Grhl3 promotes retention of epidermal cells under endocytic stress to maintain epidermal architecture in zebrafish

Epithelia such as epidermis cover large surfaces and are crucial for survival. Maintenance of tissue homeostasis by balancing cell proliferation, cell size, and cell extrusion ensures epidermal integrity. Although the mechanisms of cell extrusion are better understood, how epithelial cells that round up under developmental or perturbed genetic conditions are reintegrated in the epithelium to maintain homeostasis remains unclear. Here, we performed live imaging in zebrafish embryos to show that epidermal cells that round up due to membrane homeostasis defects in the absence of goosepimples/myosinVb (myoVb) function, are reintegrated into the epithelium. Transcriptome analysis and genetic interaction studies suggest that the transcription factor Grainyhead-like 3 (Grhl3) induces the retention of rounded cells by regulating E-cadherin levels. Moreover, Grhl3 facilitates the survival of MyoVb deficient embryos by regulating cell adhesion, cell retention, and epidermal architecture. Our analyses have unraveled a mechanism of retention of rounded cells and its importance in epithelial homeostasis.


Author summary
Developing vertebrate epidermis isolates and protects growing embryos from their surroundings. For performing such a crucial function under compromised physiological or genetic conditions, robust mechanisms allowing maintenance of epidermal integrity are warranted. However, such mechanisms are not fully explored. To investigate the mechanisms by which epidermis copes up with drastic cell-shape changes to maintain the epidermal integrity, we have used a mutant condition, goosepimples/myosinVb (myoVb), wherein epidermal cells round up due to defective intracellular membrane trafficking.
Our in vivo confocal imaging shows that this cell rounding is transient and the rounded cells are not extruded. Instead, they are retained and reintegrated. Using next generation sequencing and in situ expression analyses, we show that grainyhead-like 3 (grhl3) gene as Introduction extrusion. As reported in various Drosophila epithelia, misalignment of mitotic spindle planeinduced by ectopic expression of inscuteable-produces misplaced rounded cells. Interestingly, these cells are reintegrated into the epithelial sheet, which requires lateral adhesion mediated by Neuroglian and Fasciclin-2 [44]. However, it is not clear whether such mechanisms are present in vertebrate epithelia to reintegrate rounded up cells arising due to perturbed genetic or physiological conditions. The early developing bilayered epidermis in zebrafish consists of an outer peridermal layer and underlying basal epidermis. It is an ideal system to investigate mechanisms essential for maintaining epithelial homeostasis. Previously, using two mutant loci, goosepimples (gsp/ myoVb) and romeharsha (rhs), a direct link between plasma membrane homeostasis, cell size maintenance, and cell proliferation has been demonstrated [45,46]. In addition, it has been reported that peridermal cells round up in gsp/myoVb and rhs mutant embryos, presumably due to perturbed membrane homeostasis [45,46]. The fate of these rounded up cells and the consequences of their retention or extrusion on tissue architecture have not been investigated so far.
Our in vivo imaging analysis presented here reveals that rounded up peridermal cells in MyoVb deficient embryos are not extruded; instead, most of them are retained and reintegrated. Moreover, in the absence of myoVb function, epidermal regions bearing rounded up cells show enhanced grhl3 expression, which is essential for cell retention and maintenance of epidermal homeostasis. Mechanistically, grhl3 promotes retention of rounded up peridermal cells in MyoVb deficient embryos by regulating E-cadherin mediated cell-cell adhesion.

Peridermal cells, rounded up in the absence of gsp/myoVb function, are retained and reintegrated to facilitate phenotypic recovery
In a forward genetic mutagenesis screen aimed at identifying genes important for the maintenance of epidermal integrity and architecture, zebrafish mutants showing transient rounding up of epidermal cells, namely goosepimples (gsp) and romeharsha (rhs), were isolated [46,47]. Although it is known that the gsp locus encodes for Myosin Vb, an actin based molecular motor [45], the molecular nature of rhs has remained unknown. The gsp as well as rhs mutant embryos exhibit a profound defect in intracellular transport-characterized by increased endocytosis and perinuclear accumulation of early, recycling, and late endosomes-leading to perturbed membrane homeostasis in the peridermal cells. Consequently, surface area of the individual peridermal cell reduces and proliferation in the epidermis increases. Additionally, several peridermal cells, especially over the head, exhibit rounding up [45,46]. We have earlier shown that these rounded up cells project out of the plane of the periderm [45,46]. To better characterize the cell rounding phenotype, we carefully observed the head epidermis of individual gsp/myoVb mutant embryos between 36 to 72 hours post fertilization (hpf). Although the peridermal cell rounding phenotype was most profound at 48 hpf, it started subsiding thereafter, and cell rounding was absent by 72 hpf (Fig 1A-1J').
Given that peridermal cell rounding phenotype is transient in gsp/myoVb mutant embryos, we asked whether disappearance of rounded up cells results from cell shedding or cell reintegration. We used a specific splice site morpholino (myoVb MO) to deplete Myosin Vb levels in embryos derived from cldnB:lynEGFP transgenic line, in which peridermal cell membranes are marked by lynEGFP [45,48]). Like gsp/myoVb mutant embryos, embryos injected with myoVb MO developed a fully penetrant phenotype with almost every embryo beginning to develop clumps of the rounded up peridermal cells on the head between 36 to 48 hpf. Live confocal imaging of the myoVb morphant periderm in the dorsal head region between 48 hpf and Arrowheads indicate regions showing cell rounding. Note that cell rounding increases until 48 hpf (progression) and decreases subsequently (recovery) to finally cease by 72 hpf. Confocal time lapse images of dorsal head epidermis of cldnB:lynEGFP embryos injected with myoVb splice site morpholino (myoVb MO) during the recovery period at 48 (K), 50 (L), 52 (M), and 54 (N) hpf along with respective orthogonal projections. All cell borders as well as several intracellular membrane compartments are marked by lynEGFP (green). White arrows represent reintegrated cells, showing progressive cell shape change from round to polygonal. Red arrow and white dotted outline indicate a rounded cell undergoing apical extrusion. Scale bar = 50 μm. Traces showing changes in cell circularity values (O) for ten randomly selected cells in the above movie. Red trace represents extruding cell. Dotted line represents cell circularity = 0.9, which was used as the benchmark for classifying cells as rounded. A table (P) showing total number of cells, number of rounded cells, apical extrusions and basal extrusions (delamination) in live confocal movies of myoVb morphant embryos made between 50-53hpf. The duration of the movie is indicated in parentheses. 54 hpf revealed that most of the rounded up peridermal cells were retained in the epidermis (Fig 1K-1N and S1 Movie) and a few got reintegrated. The cell-reintegration was assessed based on visual inspection of shape transition from rounded to polygonal in X-Y plane and from more round to squamous morphology in X-Z plane (Fig 1K-1N). The analysis of cell circularity on fixed samples (presented in Fig 4U) revealed that in wild type embryos peridermal cells usually do not show circularity values above 0.9, which can be used as a threshold for classifying rounded cells. We analyzed circularities of selected cells during recovery phase in the live movie (S1 Movie), which revealed that cells with circularity higher than 0.9 persisted in the epithelium (Fig 1O). Additionally, we followed rounded cells in five movies that were made between 50 hpf to 53 hpf. We observed that apical cell-extrusion and basal cell extrusion (delamination) events were observed rarely ( Fig 1P). As reported earlier [45], the accumulation of perinuclear vesicles persisted even after the phenotypic recovery ( Fig 1N).
To conclude, the regression of peridermal cell-rounding phenotype occurred not through extrusion of rounded cells, but rather through their retention and re-integration.

Comparative transcriptomics reveals upregulation of grhl3 and cell adhesion genes in gsp and rhs mutant embryos
As shown above, the degree of peridermal cell rounding was the highest at 48 hpf, and reduced thereafter due to reintegration of rounded up cells. This analysis suggested that at 48 hpf, the factors responsible for retention and reintegration of rounded up cells begin to confer their effect. To screen for genes that might be responsible for retention and reintegration of rounded-up cells, we performed transcriptome analysis of gsp/myoVb and rhs embryos at 48 hpf using RNAseq. We reasoned that the use of rhs mutant embryos, which also exhibit peridermal cell rounding due to intracellular transport defects [46], would allow identification of the factors that are specifically involved in the epidermal response. Total mRNA was isolated from 50 gsp/myoVb and rhs mutant embryos and their wild-type siblings at 48 hpf, reverse transcribed, and sequenced using Illumina Hiseq 2000 (GEO accession: GSE173649). We obtained 1349 and 2605 differentially regulated transcripts in gsp and rhs mutants, respectively (S1 Data). We omitted transcripts whose expression was differentially regulated exclusively in either of the mutants and retained only those that were significantly up/downregulated to a similar degree in both the mutants. We obtained 514 and 133 transcripts that were up-and down-regulated, respectively, in both the mutants with 1.48-fold (1.5 σ) as the threshold (Fig 2A). Over Representation Analysis (ORA) using KEGG pathways revealed upregulation of transcripts associated with adipocytokine pathway, foxo pathway, tight junctions, and cell adhesion (P < 10 −4 ) (Fig 2B). The cell adhesion associated genes included those encoding for desmosomal, adherens junction, and tight junction components (Fig 2C). We also observed that amongst the transcription factors, the transcripts for grainyhead like 3 (grhl3) were upregulated in both gsp and rhs mutant embryos (Fig 2D). This was interesting, given the importance of Grainyhead family-transcription factors in epidermal differentiation and regulation of cell adhesion components. Importantly, while grainyhead paralogs grhl1 and grhl2a were also significantly upregulated in rhs mutant embryos, only grhl3 showed upregulation above the set threshold of 1.48-fold in the gsp/myoVb mutants (Fig 2D).
To conclude, the transcriptomic approach allowed us to identify genes that are differentially regulated in mutants showing epidermal cell rounding phenotype arising due to defects in intracellular transport. Upregulation of cell adhesion components and grhl3 may be causally linked with the restoration of epidermal architecture in the gsp mutant embryos.

Expression of grhl3 is spatiotemporally coincident with epidermal cellrounding in gsp/myoVb deficient embryos
The process of peridermal cell-rounding and reintegration occurs during a restricted time window. Given the role of Grainyhead transcription factors during epidermis development, we probed the expression of grhl3, using whole mount in situ hybridization (WISH), at 24, 36, 48, and 72 hpf. As shown earlier [30,49], control zebrafish embryos showed low levels of grhl3 expression in the epidermis at 24 hpf, which diminished by 36 hpf. After 30 hpf, the expression mostly remained restricted to the pharynx and nasal placodes ( Fig 3A). The grhl3 expression pattern in the myoVb morphant epidermis at 24 hpf was similar to that of controls, and a visible decrease in the expression was evident in the head epidermis at 30 hpf. However, grhl3 expression reappeared specifically in the aggregates of rounded up cells in the epidermis of the myoVb morphants at 36 hpf and persisted at 48 hpf (Figs 3A, 3B, S2A and S2B). By 72 hpf, grhl3 expression in MyoVb deficient embryos was highly reduced and comparable to that in the controls.
As mentioned earlier, comparative transcriptome analysis of gsp and rhs revealed up-regulation of genes encoding for tight junction components and other cell adhesion proteins. We performed WISH to assess the expression patterns of the adherens junction component cad-herin1 (cdh1), tight junction components claudin 7b (cldn7b), claudin e (cldne), and occludin a (oclna), and desmosomal component desmocollin2 like (dsc2l), in myoVb MO and control MO injected embryos. Whereas control embryos showed minimal expression of these genes in the epidermis at 48 hpf, MyoVb deficient embryos showed increased expression in and around the aggregates of rounded-up peridermal cells ( Fig 3E).
To conclude, our expression analysis revealed that grhl3 expression is spatiotemporally coincident with cell rounding in the MyoVb deficient embryos. Moreover, the adhesion components cdh1, cldn7b, cldne, oclna, and dsc2l showed elevated expression specifically in and around the rounded up peridermal cells, suggesting their potential involvement in cell retention and re-integration.

grhl3 function is essential for retaining peridermal cells in MyoVb deficient embryos
Our expression analysis pointed towards a spatiotemporally regulated function of grhl3 in MyoVb deficient embryos. We further explored whether grhl3 has a function in the resolution of aggregates of rounded up peridermal cells. For this, we knocked down grhl3 function in MyoVb deficient embryos, using a morpholino targeting the translation initiation site. The grhl3 knockdown alone resulted in no obvious morphological phenotypes at 48 hpf ( Fig 4G) except minor deformities in the median fin-fold (S1B Fig). As expected, MyoVb deficiency resulted in rounding-up of peridermal cells on the head. Interestingly, knockdown of grhl3 function in myoVb morphants did not produce rounded up cells in the periderm at 36 hpf ( Fig  4F, 4H, 4I and 4J). These embryos, deficient for both Grhl3 and MyoVb, showed significant lethality at 72 hpf as compared to that in either Grhl3 or MyoVb deficient embryos (S1A Fig). Similar lack of rounded up cells was seen in gsp/myoVb mutants injected with grhl3 MO (S1C Fig). However, the epidermis showed irregular morphology (S1C Fig) and the embryos seemed fragile, as the yolk oozed out through the epidermis during mild perturbation such as manual dechorionation at 30 hpf. The gsp/myoVb mutant embryos deficient for Grhl3 died between 36-48 hpf. Since morpholinos have been reported to exhibit off-target effects, we verified these observations using additional morpholinos (S1A Fig) published previously; one targeted against translation initiation site and two against splice sites [28,50]. To further confirm that the effect was specifically due to impaired grhl3 function, we used a hypomorphic grhl3 mutant allele generated using CRISPR-Cas9 technique. This hypomorphic grhl3 (-31) allele, contains a 31 bp deletion in exon 4, which results in a frameshift to produce a truncated peptide of 224 aa lacking the conserved DNA binding domain (S1D Fig). The grhl3 (-31/-31) mutant embryos complete epiboly, are embryonically viable, and exhibit impaired craniofacial development similar to that previously reported in grhl3 morphants (S1E Fig; [50]). MyoVb deficiency in grhl3 (-31/-31) mutant embryos resulted in the absence of rounded up peridermal cells that are characteristic of gsp/myoVb mutant or morphant embryos (Fig 4N and 4P). Analysis of cell circularity in fixed samples at 48hpf revealed that cells in myoVb MO epidermis showed a significantly higher median circularity as compared to that in control, grhl3 MO and in grhl3;myoVb MO epidermis (Fig 4Q-4U, Kruskal Wallis one-way ANOVA with Dunn's post hoc test, p<0.05). Importantly, in wild type embryos only 4% cells showed circularity value higher than 0.9 suggesting that cell circularity value occasionally exceeded beyond 0.9 in the normal epidermis. Using this as a threshold, we found that myoVb MO epidermis showed higher proportion of 29% cells with high circularity as compared to 8% in grhl3MO and 2% in grhl3;myoVb MO epidermis. These analyses suggested that Grhl3 expressing myoVb MO epidermis sustained a larger population of rounded cells.
The To quantify the cell shedding directly, we pelleted down cells that were shed in E3 medium by dechorionated embryos in every 6 h time-window between 12-36 hpf. The pelleted cells were then shortly incubated in Trypan blue, and stained (dead) cells were counted. Quantification revealed that while myoVb morphants exhibited moderate cell shedding between 12 hpf to 24 hpf, grhl3 morphants and grhl3;myoVb double morphants exhibited significantly higher cell shedding than that in myoVb morphants. Between 24 hpf to 36 hpf, grhl3;myoVb double morphants showed much higher cell shedding than that in myoVb and in grhl3 morphants (Fig 5A, One-way ANOVA with Tukey's test, p<0.05). Interestingly, this time window coincides with that of emergence of perinuclear vesicles (~24 hpf) and appearance of cell rounding (~36 hpf) in the MyoVb deficient embryos [45].  It was clear from live analysis that cell extrusion events are highly stochastic and therefore live analysis was not sufficient to allow useful comparisons between cell extrusion rates across conditions. Therefore, we quantified cell-rosettes-which form after cell extrusions [51,52]as proxies for extrusion events in fixed samples. Rosettes were defined as 5 or more cells converging at a single point. Despite their transient nature, rosettes were readily observed in grhl3 MO, myoVb MO and grhl3;myoVb MO embryos (Fig 5B'), indicating that rosettes could be captured before their resolution. Quantification of rosettes at multiple time-points revealed higher number of embryos showing extrusions as well as significantly higher number of extrusion sites per unit area in grhl3;myoVb double deficient embryos than those in either grhl3 or myoVb MO embryos beyond 30hpf ( Fig 5B, Kruskal Wallis one-way ANOVA with Dunn's post hoc test, p<0.0001). The near absence of cell-rosettes in the control embryos indicated that either the cell extrusions were absent or a few in number. It also suggested that developmental cell-rearrangements leading to rosette formation [53] were rare in the developing zebrafish epidermis.
Finally, we asked whether there is an increase in apoptosis in the absence of myoVb and grhl3 function leading to extrusion of the peridermal cells. Anti-Activated Caspase 3 antibody staining at 26 hpf, 32 hpf, 36 hpf, and 48 hpf revealed that apoptotic cells were rarely present in the periderm under all the three genetic conditions. It appeared that only cells that were undergoing terminal phase of extrusion or delamination (basal extrusion) were apoptotic (S3E Fig). At 36 hpf, none of the 135 wild-type embryos analyzed exhibited any apoptotic peridermal cell. However, in MyoVb and Grhl3 deficient conditions we observed 3 apoptotic cells in 76 embryos and 1 apoptotic cell in 82 embryos, respectively. Interestingly, in 109 double deficient embryos, we did not observe any apoptotic cell presumably due to their rapid extrusion.
Our analysis indicated that grhl3 is expressed specifically in the regions containing rounded-up peridermal cells. This points towards a region-specific requirement of grhl3 in the retention of peridermal cells. To test this possibility, we injected myoVb MO, grhl3 MO, and grhl3+myoVb MO along with EGFP mRNA as a tracer in wild-type embryos at 16-32 cell stage and calculated the proportion of embryos showing peridermal clones in the head region. Our analysis revealed that at 12 hpf, clone retention was comparable in all conditions; however, at 48 hpf, both Grhl3 deficient and Grhl3;MyoVb double deficient clones were less frequently detected than control and myoVb morphant clones (Fig 5C-5G). These data point towards a cell autonomous role of grhl3 in promoting cell retention in the periderm.
To conclude, our data suggest that although grhl3 function is important for preventing cell shedding during early epidermis development, its function becomes increasingly important in the absence of myoVb, where membrane homeostatic defects result in cell rounding.

grhl3 is important for maintaining the tissue architecture and homeostasis in the MyoVb deficient epidermis
Our data show that grhl3 function is essential for prevention of epidermal cell extrusion and becomes more relevant in the absence of myoVb function, where endocytic stress and perturbed intracellular transport leads to decreased cell size and peridermal cell rounding [45,46]. We further investigated the consequences of the increased cell extrusion in the MyoVb deficient epidermis.
Confocal micrographs indicated that grhl3;myoVb double morphants exhibited presence of larger peridermal cells than those in control and myoVb morphants (Fig 6A-6D). Further, quantification of cell height revealed that in grhl3 morphant as well as in grhl3;myoVb double morphant embryos, peridermal cells were significantly flatter than those in control and myoVb morphants ( Fig 6F). This trend was also reflected in the thickness of the periderm (Fig 6G-6J). To investigate the effect on cell-packing density in the absence of grhl3 function, we imaged the dorsal head epidermis at a lower magnification in the cldnb:lynEGFP transgenic line. The peridermal cell packing density (cell number/mm 2 ) estimated from such images at various developmental stages revealed that myoVb deficient embryos exhibit higher packing density as compared to wild type at 32, 36, and 48 hpf. However, both Grhl3 deficient and Grhl3;MyoVb double deficient embryos showed significantly lower packing density as compared to the wild type embryos at comparable time-points. Interestingly, both these conditions showed recovering trend in cell-packing density (Fig 6E, Kruskal Wallis one-way ANOVA with Dunn's post hoc test). Despite this gradual recovery, there was profound heterogeneity in peridermal cell sizes in the grhl3;myoVb double morphant suggesting localized differences in the cell-densities (S3F Fig).
We further analyzed peridermal cell proliferation to check whether it compensates for the cell extrusion and to restore the cell density in grhl3MO and grhl3;myoVb MO embryos. We performed BrdU assay at 18hpf, 24hpf, 30hpf, 36hpf, and 48hpf (Fig 6K-6S). This analysis  . Note that grhl3 morphants as well as grhl3;myoVb morphants show lower cell density than wildtype at 26 hpf. However, their cell densities approach wild-type levels by 48 hpf. In myoVb morphants, cell density increases over time and is significantly higher than that in wild-type. Scale bar = 50 μm. Asterisks revealed that peridermal proliferation was higher in Grhl3 and MyoVb deficient embryos as compared to that in control embryos starting from 30hpf. Interestingly, peridermal proliferation in the double deficient embryos was higher than control embryos from 24hpf onwards and showed higher trend than Grhl3 deficient embryos with significant difference at 24, 36, and 48hpf (One way ANOVA with Tukey's multiple comparison test, p < 0.05). These trends explain why the overall cell-packing density increases in a similar manner in both Grhl3 deficient and Grhl3;MyoVb double deficient embryos despite the extrusion being higher in latter.
To conclude, the epidermal architecture is altered in the absence of grhl3 function. The peridermal cell packing density is lower in grhl3 morphants as well as in grhl3;myoVb double morphants, especially at early stages, plausibly due to cell-extrusion. Besides, the epithelial cells are also flatter in these to genetic conditions. Increase in cell proliferation seems to aid in restoration of the packing density as the development proceeds.

Grhl3 promotes the retention of peridermal cells by regulating levels of membrane E-cadherin
Loss of grhl3 function in myoVb deficient embryos resulted in the altered epidermal architecture with localized differences in the cell-densities due to excessive cell extrusion. As discussed earlier, the transcriptome analysis followed by WISH showed a topical upregulation of adhesion molecules in the regions showing cell rounding. As the lateral adhesion may influence the process of cell retention, we asked whether grhl3 functions via regulating levels of E-cadherin.
Immunostainings with anti-E-Cadherin antibody revealed that despite the patchiness, membrane levels of E-Cadherin increase in myoVb morphant peridermal cells compared with those in control at 48 hpf (Fig 7A and 7B). The patchiness in E-cadherin localization is presumably a consequence of cell-cell rearrangements that take place at this stage, requiring junctional remodeling. However, combined loss of grhl3 and myoVb resulted in much weaker E-Cadherin localization than that in myoVb or grhl3 morphants (Fig 7C and 7D). Similarly, WISH analysis revealed that expression of e-cadherin decreases in MyoVb deficient embryos in the absence of grhl3 function (Fig 7E-7H). These data suggest that the E-Cadherin localization in myoVb morphants, at 48 hpf is grhl3 dependent. In the absence of grhl3 function, the myoVb deficient peridermal cells fail to augment E-Cadherin levels.
To test the importance of grhl3 function in MyoVb deficient embryos, we asked whether Grhl3 deficient clones are differentially retained in control versus myoVb morphant embryos. To inspect this facet, we generated grhl3 MO clones in myoVb morphant embryos, wherein all peridermal cells were MyoVb deficient and the clone was additionally deficient for Grhl3. Estimation of clone retention revealed that grhl3 MO clones had a near zero probability of being retained in the head periderm of MyoVb deficient embryos (Fig 7I-7L and 7P). Corroborating this result, we also found that in Grhl3;MyoVb double deficient embryos, control morphant clones were retained with very low frequency. These results, in addition to the preceding analysis ( Fig 5G) wherein we did not see a difference between the retention of Grhl3 deficient and indicate statistically significant difference between the wild type and the given genetic condition at a comparable time-point by Kruskal Wallis one-way ANOVA followed by Dunn's post-hoc test. Quantification of cell height (F) of peridermal cells at 48 hpf in four genetic conditions mentioned. grhl3 morphants and grhl3;myoVb double morphants show significantly reduced cell heights compared to those in control morphants and myoVb morphants, respectively (Kruskal Wallis one-way ANOVA, p<0.05). Orthogonal projections of confocal-scans showing cross-sectional view of periderm (G, H, I, J). Confocal micrographs of embryonic epidermis immunostained for Lgl/BrdU (K-R) at 18hpf (K-N) and 48hpf (O-R). Quantification of cell proliferation in the periderm at 18hpf, 4hpf, 30hpf,36hpf, and 48hpf in four genetic conditions (S). (All data presented as box plot showing median and IQR, One-way ANOVA with Tukey's test, p < 0.05). In E, F and S square brackets denote comparisons while asterisks indicate statistically significant difference. Only relevant comparisons are shown. For all one-to-one comparisons, please refer to S2 Data.
In summary, grhl3 has an essential function in preventing cell extrusion. Moreover, this function of grhl3 becomes more important in MyoVb deficient epidermis. Our data further suggest that E-cadherin cell autonomously promotes retention of cells in the combined absence of grhl3 and myoVb functions.

Discussion
Cells round up during cell division, cell-crowding and in various pathological conditions [37][38][39][40][41][42]. It is not clear whether there are mechanisms to retain and reintegrate rounded cells to maintain tissue homeostasis. Here, using the gsp/myoVb mutant, we have unraveled a Grhl3-E-cadherin dependent mechanism essential for retention of rounded up epidermal cells.
Our analyses indicate that cells rounding up due to perturbed membrane homeostasis are actively retained and reintegrated in the epidermis. Grhl3, a member of the Grainyhead family of transcription factors, is expressed specifically in the rounded up cells and plays an essential role in this process. MyoVb deficient peridermal cells exhibit higher levels of membrane Ecadherin, indicating increased cell-cell adhesion. This increased E-cadherin localization is not uniform, presumably due to junctional remodeling and cellular rearrangements in the periderm [45]. Our data indicate that the increased E-cadherin expression and localization in MyoVb deficient peridermal cells is induced by grhl3 function. E-cadherin is sufficient for the retention of Grhl3;MyoVb double deficient clones in gsp/myoVb morphant embryos. Thus, Grhl3 -E-cadherin axis plays an important role in retaining peridermal cells (Fig 7Q). Indeed, it has been shown that transcription factors belonging to Grainyhead family directly or indirectly control the expression of various adhesion components, including E-Cadherin [17,18,54,55]. Interestingly, Grhl3 has also been shown to activate E-cadherin expression via an intronic enhancer [56]. Whether Grhl3 controls E-cadherin levels directly in the MyoVb deficient periderm remains to be elucidated. Does retention of rounded cells contribute towards maintenance of epidermal homeostasis and architecture? Our analyses indicate that rounded up peridermal cells in MyoVb deficient embryos extrude in the absence of grhl3 function. Besides, grhl3 also functions in cell retention during early embryogenesis before 24 hpf. While we have not explored this developmental phenotype in detail, the cell-packing density data show a clear decrease in peridermal cell density in grhl3 morphants at 26 hpf, plausibly a consequence of the early developmental cell extrusion prior to 24 hpf. The extrusion events-based on rosette analysis and cell extrusion analysis-are minimal in grhl3 morphants after 24 hpf. A similar decrease in the cell-packing density is observed in grhl3;myoVb double morphants at 26 hpf. Interestingly, the cell packing density recovers in Grhl3 as well as Grhl3;MyoVb double deficient embryos. It is likely that the increased proliferation in both these genetic conditions is compensatory in nature and facilitates this recovery. Interestingly, however, in the double deficient embryos, there is a wave of peridermal cell extrusions post 32 hpf even as the cell density phenotype continues to recover. This wave of extrusions in double deficient embryos coincides with the occurrence of cell rounding up phenotype in MyoVb deficient embryos. The absence of cell rounding phenotype and abundant rosette formations indicate that rounded up cells might extrude rapidly rather than being retained. The time-lapse study corroborates this finding. Our analyses presented here suggest that grhl3, expressed in the rounded-up cells, is essential to facilitate their retention by up-regulating E-cadherin levels. While our data indicate that Grhl3 promotes adhesion and facilitate retention of rounded-up cells, it is not clear whether Grhl3 mediated adhesion also facilitates reintegration. In addition to this late function, it is possible that the early developmental function of grhl3-essential for the maintenance of early cell-packing density and thus presumably the tissue tension-also contributes towards the cell-retention in MyoVb deficient embryos. Further analysis using temporal inactivation will be required to tease apart the extent of early and late contribution of grhl3 function in cell retention in the myosin Vb deficient background.
In the MyoVb deficient embryos, the cell packing density increases, which is consistent with the earlier observations that peridermal cells shrink in the absence of the motor function due to perturbed plasma membrane homeostasis and the increased cell-number compensates for the decrease in cell surface area [45]. Our cell height analysis reported here further indicates that peridermal cells are taller in the MyoVb deficient embryos. While in grhl3 morphants the peridermal cells are shorter than those in the wild-type, they are even more squamous and constitute a peridermal layer that is much thinner in the double deficient embryos. In addition to cell proliferation, this increased squamation-as the peridermal cells are not being able to grow in the absence of myoVb function-possibly allows the tissue to compensate for the cell loss due to continued extrusion, albeit at the expense of the tissue getting thinner. Interestingly, despite significant restoration in packing density, the grhl3;myoVb double deficient embryos exhibit cell-size heterogeneity (S3F Fig), yielding patches of large but thin cells in the periderm, which along with reduced E-cadherin mediated adhesion, might form weak spots that potentially result in tissue tears and possibly the subsequent death of embryos.
Although termed as 'master regulators of sealing and healing' for their highly conserved role in epithelial homeostasis and repair [11], Grainyhead family transcription factors remain enigmatic, due to their diverse tissue specific functions as well as the functional redundancy between grainyheadlike paralogs. GRHL1 and GRHL2 have been reported to control cell differentiation and expression of various adhesion molecules in various epithelia to ensure proper morphogenesis, whereas GRHL3 is shown to drive peridermal differentiation and suppression of immune mediated hyperproliferation in the epidermis [22,25]. Recently, cytoplasmic function of Grhl3 has been shown to be important for acquiring mechanical properties that are essential for morphogenesis [16]. Our analyses show that grhl3 function is also essential in epithelial cell retention in a normal developing periderm. This indicates that grhl3 function might play a role in reintegration of cells undergoing mitotic rounding-up or in suppression of local crowding-induced homeostatic cell extrusion during epidermal morphogenesis; however, further investigation is necessary to explore these possibilities. Nevertheless, our analyses clearly show that grhl3 function is essential to retain cells in a stressed epithelium like myoVb morphant periderm that has a higher propensity of cell-rounding up. Incidentally, grhl3 is specifically up-regulated in psoriatic skin lesions and is essential for their healing [25]. Given the complex and multifaceted functions of grhl3, it would be interesting to test whether Grhl3 acts in a manner similar to what we have shown, to facilitate epidermal repair in skin diseases.

Fish strains
The wild-type fish strains Tübingen (Tü) and DBS (wild-type strain derived from a local pet shop), and transgenic line Tg(cldnB:lynEGFP) [57] were used for knockdown analyses, live confocal analysis, cell surface area measurements. The goosepimples/myoVb [45] and romeharsha [46] mutant lines were used for monitoring the progression of balling up and transcriptome analysis. In situ hybridization analysis was done using albino strain.

Generation of grhl3(-31) mutant via CRISPR/Cas9 mediated deletion
The target site for generation of grhl3 single guide (sg) RNA was predicted using ZiFit (http:// zifit.partners.org/ZiFiT/). Oligonucleotides (IDT) encoding the determined target site (TCGCAGAAATCATGGCAATC Reverse strand) were cloned into the plasmid pDR274, and sequence-verified plasmids were used as a template to transcribe sgRNA using the MEGAscript T7 kit (Ambion) as per manufactures recommendations. sgRNA and Cas9 nuclease (NEB M0386S) were co-injected into the cell at single-cell stage WT Tü zebrafish embryos. sgRNA was injected at concentrations ranging between 20-750 pg/nl, with greater efficiency of cutting observed at higher concentrations, as previously reported for other sgRNA [58,59], and Cas9 protein was injected at a final concentration of 320 ng/μl.
To detect genomic lesions, DNA was isolated from embryos obtained from outcrossed injected fish in 20 μl 50 mM NaOH, and heated to 95˚C for 10 min. The solution was neutralized by adding 2 μl 1 M Tris-HCl, pH 8.0. Samples were pulse centrifuged to pellet debris, and supernatant was used as a template for subsequent PCR reactions. PCR amplification of the regions flanking the target site, and subsequent sequencing using primers flanking the target site (grhl3_e4_SF 5'-CAGCTCTTCCCCTGAACTTG-3', grhl3_e4_sR 5'-ACATAAATGCGG ACCTCAGGTGT-3') were used to determine genomic changes.

Genotyping grhl3 mutant embryos
During the assessment of genetic interaction between myoVb and grhl3, genomic DNA from single 5d old larva was extracted as above after the morphological evaluation of the cell rounding phenotype. PCR amplification was performed using the above primers, and 10 μl was run on a 3% agarose gel. The following band sizes were used to identify WT and grhl3 mutant embryos; WT: 234 bp, grhl3 (-31bp): 203 bp. In addition to or instead of genotyping, early phenotype of delayed epiboly was also used to identify grhl3 mutant embryos.

Comparative transcriptomics analysis using RNAseq
Fifty representative embryos each for goosepimples and romeharsha mutants, as well as respective siblings were fixed in 2 ml RNAlater (Thermo Fischer Scientific). The fixed embryos were processed by ILife Discoveries (Gurgaon, India), where mRNA was extracted using RNAlater protocol, and cDNA library (300 bp fragments) was prepared, followed by sequencing on Illumina GA 2000 with a predicted transcriptome coverage of 10× with 100 bp paired end reads. Raw data (~30 Gbp per sample) was filtered for read quality and reads were trimmed, using FastQC and FastX toolkit. Alignment was performed using TopHat 2.0 and RPKM values were calculated using DEGseq.
For comparative analysis of transcriptomes of gsp and rhs, transcripts with very low expression levels (RPKM<100), and transcripts showing significant difference (1.48 fold; 1.5 σ) in RPKM values between the two sibling samples were filtered out, to obtain 13,633 transcripts, with 1962 and 3532 differentially regulated transcripts in gsp and rhs mutants respectively. To further refine the data, we filtered out genes that did not show similar extent of misregulation between gsp and rhs. Further, transcripts showing dissimilar trends of regulation were filtered out by discarding those showing a difference in RPKM values by > 2-fold between the two mutants. As a result, we obtained 1349 (gsp) and 2605 (rhs) differentially regulated transcripts.
Over representation analysis and GO term for the common significantly regulated transcripts was performed using DAVID.
Antisense morpholino oligos against grhl3 and myoVb, as well as the respective control morpholinos were injected at 200 μM concentration. Injections were done at 1-2 cell stage, unless specified otherwise. For double knockdown experiments, a 1:1 mixture of myoVb splice MO and grhl3 MO (400 μM each) was prepared to obtain a resultant concentration of 200 μM. Similarly, a 1:1 mixture of respective control morpholinos was used. For in situ hybridization experiments, to maintain a highly precise morpholino dose across single and double morpholino injections, instead of injection of 1:1 morpholino mixture, serial injections with myoVb splice MO and grhl3 MO were performed, using the same needle used for single knockdown injections in the same set. All injections in a single set were performed in tandem, in a single session.

Clonal morpholino knockdown and e-cadherin overexpression
For clonal knockdown of myoVb, grhl3, and combinations thereof, we used a mixture of morpholino solution containing single or double morpholinos with appropriate resultant concentrations and a GFP mRNA to locate morphant clones. Injections were done in a randomly selected cell at 16 or 32 cell stage. The zebrafish e-cadherin-mCherry construct was a gift from Prof. Erez Raz (Universität Münster). The E-cadherin mcherry mRNA was synthesised using SP6 Transcription Kit (ThermoFisher-Invitrogen, AM1340), purified using P-30 Gel Columns (BIORAD, 732-6250) and injected at 16/32 cell stage to achieve clonal overexpression.

Quantification of clone retention
Clonally injected embryos were fixed and embryos showing GFP positive clones (generally 80-100% of total embryos) were sorted on Zeiss Apotome. Any embryos showing GFP expression in the whole embryo were discarded. The embryos with GFP positive clones represent successful clonal injections. These embryos were then stained as mentioned below using anti-GFP and anti-E-cadherin antibodies. Post glycerol upgradation, heads of all embryos were dissected out and mounted. These were thoroughly screened for presence of peridermal clones in the dorsal head region, using Zeiss LSM 510 or Zeiss LSM exciter, 40× objective. Clone retention percentage was calculated as: 100 × (Number of embryos showing peridermal clones in the dorsal head region / Total number of embryos with GFP labeled clones)

Immunostaining
Immunostainings were carried out according to previously published protocol with some modifications (Sonawane et al., 2009). Briefly, embryos or larvae at the appropriate stages were fixed by overnight incubation at 4˚C in 4% paraformaldehyde (PFA) in PBS, followed by methanol up-gradation and storage at-20˚C (for cell size and cell proliferation analysis). For E-cadherin antibody staining as well as analysis of clonal knockdowns, the methanol upgradation was skipped. Embryos/larvae downgraded from methanol and those in PFA were washed with PBS followed by PBT (0.8% Triton X-100 in PBS), incubated in 10% NGS for 3 h at room temperature and followed by incubation with appropriate dilution of antibodies in 1% NGS in PBT for 4 hours at room temperature. The antibodies and their dilutions were as follows-rabbit anti-GFP

Image acquisition and processing
Imaging of immunostainings was done in the dorsal head epidermis, using either Zeiss LSM 510 Meta with Plan apochromat 639/1.40 oil or EC Plan-Neofluar 409/1.30 oil objective or Zeiss LSM 710/880 with Plan-apochromat 639/1.40 oil. Optical zoom of 63x was used for most of the images. The image acquisition was done with pinhole size of 1 AU, frame size of 1024 × 1024 pixels, with averaging of 4, and bit depth of 8 or 16. The PMT gains were adjusted to capture maximum possible dynamic range.
For live confocal imaging, larvae were embedded in 0.2% Low melting agarose (Sigma-Aldrich), either placed in petri dishes having glass coverslips at the bottom and imaged on LSM5 exciter, LSM 710 or LSM 880 with 40X Plan-apochromat objective/1.4NA with 1X/2X Zoom. For longer imaging experiments to capture cell extrusions, embryos were partially submerged in agarose, exposing the dorsal side of their head. E3 was added on top to cover the entire surface of the plate to avoid dehydration. Water was carefully replenished without disturbing the frame every 30 min. Zeiss 880 with 40x water objective was used for scanning. Bright field imaging was done using embryos mounted in methylcellulose (3%), either on Zeiss SteREO Discovery with AxioCam or Nikon SMZ18 with a Nikon DS-Ri2 CMOS camera. ImageJ (FIJI) was used for image processing and analysis.

BrdU incorporation assay and analysis of proliferation
Embryos were dechorionated at 18hpf, 24hpf, 30hpf, 36hpf, and 48 hpf, and incubated in 10 mM solution of BrdU in 2% DMSO in E3 for 3h at 29˚C. After incubation, followed by 5 min washes with E3 buffer, embryos were fixed in 4% PFA and upgraded in methanol. After downgrading to PBS and prior to immunostaining, the fixed larvae were treated with 4N HCl for 20 min at room temperature.
Quantification of proliferation in the periderm and basal epidermis was done using confocal stacks. Nuclei positive for BrdU staining, as well as total cells (based on Lgl staining demarcating cell boundaries) were counted using cell counter plugin in ImageJ. Percentage of cells positive for BrdU out of total cells was used as a proliferation index in the epidermis. Five animals per set per genetic condition and two such experimental sets were analyzed for proliferation. One-way ANOVA followed by Tukey's test was used for determining statistically significant differences between groups.

Cell height quantification
The cell heights were estimated by counting the number of stacks showing lynEGFP staining for individual cells. Height of five cells were measured from each embryo, total 15 embryos from three sets were used for the analysis. Kruskal-Wallis one-way ANOVA with Dunn's test was used for determining significant differences between groups.

Cell density and rosette analysis
Tg(cldnB:lynEGFP) embryos injected with either myoVb MO, grhl3 MO or a combination of both and were fixed and stained with anti GFP antibody. Images were acquired as described above using 40X oil objective at 0.6X zoom. Number of cells and total area occupied by them were calculated using ImageJ. Cell density was calculated as cells /mm 2 . Similarly, number of cell-rosettes were counted and expressed as rosettes/mm 2 . Between 35-60 embryos per genetic condition per stage were analyzed from three independent experiments. Kruskal-Wallis oneway ANOVA followed by Dunn's test was used for determining statistically significant differences between groups.

Cell extrusion assay
Wild-type embryos were injected with myoVb MO, grhl3 MO or a combination of both and used for quantification of cell extrusion occurring between 12-18 hpf, 18-24 hpf, 24-30 hpf, 30-36 hpf, and 36-42 hpf. Twenty (or ten for 36 hpf-43 hpf) embryos were dechorionated and transferred to Eppendorf tubes containing 2 ml E3 buffer and incubated at 29˚C for 6 h. The E3 buffer used for incubation was collected and embryos were washed thrice with 2 ml E3 buffer; the buffer used for washing was pooled with the buffer used for incubation. This was then centrifuged for 5 min at 1000×g and supernatant was carefully discarded. The pellet was re-suspended in 10μL of 0.1% Trypan blue solution (Sigma Aldrich, T6146) and incubated at 29˚C for 10 min. the solution was well mixed and mounted on a slide and dead cells were counted manually under a light microscope using a 20X objective (Zeiss LSM 880 or Zeiss Apotome). Five replicates were performed. Statistically significant differences between groups were determined using one-way ANOVA followed by Tukey's test.

Statistical analysis
GraphPad PRISM and Microsoft Excel (2019) were used for statistical analysis and plotting graphs. Along with this, an online tool http://shiny.chemgrid.org/boxplotr/ was used to create box plots. Details of all the statistical tests and comparisons are presented in S2 Data.