Mechanisms that sense and regulate epithelial morphogenesis, integrity, and homeostasis are incompletely understood. Protease-activated receptor 2 (Par2), the Par2-activating membrane-tethered protease matriptase, and its inhibitor, hepatocyte activator inhibitor 1 (Hai1), are coexpressed in most epithelia and may make up a local signaling system that regulates epithelial behavior. We explored the role of Par2b in matriptase-dependent skin abnormalities in Hai1a-deficient zebrafish embryos. We show an unexpected role for Par2b in regulation of epithelial apical cell extrusion, roles in regulating proliferation that were opposite in distinct but adjacent epithelial monolayers, and roles in regulating cell–cell junctions, mobility, survival, and expression of genes involved in tissue remodeling and inflammation. The epidermal growth factor receptor Erbb2 and matrix metalloproteinases, the latter induced by Par2b, may contribute to some matriptase- and Par2b-dependent phenotypes and be permissive for others. Our results suggest that local protease-activated receptor signaling can coordinate cell behaviors known to contribute to epithelial morphogenesis and homeostasis.
Protease-activated receptors (PARs) are G protein–coupled receptors that mediate cellular responses to extracellular proteases (Vu et al., 1991a). Site-specific cleavage of the N-terminal ectodomain of these receptors serves to uncover a tethered peptide ligand, which binds to the receptor’s heptahelical bundle to effect transmembrane signaling and G protein activation (Vu et al., 1991a,b). Among the four PARs found in mammals, PAR1, PAR3 and PAR4 mediate cellular responses to the coagulation protease thrombin. Genetic studies in mice and pharmacological studies in humans suggest that signaling via these receptors helps orchestrate physiological responses to tissue injury including hemostasis and perhaps inflammation and repair (Coughlin, 2000, 2005). The identity of the physiological activators of PAR2 and its roles in vivo are less explored.
Studies in cell culture and mice suggest that Par2 together with the protease matriptase and its inhibitors Hai1 and Hai2, all integral membrane proteins, may make up a local signaling system that regulates epithelial behavior (Takeuchi et al., 2000; Camerer et al., 2010; Szabo and Bugge, 2011; Sales et al., 2015b). Matriptase, gene symbol St14, is a type 2 integral membrane protein that displays a trypsin-like serine protease domain extracellularly. Subnanomolar concentrations of soluble matriptase protease domain can cleave and activate Par2 in cell culture (Takeuchi et al., 2000; Camerer et al., 2010); heterologous coexpression of full-length matriptase with Par2 can also promote Par2 cleavage (Camerer et al., 2010). Par2, matriptase, Hai1, and Hai2 are naturally coexpressed in most epithelial tissues (Takeuchi et al., 2000; Camerer et al., 2010; Szabo and Bugge, 2011; Sales et al., 2015b). In the mouse embryo, Par2 is expressed in ectodermal epithelium at the time and place of neural tube closure. Loss of Par2 function in certain genetic backgrounds can impair this process, as can loss of Hai2 function (Szabo et al., 2009; Camerer et al., 2010). Conditional knockout of the matriptase gene in the epithelial cells of mouse intestine or skin leads to defective barrier function in these tissues. Intestinal knockout leads to gut inflammation and skin knockout to perinatal death from desiccation (List et al., 2002, 2009). Overexpression of matriptase or the matriptase activator prostasin in mouse skin leads to hyperplasia, inflammation, ichthyosis, and pruritus, and these phenotypes are mimicked by Par2 overexpression and prevented by Par2 deficiency (Steinhoff et al., 2003; Frateschi et al., 2011; Sales et al., 2015a,b). Although matriptase has multiple substrates (Takeuchi et al., 2000; Antalis et al., 2011) and Par2 can be activated by other trypsin-like proteases (Nystedt et al., 1994; Corvera et al., 1997; Camerer et al., 2000; Zhao et al., 2014), these and other data suggest that matriptase and Par2 may comprise a local signaling system that contributes to the formation and function of epithelial tissues (Camerer et al., 2010; Szabo et al., 2014; Le Gall et al., 2016; Fig. 1 A). However, the precise cellular behaviors regulated by this system and its physiological roles remain to be fully illuminated.
Epithelial growth and remodeling, epithelial-to-mesenchymal transformation, and signaling from epithelium to other cells play central roles in morphogenesis and organogenesis during embryonic development. Epithelial barriers between tissues and between an organism and its environment play key roles in adult physiology and homeostasis by regulating the movement and delivery of molecules between compartments (Marchiando et al., 2010). To support these vital functions, epithelia must regulate cell junctions, polarity, shape, size, movement, density, and number and jettison crowded or damaged cells—all without losing barrier function (Ragkousi and Gibson, 2014; Gudipaty and Rosenblatt, 2017). Given the evidence that Par2 activation by matriptase can contribute to development and homeostasis of epithelial tissues, we sought a system in which the roles of matriptase and Par2 activation in an epithelium could be explored in detail.
The zebrafish system offers genetic tractability and external development of large numbers of optically transparent embryos amenable to video microscopy. The skin of the early zebrafish embryo is a simple bilayer epithelium composed of an outer periderm layer and an inner basal layer (Le Guellec et al., 2004). Periderm keratinocytes adhere tightly to each other and form tight junctions to create a strong epithelial barrier (Kiener et al., 2008; Kwong and Perry, 2013). Basal layer keratinocytes adhere to each other and to the periderm through E-cadherin, the extracellular component of the adherens junctions, and hemidesmosomes (Slanchev et al., 2009; Sonawane et al., 2009). Deficiency of the matriptase inhibitor Hai1a (also known as Spint1a) in zebrafish embryos causes abnormal skin development characterized by accumulation of cell clusters on the skin surface, altered cell–cell contacts, leukocyte infiltration, and other abnormalities (Carney et al., 2007; Mathias et al., 2007). Deficiency of the matriptase homologue St14a rescues these phenotypes, suggesting that they are mediated by increased matriptase activity associated with loss of its inhibition by Hai1a (Carney et al., 2007; Mathias et al., 2007). The availability of a matriptase-dependent epithelial phenotype in zebrafish embryos provided an opportunity to search for and characterize roles of matriptase-driven Par2 signaling in a vertebrate epithelium.
Par2b is coexpressed with matriptase and Hai1a in the skin of the zebrafish embryo
To determine whether a local matriptase-Hai1-Par2 system might operate in zebrafish embryonic epidermis, we first asked whether its components are coexpressed in the cell types that make up this tissue. We isolated periderm or basal layer cells by FACS from embryos collected at 24 h postfertilization (hpf) using zebrafish lines carrying a fluorescent marker for periderm (Tg (krt4:nlsEGFP)cy34) or basal layer (TgBAC (ΔNp63:Gal4FF)la213;Tg (uas:LifeActGFP)mu271). These lines are hereafter called periderm-nuclear-EGFP and basal-layer-LifeActGFP, respectively. Compared with RNA from whole embryos, RNA from the sorted periderm cell preparation was enriched ∼17-fold for the periderm markers krt5 and krt4 but showed no enrichment for the basal marker p63. RNA from the sorted basal layer cell preparation was enriched approximately fivefold for p63 but showed no enrichment for krt4 (Table S1). Thus, the sorted cell populations showed enrichment for the expected markers.
mRNAs encoding the Hai1 zebrafish homologue Hai1a, the matriptase homologue St14a, and the Par2 homologue Par2b (also known as F2rl1.2) were readily detected in both the periderm and basal layer preparations and enriched compared with whole embryo. The level of hai1a, st14a, and par2b mRNA in periderm preparations was ∼9-, ∼9-, and ∼16-fold enriched, respectively, compared with whole embryo. In basal layer, hai1a, st14a, and par2b mRNA were enriched ∼10-, ∼4-, and ∼8-fold, respectively (Table S1). These results suggest that matriptase gene st14a and the Hai1 gene hai1a are coexpressed with par2b in both the periderm and the basal layer of zebrafish embryo skin. Previous in situ hybridization studies indicated expression of hai1a in the skin of the zebrafish embryo (Carney et al., 2007).
Zebrafish matriptase can cleave zebrafish Par2b at its activation site
The Par2b N-terminal exodomain contains the amino acid sequence KNGR28/M29. Studies of mammalian matriptase substrate specificity (Takeuchi et al., 2000) suggest that matriptase should cleave this sequence at the R28/M29 peptide bond (Fig. 1 A). To determine whether zebrafish matriptase can indeed cleave zebrafish Par2b like the cognate mammalian proteins, we generated the cleavage reporter AP-Par2b in which secreted AP is joined to the N-terminal ectodomain of Par2b. Cleavage of AP-Par2b at R28/M29, its predicted activating cleavage site, should release AP into the culture medium (Fig. 1 B; Ludeman et al., 2004; Camerer et al., 2010). Trypsin efficiently cleaves mammalian PAR2 at its activating cleavage site (Nystedt et al., 1994; Camerer et al., 2010). As a positive control, we first determined whether AP-Par2b is cleaved by exogenously added trypsin. Trypsin treatment of AP-Par2b–expressing HEK293 cells released ∼150,000 arbitrary units (AU) AP to conditioned medium (Fig. 1 C). No such increase was seen with trypsin treatment of untransfected cells or cells expressing an AP-Par2b R28A/M29P mutant in which the predicted activating cleavage site was ablated (Fig. 1 C). These results suggest that trypsin can cleave AP-Par2b at the predicted KNGR28/M29 activation site and are consistent with the observation that trypsin triggers Par2b internalization (Xu et al., 2011) as well as the notion that, like mammalian Par2, zebrafish Par2b can sense trypsin-like proteases.
Cells expressing AP-Par2b alone released ∼15,000 AU AP during a 45-min sampling period. Coexpression of zebrafish matriptase with AP-Par2b was associated with release of ∼139,000 AU AP during a 45-min sampling period, a net increase of ∼124,000 AU and ninefold that released in the absence of matriptase expression (Fig. 1 D). Cells expressing the cleavage site mutant AP-Par2b R28A/M29P alone released ∼25,000 AU of AP during the sampling period. Coexpression of zebrafish matriptase with the cleavage mutant was associated with release of ∼51,000 AU AP, a net increase of ∼26,000 AU—only twofold that released in the absence of matriptase and ∼20% of the increase in AP release seen when wild-type AP-Par2b was expressed with matriptase (Fig. 1 D). Total expression of AP-Par2b and AP-Par2b R28A/M29P was similar in these experiments (Fig. 1 E). At face value, these data suggest that, during the sampling period, ∼80% of AP release associated with coexpression of zebrafish matriptase with AP-Par2b was caused by AP-Par2b cleavage at the KNGR28/M29 site, with ∼20% of cleavage occurring at another site. Other potentially matriptase-sensitive sites include K25/N26 or R47/E48. Although we cannot exclude an indirect action of matriptase (e.g., by activating an intermediate protease), these results suggested that the zebrafish matriptase can cleave zebrafish Par2b at its predicted activating cleavage site and hence serve as a Par2b activator. Subsequent genetic experiments described in the following paragraphs are consistent with this notion.
Abnormal skin morphology in hai1a morphant embryos is par2b dependent
As noted in the Introduction, loss of function for the matriptase inhibitor Hai1a by mutation or morpholino (MO) knockdown is associated with matriptase-dependent skin abnormalities characterized by clusters of cells on the surface of the skin of zebrafish embryos (Carney et al., 2007; Mathias et al., 2007; Fig. 2, A, B, and E). The results we have described suggest the possibility that matriptase-driven Par2b activity associated with removal of matriptase inhibition might contribute to the hai1a phenotype. To test this hypothesis, we removed matriptase inhibition by using hai1a MOs and replicated key findings with hai1a mutants. Mutant hai1a phenotypes were sometimes more severe than morphants but were otherwise indistinguishable. Conversely, to ensure new findings regarding the role of Par2b would not be confounded by off-target effects, we removed Par2b activity using a par2b mutant generated with transcription activator-like effector nucleases (TALENs; see Materials and Methods and supplemental data file) and replicated selected phenotypes using par2b morphants.
We first examined the effect of matriptase (St14a) or Par2b deficiency on the presence of cell clusters on the skin of hai1a morphants (Fig. 2, A–E). No skin cell clusters were detected in uninjected wild-type or par2b mutant embryos that did not receive hai1a MO. 95% of embryos injected with hai1a MO showed clusters, but only 11% of hai1a morphants coinjected with st14a MO, 9% of par2b−/− hai1a morphants, and 21% of embryos coinjected with hai1a and par2b MOs showed clusters (Fig. 2 E). The observation that knockdown of st14a expression prevents the appearance of cell clusters in hai1a morphants is consistent with prior work and with the model that increased matriptase activity contributes to this phenotype (Carney et al., 2007; Mathias et al., 2007). The observation that par2b mutation or MO knockdown also prevents the appearance of cell clusters on the skin of hai1a morphants is consistent with the hypothesis that Par2b activation by matriptase contributes to the hai1a phenotype.
In addition to cell clusters on the skin surface (Fig. 2 B), hai1a mutant and morphant zebrafish embryos have a variety of abnormalities in basal layer keratinocytes, including loss of cell–cell contacts and increased motility and proliferation (Carney et al., 2007). Periderm has not been studied in similar detail in this model. Hai1a-deficient embryos also exhibit increased expression of inflammation markers as well as leukocyte infiltration of tissues (Mathias et al., 2007; LeBert et al., 2015). Thus, Hai1a-deficient zebrafish embryos provided an opportunity to probe the role of Par2b in a variety of epithelial functions in the basal layer and in inflammatory responses and to uncover new roles in periderm.
Periderm keratinocytes undergo apical extrusion in hai1a morphants in a matriptase- and par2b-dependent manner
Toward understanding the nature of the cell clusters in hai1a morphants and the effect of Par2b deficiency, we first performed time-lapse vital microscopy using a Tg (krt4:Sce.Abp140-Venus)cy22 line, hereafter called periderm-LifeActGFP, which expresses a fluorescent marker for filamentous actin (F-actin) in periderm keratinocytes. Because abnormal skin morphology in hai1a mutants was most frequent and most obvious over the ventral yolk sac and at the junction of the yolk sac and trunk (Fig. 2 B), this region (Fig. S1) was imaged from 28 to 46 hpf in 10 embryos for each condition.
In control periderm-LifeActGFP embryos, little change in the arrangement of cells or F-actin distribution was noted over the imaging period (Video 1). Strikingly, hai1a morphant periderm-LifeActGFP embryos showed F-actin rings forming and contracting around periderm cells that were extruded apically from the epithelium. The cells surrounding an extruding cell formed rosettes (Fig. 3 A and Videos 2 and 3). Similar events occurred in the skin overlying the trunk (Fig. S2). Each extrusion event represented a single cell leaving the epithelium, but several cells in the same vicinity often underwent extrusion (Videos 2, 3, and 5). The formation of an actin ring around a cell that is “squeezed” to exit apically from an epithelial monolayer, leaving behind a rosette of surrounding cells, is characteristic of apical cell extrusion, a process normally used to eject apoptotic, damaged, or crowded cells while maintaining epithelial barrier function (Gudipaty and Rosenblatt, 2017).
To further explore this extrusion phenotype, we used the periderm-nuclear-EGFP line to determine the frequency of extrusion events and visualize nuclear morphology in extruded cells as an index of apoptosis (Videos 4, 5, and 6; and Fig. 3 B). In control embryos, <0.1% of cells in the field imaged extruded during the 28- to 46-hpf imaging period (Video 4 and Fig. 3, C and D). The rare cells that did extrude had grossly fragmented nuclei (Video 7 and Fig. 3 B), suggesting that extrusion from normal periderm reflects jettisoning of the occasional apoptotic cell. In periderm of hai1a morphants and mutants, the rate of extrusion increased to 4.3% and 4.4%, respectively, but only 0.1% and 0.4% of cells extruded with obvious nuclear fragmentation (Videos 5 and 8 and Fig. 3, C and D). However, the extruding cells in hai1a morphants and mutants often showed disappearance of nuclear EGFP fluorescence coincident with a uniform increase in fluorescence in the cytosol (Videos 5 and 8), suggesting possible disruption of the nuclear envelope late in the extrusion process (see Discussion).
Strikingly, in par2b−/− hai1a morphant embryos, only 0.3% of cells extruded during imaging (Video 6 and Fig. 3, C and D). Similarly, in st14a:hai1a double morphants, only 0.1% of cells extruded during the imaging period (Fig. 3, C and D). Overall, these results reveal matriptase- and Par2b-dependent apical cell extrusion of periderm cells in Hai1a-deficient embryos, suggesting that activation of Par2b by matriptase can directly or indirectly drive this important epithelial behavior.
Loss of cell–cell contacts, hypermobility, and swarming exhibited by basal layer keratinocytes in hai1a morphants is par2b dependent
Hai1a deficiency promotes basal cell migration (Carney et al., 2007). Accordingly, we examined the dependence of dynamics of such basal layer dynamics on par2b using the basal-layer-LifeActGFP line. Sequential images and 2-h time-lapse videos were acquired on live embryos 24–30 hpf. In control embryos, basal layer keratinocytes were stably apposed; 94% of cells maintained contact with one or more adjacent cells for >60 min (Fig. 4, A and E; and Video 9). In contrast, in hai1a morphant embryos, basal layer keratinocytes were usually not tightly apposed and extended filopodia and lamellipodia into the gaps between cells; only 9% of cells maintained stable cell–cell contacts (Fig. 4, B and E; and Video 10). Along with decreased cell–cell contact, cell movement was increased in hai1a morphants (Videos 10 and 11) compared with controls (Video 9), and centripetal movement or “swarming” of basal keratinocytes toward aggregations of the same was sometimes noted (Video 12). In hai1a:st14a double morphants, the basal layer resembled that of controls; basal keratinocytes were tightly apposed and 88% maintained stable cell–cell contacts (Fig. 4, D and E; and Video 13). In par2b−/− hai1a morphants, cells were more tightly apposed than in hai1a morphants, and 73% maintained stable cell–cell contacts compared with 9% in hai1a morphants and 94% in controls (Fig. 4, C and E; and Video 14). However, moving areas of increased LifeActGFP fluorescence suggesting increased F-actin dynamics were sometimes noted (Video 14). In the absence of hai1a MO, basal keratinocyte behavior in par2b−/− zebrafish was indistinguishable from that in wild type (Fig. 4). Collectively, these data show that loss of stable cell–cell contacts and hypermobility in hai1a morphants is matriptase-dependent and, in large part, par2b-dependent, suggesting that activation of Par2b by matriptase directly or indirectly helps drive this epithelial-mesenchymal transition-like behavior of basal layer.
Basal-layer-LifeActGFP embryos provided an opportunity to determine whether basal cells extrude in hai1a morphant embryos as periderm cells do. In contrast to hai1a morphants carrying fluorescent periderm markers, no extrusion of fluorescent cells was detected in the time-lapse studies of hai1a morphant basal-layer-LifeActGFP embryos described in Fig. 4. This observation suggests that apical cell extrusion in hai1a morphants was limited to the periderm.
Abnormal distribution of nuclei in periderm and basal layer in hai1a morphants is par2b dependent
Imaging for the periderm-nuclear-EGFP marker and immunostaining for the basal marker p63 was itself revealing. In contrast to the regular spacing of nuclei seen in controls, patches with increased numbers of nuclei per unit area were observed in both periderm and basal layers in hai1a morphants. These “crowded” patches in periderm (Fig. 5 A) were colocated with crowded patches in the basal layer (Fig. 5 B), suggesting communication between layers. The crowded patches seen in still images corresponded to the areas of apical extrusion and remodeling of the periderm. These areas presumably also corresponded to the patches of hypermobility and swarming of basal keratinocytes seen with the basal-layer-LifeActGFP line described in Fig. 5 and Video 12. Crowded patches were not seen in the periderm or basal layer of par2b−/− hai1a morphants (Fig. 5 and Fig. S3 A). The colocation of “crowding” in both layers with extrusion in periderm and hypermobility in the basal layer is again consistent with some linkage of these phenomena, and their common Par2b dependence suggests that activation of Par2b directly or indirectly drives remodeling, albeit of distinct types, in both layers.
Abnormal subcellular localization of E-cadherin in hai1a morphants is par2b dependent
Cell extrusion from the periderm and loss of cell–cell contacts and hypermotility in the basal layer suggest that cell–cell junctions are altered in both layers in hai1a morphants and mutants. E-cadherin distribution in periderm in hai1a morphants has not been studied. E-cadherin mislocalization in basal layer keratinocytes in hai1a morphants has been reported (Carney et al., 2007), but its dependence on par2b is unknown. Accordingly, we examined the pattern of E-cadherin immunostaining and its dependence on par2b in zebrafish embryos carrying the periderm-nuclear-EGFP marker and immunostained for the basal marker p63.
Immunostaining of hai1a morphants for E-cadherin revealed increased granular cytoplasmic staining in the basal layer clusters (Fig. 5 C), consistent with redistribution of E-cadherin from junctions to intracellular vesicles and in agreement with published results (Carney et al., 2007). Importantly, mislocalization of E-cadherin and abnormal clustering of basal layer cells was absent in par2b−/− hai1a morphants (Fig. 5). These data are in accord with the par2b dependence of loss of stable cell–cell contacts and increased cell motility in the basal layer of hai1a morphants (Fig. 4 and Videos 9, 10, 11, 12, 13, and 14).
E-cadherin mislocalization was also seen in hai1a morphant periderm in areas of clustering of periderm cell nuclei, and such mislocalization was largely absent in par2b−/− hai1a morphants (Fig. S3 A). Thus, E-cadherin is indeed mislocalized in a manner consistent with disruption of adherens junctions in periderm and basal layer in hai1a morphants in a par2b-dependent manner.
Opposite effects of Par2b deficiency on BrdU incorporation in periderm versus basal layer in hai1a morphants
Increased cell proliferation in the basal layer in hai1a morphants has been reported (Carney et al., 2007). Additionally, although crowded patches were absent, we noted a relatively uniform increase in the overall number of nuclei in periderm of par2b−/− hai1a morphants compared with hai1a morphants or controls (Fig. 5 A). Accordingly, we characterized cell proliferation as measured by BrdU incorporation in periderm and basal layer of hai1a morphants and its dependence on par2b.
Periderm-nuclear-EGFP embryos were incubated with BrdU from 28 to 31 hpf, washed, collected at 46 hpf, and immunostained for BrdU and p63. Colocation of BrdU staining with nuclear EGFP was used to identify periderm layer cells that had incorporated BrdU; colocation with nuclear p63 staining was used to identify basal layer cells (Fig. 6, A and B).
In basal layer, 0.4% of nuclei were BrdU-positive in controls compared with 6.7% in hai1a morphant embryos (Fig. 6, A and D). In contrast, the percentage of basal layer nuclei labeled with BrdU in hai1a morphant embryos coinjected with st14a MO or deficient for Par2b was indistinguishable from control. Thus, increased BrdU incorporation in basal layer nuclei in hai1a morphants is st14a and par2b dependent, consistent with the notion that matriptase-driven Par2b activity contributes to proliferation of basal layer cells in this setting.
In periderm, 0.8% of nuclei were BrdU positive in controls compared with 4.0% in hai1a morphants and 0.3% in hai1a morphants coinjected with st14a MO (Fig. 6 C). Thus, as in the basal layer, loss of Hai1a function was associated with a matriptase-dependent increase in the fraction of periderm keratinocytes in S-phase. Surprisingly, in contrast to the basal layer in which increased BrdU labeling in hai1a morphants was prevented by Par2b deficiency, periderm in par2b−/− hai1a morphant embryos showed a further increase in BrdU labeling, with 11% of nuclei BrdU positive compared with 4% in hai1a morphants without Par2b deficiency (Fig. 6 C). In accord with this result, there was an obvious increase in the fraction of periderm cells that underwent nuclear division in time-lapse imaging of par2b−/− hai1a morphant embryos compared with hai1a morphant without Par2b deficiency or controls (Videos 4, 5, and 6).
Of note, BrdU labeling was also increased in periderm in par2b−/− embryos compared with controls (5.0% vs. 0.8%), and injection of par2b−/− hai1a morphant embryos with an st14a MO reduced the percent of periderm nuclei labeled from 11% in par2b−/− hai1a morphant embryos to 1.7% (Fig. 6 C). At face value, these results suggest that matriptase promotes proliferation of periderm cells in hai1a morphants in a Par2b-independent manner and that Par2b, rather than promoting proliferation as it does in the basal layer, inhibits proliferation in periderm.
Leukocyte infiltration of skin and increased inflammatory marker expression in hai1a morphants is par2b dependent
Leukocyte infiltration of skin and increased expression of mmp9 and other inflammatory markers in hai1a morphant embryos has been reported (Carney et al., 2007; Mathias et al., 2007; LeBert et al., 2015). Infiltration of the fin fold with GFP-positive cells was readily seen in hai1a morphants in the Tg (MPO:GFP)uw line, which expresses GFP in neutrophils, and such infiltration was absent in par2b:hai1a double morphants (Fig. S4). Similarly, expression of mmp9, mmp13, and il1b was increased in hai1a morphants, and such increases were st14a and par2b dependent (Fig. 7 A). These results suggest that matriptase-dependent Par2b activation directly or indirectly supports leukocyte infiltration and expression of markers and mediators of inflammation in the skin of hai1a morphants.
Matrix metalloproteinase (Mmp)–dependent components of the hai1a phenotype
Mmp activity can contribute to tissue remodeling (Page-McCaw et al., 2007), and Mmp9 depletion can partially rescue the appearance of skin clusters in hai1a morphants (LeBert et al., 2015). This result and our observation that increased Mmp expression in hai1a morphants is par2b dependent (Fig. 7 A) raised the possibility that Mmp activity might contribute to par2b-dependent features of the hai1a phenotype. Accordingly, we determined the effect of MO knockdown and pharmacological inhibition of Mmp9 and Mmp13 on apical cell extrusion in periderm, altered cell–cell contacts in the basal layer, and altered BrdU incorporation in periderm and in basal keratinocytes in hai1a morphants (Fig. 7, B–E). Although completely par2b dependent (Fig. 3), apical cell extrusion from periderm in hai1a morphants was not different in the absence or presence of mmp9 MOs or Mmp9/13 inhibitor (Fig. 7 B). In contrast, exposure to Mmp inhibitors partially rescued stable cell–cell contacts in the basal layer in hai1a morphants (Fig. 7 C). Most strikingly, exposure to Mmp inhibitors mimicked the effects of Par2b deficiency on BrdU incorporation. Specifically, treatment of hai1a morphants with Mmp9/13 inhibitor prevented increased BrdU incorporation in the basal layer but augmented BrdU incorporation in periderm (Fig. 7, D and E). Mmp9 MOs had directionally similar but smaller effects. At face value, these results suggest that Mmp9/13 activity contributes to altered cell–cell interactions and increased BrdU incorporation in the basal layer and to restraint of BrdU incorporation in periderm, and given par2b-dependent mmp9 and mmp13 expression in hai1a morphants, Par2b-induced Mmp expression may contribute to these phenotypes. Our experiments do not exclude the alternative that Mmp activity is simply permissive for these phenotypes.
Effects of Erbb2 knockdown or EGF receptor (EGFR) inhibition on the hai1a phenotype
Like Hai1a-deficient embryos (Carney et al., 2007; this study), basal layer cells in lgl2 mutant zebrafish embryos show E-cadherin redistribution and increased mobility (Reischauer et al., 2009). These lgl2 phenotypes are prevented by Erbb2 deficiency (Reischauer et al., 2009). G protein–coupled receptors including Par2 can promote EGFR activation by an incompletely understood mechanism dubbed transactivation (Chung et al., 2013; Wang, 2016). Thus, it is plausible that transactivation of Erbb2 by Par2b might contribute to the hai1a phenotype. To explore this hypothesis, we investigated the effect of erbb2 MO and the EGFR inhibitor PD168393 on the hai1a phenotype. PD168393 treatment did not reduce apical extrusion of cells lacking nuclear fragmentation from the periderm in hai1a morphant embryos (Fig. 8 A), nor did it reverse E-cadherin redistribution to endosomes in periderm (Fig. S3 B). Unexpectedly, however, PD168393 treatment of hai1a morphant embryos was associated with a marked increase in apical extrusion of periderm cells with obvious nuclear fragmentation (Fig. 8 B). PD168393 treatment did not increase cell extrusion in control embryos. The dose of erbb2 MO was limited by side effects; at the dose used, its effects mimicked those of PD168393 treatment directionally but did not reach statistical significance. These results suggest that Erbb2 activation does not contribute to live apical cell extrusion in Hai1a-deficient embryos but that Erbb2 or another PD168393 target may be necessary for survival of cells in Hai1a-deficient periderm. Of note, Par2b deficiency blunted the increase in extrusion of cells with fragmented nuclei in Hai1a-deficient PD168393-treated embryos, suggesting that Par2b signaling may contribute to increased apoptosis and/or for extrusion in this setting.
erbb2 MO or PD168393 treatment had no effect on increased BrdU labeling of periderm cells in hai1a morphants, but both markedly attenuated the increased BrdU labeling seen in the periderm of par2b−/− hai1a morphants (Fig. 8, D and E). Indeed, combined Par2b deficiency and Erbb2 inhibition produced a virtually complete reversal of the hai1a periderm phenotype. Collectively, our results are consistent with the notion that Erbb2 activity supports survival of periderm cells in hai1a morphants with intact par2b function and proliferation of periderm cells in hai1a morphants lacking par2b and extrusion. Additionally, the observation that erbb2 MO and PD168393 blocked proliferation in Hai1a-deficient periderm lacking par2b suggests that Erbb2 plays a Par2b-independent role in regulating proliferation of these cells (Fig. 9 A).
In the basal layer, erbb2 MO and PD168393 treatment blunted the increased BrdU labeling in hai1a morphants (Fig. 8, F and G). PD168393 treatment also prevented loss of stable cell–cell contacts and hypermotility in the basal layer in hai1a morphants (Fig. 8 C). Thus, like Par2b, Erbb2 appears to be required for increased cell proliferation and altered cell–cell adhesion in the basal layer of hai1a morphants and may mediate or be permissive for these behaviors (Fig. 9 B).
Matriptase provides local trypsin-like serine protease activity on the surface of epithelial cells (List et al., 2009; Buzza et al., 2010). Matriptase is negatively regulated by Hai1 and Hai2 (Oberst et al., 2005; Szabo et al., 2009; Nonboe et al., 2017). We and others have provided evidence that matriptase and its regulators can modify epithelial cell behavior in part via Par2 (Bocheva et al., 2009; Camerer et al., 2010; Sales et al., 2015b; Le Gall et al., 2016). Gain-of-function for matriptase, its activators, or Par2 can trigger chronic inflammatory and proliferative responses in mouse skin, in some models leading to squamous cell carcinoma (Steinhoff et al., 2003; Bocheva et al., 2009; Frateschi et al., 2011; Cheng et al., 2014). However, the roles of Par2 in regulating epithelial cell behavior and their relationships to matriptase activity have been incompletely explored.
Hai1a-deficient skin in zebrafish embryos has been put forward as a model of chronic inflammation (Mathias et al., 2007). It can also be viewed as a model of epithelial remodeling. Using this model, we identified and characterized Par2-dependent responses. Our studies unexpectedly revealed classic epithelial apical cell extrusion as a feature of this model and demonstrated that this behavior is Par2 dependent. Roles for Par2 in regulating cell proliferation that were opposite in distinct but adjacent epithelial monolayers, and roles in regulating cell–cell junctions, mobility, survival, and expression of genes involved in tissue remodeling and inflammation, were also uncovered. These are epithelial cell behaviors that contribute broadly to embryonic development, homeostasis, inflammation and reparative responses, and because Hai1, matriptase, and Par2 are coexpressed in most mammalian epithelia (Camerer et al., 2010), the actions of matriptase-driven Par2 activity in zebrafish skin likely have counterparts in other tissues and settings.
Our observations are summarized in Table S2. As previously reported, the basal layer of Hai1a-deficient embryos showed a redistribution of E-cadherin, loss of cell–cell contacts, and increased cell motility as well as increased BrdU labeling, all of which were matriptase dependent (Carney et al., 2007). We show that all of these phenotypes are also Par2b dependent. Further, our data suggest that Hai1a, matriptase, and Par2b are coexpressed in basal layer cells and that zebrafish matriptase can cleave zebrafish Par2b at its activating cleavage site like the cognate mammalian proteins (Camerer et al., 2010). Transplantation experiments by Carney et al. (2007) suggest that loss of normal cell–cell contacts and increased movement of basal layer cells associated with Hai1a deficiency is basal cell autonomous. Par2 is an efficient activator of Rac, and hence of F-actin organization in lamellipodia, which drives cell migration (Camerer et al., 2010; Shi et al., 2013). Par2 activation is also associated with loss of E-cadherin adhesion in airway epithelial cells (Winter et al., 2006), and Par2 can promote cell proliferation in some settings (Hirota et al., 2005; Hu et al., 2013; Sales et al., 2015b). Thus, a parsimonious model is that loss of Hai1a function in basal layer cells leads to increased activity of matriptase and Par2b activation in these same cells, with Par2b then promoting E-cadherin redistribution, loss of cell–cell contact, motility, and proliferation (Fig. 9). This model does not exclude participation of other Hai1a targets or matriptase substrates. For example, matriptase can cleave EpCam in vitro, and mutation of epcam in zebrafish triggers cell shedding and proliferative defects (Slanchev et al., 2009; Wu et al., 2017).
Increased il1b, mmp9, and mmp13 expression and leukocyte infiltration of the fin fold in Hai1a-deficient zebrafish embryos were also par2b dependent. Thus, like the basal layer responses, this inflammatory response fits a simple matriptase-driven Par2b activity–dependent model (Fig. 9). Depletion of leukocytes does not prevent the gross skin phenotype associated with Hai1a deficiency (Carney et al., 2007; Mathias et al., 2007) suggesting that, although leukocyte infiltration is downstream of Par2b, it does not cause the other Par2-dependent structural and functional abnormalities seen in the skin of Hai1a-deficient zebrafish embryos. par2b-dependent induction of interleukin-1b (Fig. 7 A) and other cytokines and chemokines or par2b-dependent disruption of barrier function may trigger leukocyte infiltration. Our results in zebrafish are consistent with previous observations suggesting that matriptase and Par2 activity can promote inflammatory responses in mouse skin (Frateschi et al., 2011) and suggest that leukocyte infiltration may not be causal for proliferative and other skin phenotypes in these models.
The periderm of Hai1a-deficient embryos has not been previously studied in detail and revealed several distinct and surprising behaviors. One was apical cell extrusion that was matriptase- and Par2b-dependent. Classically, apical cell extrusion from an epithelium is triggered by apoptosis, damage, or cell crowding (Eisenhoffer et al., 2012; Gudipaty and Rosenblatt, 2017). The nuclei of periderm cells extruding in Hai1a-deficient embryos were not grossly fragmented. Extruding cells did show evidence of nuclear envelope breakdown, but this occurred late in the extrusion process. Thus, apical cell extrusion in this model is probably not initiated by apoptosis, but extruding cells may undergo anoikis.
Interestingly, treatment with Erbb2 inhibitors doubled the already high rate of apical extrusion of periderm cells in Hai1a-deficient embryos, and this increase was attributable to extrusion of cells with already grossly fragmented nuclei. Such extrusion was also Parb2 dependent. Thus, our results suggest that Par2b activity contributes to apical extrusion of both live and apoptotic cells (Fig. 9). Par2b is an efficient activator of Rac and F-actin assembly (Shi et al., 2013), and our data are consistent with a simple model in which matriptase-driven Par2b activity in periderm helps drive cell extrusion directly. Par2 is a G protein–coupled receptor (GPCR), and S1pr2, a GPCR for sphingosine-1-phosphate, can also stimulate apical extrusion in MDCK monolayers and zebrafish epidermis (Gu et al., 2011). Thus, different GPCRs may contribute to extrusion in different settings.
Our studies do not exclude participation of other matriptase substrates or less direct mechanisms of Par2b-driven extrusion. For example, cell crowding can drive apical cell extrusion (Eisenhoffer et al., 2012; Gudipaty and Rosenblatt, 2017), sites with an increased rate of extrusion in periderm were colocated with crowded patches of periderm cells, and the formation of such patches was Par2b dependent. Thus, it is possible that Par2b activity first drives periderm remodeling and cell crowding, with the latter helping to drive extrusion. However, embryos deficient in Hai1a and Par2b showed markedly less extrusion than Hai1a-deficient embryos despite increased periderm cell proliferation and a relatively uniform increase in cell density. Thus, crowding, at least as occurs in Hai1a-Par2b-deficient periderm, is not sufficient to drive extrusion.
Areas of apical cell extrusion and increased cell density in periderm were also colocated with areas of increased cell density in the basal layer, suggesting crosstalk between layers. Mmp or Erbb2 inhibition substantially reversed the basal layer phenotypes of decreased cell–cell contact and increased mobility and proliferation in Hai1a-deficient embryos without blocking apical cell extrusion or increased proliferation in periderm (Table S2). Thus, periderm phenotypes can occur in the absence of basal layer phenotypes. Similarly, previous transplantation studies suggest that basal layer phenotypes in Hai1a-deficient embryos are cell autonomous (Carney et al., 2007) and may occur in the absence of periderm phenotypes. Overall, it appears that matriptase- and Par2b-driven epithelial cell behaviors in each layer may be partly autonomous but coordinated by cross-talk. Potential mechanisms for cross-talk that promotes colocation are discussed later in this article.
In contrast to its effects in the basal layer, Par2b deficiency alone was associated with increased BrdU incorporation in periderm. Further, BrdU incorporation in periderm of embryos lacking both Hai1a and Par2b was increased compared with embryos lacking either Hai1a or Par2b alone. Interestingly, matriptase knockdown largely prevented increased BrdU incorporation in this setting. Thus, as in the basal layer, matriptase activity promotes proliferation of periderm cells but, unlike in the basal layer, Par2b appears to inhibit rather than mediate cell proliferation. These results suggest that matriptase promotes periderm cell proliferation via a substrates other than Par2b. Additionally, opposite roles for Par2b in regulating proliferation in adjacent epithelial structures raises the possibility that this system contributes to a remodeling or differentiation process that must coordinate proliferation across cell types.
Although the periderm and basal layers in zebrafish embryo skin represent distinct lineages (Le Guellec et al., 2004) and apical layers are derived from the basal layer in mouse (Simpson et al., 2011), zebrafish embryo and mouse skin both have apical layers comprised of differentiated squamous cells and basal layers that are more stem-like. Interestingly, Par2 overexpression in basal stem cells in mouse skin is associated with tumorigenesis, but Par2 overexpression in squamous cells inhibits proliferation (Rattenholl et al., 2007; Sales et al., 2015b). Thus, our findings suggest that a functionally analogous system featuring inhibition of apical squamous cell proliferation and stimulation of basal cell proliferation by Par2b is conserved across these species and settings.
Mmps contribute to the Hai1a skin phenotype (LeBert et al., 2015). We observed that induction of Mmp9 and Mmp13 expression in Hai1a-deficient embryos is Par2b dependent, and that Mmp9 and Mmp13 inhibition mimicked the effects of Par2b deficiency in the basal layer. Thus, Par2b-driven increases in Mmp9 and Mmp13 expression by basal cells may contribute to loss of cell–cell contacts and increased motility and proliferation of basal layer cells in Hai1a-deficient embryos. Our studies do not exclude a less direct mechanism or a permissive role for Mmp9 and Mmp13 activity in these phenotypes.
In periderm, Mmp9 and Mmp13 inhibition mimicked the effect of Par2b deficiency on BrdU incorporation. Given that increased Mmp expression was Par2b dependent, this observation raises the possibility that Par2b-driven increases in Mmp activity may contribute to inhibition of proliferation in periderm. In contrast, and perhaps surprisingly, Mmp9 and Mmp13 inhibition had no effect on apical cell shedding in Hai1a-deficient embryos, suggesting that activity of these proteases is unnecessary for this process.
The matriptase- and Par2-driven Erbb2-dependent phenotypes we observed in zebrafish skin may have parallels in human cancer cells. Par2 activation is associated with Mmp- and EGFR-dependent ERK activation and proliferation in a colon cancer cell line, and “transactivation” of EGFR by Par2 has been suggested to contribute to these phenotypes (Darmoul et al., 2001, 2004). Par2 has also been reported to promote survival of an adenocarcinoma-derived lung epithelial cell line by transactivation of EGFR (Michel et al., 2014). These results raise the possibility that Par2b-dependent activation of Erbb2 might contribute to Erbb2-dependent cell proliferation in basal layer and cell survival in periderm in Hai1a-deficient zebrafish skin. However, the ability of erbb2 MO and PD168393 to reverse the increased BrdU incorporation in Hai1a-deficient periderm lacking Par2b suggests that Erbb2 is acting in a Par2b-independent manner in this setting, and that Hai1a- and matriptase-dependent, Par2b-independent mechanisms of Erbb2 activation in this system remain to be uncovered.
The mechanism by which matriptase activity is regulated in normal physiology is not fully understood. In at least some polarized epithelial cell types, the matriptase activator prostasin traffics to the apical cell membrane, whereas matriptase is basolateral (Buzza et al., 2013). Because these proteases can mutually activate, loss of junctional integrity and polarity resulting in altered protease localization might trigger their activation. Given that matriptase gain of function leads to E-cadherin redistribution and loss of cell–cell junctions (Buzza et al., 2010) and our finding that these activities are Par2b dependent in zebrafish embryo skin, a positive feedback loop might exist in which matriptase and Par2b activity begets loss of junctions and polarity with altered trafficking that begets more local matriptase and Par2b activity. Further, Hai1a is partially redundant with Hai1b in zebrafish skin (Carney et al., 2007). Thus, the matriptase system in Hai1a morphant embryos is only partially activated at baseline, but it may be more sensitive to activation and amplification mechanisms. Such a sensitized self-amplifying system might allow mechanical or biochemical signals, perhaps inducing cytokines or Mmps or shed matriptase itself to act between layers and drive the colocation of patches of affected epithelium in basal and periderm layers observed in our studies.
The Par2b-dependent responses that were triggered by Hai1a deficiency in our studies—apical cell shedding in the periderm; loss of cell–cell contacts, mobility, and proliferation in the basal layer; and cytokine and Mmp production—would make sense as an initial response to injury, and the notion that activation of matriptase and Par2b in this system might normally be initiated by local injury sensed as loss of epithelial integrity or polarity would fit such a model. The zebrafish skin offers a tractable system to test these ideas by future studies to determine where and when matriptase and Par2b are activated, their cell-autonomous roles in distinct epithelial behaviors, and their importance in responses to epithelial perturbations. Use of this system to further understand the roles of Par2 may illuminate mechanisms contributing to epithelial homeostasis and disease and help guide preclinical and clinical studies of drugs targeting this receptor (Yau et al., 2013).
Materials and methods
Matriptase is also known as MTSP1 and St14. The zebrafish matriptase homologue St14a (gene symbol st14a) was studied here. In general, we refer to the protein encoded by st14a as matriptase and the gene as st14a. Hepatocyte-activator inhibitor 1 (Hai1) is also known as Spint1. The zebrafish Hai1 homologue Hai1a, also known as Spint1a (gene symbol hai1a or spint1a), was studied here. Par2 is also known as F2rl1. The zebrafish Par2 homologue Par2b, also known as F2rl1.2 (gene symbols par2b or f2rl1.2), was studied here.
par2b and st14a constructs
Zebrafish par2b and zebrafish st14a cDNAs were obtained from GE. The Par2b cleavage reporter AP-Par2b was generated in a manner analogous to previous mammalian versions (Ludeman et al., 2004; Camerer et al., 2010). In brief, zebrafish Par2b was amplified by PCR using primers 5′-GGCCGGATCCACCATGGCGGTGTCCGAGA-3′ and 5′-GGCCGCGGCCGCTCAGCAAGTGCTGGTTTCCGTGTT-3′ and inserted between the BamH1 and Not1 sites in pcDNA3.1, or using the primers 5′-GGCCGTTAACGCCCAGCCAGGCAAAAATGG-3′ and 5′-GGCCGTTAACGCAAGTGCTGGTTTCCGTGTT-3′ and inserted at the HapI site in pCMV-SEAP. A Par2b cleavage site mutant version of AP-Par2b was made using the primers 5′-GGCCGGATCCACCATGGCGGTGTCCGAGAGCTACAGGATTTTATTATTTTTGGCGTGTGTCATTTTTGCTTCTGCCCAGCCAGGCAAAAATG-3′ and 5′-GGCCTCTAGATCAGCAAGTGCTGGTTTCC-3′. The PCR product was used to generate a BamHI and XbaI fragment to replace the cognate fragment in AP-Par2b.
Zebrafish st14 was amplified by PCR using primers 5′-GGCCGGATCCACCATGGACCCTATGGATGGAGGAAT-3′ and 5′-GGCCGCGGCCGCTTACACTCCCGTCTTCTCCTT-3′ and subcloned between the BamHI and NotI sites in pcDNA3.1. Constructs were confirmed by sequencing. Enzymes were from New England Biolabs.
AP release assay
HEK 293T cells were plated in DMEM with 10% FBS at a density of 0.5 × 105 cells/well of a 24-well plate. Approximately 24 h later, cells were left untransfected or transfected with empty pcDNA3.1 or pcDNA3.1 constructs directing expression of AP-Par2b with a wild-type or mutated cleavage site and/or st14a. Approximately 24 h later, wells were washed once with 0.5 ml DMEM medium containing 0.1% BSA and 20 mM Hepes, and 300 µl medium was added to each well and incubated for 45 min at 37°C. 200 µl medium was removed and transferred to labeled tubes kept on ice. The remaining 100 µl was removed from the wells and cells washed once with 0.5 ml medium. 300 µl fresh medium containing 0.1 µg/ml trypsin (TPCK treated; Sigma-Aldrich) was added to each well and incubated for 10 min at 37°C. 200 µl of the trypsin-containing medium was removed from the wells and kept on ice. The samples from pre- and post-trypsin treatments were centrifuged at 13,000 rpm for 10 min, and 60 µl of each sample was added to 180 µl of 1× sample dilution buffer from the Tropix kit (Applied Biosystems) and heated for 30 min at 65°C. Samples were then cooled to room temperature, and 50 µl of each was added to 50 µl assay buffer in a 96-well plate in triplicate and incubated for 5 min at room temperature. 50 µl reaction buffer containing AP chemiluminescent substrate (Tropix kit) was added to each well and incubated further for 20 min. Chemiluminescence was measured in a microplate luminometer (Promega Glo Max). Where indicated, total expression of AP-Par2b constructs was assessed by lysing cells in 250 µl of 0.2% Triton X-100 24 h after transfection; 15 µl lysate was added to 185 µl Tropix buffer, which was then processed as just described to measure AP activity.
Zebrafish maintenance and strains
Zebrafish were maintained and handled in compliance with standard (http://zfin.org) and University of California, San Francisco Institutional Animal Care and Use Committee protocols. Wild-type strains used were TL, AB, and EKW. The par2b mutant line was outcrossed twice to each of these strains and to four of the transgenic lines; in all backgrounds, Par2b deficiency rescued the hai1a morphant skin phenotype as described in Results. Published transgenic lines were Tg (krt4:nlsEGFP)cy34 (Chen et al., 2011), Tg (krt4:Sce.Abp140-Venus)cy22 (Chen et al., 2011), Tg (UAS:Lifeact-GFP)mu271 (Helker et al., 2013), TgBAC (ΔNp63:Gal4FF)la213 (Rasmussen et al., 2015), Tg (MPO:GFP)uw (Mathias et al., 2006), and Hai1a Mutant Line: spint1lhi2217 (Carney et al., 2007; Mathias et al., 2007).
Injections and MOs
MOs were obtained from Gene Tools. 2 nl of MO solution was injected at the one- to two-cell stage. Sequences and concentrations were as follows: Hai1a MO: 5′-ACCCTGAGTAGAGCCAGAGTCATCC-3′, 50 mM (Carney et al., 2007); Par2b (f2rl1.2) MO: 5′-ATTCCGCTTTTTCCTCTAGTACTCC-3′, 100 mM; matriptase1a (st14a) MO: 5′-AACGCATTCCTCCATCCATAGGGTC-3′, 100 mM (Carney et al., 2007); mmp9 MO1 5′-GAATAATGTCCCACCTGTATGTGAC-3′, 50 mM (Volkmann et al., 2010); mmp9 MO2 5′-GTAAGTTTACCTCTGTTAGGGCAGA-3′, 50 mM (Volkmann et al., 2010); and erbb2 MO: 5′-GTCCGCCTCCATCGATTATTCCTCC-3′, 50 mM (Lyons et al., 2005).
Generation of par2b (f2rl1.2) mutant
The design of the TALEN effectors was done with the web tool at https://tale-nt.cac.cornell.edu/. TALEN constructs were assembled using the Golden Gate TALEN and TAL Effector kit 2.0 according to Cermak et al. (2011) (Addgene). Two Target sequences were used for exon 2: 5′-TACACCCAGCTGCCATTTACATGGGCAACCTGGCACTTGCAGACCTG-3′ and 5′-TATGGCGAGGAGATGTGCAAAGTATCAGTGGGCTTCTTCTACGGGAACATGTA-3′ (spacer sequence underlined). TALEN effector mRNAs were synthesized using the Ambion mMessage mMachine T7 Ultra kit (Applied Biosystems) and injected at the one-cell stage. The experiments were done with F2 and subsequent generations.
Mmp9 inhibitor I (444278) and Mmp9/13 inhibitor I (444252; Calbiochem/Millipore) were used at 100 µM. The EGFR inhibitor PD168393 (513033; Calbiochem/Millipore) was used at 10 µM. For all the experiments, drugs were added in embryo medium (5.0 mM NaCl, 0.17 mM KCl, 0.33 mM CaCl2, and 0.33 mM MgSO4) with 2% DMSO 6 h before the measurements or time-lapse started and remained present for the duration of the experiment.
For photos of live embryos at 30 hpf, the embryos were embedded in 2% methylcellulose (Sigma-Aldrich) in embryo medium with 0.04% tricaine (Sigma-Aldrich). For time-lapse vital microscopy, embryos were embedded in 1% low-melting agar in embryo medium with 0.04% tricaine (Sigma-Aldrich). When imaging the periderm, recording started at 28 hpf and continued for 18 h, with images acquired at 8-min intervals. A 710 laser-scanning confocal (Zeiss) and spinning disk fluorescence (Nikon) microscopes were used. When imaging the basal layer, 2-h recordings were performed between 24 and 30 hpf with images collected at 2-min intervals. A 710 confocal fluorescence microscope (Zeiss) was used. Imaging was performed with a long working distance 20× lens. All the time-lapse studies were performed at 28.5°C. Image analysis was performed using Zeiss Black or Fiji ImageJ software.
Immunofluorescence staining and BrdU incorporation
For E-cadherin and p63 staining, embryos were fixed at 30 hpf in 4% PFA in PBS overnight and postfixed in methanol at −20°C overnight. Staining was performed as previously described (Reischauer et al., 2009). Mouse E-cadherin antibody (clone 36; 610181; BD Bioscience) was used at 1:100, rabbit p63 antibody (gtx124660; Genetex) at 1:100, and secondary antibodies conjugated with Alexa Fluor 564 and 647 (Life Technologies) at 1:500. For BrdU incorporation and staining, embryos at 28 hpf were incubated with 10 mM BrdU (Life Technologies) and 2% DMSO in embryo medium for 3 h followed by three 5-min washes in embryo medium. Embryos were then incubated for 15 h and fixed in PFA 4%. The BrdU staining was performed as described (Reischauer et al., 2009). Sheep BrdU antibody (ab1893; Abcam) was used at 1:100 and mouse p63 antibody (clone 4A4; sc8431; Santa Cruz Biotechnology) at 1:200. Immunostained embryos were imaged using a 710 confocal microscope and 20× and 40× long working distance lenses. Imaging analysis was performed using Zeiss Black or Fiji ImageJ software.
Keratinocyte isolation and quantitative PCR
Periderm and basal layer keratinocytes were sorted from Tg (krt4:nlsEGFP)cy34 and TgBAC (ΔNp63:Gal4FF)la213: Tg (UAS:Lifeact-GFP)mu271 zebrafish embryos, respectively, as previously described (Manoli and Driever, 2012). In brief, 100 embryos of each genotype were manually dechorionated at 24 hpf. Cells were dispersed from embryos, and GFP-labeled cells were sorted using a FACS Aria2 (BD Bioscience) and lysed in Trizol (Sigma-Aldrich); RNA was purified using the RNeasy micro-kit (Qiagen) and converted to cDNA library using the SuperScript-Vilo kit (Life Technologies). Quantitative PCR was performed using TaqMan gene expression master mix (Life Technologies) according to the manufacturer’s instructions. TaqMan-Fam Probes were ordered from Life Technologies. The reaction was performed with 7900HT Fast Real-Time PCR system (Applied Biosystems). Probes and assay IDs were as follows: reference gene Eef1a1l, assay ID: DR03432748_m1; Il1b, assay ID: DR03114367_g1; Mmp2, assay ID: DR03076189; Mmp9, assay ID: Dr03139883_g1; Mmp13a, assay ID: Dr03438514_g1; St14a, assay ID: AIS09DC; Krt4, assay ID: DR03093320_gh; P63, assay ID: Dr03131730_m1; Spint1a, assay ID: DR03139912_m1; and F2rl1.2, assay ID: Dr03168088.
One-way ANOVA with Bonferroni posttest or two-way ANOVA with Tukey posttest were used to identify significant differences between treatment groups. Sample sizes and p-values are indicated in figure legends. Asterisk in figures indicates significant differences after correction for multiple comparisons with p-value indicated in the legends. Calculations were done using Prism statistical software (GraphPad), and the level of statistical significance was set at P < 0.05 for initial ANOVA before any post-tests were permitted.
Online supplemental material
Fig. S1 shows photos of a mounted zebrafish embryo indicating the region selected for time-lapse imaging. Fig. S2 shows apical extrusion of cells from the trunk periderm cells in hai1a morphants. Fig. S3 shows distribution of nuclei and subcellular distribution of E-cadherin in periderm in control and hai1a morphants and effect of inhibition of mmp9/mmp13 (by drugs and MOs), erbb2 (by MO), and EGFR (by inhibitor PD168393). Fig. S4 shows leukocyte infiltration of skin in hai1a morphant and rescue by Par2b depletion. Video 1 show periderm in control with LifeAct-GFP marker. Video 2 shows periderm in hai1a morphant showing cell extrusion with LifeAct-GFP marker. Video 3 shows periderm in hai1a morphant showing multiple single cell extrusions from one site with LifeAct-GFP marker. Video 4 shows periderm in control using nuclear-GFP marker. Video 5 shows periderm in hai1a morphant showing clustering of extrusion events in an area of apparent cell crowding with nuclear-GFP marker. Video 6 shows periderm in par2b mutant/hai1a morphant showing increased nuclear division with nuclear-GFP marker. Video 7 shows rare extrusion of a periderm keratinocyte with nuclear fragmentation in a control embryo with nuclear-GFP marker. Video 8 shows a common event in hai1a morphants—extrusion of cells with an initially intact nucleus that appears to undergo nuclear envelope breakdown late in the process (with nuclear-GFP marker). Video 9 shows the basal layer in control showing stable cell–cell contact with LifeAct-GFP marker. Videos 10 and 11 show the basal layer in hai1a morphants showing lack of stable cell–cell interaction, filopodia and lamellipodia formation, and increased cell movement. with LifeAct-GFP marker Video 12 shows the basal layer in hai1a morphant showing swarming behavior with LifeAct-GFP marker. Video 13 shows basal layer in st14a/hai1a morphants showing restoration of stable cell–cell contact with LifeAct-GFP marker. Video 14 shows the basal layer in par2a mutant/hai1a morphant showing restoration of stable cell–cell contacts with LifeAct-GFP marker. Table S1 shows quantitative PCR data showing expression of mRNA for hai1a, st14a, and par2b in periderm and basal layer keratinocytes sorted from 24-hpf zebrafish embryos. keratin4 and tp63 are included as periderm and basal keratinocyte markers. Table S2 is a summary of effects of Par2b and St14a deficiency on Hai1a phenotypes. The supplemental data file shows nucleotide and predicted amino acid sequences of wild-type and TALEN mutant zebrafish par2b.
We thank professors Takashi Mikawa and Orion Weiner for their critical reading of this manuscript.
A. Schepis was supported by National Institutes of Health T32 HL007731. This work was supported by National Institutes of Health HL R35HL135755 and R01 HL054737 to S.R. Coughlin.
The authors declare no competing financial interests.
Author contributions: A. Schepis designed studies with S.R. Coughlin, performed most of the experiments, and wrote the manuscript with S.R. Coughlin. A. Barker performed early foundational experiments. Y. Srinivasan did the AP-Par2b studies with E. Balouch. Y. Zheng designed and made several of the constructs. I. Lam contributed genotyping and quantitative PCR. H. Clay provided important early experiments with A. Barker. C.-D. Hsiao provided important zebrafish lines.