Cytoplasmic localization of GRHL3 upon epidermal differentiation triggers cell shape change for epithelial morphogenesis

Epithelial cell shape change is a pivotal driving force for morphogenesis of complex three-dimensional architecture. However, molecular mechanisms triggering shape changes of epithelial cells in the course of growth and differentiation have not been entirely elucidated. Grhl3 plays a crucial role as a downstream transcription factor of Wnt/β-catenin in epidermal differentiation. Here, we show Grhl3 induced large, mature epidermal cells, enriched with actomyosin networks, from embryoid bodies in vitro. Such epidermal cells were apparently formed by the simultaneous activation of canonical and non-canonical Wnt signaling pathways. A nuclear transcription factor, GRHL3 is localized in the cytoplasm and cell membrane during epidermal differentiation. Subsequently, such extranuclear GRHL3 is essential for the membrane-associated expression of VANGL2 and CELSR1. Cytoplasmic GRHL3, thereby, allows epidermal cells to acquire mechanical properties for changes in epithelial cell shape. Thus, we propose that cytoplasmic localization of GRHL3 upon epidermal differentiation directly triggers epithelial morphogenesis.

pithelial morphogenesis directed by multiple cellular processes, such as cell shape changes, proliferation, and migration, involves the formation of complex threedimensional architecture as seen, for example, in the formation of the mammalian neural tube 1 . Coordinated cell shape changes, including contraction and elongation along the apical-basal axis, play a pivotal role in epithelial morphogenesis 2,3 . Such cell deformations are controlled by cellular mechanical stress and tension at the cell surface primarily through intracellular contractile actomyosin networks 1,[4][5][6][7] . However, the mechanisms initiating the morphogenesis of epithelial cells in coordination with the timing of specification, i.e. cell growth or differentiation during development, are relatively unknown.
The epidermis constitutes the outermost epithelial layer that wraps the entire body and changes the body form. During epidermal development, the single ectodermal sheet is largely specified into neural and surface ectoderm (SE), which is an immature embryonic state of epidermis 8 . Such temporal SE cells subsequently commit into periderm and finally form mature epidermis, the outer component of the skin 8 . Epidermal specification is initiated by the signaling of several growth factors. Bone morphogenetic protein signaling has been shown to direct epidermal specification in frog and zebrafish embryos 9 . Wnt signaling has also been suggested to be involved in epidermal specification in chick and mouse embryos [10][11][12][13] . Notably, we have found that during neural tube closure the canonical Wnt signaling pathway progressively specifies SE fate at the neural plate border, where neither surface nor neural cells are specified as uncommitted ectodermal progenitors 14 . These findings have led to the hypothesis that cell fate specification of SE during neurulation may be intimately linked to the epithelial morphogenesis of primary neurulation, which is governed by the non-canonical Wnt pathway involving planar cell polarity (PCP) genes 15 . However, little is known about what, when, and how molecular mechanisms control the coordination of epidermal fate decision and PCP-mediated epithelial morphogenesis.
The Grainy head family of transcription factors plays a highly conserved role in epithelial tissue development and remodeling in the animal kingdom 16,17 . Since epithelial morphogenesis is one of the major driving forces of neurulation [18][19][20] , the mammalian Grainy head family, encoding Grainyhead-like (Grhl) factors, is crucial for neural tube formation 21,22 . During mouse primary neurulation, Grhl3 acts as a downstream effector of Wnt/βcatenin signaling to direct the specification of SE, a developmentally temporal structure of epidermis 8,14,23,24 . However, molecular mechanisms underlying Grhl3-dependent epidermal differentiation in any subsequent epithelial morphogenesis remain poorly understood.
In this study, we demonstrate how the differentiation of epidermal cells intimately links to cell shape changes through Grhl3. Mechanistically, GRHL3, which has been considered as a transcription factor, appears to localize in the cytoplasm upon differentiation over time to activate non-canonical Wnt signaling. Such GRHL3-induced epidermal cells acquire mechanical properties necessary for cell shape change through actomyosin networks. This finding provides unique insights into the temporal mechanisms involved in the initiation of epithelial morphogenesis in coordination with cell specification during animal development.

Results
Grhl3 induces large and mature epidermal cells. During primary neurulation, Grhl3 can specify cellular fate into SE from ectodermal progenitor cells in neural folds 14 . To analyze the precise molecular mechanisms underlying epidermal differentiation by Grhl3, a two-step in vitro culture system was exploited (Fig. 1a) 25 . Embryonic stem (ES) cell-derived embryoid bodies (EBs) consisted of central undifferentiated progenitor cells and TROMA-1 (keratin 8) positive peripheral epidermal cells, which are formed without any additional factors (Fig. 1b). In this system, the overexpression of Grhl3 cDNA induced epidermal cells within central EBs that were distinct from those in the periphery of EBs by control vector (pgk-neo) transfectants (Fig. 1b, c, g; n = 33/33 100%). To explore the role of Grhl3 in more detail, Grhl3-induced epidermal cells were characterized (Fig. 1h, i, Supplementary Fig. 1a-g); histological examination revealed these to be large and multinucleated (Fig. 1h, i). Molecular analyses using epidermal markers indicated that such large multinucleated epidermal cells were more mature than epidermal cells in the periphery of EBs ( Supplementary Fig. 1a-h). In this regard, a marker for the neural crest, SLUG, was not expressed in large, mature (LM)-epidermal cells ( Supplementary Fig. 1i, j). Furthermore, LM-epidermal cells were highly enriched with actomyosin as well as keratin bundles ( Supplementary Fig. 1g, l, m). These findings indicate that Grhl3 can induce distinct LMepidermal cells in the central region of EBs.
To identify whether Grhl1, another member of the Grhl family, could induce LM-epidermal cells in a similar manner, we overexpressed Grhl1 cDNA in EBs and found that Grhl1 did not induce LM-epidermal cells efficiently (Fig. 1d, g). However, Grhl1 cDNA appeared to induce defective types of LM-epidermal cells, designated as "solitary" and "scattered" epidermal cells, in EBs ( Supplementary Fig. 1n, o). Afterward, we defined these two epidermal cell types cytomorphologically as follows: "solitary" epidermal cells consisted of a single epidermal cell in isolation but not multinucleated in EBs, while "scattered" epidermal cells formed as aggregates in EBs but not in multinucleated cells ( Supplementary Fig. 1n, o). "None" means that TROMA-1-positive epidermal cells were found in EBs but not in the periphery ( Supplementary Fig. 1k, p, Fig. 1b). Concurrently, a dominant-negative form of Grhl3 cDNA (DN-Grhl3), which can inhibit the activity of the Grhl family 26 , reduced the number of LM-epidermal cells induced by Grhl3 cDNA (Fig. 1e-g). These findings indicate that Grhl3 has a distinct role in inducing "LMepidermal cells" in EBs in vitro.
Since the transcriptional activity of CP2 transcription factors, which include the Grhl family, may be regulated according to their subcellular localization 27 , we hypothesized that the activity of GRHL3 in the formation of LM-epidermal cells may also similarly be controlled according to its subcellular localization. To explore this idea, we examined GRHL3 expression and found that GRHL3 protein was located in the cytoplasm of LM-epidermal cells (Fig. 3a-c, Supplementary Fig. 2  (d, f) at E4.5 (c, d), and E5.5 (e, f), respectively. Hematoxylin-eosin sagittal sections. g-l Immunohistochemical analyses of whole embryos at E3.5 in wild type (g, i, k) and Tg (CAG-Grhl3) (h, j, l). Phalloidin (magenta; g, h), CLAUDIN4 (magenta; i, j) and TROMA-1 (magenta; k, l), and DAPI (cyan; g-j, blue; k, l) stains. Multinucleate cells are seen in the inner cell mass region (arrows; h, j). m Schematic protocol for the electroporation of vectors, CAG-EGFP with or without CAG-Grhl3, into the embryo proper at E8.5. n, o Expression of CAG-EGFP vector in the surface ectoderm (SE) after 16 h culture in vitro. Anti-EGFP (green) and TROMA-1 (magenta). p-u Immunohistochemistry of electroporated embryos after 24 h culture in vitro. EGFP (green), phalloidin (magenta), and DAPI (blue). Cell size enlargement and F-actin accumulation are seen in the CAG-Grhl3 electroporated SE (s-u). Representative images from more than two independentexperiments. Scale bars represent 10 (p-u), 50 (b, g-l), 100 (c-f), and 200 μm (n) localization during epidermal maturation, a HaloTag system was designed to image protein localization in living cells (Fig. 3a-c) 28 . After a 48 h culture, HaloTag-GRHL3, as visualized by TMRligand, was found mainly localized within the nucleus (Fig. 3a, c, Supplementary Fig. 2). Strikingly, a further 24 h culture (72 h in total) shifted GRHL3 expression from the nucleus to the cytoplasm (Fig. 3b, c, Supplementary Fig. 2).
To understand the mechanisms involved in the localization of GRHL3, we analyzed the subdomains directing cytoplasmic localization using a GRHL3-EGFP fusion protein with N-  terminus, middle region, and C-terminus, respectively (Fig. 3d, Supplementary Fig. 3a, b). We found GRHL3 was able to localize to the cytoplasm through the N-terminus, CP2 domain, and Cterminus, and to the nucleus through the CP2 domain and Cterminus ( Supplementary Fig. 3a, b). Immunohistochemistry with specific polyclonal antibodies against mouse GRHL3 confirmed that GRHL3 localized to both the nucleus and cytoplasm in LMepidermal cells ( Supplementary Fig. 3c-j). The above findings together support the hypothesis that GRHL3 localization shifts from the nucleus to the cytoplasm during epidermal maturation ( Fig. 3a-c).
To examine if the subcellular localization of GRHL3 is crucial to the induction of LM-epidermal cells, we analyzed deletion constructs of mouse Grhl3 lacking C2, transcriptional activation (TA), and CP2 domains as well as Ub-like folding domains ( Fig. 3d-f) and transfected these into ES cells. When the C2 (7-68 aa) or TA domain (30-96 aa) was deleted, induced LM-epidermal cells were significantly reduced in number compared to those induced in response to intact GRHL3 ( Fig. 3d-f; 29.4% [n = 5/ 17], 25.0% [n = 3/12], respectively). These findings suggest that the N-terminus is necessary for the induction of LM-epidermal cells. Additionally, the removal of the Ub-like folding domain also reduced the emergence of LM-epidermal cells (Fig. 3d-f). Furthermore, CP2 deletion severely reduced the induction of all three types of epidermal cells: LM-, scattered, and solitary epidermal cells (Fig. 3e, f). Collectively therefore, in addition to the CP2 DNA-binding domain, N-(C2 and TA) and C-termini (Ub-like folding) that involve cytoplasmic localization are necessary for the formation of LM-epidermal cells.
To estimate whether the subcellular localization of GRHL3 itself can directly control the induction of LM-epidermal cells, we modified GRHL3 localization by the inclusion of nuclear import and export signals (nuclear localization signal [NLS] and nuclear export signal [NES], respectively) in the GRHL3 protein and then analyzed the formation of LM-epidermal cells from EBs ( Fig. 3e, g-i) 29 . By adding an NLS sequence, the induction of LMepidermal cells was markedly reduced (Fig. 3g, i). Conversely, GRHL3 carrying the NES sequence did not affect the formation of LM-epidermal cells (Fig. 3h, i). The above results together support the hypothesis that cytoplasmic GRHL3 (C-GRHL3) is crucial for the induction of LM-epidermal cells.
Canonical and non-canonical Wnt pathways in LM-epidermal induction. The previous finding that Grhl3 acts as a downstream effector of Wnt/β-catenin signaling 14 led us to study whether this canonical Wnt pathway is involved in the induction of LMepidermal cells (Figs. 1, 2, 3). To test this hypothesis, the constitutively active form of β-catenin (β-cateninS37A) was transfected into ES cells (Fig. 4a). However, β-cateninS37A failed to induce LM-epidermal cells ( Fig. 4a; n = 0/31). Additionally, chemical modulators of canonical Wnt signaling, a Wnt agonist and an inhibitor were also employed. First, the Wnt agonist was not able to induce LM-epidermal cells ( Fig. 4b; n = 0/9). Next, the effects of a combination of chemical reagents and cDNA constructs (Grhl3 cDNA) on the induction of LM-epidermal cells were analyzed. Strikingly, neither an activator nor an inhibitor of the canonical pathway (Wnt agonist and FH535, respectively) affected the emergence of LM-epidermal cells (Fig. 4e, f; n = 15, n = 17, respectively). Collectively, these findings suggest that canonical Wnt signaling is neither a downstream nor parallel pathway of Grhl3 for the induction of LM-epidermal cells. This led us to explore additional signaling pathways that may be necessary for the induction of LM-epidermal cells.
We subsequently tested the non-canonical Wnt pathway by means of chemical modulators such as an activator (Rho/Rac/ Cdc42 activator I) and inhibitors (H-89, SP600125, and NSC23766; Fig. 4c, d, g-i). Notably, activation of the noncanonical Wnt pathway significantly induced much larger sized LM-epidermal cells compared to those induced by Grhl3 alone (p < 0.05; Fig. 4c, d, g, h; n = 46). Moreover, inhibition of the noncanonical pathway repressed the formation of LM-epidermal cells induced by Grhl3 (Fig. 4i). To explore the involvement of the non-canonical Wnt pathway, the dominant negative form of Dsh (CAG-Dsh-DEP), which specifically repressed the PCP pathway, was exploited ( Fig. 4i). Consistently, the Grhl3-dependent formation of LM-epidermal cells was blocked by CAG-Dsh-DEP ( Fig. 4i). Thus collectively, the non-canonical Wnt pathway appears to be involved in the induction of LM-epidermal cells as downstream or parallel pathways of Grhl3.
Finally, the above findings led us to hypothesize that simultaneous activation of canonical and non-canonical Wnt pathways may lead to the formation of LM-epidermal cells. To test this concept, two cDNA vectors, harboring β-cateninS37A and Rac1, respectively, or the constitutively active form of Disheveled associated activator of morphogenesis 1 (Daam1) 19,30-32 , which links the PCP pathway to PDZ-Rhogef, were transfected simultaneously into ES cells ( Fig. 4j-m). Consequently, both transfected EBs were able to form LM-epidermal cells without Grhl3 cDNA (Fig. 4k, m; n = 3,8, respectively). Similarly, the simultaneous activation of both canonical and non-canonical pathways using chemicals was able to induce the formation of LM-epidermal cells ( Fig. 4n; n = 7). These results indicate that the simultaneous activation of these two pathways was involved in the formation of LM-epidermal cells. Moreover, these findings imply that the canonical Wnt pathway facilitates the differentiation of epidermal cells while activation of the non-canonical Wnt pathway facilitates cell shape changes such as increased cell size, multi-nucleation, and the enrichment of actomyosin networks.
To explore how GRHL3 activated the non-canonical Wnt pathway, the expression of VANGL2, a PCP component, was analyzed using a HaloTag system (Fig. 5a). The expression of HaloTag-Vangl2 was localized to the cytoplasm of an LMepidermal cell (Fig. 5a). Moreover, considerable amounts of cytoplasmic VANGL2 apparently co-localized with GRHL3 ( Fig. 5b, asterisks). To confirm the co-localization of GRHL3 Fig. 3 Identification of the GRHL3 structure contributing to LM-epidermal formation. a, b Embryonic stem (ES) cells transfected with CAG-HaloTag-GRHL3 were incubated with medium containing TMR-ligand (red). Hoechst (blue) nuclear stain. Wheat germ agglutinin (WGA, green). c The frequency of nuclear or cytoplasmic (cyto) localization of the GRHL3 protein observed in LM-epidermal cells using the HaloTag system after 48 h and 72 h culture. d Schematic illustration showing the protein domain organization of mouse GRHL3. e Schematic structures of GRHL3 with deletion mutations, or the insertion of NLS (nuclear import, PKKKRKV) or NES (nuclear export, LALKLAGLDI) signals are shown. f Summary of the induced frequency of epidermal cells after transfection of ES cells with various deletion constructs. TROMA-1-positive central epidermal cells were classified into three types: LM-, solitary, and scattered epidermal cells. **p < 0.01 (χ 2 test) g, h Examples of epidermal cells induced by CAG-NLS-Grhl3 (g) and CAG-NES-Grhl3 (h). Immunofluorescence with TROMA-1 (magenta) merged with DAPI staining (nucleus; blue). i The proportions of the formation of LM-epidermal cells after transfection of ES cells with CAG-Grhl3, CAG-NLS-Grhl3, or CAG-NES-Grhl3 plasmids. "n" indicates the number of EBs analyzed. **p < 0.01 (χ 2 test). Representative images and frequency from more than three independent experiments. Scale bars represent 50 (a), and 200 μm (g, h) and VANGL2, we analyzed protein-protein interaction in NIH3T3 cells using fluorescent protein fragments of monomeric Kusabira-Green (mKG; n = 17; Supplementary Fig. 4). Such expression studies demonstrated that GRHL3 was located in close proximity to VANGL2 so that GRHL3 may indirectly or directly interact with non-canonical Wnt molecules in LM-epidermal cells.
In addition, expression of Scrib, another PCP protein 33 , was specifically induced in the LM-epidermal cells (Fig. 5c). Coincidentally, the phosphorylation of non-muscle myosin light chain (pMLC), a marker for stress fiber formation 34 , was exclusively found in LM-epidermal cells (Fig. 5d). However, although the expression of HaloTag-β-catenin was localized to both the nucleus and cytoplasm of LM-epidermal cells, cytoplasmic βcatenin did not co-localize with GRHL3 (Fig. 5e, f). The aforementioned studies support the notion that GRHL3 may contribute to non-canonical Wnt signaling partly through expression of PCP proteins in the cytoplasm.     Cre recombinase, internal ribosomal entry site (IRES), and nuclear localized β-gal were knocked in Grhl3 in the frame of the ATG start site in the Grhl3 cre allele. Arrows indicate multinucleate surface ectoderm (SE) cells (a-c) in a Grhl3 cre/+ embryo. β-gal-GRHL3 (magenta) and VANGL2 (green) in a Grhl3 cre/+ embryo (e, f). g-p VANGL2 (green; g, h), CELSR1 (green; i, j), RHOA (green; k, l), SCRIB (green; m, n), phospho non-muscle myosin light chain (pMLC green; o, p) and DAPI (blue; g-p) in wild-type (g, i, k, m, o), and Grhl3 cre/ cre (h, j, l, n, p) embryos at E8.5. RHOA expression during cell division appeared to be unchanged in a Grhl3 cre/cre embryo (l). q-t Whole-mount in situ hybridization of Vangl2 mRNA in wild-type (q, s) and Grhl3 cre/cre (r, t) embryos at E8.5. u-x Molecular marker analyses in Tg(CAG-Grhl3) embryos at E3.5. Immunohistochemical analyses of RHOA (magenta; u, v), CELSR1 (magenta; w, x), and DAPI (cyan) in wild type (u, w) and Tg(CAG-Grhl3) (v, x), respectively. Representative images from more than two independent experiments. Scale bars represent 20 (u-x) and 50 μm (a-r) galactosidase antibody in Grhl3 cre/+ embryos, were evident in neural folds prior to neural tube closure and thereafter in dorsal SE cells 14 . In contrast, Grhl3-negative SE cells were distributed to the more lateral side of the epidermal layer (Fig. 6a-c) 14 . Further, F-actin and pMLC staining revealed that Grhl3-positive cells possessed assemblies of actomyosin bundles (Fig. 6a-d, o). Additionally, endogenous GRHL3 apparently localized to the nucleus and cytoplasm, as well as the cell membrane (Supplementary Fig. 5). Furthermore, multinuclear cells were sometimes found in Grhl3-positive (β-gal positive) SE cells (Fig. 6a-c; white arrows). Such studies suggest that Grhl3-positive dorsal epidermal cells during neurulation may be compatible with LM-epidermal cells derived from EBs in vitro (Figs. 1-4). Given that GRHL3 appeared to co-localize with VANGL2 in LM-epidermal cells in vitro (Fig. 5a, b), it was reasonable to postulate that Grhl3 would genetically interact with Vangl2 during neurulation (Fig. 6e-h). When we analyzed VANGL2 localization, we found that VANGL2 protein was detected in dorsal SE cells, specifically in the cell membrane and partly in the cytoplasm, but not in the nucleus (Fig. 6e-g). Similarly, expression of other PCP components, CELSR1, RHOA, and SCRIB, were also evident in the cell membrane of dorsal SE cells (Fig. 6i, k, m). In Grhl3 cre/cre mutant embryos, however, membrane-associated VANGL2 expression was not evident (Fig. 6h) despite Vangl2 transcripts not being dysregulated in Grhl3 cre/cre embryos ( Fig. 6q-t). Concordantly, the expression of other PCP components, CELSR1, RHOA and SCRIB, was reduced in Grhl3 cre/cre embryos, suggesting that expression of PCP components was down-regulated by the Grhl3 mutation (Fig. 6j, l, n). These findings suggest that GRHL3 is necessary for the correct membrane-associated localization of PCP components at the protein level. Consistent with a reduction of PCP-related molecules, pMLC expression was also down-regulated in Grhl3 cre/ cre embryos (Fig. 6o, p), suggesting the crucial role of Grhl3 in the formation of actomyosin networks through PCP molecules. Concurrent with the above Grhl3-deficient phenotypes, expression of the PCP molecules, RHOA and CELSR1, was markedly up-regulated at E3.5 in Tg(CAG-Grhl3) transgenic embryos ( Fig. 6u-x; n = 6, 2, respectively). These findings indicate that GRHL3 is necessary for the correct expression of PCP components.
To evaluate whether cytoplasmic GRHL3 was involved in epithelial morphogenesis, we generated genetically modified mice (Grhl3 NLS ), in which GRHL3 localized to the nucleus rather than the cytoplasm, by inserting an NLS sequence into the translational start site of the Grhl3 locus (Fig. 7a, b, Supplementary Fig. 6). Grhl3 transcripts were normally generated and full-length GRHL3 was translated in Grhl3 NLS/NLS embryos ( Fig. 7c-g,  Supplementary Fig. 7,8). Additionally, NLS-fused GRHL3 was mainly localized in the nucleus of SE cells in Grhl3 NLS/NLS embryos as compared with GRHL3 in the wild type (Fig. 7h-k). Strikingly, Grhl3 NLS/NLS embryos displayed neural tube defects (NTDs) at E12.5 (Fig. 8c). Moreover, differences in the degree of abnormalities were apparent between Grhl3 null (Grhl3 Δex2/Δex2 ) and Grhl3 NLS/NLS embryos (Fig. 8a-c). While spina bifida developed in all Grhl3-deficient embryos (Fig. 8b), the spina bifida phenotype of Grhl3 NLS/NLS appeared to be milder than that of the Grhl3-deficient mutation; a much lower sacral spina bifida emerged in 57.1% of Grhl3 NLS/NLS embryos (n = 12/21); a curved tail phenotype alone was present in 19%, which manifested as incomplete penetrance (Fig. 8c, o). Therefore, collectively, cytoplasmic GRHL3 is essential for correct neural tube closure.
In order to verify whether defects in Grhl3 NLS/NLS embryos were a result of canonical Wnt-dependent epidermal differentiation or non-canonical Wnt-dependent epithelial morphogenesis, we analyzed Grhl3 NLS/NLS embryos more closely (Fig. 8d-n). In E9.25 wild-type embryos, ectodermal progenitor cells that did not express epidermal or neural markers were seen at the neural plate border ( Fig. 8d; arrowheads) 14 . Conversely, in Grhl3-deficient embryos, progenitor cells were abolished (Fig. 8e) 14 . However, in Grhl3 NLS/NLS embryos, such progenitor cells appeared to be present and were similar to those in wild-type embryos ( Fig. 8f; arrowheads). These results indicate that epidermal specification occurred normally in Grhl3 NLS/NLS embryos. Next, we found that expression of PCP molecules and pMLC were abolished in Grhl3 NLS/NLS embryos as found in Grhl3 cre/cre embryos (Fig. 5,  Fig. 8g-n). Additionally, in agreement with the demonstration that Grhl3 is essential for wound repair (Supplementary Fig. 9) 35 , Grhl3 NLS/NLS embryos did not show repair of the hind-limb amputation unlike wild-type embryos (Supplementary Fig. 9; n = 5). These findings suggest that defects in Grhl3 NLS/NLS embryos are brought about by failure of cytoplasmic GRHL3-dependent epithelial morphogenesis involving the non-canonical Wnt pathway, but not nuclear GRHL3 (N-GRHL3)-dependent transcription involving the canonical Wnt pathway.
Since Grhl3 is known to activate the PCP pathway at the transcriptional level via Rhogef19 expression for promoting actin polymerization 35 , we verified whether failure in epithelial morphogenesis observed in Grhl3 NLS/NLS embryos was due to a reduction of Rhogef19 expression at the level of transcription or not (Fig. 8p-u). At E9.25 and E11.5, Rhogef19 transcripts were found in the SE (Fig. 8p, s). In agreement with the previous report, Rhogef19 transcripts were reduced in the SE of Grhl3deficient embryos (Fig. 8q, t). Conversely, transcripts were not reduced in the SE of Grhl3 NLS/NLS embryos (Fig. 8r, u). These additional findings support the concept that defects of epithelial morphogenesis in Grhl3 NLS/NLS embryos were not brought about by down-regulation of Rhogef19 transcripts through the transcriptional activity of N-GRHL3, but rather by the downregulation of PCP components at the protein level through cytoplasmic GRHL3. Fig. 7 Generation of an allelic series of mutations at the Grhl3 NLS knock-in locus. a Strategy for the production of an allelic series of mutations at the Grhl3 nuclear localization signal (NLS) mutant locus. A schematic representation of the Grhl3 wild-type allele, targeting vector, Grhl3 NLS-neo and Grhl3 NLS alleles. b Identification and characterization of mutant alleles by southern blot analysis. Genomic DNA samples from wild-type (Wt) and mutant mice were digested with NcoI and hybridized with the DNA probe (from +551 to +1194) to examine the correct 3′ recombination and excised pgk-neo cassette. The 3.9-kb and 2.5-kb bands represent wild-type (Wt) and targeted NLS-neo alleles. The mutant 2.5-kb band shifted up to 3.9 kb after Flpe excision in the NLS allele. c The Grhl3 NLS allele produced Grhl3 transcripts from the inserted site of NLS signal. Reverse transcription (RT)-PCR analysis with cDNA from wild-type, Grhl3 ΔEx2/ ΔEx2 and Grhl3 NLS/NLS mutant embryos at E11.5. The 171-bp products correspond to wild-type transcripts, whereas the 192-bp products correspond to mutant transcripts that contain an additional NLS signal sequence. d-f Whole mount in situ hybridization analysis of Grhl3 transcripts in wild-type (d), Grhl3 ΔEx2/ΔEx2 (e), and Grhl3 NLS/NLS (f) mutant embryos at E8.0. Grhl3 transcripts were clearly detected at the boundary between surface ectoderm and neuroectoderm in wild-type (d, arrows) and Grhl3 NLS/NLS mutant (f, arrows) embryos while they were undetected in the Grhl3 cre/cre embryo (e). g Western blots using the GRHL3 antibody aa195-211 ( Supplementary Fig. 3g, 5c) with protein extracts of embryonic nucleus. Bands of around 67 kDa, corresponding to full-length GRHL3 products, were evident in wild-type and NLS alleles but not Δex2 alleles. h-k Identification of nuclear-localized GRHL3 protein in Grhl3 NLS/NLS mutant embryos. Immunohistochemistry of GRHL3 (anti-GRHL3 ab; aa478-aa493; Supplementary Fig. 5e; green) and DAPI (magenta). GRHL3 protein was mostly distributed in the nuclei of SE cells of a Grhl3 NLS/NLS embryo (j, k), while it localized to the cytoplasm mainly and partly nucleus in SE cells of a wild-type embryo at E8.5 (h, i). Representative images from more than two independentexperiments. Scale bars represent 50 (h, j) and 200 μm (d-f) GRHL3 confers mechanical properties on epidermal cells. To explore if GRHL3-positive epidermal cells acquire mechanical cues necessary for epithelial morphogenesis, elastic properties of epidermal tissues were measured using micropipette aspiration experiments (Fig. 9a, b, Supplementary Fig. 10). Wild-type or Grhl3 heterozygous dorsal SE tissues, in which Grhl3 was expressed, were aspirated under 1.0 or 2.0 kPa with the same glass micropipette (100 μm diameter) for 1 min, and tissue deformation into the micropipette recorded (Fig. 9a, b, Supplementary  Fig. 10). Consequently, plotted total lengths of aspirated SE aggregates indicated that the elastic property of Grhl3-deficient SE tissues was less stiff than that of the wild type (Fig. 9b).
Next, we measured a Young's modulus on SE tissues during neurulation using an atomic force microscope (AFM; Fig. 9c-f). We determined that the Young's modulus of SE sheets of wild-type embryos was an average of 8.328 kPa (embryos n = 4; areas n = 10), while that of Grhl3 cre/cre mutant embryos was an average of 2.847 kPa (embryos n = 2; areas n = 6; Fig. 9c-f). Additionally, Young's modulus of SE cells in Grhl3 NLS/NLS embryos was an average of 5.796 kPa (embryos n = 3; areas n = 8). These data indicate that SE cells expressing Grhl3 or C-GRHL3 protein were stiffer than Grhl3-negative SE cells. Taken together, these results suggest that Grhl3 is essential for ensuring mechanical properties in epidermal cells against the tensile force required for epithelial morphogenesis.
Finally, to directly evaluate the cortical tension of LMepidermal cells induced by Grhl3 cDNA in vitro, a Young's modulus of TROMA-1-positive epidermal cells was measured with an AFM (Fig. 9g-l) 36 Fig. 8 Morphological and molecular marker analyses in Grhl3 NLS/NLS mutant embryos. a-c Gross morphology in wild-type (a), Grhl3 Δex2/Δex2 (b), and Grhl3 NLS/NLS embryos (c) at E12.5. d-n Whole-mount immunohistochemistry of wild-type (d,g,i,k,m), Grhl3 Δex2/Δex2 (e) and Grhl3 NLS/NLS embryos (f, h, j, l, n), at E9.25. TROMA-1 (magenta), N-cadherin (blue) and DAPI (green) in d-f. VANGL2 (green in g, h), CELSR1 (green; in i, j) RHOA (green in k, l), nonmuscle myosin light chain (pMLC; green in m, n) and DAPI (blue in g-n). Arrowheads indicate ectodermal progenitor cells that did not express epidermal or neural markers at the neural plate border. o Variations of neural tube defect phenotypes in Grhl3 NLS/+ crossed with Vangl2 Lp/+ or β-catenin +/− mutant mice from E11.5 to E14.5. The loop-tail (Vangl2 Lp ) allele has a strong planar cell polarity (PCP) phenotype during epithelial morphogenesis. Ct Curved tail. pu Expression analysis of Rhogef19 by whole-mount in situ hybridization in wild-type (p, s), Grhl3 cre/cre (q, t) and Grhl3 NLS/NLS (r, u) embryos and their transverse sections (s-u). Rhogef19 transcripts were detected in the surface ectoderm (SE) region of wild-type (s, arrows) and Grhl3 NLS/NLS (u, arrows) embryos but were absent in the surface region of the Grhl3 cre/cre embryo (t, white dotted lines). Representative images and frequency from more than two independent experiments. Scale bars represent 100 (k, l) and 200 μm (g-j, m, n, p-u) The stiffness of LM-epidermal cells induced by Grhl3 cDNA was therefore significantly higher than that of epidermal cells formed in the periphery (p < 0.05; one-tailed t-test). Thus, concurrent with the above in vivo study, Grhl3 is able to confer proper physical properties on epidermal cells.

Discussion
Our current findings indicate that N-GRHL3 directs epidermal differentiation from ectodermal progenitors and that this specification process is controlled by canonical Wnt signaling (Fig. 10) 14 . Upon epidermal differentiation, GRHL3 is localized in the cytoplasm and cell membrane, and is necessary for non-canonical Wnt-dependent cell shape changes (Fig. 10). Consequently, upregulation of PCP molecules generates biomechanical forces primarily mediated by actomyosin networks during epithelial morphogenesis, i.e. neurulation (Fig. 10). Thus, GRHL3 functions as a unique signaling trigger for both canonical Wnt-dependent epidermal differentiation in the nucleus and non-canonical Wntdependent epithelial morphogenesis in the cytoplasm and cell membrane. Cytoplasmic GRHL3 confers mechanical properties on epidermal cells for cell shape changes in coordination with cellular differentiation. Mechanical forces contribute to apical constriction and convergence toward the midline of the neural plate through actomyosin networks 37 . In this manner, mechanical forces drive neural tube closure processes such as bending and folding of the neural plate 15,38 . Therefore, it is highly probable that activation of PCP molecules directed by cytoplasmic GRHL3 facilitates actomyosin networks in epidermal cells and the resultant mechanical properties can direct epithelial morphogenesis and remodeling for neurulation.
We have identified two separate functions for GRHL3: as a nuclear transcription factor and as a cytoplasmic transducer of cell shape changes that may be mediated by distinct protein domains, transcriptional activation, DNA-binding CP2 domains, and Nand/or C-terminal regions, respectively. The two protein functions of GRHL3 are apparently controlled during the course of embryonic development. For many years, the processes of specification and morphogenesis, crucial during development, were thought to be controlled independently by different sets of molecules. In this respect, GRHL3 may be a unique factor that plays two different roles in epidermal specification and cell shape changes at the same time through the different subcellular localization of a single protein molecule. In this manner, GRHL3 can transmit and convert those signals used in epithelial differentiation to be useful for cell shape changes during development. To date, numerous lines of evidence have revealed that specification processes are initiated by extracellular signaling, with such triggering mechanisms intensively analyzed. In contrast, molecular entities that trigger cell shape changes have not yet been clearly identified.
With respect to the polyploidy-like phenotype induced by GRHL3, it is well known that the epidermis of mammalian skin becomes polyploidy during homeostasis [39][40][41] . In addition, wound-induced polyploidization of the epidermis has also only recently been discovered and analyzed in the mouse; 42 epidermal cells after injury are multinucleated and become enlarged in size in order for epithelial structure to recover quickly. As Grhl3 is also crucial for skin development as well as wound repair processes 35 , it may be rational that GRHL3 contribute to multinucleation and the enlargement of cell size during normal development as well as wound repair.  Fig. 10 Cytoplasmic localization of GRHL3 upon epidermal differentiation triggers cell shape changes. GRHL3 exerts two different functions related to epidermal fate specification and epithelial morphogenesis during the course of epidermal development. The former function is mediated by N-GRHL3 as a downstream transcriptional factor of β-catenin. The latter function is mediated by C-GRHL3 involving the non-canonical Wnt pathway, e.g. planar cell polarity molecules. The non-canonical Wnt pathway facilitates actomyosin networks and, consequently, alters physical properties such as the mechanical stiffness of cells necessary for epithelial morphogenesis. The cytoplasmic localization of GRHL3 during epidermal differentiation appears to be controlled by CP2, C-terminus, and N-terminus domains of GRHL3. Thus, this cytoplasmic localization of GRHL3 triggers cell shape changes in epithelial morphogenesis from epidermal differentiation during embryonic development. SE Surface ectoderm, NE Neural ectoderm At this stage, it remains to be determined how LM-epidermal cells induced by Grhl3 in vitro are relevant to in vivo epidermal cells. Phenotypes of transgenic mice and electroporated cells misexpressing Grhl3 in vivo may not be completely identical with those of in vitro LM-epidermal cells. However, clear similarities are present between in vitro and in vivo cells. For example, multinucleation and the up-regulation of F-actin appear to be common (Fig. 2). Additionally, the expression of PCP molecules was also up-regulated both in vivo and in vitro (Figs. 5, 6). To identify in vivo counterpart of LM-epidermal cells, further comprehensive studies between LM-epidermal cells in vitro and epidermal cells in vivo during mouse development would be necessary.
This study suggests that GRHL3 may trigger both cell differentiation and shape change, a hypothesis that may provide unique insights into why cell specification is coordinated with morphogenetic processes to construct a three-dimensional complex structure during animal development. The grainy head family plays a phylogenetically conserved role in epithelial morphogenesis in the animal kingdom 16,17,43 . The capability of dual functions in the form of a nuclear transcription factor and a cytoplasmic PCP transducer of GRHL3 may also be conserved in other species. In summary, our study sheds light on some of the processes involved in the initiation of epithelial morphogenesis and its coordination with cell specification during development.
Cell lines. G4 ES cells were obtained from Dr. Andras Nagy (Lunenfeld-Tanenbaum Research Institute at Mount Sinai Hospital in Toronto) and NIH3T3 cells were obtained from the Japanese Collection of Research Bioresources (JCRB, Japan).
Immunohistochemistry and in situ hybridization. Information of primary antibodies used in this study is described in Supplmentary Table. Specifically, regarding the antibody against mouse GRHL3, since the aa478-493 antibody displayed less background staining and recognized cytoplasmic GRHL3, the aa478-493 antibody is suitable to detect the translocation of GRHL3 from the cytoplasm into nucleus, i.e. it is appropriate to examine how much cytoplasmic GRHL3 is reduced or not. Consequently, the aa478-493 antibody was used to analyze the distribution of GRHL3 in Grhl3 NLS/NLS mutant embryos. In situ hybridization involving digoxigenin-labeled probes was conducted in a manner identical to that of Wilkinson 49 .
Generation of a nuclear localization signal knock-in mouse. All constructs used this study were generated by standard molecular cloning techniques. In brief, a fusion gene consisting of the 5′ noncoding region of the nuclear localization signal (NLS) 29 , the translational start site of the Grhl3 gene, and a neo cassette flanked with loxP was constructed (detailed procedures for this construction are available upon request). In the targeting vector, the lengths of the homologous region were 6.3 and 2.5 kb at the 5′ and 3′ sides of the insert, respectively. Homologous recombinant G4 ES cells and chimeric mice were generated by microinjection. To delete the loxP-flanked neo cassette, male chimeras were mated with C57Bl/6 background females carrying a Flpe gene driven by the chicken β-actin promoter 50 .
Generation of CAG-Grhl3 transgenic mice. Mouse Grhl3 cDNA fused to the CAG promoter was ligated to the LacZ cassette flanked by two loxP sequences 51 . The resultant construct, CAG-loxP-lacZ-loxP-Grhl3 cDNA (5 ng μl −1 ) was microinjected into the pro-nucleus of the fertilized eggs from superovulated CD-1 female mice and the injected eggs were cultivated in KSOM overnight then transferred to the pseudopregnant CD-1 44 . The resultant offsping are genotyped to identify Tg (CAG-lacZ-Grhl3) founders. Tg(CAG-Grhl3) embryos were obtained by crossing Tg (CAG-lacZ-Grhl3) mice with β-actin Cre mice 14 .
Mouse genotyping. Grhl3 NLS knock-in founders and their progenitors were routinely determined by PCR and confirmed, when necessary, by southern blots of genomic DNA samples prepared from mouse tails or yolk sacs. In PCR analyses, primers and lengths of products were as follows: knock-in NLS sequence in the Grhl3 allele Grhl3-cre (5′-AATTAAGAGACGAGTGGTCAGCAGCAGCGCCTG-3′) and Grhl3-cre wt rev (5′-ACCCTTACAAATTGCCGTGTGAATCCGGGC-3′), yielding 213 bp as the wild-type allele and 234 bp as the NLS knock-in allele.
Whole embryo electropolation ex utero. Ex utero whole-embryo (E8.75) electroporation was conducted as follows 56 . In brief, dissected tissues of ICR strains of mice were soaked in culture medium (50% rat serum and 50% Tyrode's saline buffer). To prepare embryos for electroporation, decidua was trimmed away and two types of DNA solution, CAG-EGFP (1 μg mL -1 ), with or without CAG-Grhl3 (1 μg mL -1 ), were injected into the amniotic cavity using a glass capillary. Electroporation was conducted at a voltage of 70 (5 cycles of 50 msec pulse length and 950 msec interval) in Tyrode saline buffer using a CUY21EDITII electroporater and LF650P3 electrode (BEX, Tokyo, Japan). After electroporation, embryos were cultured in the above culture medium at 37°C in a 5% CO 2 incubator.
Histology. For histological analysis, embryos and EB cells were fixed in Bouin's fixative, dehydrated, and embedded in paraplast. Serial sections were generated and stained with hematoxylin and eosin.
Western blot analysis. Nuclear proteins from SE tissues of E16.5 were extracted using a Subcellular Protein Fraction Kit (ThermoFisher, cat. No.78840). Nucleoprotein extracts were quantified using a Qubit 3 Fluorometer (ThermoFisher), after which 40 μg of each extract was loaded onto a 10% bis-acrylamide gel and subjected to SDS-polyacrylamide gel electrophoresis (PAGE). Proteins were transferred to a polyvinylidene fluoride membrane using a Trans-Blot turbo transfer system (BioRad). The membrane was blocked using Block Ace Powder (DS Pharma Biomedical) for 2 h at RT. After primary incubation using blocking solution with anti-GRHL3 antibody (1:1000 antigen aa195-211) overnight at 4°C, a second incubation was performed using blocking solution with anti-rabbit HRP antibody (Promega). Finally, the signal was revealed using a Western Lightning ECL Pro Kit (Perkin Elmer, Waltham, MA, USA; cat. No. NEL120001EA).
Atomic force microscopy. Young's modulus of epidermal cells in vitro was measured using atomic force microscopy (AFM). AFM and the following statistical analysis have been described previously 57,58 . Epidermal cells differentiated from ES cells via EBs (embryoid bodies) were transfected with a Grhl3 plasmid and cultured on a glass substrate. AFM was used to measure Young's modulus of epidermal cells. After the AFM measurement, cells were fixed and stained with TROMA-I-specific antibody. TROMA-I-positive cells were considered epidermal cells and used for statistical analysis. Cells with an area smaller than 3000 μm 2 were considered peripheral epidermal cells, and cells with an area larger than 10,000 μm 2 were considered large and mature (LM)-epidermal cells.
Young's modulus of mouse embryonic SE cells was measured by AFM using a NanoWizard IV imaging system (JPK Instruments) combined with an inverted optical microscope (Olympus; IX73). AFM was operated in the QI mode in liquid by adjusting the contact force during imaging with a SQUBE type cantilever (type CP-cont-BSG-B). Post-processing was conducted using JPK Image Processing; in any case, the post-processing was minimized so to reduce processing artifacts.
Micropipette aspiration experiments. In the micropipette aspiration assay, two opposing glass micropipettes (about 100 μm diameter) were connected with a pressure control system through a manipulator (Narishige; cat. no. MMO-202ND and IM-11-2) and held against the dorso-ventral region of each embryo. Aspirations, (1 kPa and an additional 1 kPa to total 2 kPa) measured by digital pressure gauge (Krone, KDM30), were applied to the dorsal region of embryos to stably hold SE tissues. After 1 min aspiration, the height of absorbed SE tissues was recorded with an Olympus FluoView TM FV1000 and × 40 zoom, and calculated by Olympus software.
Embryonic wound repair assays. Analysis of wound healing repair was performed on E12.5 wild type, Grhl3 cre/cre and Grhl3 NLS/NLS mutant embryos 59 . After amputation of hind-limb, wound repairs were judged by the repaired epidermal region covered on the original wound diameter marked by TROMA-I expression with immunohistochemistry and scanning electron microscope (SEM) (Miniscope TM3030Plus Hitachi).