Cell extrusion drives neural crest cell delamination

Significance

Neural crest cells contribute to the development of nearly every tissue and organ throughout the body. Born in the neural epithelium during early embryogenesis, they are thought to become migratory through a process known as epithelial–mesenchymal transition. But is that the only way? We identified, through live imaging of embryo development, that neural crest cells can also become migratory via cell extrusion, a mechanism thought to also be important in cancer metastasis. This process is regulated by the mechanosensitive ion channel, PIEZO1, which monitors tissue pressure and tension to help maintain homeostasis. We have therefore uncovered cell extrusion as a mechanism driving neural crest cell development, which has broad implications for our understanding of craniofacial morphogenesis and cancer metastasis.

Abstract

Neural crest cells (NCC) comprise a heterogeneous population of cells with variable potency that contribute to nearly every tissue and organ throughout the body. Considered unique to vertebrates, NCC are transiently generated within the dorsolateral region of the neural plate or neural tube during neurulation. Their delamination and migration are crucial for embryo development as NCC differentiation is influenced by their final resting locations. Previous work in avian and aquatic species revealed that NCC delaminate via an epithelial–mesenchymal transition (EMT), which transforms these progenitor cells from static polarized epithelial cells into migratory mesenchymal cells with fluid front and back polarity. However, the cellular and molecular mechanisms facilitating NCC delamination in mammals are poorly understood. Through time-lapse imaging of NCC delamination in mouse embryos, we identified a subset of cells that exit the neuroepithelium as isolated round cells, which then halt for a short period prior to acquiring the mesenchymal migratory morphology classically associated with delaminating NCC. High-magnification imaging and protein localization analyses of the cytoskeleton, together with measurements of pressure and tension of delaminating NCC and neighboring neuroepithelial cells, revealed that round NCC are extruded from the neuroepithelium prior to completion of EMT. Furthermore, cranial NCC are extruded through activation of the mechanosensitive ion channel, PIEZO1. Our results support a model in which cell density, pressure, and tension in the neuroepithelium result in activation of the live cell extrusion pathway and delamination of a subpopulation of NCC in parallel with EMT, which has implications for cell delamination in development and disease.

Neural crest cells (NCC) are a heterogeneous migratory stem and progenitor cell population unique to vertebrates that are responsible for generating most of the craniofacial skeleton and peripheral nervous system among many others (1, 2). NCC development can be divided into four distinct phases: formation, delamination, migration, and differentiation (2–6). Delamination, the separation of individual cells or a tissue from a surrounding tissue layer, occurs repeatedly throughout development, especially during gastrulation and neurulation, and in disease pathogenesis, such as in cancer invasion (7). Given the diversity of situations in which a cell exits a tissue during development and disease, a number of cellular delamination mechanisms have been identified, including epithelial–mesenchymal transition (8), (which is frequently interchangeably described as ingression when referencing gastrulation), cell extrusion and asymmetric cell division (9–15). Delamination is required for NCC to detach from the dorsolateral border region of the neural plate in which they form, and this occurs through an epithelial–mesenchymal transition (EMT) (8, 16, 17).

EMT is a cell-autonomous process in which an epithelial cell becomes mesenchymal. Epithelial cells are apical–basal polarized and tightly adhered to each other via strong cell–cell junctions in one confluent tissue, which provides structure and often acts as a barrier or the initial defense against extrinsic factors (18). In contrast, mesenchymal cells exhibit front–back polarity, are loosely organized, and often exhibit migratory capacity (18, 19). NCC progenitors are epithelial when first specified in the dorsolateral border region of the neural plate, a pseudostratified single-layer epithelium. NCC then form by the loss of static apical–basal polarity, and adherens and tight cell–cell adhesive junctions, in combination with the acquisition of a dynamic front–back polarity, focal adhesions, and mesenchymal cell motility, consistent with EMT (9, 15, 18, 20, 21). Zeb2, Snai1/2, and Twist1 are considered master regulators of EMT because of their universal roles in both development and disease-associated EMT (9, 22).

Decades of research in avian and aquatic species have provided a rich understanding of the signals and gene regulatory networks governing EMT during NCC delamination. Briefly, EMT master regulators Zeb2, Snail1/2, and Twist1 are induced by WNT, NOTCH, TGFb, and Hypoxia signaling pathways (15, 20, 23–28). In Xenopus, snai1/2 are required and sufficient for NCC EMT and delamination. Similarly, both Snai2 and Zeb2 are required for NCC EMT in chicken embryos (23, 24, 29–33). Expression of the EMT master regulators results in repression of the epithelial cell–cell adhesion protein E-cadherin, and upregulation of the mesenchymal intermediate filament protein vimentin (20).

However, Zeb2, Snail1/2, and Twist1 loss-of-function does not result in the perturbation of NCC delamination in mouse embryos (34–38). This difference in the regulation of NCC EMT in mammals is surprising but may be attributable in part to variation in species-specific spatiotemporal signaling (39). For example, unlike in frog and chicken embryos, Snai2 is not expressed in premigratory NCC in the mouse (40). In addition, Twist1 expression in mouse NCC begins only after delamination has already occurred (SI Appendix, Fig. S1) (34, 41, 42), but is absent from NCC in chicken embryos (43). Another contributing factor may be functional redundancy between these key EMT regulators in mammalian embryos. However, redundancy has been difficult to discern in the mouse since global knockout of Snai1 results in embryonic lethality during gastrulation, which is prior to NCC formation and delamination (25). Nonetheless, in conditional Snai1/2 double mutant mice, although left–right patterning is disrupted, NCC formation and delamination are not affected (37). It is also important to note, however, that Wnt1-Cre, which is the gold standard in the field for conditional NCC lineage and function studies, is likely activated too late to affect NCC EMT and delamination and this could explain the absence of a NCC EMT-associated phenotype following conditional deletion of Snai1, Zeb2, and Twist1 (44). This leaves open the question of what drives NCC delamination in mouse embryos.

One of the limitations to a better understanding of the cellular dynamics of NCC delamination in mammals is the difficulty of real-time visualization of early head fold stage embryo development. To address this gap in knowledge, we developed methods for live imaging of the dynamics of NCC induction, delamination, and migration at single-cell resolution (45). Consequently, we uncovered a unique population of delaminating NCC that were round and lacking in polarity. These round cells differed considerably from the majority of delaminating NCC which exhibit a classic elongated morphology with front and back polarity. The lack of polarity and round morphology suggested these cells were no longer epithelial, but also not yet mesenchymal. Staining for mesenchymal marker expression confirmed these round cells were not yet in a mesenchymal state. Previous work has demonstrated that EMT proceeds through a dynamic series or spectrum of transitory phases, and we have shown these intermediate transition states take place while the NCC are within the neuroepithelium, prior to their exit (46). Considering the round cells are neither epithelial nor mesenchymal in character, they had not completed EMT. We therefore hypothesized these round NCC delaminated via an alternative mechanism, specifically cell extrusion, in parallel with NCC that delaminate by classic EMT.

Cell extrusion has been shown to reduce tissue stress associated with overcrowding, by facilitating cell delamination. During cell extrusion, a cell is forcefully expelled from an epithelium by the active contraction of neighboring cells (10, 47–49). First identified and described by this unique non-cell-autonomous activity (10), cell extrusion has unfortunately been frequently misrepresented as a cell-autonomous step of EMT. However, cell extrusion requires unique structural rearrangement of the cytoskeleton in both the delaminating and neighboring cells (10, 47–49). In addition, the epithelial cell–cell junctions remain in place to facilitate the expulsion of an extruding cell by the contraction of the surrounding cells, whereas the loss of apical attachment and cell–cell junctions are critical steps of EMT (50–56). The resulting round nonpolar extruded cells then undergo cell death or adapt, survive, and invade their new surrounding environment (10, 47–49). Thus, not every cell that undergoes EMT does so via cell extrusion, and not every cell that is extruded undergoes EMT.

Through electron microscopy and immunostaining, we determined cell extrusion-specific cytoskeleton structures were present in round NCC and neighboring cells during delamination. Measurements of cell density in the neuroepithelium and mechanical stress in delaminating NCC further supported a role for cell extrusion. Activation of the cell extrusion pathway by mechanical induction is mediated through the mechanosensitive ion channel, PIEZO1 (47, 48), and single-cell RNA sequencing of the cranial region of mouse embryos exhibiting NCC delamination and migration together with immunostaining, demonstrated PIEZO1 is expressed by NCC during their delamination (46, 57). To elucidate whether NCC can delaminate by PIEZO1-mediated cell extrusion, we cultured early head fold stage mouse embryos in the presence of an antagonist (GsMTx4) or agonist (Yoda1) of PIEZO1. Fewer delaminated NCC were observed following PIEZO1 inhibition, and furthermore, these embryos lacked the round NCC population characteristic of cell extrusion. In contrast, when PIEZO1 was activated, the total number of delaminated NCC increased in association with the presence of increased round NCC.

Altogether, our data have revealed cell extrusion to be a distinct mechanism by which ~20 to 30% of cranial NCC delaminate. These findings further our understanding of mammalian NCC development and have the potential to inform our knowledge of delamination in other biological contexts such as in cancer metastasis.

Results

A Subpopulation of NCC Delaminate as Distinct Round Cells.

To better define the cellular dynamics underpinning mammalian NCC delamination, we performed time-lapse imaging of cultured whole Wnt1-Cre;R26R-mTmG transgenic mouse embryos. The Wnt1-Cre;R26R-mTmG transgenic line allows for lineage tracing of NCC beginning at their premigratory stage (58). In these mice, Cre-mediated excision of a stop codon results in membrane-tagged GFP expression and thus tissue-specific visualization of cellular morphology (59). NCC delamination occurs in a temporal wave beginning at around embryonic day (E) 8.5, or the 4-5-somite stage (60, 61), in the midbrain, and then expands anterior and posteriorly along the neural axis as the embryo develops. Therefore, to capture morphological changes from the beginning of delamination, we focused our imaging on the region immediately posterior to the most recent domain of NCC delamination in the head of E8.5 embryos. A z-stack was acquired every 10 min for 130 min. The first cells to exit the neuroepithelium exhibited a round morphology that appeared to be maintained for ~30 min (Fig. 1A and Movie S1). This nonmigratory round morphology was surprising as cells that undergo EMT typically rapidly acquire a front–back polarity and form filopodial protrusions that immediately guide and drive migration (15, 20). Following this ~30-min morphological transition, round NCC that delaminated did not apoptose, but ultimately elongated and produced protrusions indicative of the acquisition of mesenchymal morphology (Fig. 1A and Movie S1). The data provided are from one embryo; however, the presence of round NCC and their changes in morphology were observed in three additional live imaging experiments.

Fig. 1.

A subpopulation of NCC delaminate as distinct round cells lacking polarity. (A) Montage from time-lapse imaging of an E8.5 Wnt1-Cre;R26R-mTmG shows the first NCC to delaminate from the neuroepithelium do so as distinctly round cells. At 50 min, the cells are elongating and forming protrusions. At 90 min, the elongated morphology is the only phenotype observed. (B) Immunostaining for GFP (Cyan) on transverse histological sections of E8.5 Wnt1-Cre;R26R-mTmG embryos [not cultured] reveal the same cellular dynamics of delaminating NCC as compared to time-lapse imaging. Images are compiled from cranial tissue sections from Wnt1-Cre;R26R-mTmG 6-8-somite stage embryos (n = 2 embryos). (C) Transmission electron microscopy on sagittal histological sections of E8.5 CD1 (wildtype) embryos. Comparatively posterior regions of the neuroepithelium exhibited a population of round cells immediately outside of the basal edge. In contrast, in more anterior regions where delamination had proceeded for a longer time, mesenchymal-like cells were evident outside of the neuroepithelium. (A–C) Arrow heads label the round subpopulation, asterisks label presumptively delaminating mesenchymal NCC, and the dashed line outlines the basal edge of the neuroepithelium. (D) Quantification of the total number of round vs. elongated NCC delamination events observed in cranial tissue sections from E8.5 Wnt1-Cre;R26R-mTmG embryos, specifically at the 5- to 6-somite stage (n = 6). The values are shown as a percentage of the total number of observed delaminated NCC. Roughly 20 to 30% of NCC delaminate as round cells, whereas 70 to 80% exhibit classic elongated EMT morphology.

To determine whether the round cells were indeed a discrete delaminating population of NCC and to rule out an artifact of culture and imaging conditions, we next sought to identify these round cells in noncultured embryos in situ. Immunostaining for GFP on transverse histological sections of E8.5 Wnt1-Cre;R26R-mTmG embryos revealed a similar spectrum of cell morphologies (Fig. 1B). Round NCC could be identified in select sections alongside elongated NCC, which comprised the predominant population of delaminated cells (Fig. 1B). Evaluation of electron microscopy images from sagittal sections of E8.5 CD1 embryos also revealed the presence of round NCC outside the neuroepithelium (Fig. 1C). Notably, these round cells lacked intracellular polarity (Fig. 1C). In more anterior (mature) locations, delaminating cells exhibited the classic mesenchymal NCC morphology with distinct front–back polarity (Fig. 1C). Altogether, our time-lapse and static high-magnification imaging of the dynamics of NCC delamination has uncovered a subpopulation of cells delaminating with a round morphology, suggestive of a nonclassic EMT mechanism.

Given these round NCC appeared to delaminate via a mechanism distinct from classic EMT, we investigated how often this nonclassical delamination occurred. To determine how frequently NCC delaminated in this manner, we quantified the number of round or elongated NCC that were perceived to be delaminating or had recently delaminated from the neuroepithelium across transverse tissue sections that spanned the forebrain through the hindbrain. Our time-lapse imaging demonstrated that the round morphology was transient until the NCC began migrating away from the neuroepithelium, making it impossible to distinguish between delamination mechanisms if a cell had moved any distance away from the neuroepithelium. Therefore, a cell was counted if the cell membrane was still partially connected to the neuroepithelium and its nucleus was already beyond the basal border of the epithelium or if the cell was immediately outside the neuroepithelium as this would mean delamination was very recent. A total of 6 × E8.5 Wnt1-Cre;R26R-mTmG embryos (5- to 6-somite stage) from multiple litters were evaluated. NCC were observed to delaminate with a round morphology ~20 to 30% of the time, with the remaining 70 to 80% of NCC exhibiting the classic elongated EMT morphology (Fig. 1D).

Cytoskeleton Protein Localization in Round Delaminating Cells Is Consistent with Cell Extrusion.

We next sought to understand why the dynamic morphology of these round NCC differed from the expected morphological changes typical of delamination by classic EMT. Breakdown of the basement membrane precedes NCC delamination and is necessary for emigration (62–65). To determine whether basement membrane breakdown had occurred when the round NCC exited the neuroepithelium, we evaluated the expression of laminin, which is a major component of the basal lamina. Immunostaining for laminin indicated that the basement membrane had begun breaking down in concert with round NCC delamination, as evidenced by lamina fragmentation near the cells (SI Appendix, Fig. S2). In fact, there appeared to be more fragmented laminin in regions of elongated cells, consistent with subsequent EMT-mediated cell delamination and further progression of basement membrane breakdown (SI Appendix, Fig. S2).

Since the basement membrane was broken, we then examined whether these round cells completed EMT and became mesenchymal. Epithelial cells are characterized by their apical–basal polarity, and within the neuroepithelium, they have a distinctly columnar shape (18). Mesenchymal cells are defined by their front–back polarity and formation of lamellipodial and filopodial protrusions (19). The round NCC do not resemble either of these cellular states. Recent research has shown that EMT is not a direct binary switch between epithelial and mesenchymal states, but rather occurs progressively through a contiguous spectrum of transitory intermediate states with varying degrees of both epithelial and mesenchymal character, until the transition is complete (9, 46). Our previous analysis of mammalian NCC delamination suggested the EMT transition states occur within the neuroepithelium (46). To elucidate the cellular state of these round NCC that appear neither epithelial nor mesenchymal, we evaluated the expression of vimentin, an intermediate filament protein which is commonly used as a mesenchymal marker, and E-cadherin, a component of adherens junctions in epithelial cell populations. If the round cells were partially mesenchymal, we would expect some expression of the mesenchymal marker vimentin. In 48 histological sections from two different embryos, we observed numerous elongated NCC that morphologically aligned with EMT delamination. Immunostaining confirmed vimentin was present at the leading edge, consistent with completion of their transition to a mesenchymal state and delamination by EMT (Fig. 2A). In contrast, round NCC completely lacked vimentin expression, demonstrating that these cells were not yet mesenchymal (Fig. 2A). We quantified the average intensity of vimentin staining in the round (extruded) NCC, in cells of the neuroepithelium and in migratory NCC (mNCC) (Fig. 2B). The average intensity was quantified in FIJI by drawing a region of interest (ROI) along each cell’s membrane to measure the average vimentin intensity which was then normalized to the cell’s average DAPI intensity. Neuroepithelial cells should not express vimentin, therefore we used these cells as our negative control and found that the round (extruding) NCC did not have a statistically significant difference in average vimentin expression (Fig. 2B). Because these measurements were performed on average projected z-stack images, some expression of vimentin from the extracellular environment is captured along with expression in the cell, likely accounting for the slight increase seen between the extruded and neuroepithelial populations. Migratory NCC should express vimentin, and when comparing the average intensity between migratory NCC and the round extruded NCC, we observed a statistically significant difference between these two populations (P = 0.0495). We then evaluated the expression of the epithelial marker E-cadherin, which was strongly expressed in the ectodermal populations, but in contrast only at background levels along the neuroepithelium (SI Appendix, Fig. S3A). When comparing the average intensity levels of E-cadherin expression in the neuroepithelium to the round NCC, no difference was detected (SI Appendix, Fig. S3B). It is therefore hard to say whether the round cells have lost the expression of E-cadherin compared to cells of the neuroepithelium, but they certainly have less than the ectoderm. Regardless, these results indicate that the round NCC must delaminate by a nonclassic EMT mechanism.

Fig. 2.

Cytoskeletal composition indicative of cell extrusion. (A) Immunostaining for Vimentin and GFP together with DAPI on transverse histological sections of E8.5 Wnt1-Cre;R26R-mTmG embryos. The box indicates the location of the original tissue section from which the high-magnification Inset image was taken. The cells of interest are outlined in white with the top set of images showcasing NCC delaminating by EMT and the bottom set of images depicting an example of a round delaminating cell. The arrow is pointing out the expression of vimentin in the mesenchymal NCC, which was not present in the round cell population. (B) Quantification of the average intensity of Vimentin in extruded NCC, neuroepithelial cells, and mNCC normalized to the average expression of DAPI within the cell. There was no statistical difference between extruding NCC, and cells of the neuroepithelium, which do not express Vimentin. In contrast, there was a statistically significant (P = 0.0495) difference in the average intensity between mNCC and extruding cells, demonstrating that extruding NCC do not yet express Vimentin. Statistical significance was determined using an unpaired t test, with seven cells in each population. (C) Immunostaining for Actin, Tubulin, and GFP on transverse histological sections of E8.5 Wnt1-Cre;R26R-mTmG embryos. Polyline kymograph analysis is presented in the graphs next to the cell from which the measurement was performed. The dashed line arrow indicates the region and direction the polyline was drawn across the cell of interest. T labels the trailing edge and L labels the leading edge. The bracket highlights the localization of actin and tubulin found in the trailing edge of the extruding cell which was not present in the mesenchymal delaminating NCC.

Asymmetric cell division has previously been proposed as an alternative mechanism by which NCC could delaminate from the neuroepithelium (53, 66, 67). Furthermore, cells are known to acquire a round morphology when undergoing mitosis (68). Cells undergoing division can be identified with electron microscopy by the condensation and alignment of chromosomes, and those that recently divided can be denoted by a reforming nuclear envelope. Examples of these stages of cell division were identifiable in various mesodermal locations in our electron microscopy images (SI Appendix, Fig. S4A). However, the round NCC delaminating from the neuroepithelium did not exhibit chromatin condensation (Fig. 1C), suggesting these cells were not round or delaminating in association with cell division. To further discern whether the round cells were undergoing mitosis following delamination, we performed immunostaining with the G2/M marker phospho-histone H3 (pHH3) on 18 sections from two different embryos. pHH3 was not observed in the round cell population indicating that the round morphology was not a result of, or associated with, cell division (SI Appendix, Fig. S4B). Some of the round cells observed in our time-lapse did appear to eventually undergo cell division, although it is important to note these divisions were not immediate as the delamination of the round NCC had long since taken place (Movie S1).

Cell extrusion is another form of delamination in which tissue stress triggers neighboring cells to expel adjacent cells from the epithelium in order to maintain tissue homeostasis (14, 47–49, 69–71). To facilitate delamination by cell extrusion, neighboring cells and the extruding cell undergo cytoskeletal reorganization. The cytoskeleton changes include formation of an actomyosin ring and orientation of microtubules along the cell membrane in the direction of expulsion (10, 14, 48, 49, 72). The orchestration of inter- and intracellular cytoskeleton changes during cell extrusion contrasts with EMT, which is considered a cell-autonomous process. To determine whether these cytoskeletal differences could be identified, we analyzed the localization of actin and tubulin in the round and elongated NCC populations in cells that were still partially attached to the epithelium. A polyline kymograph analysis was performed by drawing a line from the trailing edge of the cell, the last edge of the cell to leave the epithelium, to the leading edge of the cell, the first edge to leave the epithelium. Actin and tubulin were localized to the site of delamination, or trailing edge in round NCC as compared to other regions of the cell (Fig. 2B). In NCC which delaminate by EMT, actin and tubulin were broadly distributed across the cell (Fig. 2B). These cytoskeletal features and actin and tubulin distribution were consistently observed in 12 round NCC (localization found in 8/12 cells, SEM = 0.1361) and 13 EMT NCC (broad distribution indicating no localization at the trailing edge in 11/13 cells, SEM = 0.1001) compiled from four different embryos, and the difference as determined by t test analysis was statistically significant (P-value of 0.0077). Collectively, these data demonstrate that some NCC delaminate by cell extrusion in parallel with classic EMT.

The Neuroepithelium Contains Higher Tissue Stress in Regions of Delamination.

The cell extrusion pathway is activated by mechanical induction from physical tissue stress such as internal pressure and edge tension caused by overcrowding in an epithelium (47, 69–71, 73). During neurulation, the neuroepithelium, which is a single-cell-layered pseudostratified epithelium, elevates and folds to form the neural tube. At the same time, in concert with embryo growth and elongation, the neuroepithelium rapidly proliferates. These dynamic changes would be expected to generate high levels of tissue stress and overcrowding that need to be relieved to maintain tissue homeostasis (74, 75). NCC delamination occurs in an anterior–posterior wave along the neuraxis, and we reasoned that perhaps overcrowding within the neuroepithelium drove the delamination of some NCC via cell extrusion in parallel with classic EMT. We therefore calculated cell density as a readout of overcrowding within three distinct regions of the neuroepithelium, in cranial tissue sections from 5- to 6-somite stage embryos (n = 6), when the wave of delamination in the cranial region has just commenced. The three regions were defined as the dorsal most portion of the neuroepithelium from which NCC delaminate, the ventral region closest to the midline and the medial region that lies in between (Fig. 3 A and B). Cell density was measured in each region by quantification of cells based on DAPI staining of nuclei using Cellpose (76). Only the first 20 slices of the z-stack were used in each image to prevent overcounting. The total number of cells in the region was then divided by the volume of the 20-slice z-stack for normalization. The neuroepithelium is axially symmetrical so two dorsal, two ventral, and two medial regions were measured per section and then averaged. When plotted, the dorsal and medial regions were shown to be higher in density compared to the ventral regions, consistent with a crowding in the dorsal and medial regions during neurulation (Fig. 3C). These results suggest that cell density does vary along the neuroepithelium during delamination. The observation of elevated cell density in the medial region is likely associated with the medial hinge points where the neuroepithelium is bending during neural tube closure.

Fig. 3.

Cell density levels vary in the neuroepithelium and between somite stages during NCC development consistent with the wave of delamination. (A) A max projection of transverse histological sections from an E8.5 Wnt1-Cre;R26R-mTmG embryo used for cell density measurements throughout the neuroepithelium. The purple ROI demarcates dorsal, red ROI demarcates medial, and green ROI demarcates ventral. (B) High-magnification images of the ROIs from which the cell densities were calculated. (C) Quantification of cellular density in each region of the neuroepithelium (Dorsal, Medial, Ventral) of 5- to 6-somite stage embryos (n = 6). The dorsal and medial regions of the neuroepithelium exhibit higher density than the ventral region. Given the variance that can occur between tissue sections and embryo stages, we performed a matched One-Way ANOVA between regions of the same tissue section and determined the differences in cell density to be statistically significant (P-value of 0.0122). Tukey’s multiple comparisons test was used between each region (Medial vs. Ventral P-value of 0.088). (D) Quantification of cellular density throughout the neuroepithelium of 4-, 5-, 6-, and 8-somite stage embryos. At the 4-somite stage, delamination is about to begin and the overall density of the neuroepithelium is comparatively low. At the 5- and 6-somite stage, delamination has commenced in the cranial region and the cellular density in the neuroepithelium is much higher and peaks, such that by the 8-somite stage, as delamination continues in the cranial region, and the wave of delamination progresses toward the trunk, cellular density has diminished compared to that in the 5- and 6-somite stage embryos, but still remains higher compared to 4-somite stage embryos. Nested One-Way ANOVA revealed a statistically significant difference across the 4 embryo stages (P-value of 0.0010). Tukey’s multiple comparisons test was used between each stage (4SS vs. 5SS P-value of 0.0012; 4SS vs. 6SS P-value of 0.0015; 4SS vs. 8SS P-value of 0.0256).

Delamination takes place in an anterior–posterior wave, and we next reasoned that cell density might also be changing over developmental time in accordance with the progression of delamination. We therefore compared the cellular density of the neuroepithelium in cranial tissue sections of 4-, 5-, 6-, and 8-somite stage embryos (Fig. 3D). At the 4-somite stage, delamination is just beginning in the head but is well underway at the 5- and 6-somite stages. Consistent with the wave of delamination, we observed an increase in cell density between the 4-somite stage to the 5- and 6-somite stages, where it peaked, before beginning to decrease at the 8-somite stage (Fig. 3D). These changes in cell density across the neuroepithelium over developmental time correlate with crowding and release or relief that is brought about by NCC delamination.

If cell extrusion can orchestrate NCC delamination as suggested by the cytoskeletal and cell density experiments, then higher pressure or tension should be evident in delaminating NCC. The Cellular Force Inference Toolkit (CellFIT) enables the mapping of relative internal pressure and edge–edge tension within cells of a confluent tissue and has previously been used to study the role of tissue stress and cell extrusion in wound healing (71, 77). We used Tissue Analyzer to segment cells in transverse histological sections of 4-8-somite stage Wnt1-Cre;R26R-mTmG embryos (SI Appendix, Fig. S5A) (78), and the segmented cells were then analyzed with CellFIT to calculate tissue stress. Higher internal pressure and edge–edge tension were consistently observed within NCC partially attached to the neuroepithelium, i.e., actively delaminating (SI Appendix, Fig. S5B). We consistently observed this regionalization of tension and pressure in 28 sections acquired from four different embryos. Therefore, this suggests that mechanical tissue stress, which is required to induce cell extrusion, is present during NCC delamination.

Regulators of the Cell Extrusion Pathway Are Expressed in Premigratory NCC.

Activation of the cell extrusion pathway by physical tissue stress is facilitated by the mechanosensitive ion channel, PIEZO1 (14, 47–49). Although the complete downstream pathway following PIEZO1 activation remains to be elucidated (14), during apical extrusion, it is known that activation of PIEZO1 ultimately results in the production of sphingosine-1-phosphate (S1P) which then binds to its receptor S1P2 on neighboring cells (47–49, 79). S1P2 activates Rho signaling through ARHGEF1 to target the localization of microtubules, actin, and myosin within the neighboring cells (72, 79). To determine whether these PIEZO1 signaling components are expressed during NCC delamination, we analyzed our previously acquired single-cell RNA sequencing data of E8.5 mouse embryos which covered NCC development from premigratory to migratory stages (46, 57). Piezo1, S1p2, and Arhgef1 are expressed throughout NCC development (SI Appendix, Fig. S6 A–D), including during delamination (SI Appendix, Fig. S6E, specifically clusters 10 and 2), suggesting a potential role for PIEZO1-mediated mechanotransduction in NCC delamination.

We next validated the expression of the extrusion pathway regulator, Piezo1, in premigratory cells within the neuroepithelium. We focused on Piezo1 because its expression is specific to the extruding cell and live cell extrusion. Immunostaining demonstrated PIEZO1 was expressed in premigratory cells within the neuroepithelium prior to delamination (Fig. 4). PIEZO1 was dynamically expressed in different regions of the same embryo, aligning with cell extrusion playing a temporally specific role in delamination (Fig. 4A). Furthermore, PIEZO1 is known to localize in a cell at regions of cellular stress, and PIEZO1 was observed in the basal region of cells along the basal side of the neuroepithelium consistent with the detection of stress by CellFIT analysis (Fig. 4B) (48, 80–83). It is important to note that these images are max projected images from a Z-stack, therefore any expression above or below the nucleus would still be brought forward by max projection and the nucleus is not blocking visualization of expression. A total of 19 sections from three different embryos were immunostained for PIEZO1. Overall, these data indicate the components of the cell extrusion pathway are expressed in NCC during delamination.

Fig. 4.

PIEZO1 is dynamically expressed in NCC during delamination. Immunostaining for PIEZO1 and GFP together with DAPI on transverse histological sections of an E8.5 Wnt1-Cre;R26R-mTmG embryo. (A) The box indicates the location of the original tissue section from which the higher magnification Inset image on the Right was taken. The expression of PIEZO1 can be found within GFP-positive cells in the neuroepithelium as seen in the more anterior section. In the more posterior section where delamination has not begun, PIEZO1 is not expressed in the neuroepithelium suggesting the expression of PIEZO1 is highly dynamic. (B) PIEZO1 expression is noticeably localized to the basal side of the cell within the neuroepithelium during NCC delamination. These images are from the more anterior section of A as denoted by the dotted line box. A cell is outlined based on GFP expression to better display the basal localization. Importantly, these are max projected images so the nuclei are not visually blocking any underlying expression of PIEZO1 in anterior regions of the cells. The arrow points toward the expression of PIEZO1 within the outlined cell. The schematic provides an additional representation of expression matching the outlined cell in the tissue section image.

The Cell Extrusion Pathway Modulates Delamination of the Round NCC Population.

If PIEZO1-mediated cell extrusion is a mechanism by which NCC can delaminate, then modulating PIEZO1 activity should affect their delamination and consequently the number of migratory NCC. To test this model, CD1 embryos were cultured in the presence of a commonly used inhibitor of PIEZO1, GsMTx4, for 4 h beginning at E8.25, which is prior to the start of cranial NCC delamination. Following embryo culture, the total number of SOX10-positive cells, a marker of migratory NCC, was quantified to determine how many NCC had delaminated (Fig. 5A). The SOX10 population was then normalized to the total number of cells in the head as determined by DAPI staining. The GsMTx4-treated embryos displayed a significant decrease in NCC delamination compared to vehicle-treated stage-matched littermates (Fig. 5B). To confirm this decrease in delamination was due to perturbation of cell extrusion and thus loss of the round NCC population, we repeated this experiment with Wnt1-Cre;R26R-mTmG embryos. Treating E8.25 transgenic embryos with GsMTx4 diminished the number of round NCC outside of the neuroepithelium (Fig. 5 C and D). To ensure the decrease was not caused by an elevation in cell death brought about by GsMTx4 treatment, we performed TUNEL staining on tissue sections from the cultured Wnt1-Cre;R26R-mTmG embryos (SI Appendix, Fig. S7A). We did not detect any differences in TUNEL staining between the control and GsMTx4-treated embryos (SI Appendix, Fig. S7A). Quantification of TUNEL intensity at the dorsal region of the neuroepithelium where the NCC delaminate confirmed this assessment (SI Appendix, Fig. S7B). Altogether, these experiments demonstrate that PIEZO1-mediated mechanotransduction is required for NCC delamination by extrusion. Notably, treatment of embryos with GsMTx4 did not completely inhibit NCC delamination, and those NCC that were found outside of the neuroepithelium had an elongated morphology and protrusions, suggesting PIEZO1 does not affect classic EMT delamination.

Fig. 5.

Inhibiting cell extrusion regulator, PIEZO1, results in loss of the extruded cell population and fewer migratory NCC. (A) Immunostaining for SOX10 and DAPI staining of E8.5 CD1 embryos following 4-h culture with drug vehicle (control) or 1.5 µM GsMTx4. The experiment was performed on three different litters. (B) Quantification of the total number of migratory NCC (SOX10+) normalized to the total number of cells (DAPI) in the cranial regions of immunostained CD1 embryos exemplified in (A). Embryos treated with 1.5 µM GsMTx4 had fewer migratory NCC compared to vehicle-treated control embryos suggesting PIEZO1 inhibition results in a decrease in the number of NCC that undergo delamination. Culture and analysis were performed using embryos from 3 litters. We performed a paired t test analysis of stage-matched littermate embryos to account for any experimental and developmental variability (P-value of 0.028). (C) Immunostaining for GFP and DAPI staining of E8.5 Wnt1-Cre;R26R-mTmG embryos following 4-h culture with drug vehicle (control) or 1.5 µM GsMTx4. Arrows indicate the extruded cells present in the control embryos, whereas no extruded cells were present in the PIEZO1 inhibitor–treated embryos. (D) Quantification of the total number of extruded NCC normalized to the total number of delaminating NCC in the cranial ROI (box) of immunostained Wnt1-Cre;R26R-mTmG embryos, as exemplified in (C). Embryos treated with 1.5 µM GsMTx4 had fewer extruded NCC compared to vehicle-treated control embryos suggesting PIEZO1 inhibition leads to a decrease in the number of NCC that undergo cell extrusion.

Consistent with this model, we hypothesized that activation of PIEZO1 would conversely enhance cell extrusion-mediated NCC delamination. We therefore treated E8.25 CD1 embryos with the PIEZO1-specific chemical agonist, Yoda1 (84) for 4 h in culture, and we observed a significant increase in delaminated NCC as compared to vehicle-treated stage-matched littermates (Fig. 6 A and B). Furthermore, when E8.5 Wnt1-Cre;R26R-mTmG embryos were cultured in the presence of Yoda1, more round NCC were observed outside the neuroepithelium at the most recent position of delamination (Fig. 6 C and D). To ensure the increase was not caused by an elevation in proliferation as a consequence of Yoda1 treatment, we performed pHH3 staining on the cultured Wnt1-Cre;R26R-mTmG embryos (SI Appendix, Fig. S7C). We did not detect any differences in proliferation between the control and Yoda1-treated embryos (SI Appendix, Fig. S7D). Quantification of the number of cells expressing pHH3 in the head and more specifically pHH3 in the NCC, confirmed proliferation was not affected by Yoda1 treatment (SI Appendix, Fig. S7D). Taken together, GsMTx4 antagonism and Yoda1 agonism of PIEZO1 activity indicate that some NCC can delaminate from the neuroepithelium by PIEZO1-induced cell extrusion, which can be identified and quantified by the presence of transiently round NCC.

Fig. 6.

Activating cell extrusion regulator, PIEZO1, increases the total number of migratory NCC. (A) Immunostaining for SOX10 and DAPI staining on E8.5 CD1 embryos cultured for 4 h with drug vehicle (control) or 75 µM Yoda1. (B) Quantification of the total number of migratory NCC (SOX10+) normalized to the total number of cells (DAPI) in the cranial regions of the immunostained CD1 embryos exemplified in (A). Embryos treated with 75 µM Yoda1 had an increased number of migratory NCC compared to vehicle-treated control embryos suggesting PIEZO1 activation leads to increased NCC delamination. Culture and analysis were performed on embryos from 3 litters. We performed a paired t test analysis of stage-matched littermate embryos to account for any subtle experimental and developmental variability (P-value of 0.040). (C) Immunostaining for GFP and DAPI staining of E8.5 Wnt1-Cre;R26R-mTmG embryos cultured for 4 h with drug vehicle (control) or 75 µM Yoda1. Arrows indicate the extruded cells present in the control and PIEZO1 agonist–treated embryos. (D) Quantification of the total number of extruded NCC normalized to the total number of delaminating NCC in the cranial ROI (box) of immunostained Wnt1-Cre;R26R-mTmG embryos, as exemplified in (C). Embryos treated with 75 µM Yoda1 had more extruded NCC compared to vehicle-treated control embryos suggesting PIEZO1 activation leads to an increase in the number of NCC that undergo cell extrusion.

S1P and S1P2 signaling are considered downstream regulators of cell extrusion and we detected S1P kinase (SphK1) and S1P2 expression in NCC in our scRNA-seq data of early NCC formation, delamination, and migration (46, 57). However, it has previously been suggested that S1P signaling and S1P2 do not regulate basal cell extrusion events like NCC delamination. Cells that extrude basally have reduced levels of S1P compared to apical extruding cells (79). In addition, inhibiting S1P2 does not appear to prevent or induce basal extrusion (79, 85). Therefore, to test this idea and determine whether S1P signaling functions in NCC delamination, we cultured E8.25 CD1 embryos for 4 h in the presence of the S1P2-specific inhibitor, JTE-013 (SI Appendix, Fig. S8A). However, no measurable differences in the number of SOX10+ migrating NCC were found in comparisons between inhibitor-treated and vehicle-treated stage-matched littermates (SI Appendix, Fig. S8B). Therefore, S1P2 does not facilitate NCC delamination, consistent with previous suggestions that S1P signaling does not regulate the process of basal cell extrusion. There are, however, five S1P receptors, and to evaluate whether another S1P receptor could function in this process, we interrogated our scRNA-seq data (46, 57). Analysis of the expression of all five S1P receptors determined no other receptors were expressed within the NCC population (SI Appendix, Fig. S8C). Thus, S1P signaling does not function in this model of basal delamination.

Discussion

Our results have elucidated an additional mechanism by which NCC delaminate during mammalian development. Typically, NCC undergo EMT and exit the neuroepithelium as mesenchymal cells; however, we have identified a subpopulation of NCC that delaminate via PIEZO1-mediated cell extrusion, prior to transitioning to a mesenchymal state. Altogether, our data suggest the following cellular dynamics occur during this type of NCC delamination in mammals: 1. An increase in cell density in the neuroepithelium results in a build-up of tissue stress. 2. The ensuing crowding and tissue stress trigger PIEZO1 to induce the extrusion of ~20 to 30% of cranial NCC as round cells. 3. After a short period, the round extruded cells complete the transition to a mesenchymal state and begin to migrate. 4. At the same time, the remaining premigratory NCC transition via classic EMT to a mesenchymal state in concert with their delamination from the neuroepithelium (Fig. 7).

Fig. 7.

NCC delaminate by cell extrusion in parallel with classic EMT. An illustration of the proposed model for NCC delamination in mammalian development beginning with younger stages of development at the top and progressing to older stages while focused on the same region of the neuroepithelium. 1. An increase in cell density in the neuroepithelium results in a build-up of tissue stress. 2. The ensuing crowding and tissue stress trigger PIEZO1 to induce the extrusion of ~20 to 30% of cranial NCC as round cells. 3. After a short period, the round extruded cells complete the transition to a mesenchymal state and begin to migrate. 4. At the same time, the remaining premigratory NCC transition via classic EMT to a mesenchymal state in concert with their delamination from the neuroepithelium.

While our data indicate that NCC can delaminate by two different mechanisms, our previously published scRNA-seq data suggest the capacity for differentiation into NCC derivatives is not restricted by whether the NCC delaminates via extrusion or classic EMT (46). We previously identified two intermediate transition populations of NCC during their delamination, which were denoted as clusters 10 and 2 (46) (SI Appendix, Fig. S6E). Piezo1 is specifically expressed in cluster 2, suggesting cluster 2 contains the extruding NCC population (SI Appendix, Fig. S6E). In agreement with cluster 2 containing the extruding NCC population, Zeb2 expression was absent from this cluster (SI Appendix, Fig. S1D). Pseudotime trajectory analysis performed on this scRNA-seq data determined that these intermediate transition populations are not lineage restricted (see figure 3 in ref. 46). Therefore, it is unlikely that the extruding NCC population is limited to specific fates at this developmental point, although this remains to be definitively tested.

Time-lapse imaging and in situ identification of round cells at the onset of NCC delamination in a given axial region, implicated cell extrusion as playing a potentially pivotal role in the initial exodus of NCC. These extruded NCC appear to be some of the first cells to break through the basement membrane and could therefore act as leader cells that modify the surrounding extracellular matrix and communicate with follower cells. In the future, it will be interesting to define whether the extruded NCC operate as leader cells once migration commences.

Our data have demonstrated that cell extrusion occurs during cranial NCC development in the mouse, but the question remains as to whether extrusion is specific to mammalian development. We hypothesize NCC extrusion occurs in other model organisms based on the presence of round cells in other studies (86–88). For example, experiments in chicken embryos that focused on the role of FGFR1 in early trunk NCC migration noted nonpolar round NCC that stall due to a loss of FGFR1-directed orientation (87). Although this was not explored further, the timing between delamination and polarization of trunk NCC by FGFR1 signaling is consistent with the stalling of extruded cells observed in our time-lapse analysis. In zebrafish, Prickle1 loss-of-function results in persistent E-cadherin expression in migratory NCC, which is indicative of incomplete or hybrid EMT (86). Although these NCC expressed E-cadherin, delamination took place with a round morphology followed by blebbing (86). It is therefore possible these cells delaminated by extrusion, which does not require complete downregulation of E-cadherin (51, 52, 56). While these studies in chicken and zebrafish align with our data on NCC extrusion, this cellular phenomenon needs to be evaluated more thoroughly to be conclusive and considered an evolutionarily conserved mechanism.

In summary, we have determined that cell extrusion plays an important role in mammalian cranial NCC delamination, implying it may impact craniofacial development and the pathogenesis of craniofacial disorders. Understanding the overall significance of NCC extrusion for craniofacial development is challenging because the round morphology is transient and there is no known marker to differentiate extrusion populations for long-term lineage tracing. Based on the percent of round cells identified in our morphology analysis (Fig. 1D), we estimate ~20 to 30% of NCC delaminate via cell extrusion which aligns with the number of NCC expressing Piezo1 in the transition state cluster 2 (SI Appendix, Fig. S6E). Consistent with the idea that cell extrusion is critical for craniofacial development, previous research has shown cell extrusion facilitates secondary palatal shelf fusion (89). Furthermore, two recently identified patients with biallelic mutations in Piezo1, present with distinct craniofacial anomalies including a flat facial gestalt and infraorbital hypoplasia (90). These phenotypes are suggestive of a defect in neural crest cell development, perhaps a deficiency in cell extrusion leading to a reduction in the number of migratory NCC as was observed in our cultured mouse embryos in which PIEZO1 activity was inhibited. In agreement with these findings, it was recently published that conditional knockout of Piezo1 with the Wnt1-Cre transgenic line resulted in craniofacial abnormalities including malocclusion and domed heads (91). While it is exciting to see these craniofacial differences in the conditional knockouts, the effect of Piezo1 loss-of-function on NCC extrusion and delamination was not examined. Furthermore, it must be noted that Wnt1-Cre mediates the excision of genes in the NCC population after specification and formation have taken place (44). Additional studies are therefore needed to facilitate a better understanding of how Piezo1 and cell extrusion fully impact NCC and craniofacial development.

Finally, cell extrusion is widely studied in the pathogenesis of cancer metastasis (14, 49). However, until now, the various extrusion models utilized in the field have focused on apical extrusion and the loss of key signaling components that result in a switch to basal extrusion (14, 49, 72, 79, 92, 93). NCC delamination occurs via basal extrusion, providing an excellent model for further elucidation of basal-specific extrusion signaling mechanisms. In summary, we have uncovered cell extrusion to be an important additional mechanism driving mammalian NCC delamination, which furthers our understanding of NCC development with implications for craniofacial morphogenesis and cancer research.

Materials and Methods

Animal Husbandry.

All animal experiments were performed in accordance with Stowers Institute for Medical Research Institutional Animal Care and Use Committee approved protocol #2022-143. Animals were housed in designated facilities on a 14-h light and 10-h dark cycle. Strains and transgenic lines of mice used included CD1, FVB, and Wnt1-Cre;R26R-mT/mG. Wnt1-Cre (H2afv^{Tg(Wnt1-cre)11Rth}Tg(Wnt1-GAL4)11Rth/J) mice were obtained from The Jackson Laboratory (stock #003829) and maintained as male heterozygotes. Males were mated to female R26R-mTmG [Gt(ROSA)^{26Sortm4(ACTB-tdTomato, -EGFP)Luo}/J] mice obtained from The Jackson Laboratory (stock #007576) to generate Wnt1-Cre;R26R-mT/mG embryos.

Embryo Collection.

Timed mated females were checked for the presence of a vaginal plug and denoted embryonic day (E) 0.5. Embryos were collected on the designated day for the required stage, following cervical dislocation of the mother. A detailed description of embryo dissection can be found in Chapter 18 “Live Imaging of the Dynamics of Mammalian Neural Crest Cell Migration” of Methods in Molecular Biology: Craniofacial Development (45).

Time-Lapse Imaging.

Time-lapse imaging and image analysis were performed as previously described (45). In brief, embryos were cultured in 75% rat serum, 25% Fluorobrite DMEM with GlutaMAX supplemented with 1% penicillin-streptomycin at 37 °C in a 5% O2/5% CO2/N2 environment. Images were acquired on an inverted spinning disk confocal microscope (Nikon CSU-W1) with an sCMOS camera and piezo stage. A z-stack was acquired every 10 min in the data shown. Analysis was performed in Fiji and the 1 µm z slices taken in each z-stack were max projected. StackReg was used to control for any drift between z-slices and time points.

All other descriptions of histology, immunostaining, fixed tissue imaging, electron microscopy, quantification of delamination morphologies, analysis of cytoskeleton protein localization, quantification of average intensity of vimentin, E-cadherin, and TUNEL, measuring relative tissue stress, measuring cellular density of the neuroepithelium, single-cell RNA sequencing and bioinformatic analyses, roller culture and PIEZO1 agonist and antagonist treatment, quantification of migratory NCC in PIEZO1 agonist and antagonist treated embryos, quantification of extruding NCC in PIEZO1 agonist and antagonist treated embryos, and quantification of proliferation in Yoda1-treated embryos are provided in SI Appendix, SI Methods.

Supplementary Material

Appendix 01 (PDF)

Movie S1.

Timelapse imaging of NCC exiting the neuroepithelium in a Wnt1-Cre;R26R-mTmG embryo at E8.5 reveals a unique mechanism for delamination that resembles cell extrusion. A z-stack was acquired every 10 minutes for 130 minutes. Arrows point out the isolated round (extruding) cells. The neuroepithelium is labeled at the top and the orientation of the embryo is denoted by the letters at the bottom right. D = Dorsal; V = Ventral; C = Caudal; and R = Rostral.

Data, Materials, and Software Availability

All original data underlying this manuscript can also be accessed from the Stowers Institute for Medical Research Original Data Repository at https://www.stowers.org/research/publications (94). All other data are included in the article and/or supporting information.