGK974-treated placodes ectopically expressed Sox9 while remaining centrally positioned, suggesting they had been transformed into an outer-like fate (Figure S4B–C). Consistent with a loss of inner cell identity, Lhx2 expression was strongly reduced upon LGK974-treatment (Figure S4D–E).

To determine how loss of inner cell identity impacts collective movements within the placode, we performed live imaging on skin explants grown in the presence of LGK974 or DMSO. Segmentation and cell tracking revealed that, compared to control or Smo cKO placodes, Wnt-inhibited placodes were relatively static (Figure 5J–K, Video S6). Placode cells underwent few rearrangements and retained their relative positions through time (Figure 5J, cells 1–4 and 5–8). This phenotype resembles the loss of cell motion observed in PCP mutants and when Rho-kinase or myosin II activities are inhibited¹⁰. Because Wnt-family ligands have been implicated in PCP⁴⁰, we tested whether Porcupine inhibition interfered with PCP by measuring the asymmetry of the core PCP protein, Celsr1, an established readout of PCP function in the skin^{15,16,41–44}. In control explants, Celsr1 localized asymmetrically to anterior and posterior junctions of basal cells, as expected, and neither the direction nor magnitude of polarized Celsr1 localization was altered in LGK974-treated explants (Figure S4F–G), indicating that the failure of Wnt-inhibited placodes to polarize was not due to a loss of PCP. We conclude that by specifying and maintaining inner cell fate, Wnt signaling promotes epithelial motility within the placode, which is prerequisite for PCP-mediated polarized rearrangements. Together, results from our live imaging experiments demonstrate that inner and outer cells of the placode provide distinct contributions to polarized morphogenesis. Wnt-dependent inner cells generate cell motion whereas Shh-dependent outer cells help to confine and organize cell motion into a stereotyped pattern.

To explore the molecular signatures that distinguish inner and outer cells and may contribute to their distinct morphogenetic behaviors, we analyzed a recently published ssRNAseq dataset of developing hair placodes¹². Interestingly, this study identified four distinct cell populations in early hair placodes (Placode clusters I-IV) that, based on in situ hybridization of select markers in each cluster, appear to spatially map to placode cells at different radial positions¹². We find that two transcriptional signatures, clusters I and IV, spatially map to inner cells (Shh+, Sox9−) and outer cells (Shh−, Sox9+), respectively. We performed a GO term analysis of the marker genes defining cluster I and cluster IV (Figure S5). This revealed many GO terms associated with cell migration, protrusion formation and actin organization enriched in cluster I (inner cells) that are not represented in cluster IV (outer cells), consistent with our experimental data that Wnt-dependent inner cell specification promotes epithelial motility.

Classical cadherins are expressed in complimentary, radial gradients across the placode

To understand the molecular changes downstream of radial patterning that contribute to cell movements driving placode polarization, we focused on the classical cadherins, whose differential expression enables cell sorting and compartmentalization of different cell types within tissues^45–48. E- and P-Cadherin are known to be differentially expressed in hair placodes and the IFE^15,49–51, so we asked whether classical cadherin expression was radially patterned across the placode and if so, what is the contribution of differential cadherin expression to placode morphogenesis. To quantify E- and P-Cadherin expression across the placode we measured the mean edge intensity of P- and E-Cadherin per cell and correlated these values with cell position and fate. This revealed that E-Cadherin and P-Cadherin levels were not binary but, rather, were inversely graded across the placode. Inner cells express the highest P-Cadherin and lowest E-Cadherin levels, outer cells express intermediate levels of both P- and E-Cadherin, and IFE cells express low P-Cadherin but high levels of E-Cadherin (Figure 6A–B, D–E; Figure S6A–B). Plotting cadherin intensities per cell relative to their distance from the placode center showed that opposing gradients of E- and P-Cadherin were highly reproducible across multiple placodes and correlated well with cell fate (Figure 6C, F).

Figure 6. Opposing gradients of P-Cadherin and E-Cadherin expression in early hair placodes.

(A,D) Planar views of early hair placodes from E15.5 Shh-GFP embryos labeled with P-Cadherin (A) or E-Cadherin (D). Scale bars, 10μm.

(B,E) Heat maps illustrating P-Cadherin (B) or E-Cadherin (E) intensity. Outlines mark the inner and outer cell fate borders based on markers shown in Figure S6. Color code denotes mean edge intensity per cell.

(C,F) Scatter plots of normalized P-Cadherin (C) or E-Cadherin (F) intensity values versus cell position relative to placode center. Each point represents a cell, color coded by its fate as determined by Shh-GFP and Sox9 expression (see Figure S6). n=6 placodes across 2 embryos for each. Scale bars, 10μm.

(G-H) Scatter plots of normalized P-Cadherin (G) and E-Cadherin (H) intensity versus cell position in mosaic β-cat cKO placodes from K14-Cre; β-cat fl/fl; mTmG embryos. Each point represents one cell with gray denoting wild-type (mT) cells and blue denoting mutant (mG) cells (see Figure S6). Data from n=4 placodes across 3 embryos.

(I-J) Scatter plots of normalized P-Cadherin (I) and E-Cadherin (J) intensity versus cell position in Smo cKO placodes from K14-Cre; Smo fl/fl; mTmG embryos. Data from 4 placodes across 3 embryos.

To test whether cadherin gradient formation depends on cell fate acquisition, we quantified P- and E-Cadherin levels in placodes that either lacked epithelial Smoothened (K14-Cre; Smo fl/fl; mTmG), or had β-catenin deleted mosaically (K14-Cre; β-catenin fl/fl; mTmG), In β-catenin cKO mosaic placodes, P- and E-Cadherin levels correlated strongly with cell genotype as well as cell position. Wild-type cells were positioned in the placode center and expressed high levels of P-Cadherin and low levels of E-Cadherin. Mutant cells were positioned at the periphery and expressed P- and E-Cadherin in the opposite pattern (Figure 6G–H; Figure S6C–D). The P-Cadherin gradient fell sharply in β-catenin cKO mosaic placodes, and was nearly binary at the boundary between wild-type and mutant cells (Figure 6G), suggesting that graded Wnt signaling normally contributes to the smooth inverse gradients of P- and E-Cadherin expression. To test if Wnt signaling is indeed graded, we examined the expression of Axin2 and LEF1, two well-established regulators and reporters of Wnt/β-catenin signaling^38,52,53. Embryonic skin explants from the Axin2 reporter mouse strain Axin2^{P2A-rtTA3-T2A-3xNLS-SGFP2} were immunostained for LEF1⁵⁴. The levels of both nuclear GFP driven by the Axin2 promoter and LEF1 protein were graded from the placode center outward in a pattern similar to that of P-Cadherin (Figure S6E–G). Consistent with Wnt signaling being the primary inducer of graded cadherin expression, E- and P-Cadherin gradients were established normally in Smo-deficient placodes (Figure 6I–J; Figure S6H–I). We conclude that a gradient of Wnt/β-catenin signaling promotes radial patterning of classical cadherins.

Differential cadherin expression is necessary for efficient cell rearrangements during placode polarization

In classical cell sorting experiments, cells expressing similar types and/or levels of cadherins preferentially adhere to one another, sorting out from cells expressing different cadherins^46,47,55. We therefore hypothesized that opposing gradients of E- and P-Cadherin in the placode might enable inner and outer cells to slide past one or, alternatively, prevent the two cell types from intermixing as they rearrange. To investigate these hypotheses, we took advantage of the fact that upon deletion of E-Cadherin, there is compensatory upregulation of P-Cadherin^51,56, which should flatten the P-Cadherin gradient across the skin. As expected, conditional ablation of E-Cadherin from the skin epithelium using K14-Cre led to compensatory upregulation of P-Cadherin throughout the basal layer, which leveled its expression across the placode and IFE (Figure 7A–D). Loss of E-Cadherin and the resulting flattened P-Cadherin gradient did not affect radial patterning into inner and outer fates (Figure 7E–F), but placode polarization appeared to be delayed. Using the ratio of posterior-to-anterior (P:A) Sox9 intensity levels as a measure of placode asymmetry, we found that E-Cad cKO placodes did polarize, but only when they had grown significantly deeper into the dermis compared to controls (Figure 7G–K). To determine the cellular basis of this phenotype we performed live imaging and cell tracking of E-Cad cKO embryonic skin explants and found that although E-Cad cKO placodes grew deeper into the dermis, very few cell rearrangements occurred over the first 10 hours of imaging (Figure 7L, Video S7). Only when the base of the follicle had budded well below the level of the IFE did we observe the characteristic anterior displacement of inner cells (Figure 7L–M, Video S7). By contrast, in wild-type placodes, counter-rotational movements begin when the epithelium is still roughly level with the IFE (Figure 5A). Although it is possible that E-Cadherin has a specialized function in outer cells that P-Cadherin cannot be replace, we favor the interpretation that differential cadherin expression across the placode enables cell rearrangements to occur within the confines of a non-motile IFE, perhaps by reducing adhesion or altering the interfacial tension between inner and outer cells (or outer cells and the IFE). When cadherin levels are flattened, rearrangements are delayed until placodes have budded to a depth where rearranging cells are positioned below rather than encircled by their non-motile IFE neighbors.

Figure 7. Flattening the cadherin gradient delays placode polarization.

(A, C) Planar views of early hair placodes from E15.5 control (K14-Cre; mTmG) and E-Cad cKO (K14-Cre; E-Cad fl/fl; mTmG) embryos stained for P-Cadherin. Scale bars, 10μm.

(B, D) Scatter plots of P-Cadherin scaled intensity values versus cell position relative to the placode center for control (B) and E-Cad cKO (D). Each point represents a cell from 3 placodes (control) or 4 placodes (E-Cad cKO) across 2 embryos.

(E-F) Planar views of Ecad cKO placodes stained for inner (Lhx2, green, E) or outer (Sox9, magenta, F) fate markers. Scale bars, 10μm.

(G-J) Planar (G-H) and sagittal (I-J) views of placodes at the polarized stage from control (G, I) and ECad cKO (H, J) embryos at E15.5. Membrane-GFP (green) and Sox9 antibody (magenta) labels are shown. Scale bars, 10μm.

(K) Scatter plot of posterior:anterior Sox9 ratio versus depth (μm) in control (black) and Ecad cKO (blue) hair follicles. Each point represents one placode. Linear regression with 95% confidence interval is shown. n=47 (control) and n=50 (Ecad cKO) placodes from 4 embryos.

(L) Live imaging of placode morphogenesis in E-Cad cKO (K14-Cre; E-Cadfl/fl; mTmG) E15.5 skin explants over 14 hours (0–10hr top, 10–14hrs bottom). Scale bars, 10μm. Compare to WT in Fig. 5A

(M) Smoothed trajectories of centroid positions over first 10 hours (top) and last 4 hours (bottom) of live imaging E-Cad cKO placode shown in (H). Compare to WT in Fig. 5B.

Discussion

How spatial patterning instructs morphogenetic processes that sculpt tissues into their ultimate form is a central question in developmental biology, and is crucial for understanding how to generate functional organs in vitro. Here, using the polarized budding of nascent hair placodes as a complex morphogenetic output, we demonstrate how radial patterning by Wnt and Shh morphogens contributes to the three-dimensional shape changes and movements underlying placode morphogenesis. Despite decades of studies on hair follicle formation, only recently has it come to be appreciated that these structures form from a placode that is radially patterned^10–12. Using live imaging, it was shown that a cell’s position along the follicle’s proximal-distal axis can be traced back to its initial, radial position in the placode primordium. Distally positioned cells (hair matrix progenitors) derive from central cells of the placode, whereas cells that populate the upper portion of the follicle and become hair follicle stem cells derive from the placode’s ring of outer cells^10,11. Moreover, a recent single cell-transcriptomic analysis distinguished four different cell types within early hair placodes, and select genes representing each cluster spatially mapped to cells at different radial positions¹². Thus, the radial pattern of distinct cell types established at the placode stage relates directly to the lineage, position, and function of cells within the adult structure. Our study further highlights the importance of radial patterning by demonstrating how Wnt and Shh signaling generate the pattern, showing how inner and outer cells each contribute to the morphogenetic processes that shape the epithelial bud, and identifying a radially patterned adhesion code needed for placode polarization. Radial patterning is also observed in other organ primordia including Drosophila salivary gland and leg imaginal discs, as well as mouse tooth and mammary placodes^57–63. It is likely that differential gene expression at different radial positions in these flat, two-dimensional organ primordia gives rise to patterned cell behaviors needed for their transformation into elongated epithelial tubes.

It has long been known that Wnt and Shh are essential for placode formation^24,32, and our data clarify their roles in specific morphogenetic outcomes. Wnt-dependent inner cells initiate placode morphogenesis, as they are first to condense and to initiate planar rearrangements. Inhibiting Wnt secretion is sufficient to block cell rearrangements, and we propose that Wnt signaling initiates and sustains a transcriptional program that promotes inner cell motility. This is supported by single-cell transcriptomic data demonstrating that a placode subcluster which spatially maps to inner cells is highly enriched for transcripts associated with cell motility and actin regulation¹². The orientation of cell movement, in contrast, is directed by planar cell polarity (PCP), where junctional PCP proteins asymmetrically localize within the epithelial plane, providing each basal epidermal cell with an intrinsic anterior-posterior polarity^10,15,43. We show that Shh-dependent outer cells bend inward and attach to a suprabasal cluster forming a dome-like compartment that encircles inner cells, and without which rearrangements are highly erratic causing cells to become twisted and intertwined. The attachment of outer cells to a suprabasal cluster is reminiscent of tooth placodes where a contractile canopy of suprabasal cells pulls on the outer ring of basal cells to bend the epithelium into a bud⁶¹. The dome-like structure we observe may be the precursor to the suprabasal canopy observed in later staged hair germs⁶¹. Although we have not explored whether suprabasal cells are contractile or if they are required for hair follicle budding, we suggest that by serving as central, apical anchorage point for outer cells, the suprabasal cluster is a key component of the physical compartment within which motile inner cells rearrange. Notably, Wnt-activated cells are slower cycling than their Shh-activated neighbors²⁰, and these differences may contribute to the morphological behaviors we describe here. For example, proliferation can promote tissue fluidity^64–66, and the faster cell cycle of outer cells may enable them to be passively swept posteriorly during placode polarization. Conversely, the slower cell cycle of Wnt-activated inner cells may allow them to sustain directed, anterior-ward migration.

Finally, we show how Wnt-dependent radial patterning generates opposing gradients of cadherin expression across the placode primordium. Differential interfacial tension, which relates to the both the type and amount of adhesion molecules present on the cell surface, underlies cell sorting and the formation of tissue boundaries^{46–48,55,67,68}. In some contexts, differential cadherin expression prevents cells of different fates from becoming intermixed, as was shown for distinct neural progenitors in the developing neural tube⁶⁹. In the epidermis, however, replacing the E-Cadherin gradient with uniform P-Cadherin did not lead to appreciable intermixing of inner and outer cells. Instead, cell rearrangements were delayed, suggesting differential E- and P-cadherin expression in the placode allows cells to efficiently exchange neighbors. In vitro, E- and P-Cadherin show promiscuous heterophilic binding, and their heterotypic affinities appear to be similar to homophilic binding⁵⁵. Thus, it is unlikely cadherin gradients simply reduce the affinity of contacts between inner cells, outer cells and the IFE. Rather, the two cadherins might signal to the cytoskeleton in slightly different ways, altering cortical tension through actomyosin^70–72. The resulting differential interfacial tension between inner and outer cells (or outer cells and the IFE) could promote rearrangements through a process similar to cell sorting, where cells slide past one another to maximize their homophilic contacts and reduce their interfacial surface tension. Alternatively, E-Cadherin may perform a function in outer cells that P-Cadherin cannot replace. Eventually, however, once placodes bud to a level where they are no longer encircled by the IFE, directed cell rearrangements commence and follicle base is displaced anteriorly. Together, our data illustrate how radially-patterned morphogen signaling and fate acquisition coordinate with differential adhesion to ensure robust budding and polarized morphogenesis of a classic model organ primordium.

Limitations of the Study

Our study defines the roles of Wnt and Shh morphogens in the establishment of two radially-patterned cell identities in hair follicle primordia, but the process is likely more complex. Transcriptomic evidence suggests the presence of at least four cell states in hair placodes¹², and removal of epithelial Smo does not completely abolish markers of outer fate. Additionally, although we have defined the different morphogenetic contributions of two early cell lineages in placode formation, identifying the functional transcriptional changes induced by Wnt and Shh that are required for their differential cell behaviors will be important to resolve in the future.

STAR METHODS

RESOURCE AVAILABILITY

Lead Contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Danelle Devenport (danelle@princeton.edu).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• All data reported in this paper will be shared by the lead contact upon request.

• Original code has been deposited at https://github.com/abiswas-odu/CellVolViewer and is publicly available as of the date of publication.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Mouse lines and breeding

The laboratory mouse, Mus musculus, was the primary experimental model used in this study. Mice were bred an maintained under standard laboratory conditions receiving food and water ad libitum in an AAALAC-accredited facility in accordance with the NIH Guide for the Care and Use of Laboratory Animals All procedures involving mice were approved by Princeton University’s Institutional Animal Care and Use Committee (IACUC). This study was compliant with all relevant ethical regulations regarding animal research. E15.5 embryos from mixed backgrounds were used for all experiments. Sex was not determined in embryos, and all embryos of the desired genotype were used regardless of sex. Genotypes were determined by PCR analysis of each allele. The presence of mTmG was validated by screening embryos for red and/or green fluorescence. E-Cadherin protein depletion was validated by staining skin explants with E-Cadherin antibodies. A complete list of mouse genotypes used for each experiment organized by figure is provided in Supplemental Table 1.

METHOD DETAILS

Whole-mount immunostaining

E15.5 embryos were dissected in PBS with Ca²⁺ and Mg²⁺ (PBS++) and fixed in 4% paraformaldehyde/PBS++ for 1 hour at room temperature. Back skins were then dissected and blocked for 1–4 hours at room temperature (RT) or overnight at 4°C. Blocking solution consisted of 2% normal goat serum, 2% normal goat serum, 1% bovine albumin, 1% fish gelatin in either PBT2 (0.02% Triton in PBS) or TBT2 (0.02% Triton in TBS). Primary antibodies were added to blocking solution and incubated overnight either at 4°C or RT. Skins were washed and then secondary antibodies were added to PBT2. Skins were mounted in Prolong Gold or VectaShield. For P-Cadherin antibody, TBT2 was always used. The following primaries were used: P-Cadherin (rat, 1:250, Clontech #M109, clone PCD-1), Sox9 (rabbit, 1:1000, Millipore # AB5535), GFP (chicken, 1:1000, Abcam # ab13970), E-Cadherin (rat, 1:1000, Invitrogen # 14-3249-82, clone DECMA-1), Sox2 (rabbit, 1:500, Millipore # AB5603), Lhx2 (rabbit, 1:250, Abcam # ab184337), LEF1 (rabbit, 1:500, Cell Signaling #2230S). Appropriate Alexa Fluor-488, 555, or 647 (Invitrogen or Jackson Laboratory) were used at 1:1000. Hoechst (Invitrogen #H1399) was used at 2μg/mL added with secondary. 405-Phalloidin was added with secondary when mTmG reporter was not in the background (1:1000, Abcam, ab176752). For mounting in VectaShield, secondaries were used at 1:500. Z-stack images were acquired on an inverted Nikon A1R confocal microscope with a field number of 25 controlled by NIS Elements software using a Plan Apo 60/1.4NA or Plan Apo 100/1.45NA oil immersion objective, or a Plan Apo Lambda 40/1.25NA silicon immersion objective or a Plan Apo 20/0.75NA air objective.

Explant culture in pharmacological inhibitors

Full thickness skin explants from E15.5 ShhCre; mTmG embryos were placed on 13mm Nucleopore membranes with a pore size of 8um (Fisher, 90-300-57) and cultured at 37°C for a total of 24 hours on F-media supplemented with 10% FBS and either DMSO or 1uM LGK974, a PORCN inhibitor that interferes with Wnt ligand secretion. Skins were transferred to new media with fresh inhibitor or DMSO half way through the incubation. After 24 hours incubation, skins were fixed and processed for immunofluorescence and confocal imaging. For live imaging of cultured explants, 1uM LGK974 was added to the F-media-agarose gel pads, and skin explants were sandwiched between the pad and Lummox membrane as described.

Live imaging

Live imaging was performed as previously described in Cetera et al., 2018¹⁰. A 1% agarose gel with F-media supplemented with 10% fetal bovine serum was used to culture full thickness embryonic skin explants. Embryos at E15.5 were screened for fluorescence prior to dissections. Back skins were then dissected and placed on the agar pad. The gel was sandwiched on a 35mm lummox dish (Starsted, product number: 94.6077.331) and imaged with an inverted Nikon Ti-E Spinning Disc with Perfect Focus using a Plan Apo 20/0.75 NA air object with a 1.5x optical zoom. Explants were kept in a humidified chamber with 5% CO₂ at 37°C over the course of imaging. Timepoints were acquired every 20 minutes with a 2um step size for a total of 13 planes for at least 14 hours.

Movie processing

FIJI MultiStackReg Plugin was used to correct for xy drift⁷⁵. The basal layer was used to align the entire timelapse. Using the orientation of the nearest guard hair in each region for reference, positions were rotated so that anterior was pointing toward the left. A single z-plane time course was reconstructed following the basal cells of the placode to be used for cell tracking. Z-stacks were used to confirm cell identity as not all placode cells are visible in the same z-plane.

Cell segmentation

Cell edges were segmented using the membrane labeled channel from single z-planes through the basal layer. For cell edge segmentation in 2D, FIJI plugin Tissue Analyzer^73,74 was used followed by manual hand correction to generate cell masks. For morphometric analysis in 2D, cell masks generated in Tissue Analyzer were imported into QuantPolarity³³ for cell area and eccentricity measurements. For 3D segmentation, z-planes spanning the apical to basal surfaces of basal cells were used. Suprabasal cells were excluded from analysis. A custom segmentation code was written for segmentation in 3D, as well as for time-lapse images of Smo cKO placodes whose irregular cell shapes were not properly detected by Tissue Analyzer. This custom code is based on a machine learning neural net system which was trained on our previously published live imaging datasets^10,43. Post-segmentation hand corrections were performed in Tissue Analyzer. Nuclear segmentation was performed using the FIJI plugin StarDist⁷⁷ on the LEF1 labeled channel after image processing with the Despeckle noise correction in ImageJ. Mean nuclear intensities within segmented ROIs were measured on raw, unprocessed images.

Time-lapse cell tracking

Following moving processing and 2D cell segmentation, placode cells were tracked through time using FIJI plugin Tissue Analyzer. Cells were identified as placode and selected for tracking based on their condensed morphologies and downward movement into the dermis over the course of the time-lapse series. Cells that had budded by the end of the time-lapse were marked as placode and tracked backward from the start of the imaging. To distinguish inner and outer cells in a label-free manner, we used three criteria: 1) cell position and number, as placodes consist of ~35 inner cells spanning 5–6 cell diameters and ~100 outer cells spanning 3–4 cell diameters; 2) cell morphology, as inner cells remain compact and isometric whereas outer cells elongate circumferentially; 3) cell position by depth, as inner cells move deepest into the dermis and are displaced anteriorly. Tracked cells were color-coded either by cell fate, or in a rainbow pattern of vertical stripes prior to placode polarization for visualization of cell rearrangements. Cell trajectories were produced in Tissue Analyzer using the centroid position of cells at 20 minute intervals. Maximum projections of cell tracks were produced in ImageJ to show cell movements across specified periods of time. Tracks were smoothed to show overall trajectories.

3D reconstruction

To construct the 3D view of basal epithelial cells, 2D masks were generated and hand corrected for each frame in the z-stack. Cells were first identified in each of the frames using the MATLAB function bwlabeln. The regionprops function was used to obtain the centroid of the cells in each frame. We implemented a distance based heuristic algorithm to connect the centroids of the cells across the z-frames and create the 3D track of each cell. These centroid connected cells identified by the algorithm were color coded with the same color across frames, labeled and viewed in MATLAB’s sliceViewer for manual verification and correction. The hand corrected cell tracking across the frames was used for computation of the number of neighboring cells, cell area, and the cell height. A cell’s neighbors were detected by dilating the cell in each frame and finding the cells that overlap with the dilated area. Based on 3D tracking, the duplicates in the neighborhood detection were removed and the total counted. The apical and basal cell area was computed using the average of the two apical-most or basal-most segmented planes, respectively, and the apical:basal ratio was calculated. Cell height was computed by calculating the Euclidean distance between the centroids of the first and last frame in which the cell appears. To produce a smooth 3D reconstruction, we added 10 interpolation frames between each consecutive real frame of the z-stack. The interpolated frames smoothly bridge the gap between the shape changes of each cell as they are tracked from frame to frame. The real frames and the interpolated frames were merged to produce a 3D matrix of the whole image. The 3D view was constructed by plotting the isosurface of each cell. The cell volume was computed by summing up the total number of pixels in the 3D reconstruction and scaling it by the ratio of the real dimensions of the image and the pixel dimensions to account for the added interpolation frames. For rainbow coloring 3D views, each cell was assigned a rainbow color based on the centroid of the placode/IFE region where the central most cells were assigned red. Cells were sorted by the distance between their centroids and the centroid of the placode/IFE region and the sorted order was used to assign each cell a color from a Jet colormap of the same length as the number of cells. The code is publicly available at https://github.com/abiswas-odu/CellVolViewer.

QUANTIFICATION AND STATISTICAL ANALYSIS

Quantification of cadherin gradients

Back skins dissected from E15.5 Shh-GFP heterozygous embryos were cut in half, stained for GFP to mark inner cells, Sox9 to mark outer cells and F-actin to mark cell edges. One flank was stained for E-Cadherin while the other was stained for P-Cadherin. Confocal images of placodes at the budding stage were acquired as above using a Plan Apo 100/1.45NA oil immersion objective. Single z-planes through the basal layer of the placode and surrounding epidermis were segmented using FIJI Plugin Tissue Analyzer with hand corrections as necessary. Merged F-Actin and cadherin channels were used to segment cell edges. Masks marking inner, outer, and IFE cell fates were created in Adobe Photoshop using Shh-GFP for the inner cell region, Sox9 for the outer cell region. The IFE region were the cells that remained. Individual channels were blurred to be able to select entire region.

A custom Matlab code was used to measure the cadherin intensity at the edges of each segmented cell and generate a heat map based on intensity values. First, the number of pixels which is considered the cell boundary is calculated by successively dilating the initial 1-pixel boundary of the segmented cell by a 1-pixel layer towards the center of the cell, if the following layer’s overall sum of pixel intensity increased by 16% of the of the current sum. This allows for initial pixels close to the membrane to be added easily, but as the boundary moves further from the initial segmentation boundary, the cost to add another layer increases. The average intensity of this boundary region was calculated by the sum of intensity in the boundary divided by the area. The intensity value of each cell, color coded by cell fate, was plotted relative to distance from the placode center. To combine cells between multiple placodes, each placode was scaled from 0–100 and the distance from the placode center was calculated. Matlab code available upon request.

Quantification of cell fates in β-catenin cKO mosaic epidermis

Confocal images of K14-Cre; mTmG; β-catenin fl/fl placodes containing a mix of wildtype (mT) and mutant (mG) cells were selected for segmentation. mTomato and mGFP channels were segmented separately to identify wildtype and mutant cells. Cell fate masks were created in Photoshop using the appropriate markers (Lhx2 for inner cells; Sox9 for outer cells) and a custom Matlab code was used to identify each cell’s genotype (based on mTomato or mGFP intensity) and its fate (based on cell fate masks). Since mT fluorescence turnover is not immediate following Cre activity, manual correction was used to exclude cells that were both mTomato+ and mGFP+. The total number of inner or outer cells was computed and the percentage of these that were either mGFP+ or mTomato+ was calculated. Due to the exclusion of mGFP+ and mTomato+ cells, the percentages do not always add up to 100%.

Quantification of hair placode polarity

The ratio of posterior-to-anterior Sox9 fluorescence intensity was used as a measure of hair placode asymmetry. Rectangular ROIs drawn around each follicle were bisected along the anterior-posterior axis. The Sox9 mean pixel intensity in each half was measured in FIJI and polarity calculated as the ratio of posterior:anterior mean Sox9 intensity. Hair follicle stages were classified according to the levels of Shh>mGFP expression and placode morphology.

Quantification of Celsr1 polarity

Celsr1 polarity was quantified as previously described^10,78 using the Tissue Analyzer FIJI plug-in, an updated version of Packing Analyzer v2 software. E-Cadherin was used to mark cell edges in interfollicular regions of the epidermis. Edge segmentation and polarity analysis were both performed in Tissue Analyzer. The axis and magnitude of Celsr1 junctional asymmetry were plotted on circular histograms using the polar plot function in Matlab.

Quantification of inner and outer cells

Confocal slices of budding-stage placodes labeled with Shh-GFP (Figure 1B) or Lhx2 (Figure 1F) were selected to quantify inner cell number. GFP+ or Lhx2+ nuclei were counted manually. Note that the numbers of inner cells between the graphs in Figure 1B and 1F are not identical as Lhx2 is expressed slightly later than Shh-GFP (see Figure S1). For quantification of outer cell fate in Smo cKO placodes, mean Sox9 intensity was measured in two circular ROIs encompassing inner cells (26um circular ROI) or outer cells (25um wide ring ROI around the inner circle ROI) in FIJI. Outer cell fate specification was represented as the ratio of outer ring mean Sox9 intensity: inner circle mean Sox9 intensity.

Quantification of placode depth

To quantify placode depth, confocal z-stacks acquired in the planar view were resliced with interpolation in FIJI to generate sagittal views through each placode. Depth measurements were made by drawing a line from the bottom of the basal layer of the interfollicular epidermis to the base of the lower, posterior end of the placode epithelium.

Orthogonal intercalations

Fully segmented and tracked time-lapse image series were used to quantify cell intercalations orthogonal to the epithelial plane. For each timepoint, the number of new cells appearing (identified as a new cell ID in TissueAnalyzer) in that frame compared to the previous was manually counted. It was noted based on cell shape and the surrounding planes whether the cell appeared in plane due to a cell division (cell rounding) or an orthogonal intercalation (cell visible in apical or basal plane). An orthogonal intercalation was considered transient if that cell remained in the plane for less than 3 hours.

Single-cell RNA sequencing GO term analysis

An existing single-cell transcriptomic dataset of E14.5 hair placodes¹² was interrogated for gene signatures characteristic of inner cells (Shh+, Sox9−) and outer cells (Sox9+, Shh−) as defined by this study. Biological process GO terms of genes expressed in the placode I and placode IV clusters as defined by Sulic et al. (2023) with a log2 fold change >1 were identified using g:Profiler software⁸⁰ with g:SCS thresholding and a user set significance threshold of 0.01.

Statistical analysis

Statistical analysis and graphical representations were performed in Microsoft Excel and Prism 10 (GraphPad). All statistical details of experiments can be found in the figure legends, including the statistical tests used, exact p vales, exact value of n, what n represents, and dispersion and precision measures (e.g., mean, median, SD, SEM, confidence intervals). Significance was defined as p<0.01.

Supplementary Material

Video S1

Video S1. (Related to Figure 2F–H) Live imaging of placode morphogenesis. Time-lapse movie of hair placode formation in E15.5 embryonic skin explant expressing mTomato imaged for 20 hours. Inner and outer cells were inferred and false colored green and magenta, respectively. Placode moves 6μm deeper in z into the dermis over the course of imaging. Scale bar, 10μm.

Video S2

Video S2. (Related to Figure 3A) Volumetric views of the IFE and hair placodes. Rotation of 3D reconstructions. IFE (top), hair placode at the specification stage (middle), hair placode at the polarizing stage (bottom). Anterior is to the left and apical is on the top at the start. Cells are color coded in a rainbow pattern from the center out.

Video S3

Video S3. (Related to Figure 4E) Volumetric views of Smo cKO hair placodes. Rotation of 3D reconstructions of Smo cKO placodes at the specification (top) and polarization (bottom) stages. Anterior is to the left and apical is on the top at the start. Cells are color coded in a rainbow pattern from the center out.

Video S4

Video S4. (Related to Figure 5A–D) Live imaging of cell rearrangements during placode morphogenesis. Time-lapse movie of placode polarization in E15.5 embryonic skin explant expressing mTomato shown over the course of 10 hours (left). Cells are false colored in a rainbow pattern at the start of cell rearrangements (middle). Cell centroids are plotted as dots over the original (right). Placode moves 4μm into the dermis over the course of imaging. Scale bar, 10μm.

Video S5

Video S5. (Related to Figure 5E–F) Live imaging of placode morphogenesis in Smo cKO. Time-lapse movie of placode polarization in E15.5 Smo cKO embryonic skin explant (K14-Cre; Smo fl/fl ; mTmG) shown over the course of 10 hours. mGFP channel shown (left). Cells are false colored in a rainbow pattern at the start of cell rearrangements (middle). Cell centroids are plotted as dots over the original (right). Scale bar, 10μm.

Video S6

Video S6. (Related to Figure 5J–K) Live imaging of placode morphogenesis after LGK974 treatment. Time-lapse movie of placode polarization in E15.5 Shh-Cre;mTmG embryonic skin explant treated with LGK974 over the course of 10 hours. mGFP channel is shown (left). Cells are false colored in a rainbow pattern at the start of cell rearrangements (middle). Cell centroids are plotted as dots over the original (right). Placode moves 6μm into the dermis over the course of imaging. Scale bar, 10μm.

Video S7

Video S7. (Related to Figure 7L–M) Live imaging of placode morphogenesis in Ecad cKO. Time-lapse movie of placode polarization from E15.5 Ecad cKO embryonic skin explant (K14Cre; E-Cadfl/fl; mTmG) shown over the course of 10 hours. mGFP channel is shown (left). Cells are false colored in a rainbow pattern at the start of cell rearrangements (middle). Cell centroids are plotted as dots over the original (right). Placode moves 8μm into the dermis over the course of imaging. Scale bar, 10μm.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Alexa Fluor 488 Goat Anti-Chicken IgY	Invitrogen	Cat#A11039
Alexa Fluor 488 AffiniPure Donkey Anti-Rabbit IgG	Jackson ImmunoResearch	Cat#711-545-152
Alexa Fluor 555 Donkey Anti-Rabbit IgG	Invitrogen	Cat#A31572
Alexa Flour 647 Donkey Anti-Rabbit IgG	Invitrogen	Cat#A31573
Alexa Fluor 488 Donkey Anti-Rat IgG	Invitrogen	Cat#A21208
Alexa Fluor 555 Goat Anti-Rat IgG	Invitrogen	Cat#A21434
Alexa Fluor 647 AffiniPure Donkey Anti-Rat IgG	Jackson ImmunoResearch	Cat#712-605-153
Phalloidin-iFluor 405 Conjugate	Abcam	Cat#ab176752
Hoechst 33342	Invitrogen	Cat#H1399
Rat monoclonal anti-P-cadherin (clone PCD-1)	Takara Bio USA	Cat#M109
Rabbit polyclonal anti-Sox9	Sigma-Aldrich	Cat#AB5535; RRID: AB_2239761
Chicken polyclonal anti-GFP	Abcam	Cat#ab13970; RRID: AB_300798
Rat monoclonal anti-E-Cadherin (clone DECMA-1)	Invitrogen	Cat#14-3249-82; RRID: AB_1210458
Rabbit polyclonal anti-Sox2	Sigma-Aldrich	Cat#AB5603; RRID: AB_2286686
Rabbit polyclonal anti-Lhx2	Abcam	Cat#ab184337
Rabbit polyclonal anti-Lef1	Cell Signaling	Cat#2230S
Chemicals, peptides, and recombinant proteins
LGK974	Cayman Chemicals	Cat#14072
VectaShield
Experimental models: Organisms/strains
Mouse: ShhCre: B6.Cg-Shhtm1(EGFP/cre)Cjt/J	The Jackson Laboratory	JAX: 005622
Mouse: mTmG: B6.129(Cg)-Gt(ROSA)26Sortm4(ACTB-tdTomato,-EGFP)Luo/J	The Jackson Laboratory	JAX: 007676
Mouse: K14Cre: B6.Cg-Tg(KRT14-cre)1Efu/J	Elaine Fuchs	N/A
Mouse: Smo fl/fl: STOCK Smotm2Amc/J	The Jackson Laboratory	JAX: 004526
Mouse: β-catenin fl/fl: B6.129-Ctnnb1tm2Kem/KnwJ	The Jackson Laboratory	JAX: 004152
Mouse: E-cadherin fl/fl: B6.129-Cdh1tm2Kem/J	The Jackson Laboratory	JAX: 005319
Mouse: Axin2 reporter: B6.Cg-Axin2tm1.1Rva/J	The Jackson Laboratory	JAX:036013
Mouse: B6: C57BL/6J	The Jackson Laboratory	JAX: 000664
Software and algorithms
FIJI		https://imagej.net/software/fiji/
Tissue Analyzer FIJI Plugin	Aigouy et al.73,74	https://github.com/baigouy/tissue_analyzer/blob/main/TA_tutos/install.md
MultiStackReg FIJI Plugin	Thevenaz et al.75	http://bradbusse.net/downloads.html
QuantifyPolarity	Tan et al.76	https://github.com/Sara-Tan/QuantifyPolarity
StarDist FIJI Plugin	Schmidt et al.77	https://github.com/stardist/stardist-imagej
g:Profiler	Kolberg et al.80	https://biit.cs.ut.ee/gprofiler/
Prism9	GraphPad	https://www.graphpad.com/
Matlab2020	MathWorks	https://www.mathworks.com/products/matlab.html
CreativeCloud (Illustrator 2022; Photoshop 2022)	Adobe	https://www.adobe.com/products/catalog.html#category=creativity-design
Other
lumox® dish 35	SARSTEDT	Cat#94.6077.331
Nuclepore Membranes, 8.0μm Pore Size, Dia. 13mm	Thermo Fisher Scientific	Cat#09-300-57

Highlights.

• Cell fates within hair follicle primordia emerge in radial pattern

• Wnt gradients and Shh radially pattern cell fate, shape, adhesion and motility

• Radial position and fate determine cell morphology and morphogenetic behavior

• Radial adhesion gradients promote cell rearrangements driving placode polarization