The Epithelial Cell Adhesion Molecule EpCAM Is Required for Epithelial Morphogenesis and Integrity during Zebrafish Epiboly and Skin Development

1Like in mammalian gestation embryos, the epidermis of teleost larvae is bi-layered, consisting of an outer enveloping cell layer (EVL), which morphologically and functionally resembles the periderm of mammalian embryos,, and a basal layer of keratinocytes. The function of the mouse periderm is poorly understood, however, recent genetic evidence points to pivotal roles during skin formation and other developmental processes. Furthermore, the zebrafish EVL might serve as a model for other, medically more relevant simple epithelia, such as the epithelial tubules of the developing zebrafish kidney, which during nephron morphogenesis display several crucial cellular features similar to those of the EVL described here.

2Zebrafish EVL cells segregate from deep cells during blastula stages and cover the embryo during further development, while it remains unclear whether during metamorphosis, when the zebrafish epidermis becomes stratified, they are replaced by cells derived from basal keratinocytes. During gastrulation, EVL cells undergo epiboly movements to progressively spread over the yolk, tightly coordinated with the simultaneous vegetal-wards displacements of deep cells and the yolk syncytial layer (YSL) –. Epiboly movements start approximately 4 hours after fertilization (hpf), when only 30% of the yolk cell is covered by the blastoderm (30% epiboly), and is completed at 10 hpf, when the yolk is entirely surrounded by deep and EVL cells (100% epiboly). The molecular mechanisms underlying EVL epiboly have just started to be elucidated, whereas the Ca2+-dependent cell adhesion molecule and adherence junction (AJ) component E-cadherin (Ecad; Cdh1) has been shown to be specifically required for epiboly of deep cells,. The major force of deep cell epiboly is their polarized intercalative displacement from inner to outer layers. According to one report, Ecad drives these directed intercalations by forming an adhesion gradient within the deep layers themselves, whereas according to another report, it is required for proper attachment of deep cells to the overlying EVL.

3In addition, the EVL serves as a primary “skin”, constituting a barrier between the embryo proper and the fresh water environment. In contrast to basal keratinocytes, EVL cells are sealed to each other via apical junctional complexes (AJCs), which consist of tight junction (TJ) and, possibly, AJ sections. TJ proteins like Tjp/Zo1 and Tjp/Zo3 already accumulate at the lateral/apical sides of EVL membranes during early blastula stages, while loss of Tjp/Zo3 function results in increased surface permeability and compromised osmoregulation. In addition, EVL cells form desmosome-like junctions between each other and with underlying cells,.

4EpCAM was first described in 1979 as a 40 KDa cell surface cancer-associated antigen,. In addition, it was isolated in several other contexts, which resulted in a plethora of synonyms such as Tacstd1 or Trop1, recently unified under the name Epithelial cell adhesion molecule (EpCAM) or CD326 (reviewed in,). EpCAM is a type I single span transmembrane glycoprotein, with extracellular epidermal growth factor-like (EGF-like) and thyroglobulin (TY) motifs,,, and an intracellular domain containing an internalization motif and several α-actinin binding sites. In mammals, EpCAM is present on the baso-lateral surface of most developing epithelia –. Expression is usually down-regulated as epithelial cells terminally differentiate. For example, progenitor cells of human skin epithelium express EpCAM, whereas differentiated keratinocytes do not. However, EpCAM levels often rise again during regeneration or neoplastic transformations,,, –. Strikingly, EpCAM is only found in epithelial-derived cancers, i.e., carcinomas, but not in others, such as sarcomas, melanomas, or lymphomas.

5Despite exhaustive in vitro studies, the exact roles of EpCAM during tumor progression and the molecular and cellular mechanisms of its functions are not fully understood, with several controversial findings, implicating EpCAM with adhesion, migration, metastasis, proliferation, differentiation, signaling and metabolism (reviewed in,). Whereas according to some data, EpCAM acts as a homophilic cell-cell adhesion molecule with positive effects on cell adhesiveness and negative effects on cell motility and metastasis, –, other data are in line with anti-adhesive and migration-promoting functions of EpCAM. Thus, in the presence of classical cadherins, EpCAM can reduce cell-cell adhesion, possibly by interfering with the interaction between cadherins and the cytoskeleton,. In addition, EpCAM physically interacts with the metastasis-promoting cell surface receptor CD44v4-V7 and the tight junction component Claudin7, possibly blocking Claudin function during invasion and metastasis of several carcinomas,. Furthermore, EpCAM enhances proliferation rates of carcinoma cells,, presumably mediated via direct nuclear signaling of its proteolytically cleaved intracellular domain EpICD, and via c-Myc and the cell cycle regulators cyclin A and E –.

6Most of these functional data were obtained in cell or tissue culture systems and via EpCAM overexpression. In contrast, in vivo and loss of function studies are scarce. No EpCAM mouse mutants have been reported so far. In zebrafish, EpCAM (Tacstd) mutants were generated via random retroviral insertions, characterized by delayed otolith formation in the developing inner ears. In this study we generate maternal-zygotic zebrafish EpCAM mutants and chimeric embryos, revealing essential roles of EpCAM in the EVL for proper epithelial morphogenesis integrity during epiboly and skin development. Some of these roles are fulfilled in partial or complete redundancy with E-cadherin, whereas for others, EpCAM is absolutely indispensable. The molecular mechanisms underlying these in vivo functions are not completely clear. However, mutant cells display basal extensions of TJs and increased membrane levels of TJ components, coupled with reduced Ecad membrane localization and reduced protrusive basal activity. After skin formation, mutants also display hyper-proliferation of EVL and the underlying epidermal cells, which, however, seems to be a secondary consequence of the epithelial defects. We conclude that EpCAM acts as cell-cell adhesion molecule and a partner of E-cadherin, promoting both epithelial integrity and epithelial morphogenesis.

7To identify genes with essential functions during zebrafish skin development, we performed an antibody-based screen on a previously described bank of retroviral insertional mutants, staining larvae at 120 hours post fertilization (hpf) for the basal keratinocyte-specific transcription factor ΔNp63, which is required for epidermal development in both fish and mammals –. The bank contained two mutant alleles of epcam (ZFIN: tacstd; GenBank accession number NM212175), hi2151 and hi2836, which were described to have delayed otolith development (Figure 1A and 1B). In the case of hi2836, the retroviral cassette is inserted upstream from exon 1 of the EpCAM gene, whereas the insertion in hi2151 is within exon 2, causing a frame shift and premature termination of the protein that removes all annotated functional domains (Figure 1D and 1E). In our assay, both alleles displayed aggregates of basal keratinocytes of undistinguishable strengths (Figure 1A and 1B; and data not shown). Similar aggregates, as well as the characteristic delay in otolith development, were observed in embryos injected with an antisense morpholino oligonucleotide (MO) targeting the translational start site of epcam mRNA (Figure 1C).

8epcam mutants kept under standard conditions (28°C) usually died between day 5 and 7 of development, whereas they were sub-viable when kept under semi-sterile conditions and at lower temperature (25°C). However, keratinocyte aggregates of survivors remained visible throughout the first three weeks of development, and fish grew more slowly (Figure S1B, S1C). Adult homozygous mutants were fertile and appeared morphologically normal, even in histological studies (data not shown).

9epcam was previously shown to be expressed in migrating neuromast primordia, otic vesicles and olfactory placodes –. Here, we have extended these analyses with special focus on the skin. Maternally provided epcam mRNA was uniformly distributed in all cells of cleavage and early blastula stage embryos (Figure 2A). After the onset of epiboly, the first of the morphogenetic movements of gastrulation, during which the yolk becomes progressively overgrown by the blastoderm, epcam mRNA was restricted to the enveloping layer (EVL), whereas deep cells had become epcam-negative (Figure 2B–2D). In offspring of homozygous hi2151 or hi2836 mothers and heterozygous fathers, all embryos lacked epcam transcripts at cleavage stages (compare Figure 2F with Figure 2E), whereas at early gastrulation, only 50% (most likely M−/−,Z−/−; see below) remained negative, while the other half (M−/−,Z+/−) had gained normal EVL expression (Figure 2G). This indicates that zygotic epcam expression starts at blastula stages. In addition, it suggests that the hi2836 and hi2151 mutations cause a transcriptional blockade or mRNA instability, respectively, and that both alleles are null mutations.

10At 24 and 48 hours post fertilization (hpf), in addition to the previously described epithelial structures (see above; Figure 2H and 2K), we noticed persistent epcam expression in the EVL and expression in the basal epidermis (Figure 2I and 2J), a derivative of the ventral ectoderm which during gastrula and early segmentation stages had been epcam-negative (Figure 2C). This expression persisted throughout the investigated larval stages (until 120 hpf). Also, according to RT-PCR analyses, epcam RNA was present in the skin of adult zebrafish (data not shown). To investigate the subcellular distribution of EpCAM protein, we tried immunostainings with different antibodies against human or mouse EpCAM (see Materials and Methods), however, none of them gave specific signals (data not shown). Therefore, we injected early zebrafish embryos with mRNA encoding full-length zebrafish EpCAM fused to Green fluorescent protein (GFP). The fusion protein was detected at the cell membrane of both EVL and deep cells (Figure 2L and 2M).

11Early zebrafish development is regulated by a combination of maternal factors (deposited in the egg during oogenesis) and zygotic factors generated by the embryo itself. Crossing homozygous or heterozygous epcam mutant females with heterozygous males, we could generate mutants of three different genotypes: M+/−,Z−/−, which still have the maternally supplied epcam gene products, but lack the embryonic contribution (zygotic effect), M−/−,Z+/−, which lack the maternal, but contain embryonic gene products (maternal effect), and M−/−,Z−/−, which lack both the maternal and the embryonic supply (maternal-zygotic effect; MZ−/−). By all means M−/−,Z+/− mutants had wild-type appearance, indicating that zygotic epcam gene products are sufficient for normal development. Nevertheless, defects of MZ−/− embryos were more severe and developed earlier than in zygotic mutants, indicating that maternally supplied gene products can partly take over the role of zygotic epcam: whereas skin aggregates in zygotic mutants only developed during the second day of development (data not shown), in MZ−/− mutants, they were already apparent at mid-segmentation (16 hpf; Figure 3A and 3B), while later, skin aggregates were larger (Figure S1A), and high numbers of shed skin cells were found floating in the chorion (starting at approximately 16 hpf; see Figure 3G for 36 hpf). At 16 hpf, aggregates often occurred in the tailbud region, where the tail grows out. Here, the skin is most likely exposed to highest mechanical pressure and/or undergoes most dramatic epithelial morphogenesis (Figure 3B). To investigate cell aggregates in greater detail, we took advantage of a transgenic zebrafish line in which EVL cells are labeled by GFP. Our analysis showed that the early skin aggregates in MZ−/− embryos primarily consisted of EVL cells, which had acquired a roundish shape and had piled up on each other. In contrast, p63-positive basal cells underneath the foci seemed unaltered (Figure 3C–3E), and only formed aggregates much later (see Figure S1A for 48 hpf). At these later stages, epcam is expressed both in EVL and basal cells (see above; Figure 2J). To distinguish whether the late aggregation of basal cells is caused by a loss of EpCAM in basal cells themselves, or by its loss in the overlying EVL cells, we generated chimeric embryos, transplanting basal MZ−/− cells into wild-type hosts and vice versa. Strikingly, even largest clones of mutant basal cells were organized normally when present in a wild-type environment (compare Figure 3I with Figure 3H), whereas clones of wild-type basal cells in mutant hosts formed aggregates indistinguishable from those in non-chimeric mutants (Figure 3J). In sum, this indicates that EpCAM from EVL cells is required for proper epithelial organization both within the enveloping layer per se and in the underlying layer of basal keratinocytes.

12Elevated EpCAM levels in transformed tissues are thought to promote cell proliferation, contributing to carcinoma progression,. To determine proliferation rates in zebrafish epcam mutants, we carried out Bromodeoxyuridine (BrdU) incorporation studies in combination with EVL or epidermal-specific markers. During somitogenesis (16 hpf), when EVL aggregation and cell shedding is already apparent, we could not detect any difference in the number of BrdU labeled cells between mutants and their WT siblings (Figure 4A and 4B). At 24 hpf, proliferation rates in both the EVL and the basal layer of epcam mutants were slightly elevated, while at 48 hpf, this difference only persisted in the EVL, but not in the basal layer (Figure 4A and 4B). Strikingly, aggregates of mutant EVL or basal cells displayed BrdU incorporation rates as in non-affected regions of the skin (Figure 4C; and data not shown). Together, this suggests that cell aggregates in epcam mutants are not due to hyper-proliferation and that, in turn, hyper-proliferation might be a secondary consequence of compromised epithelial integrity.

13At 24 hpf and later, we also observed increased numbers of leukocytes in the skin of MZepcam mutants (Video S1, Video S2; and data not shown), similar to the recently described defects in zebrafish mutants lacking the Hepatocyte growth factor activator inhibitor Hai1,. However, genetic ablation of the myeloid lineage in MZepcam mutants with pu.1 antisense MOs did not ameliorate the skin defects, ruling out that they are secondary consequences of enhanced skin inflammation (data not shown). In hai1 mutants, skin invasion by innate immune cells might be triggered by the apoptosis of skin cells. However, according to TUNEL stainings, even the skin aggregates of EpCAM mutants only displayed moderately increased numbers of apoptotic cells after 48 hpf (Figure 4D; data not shown), making such a mechanism rather unlikely. To test whether leukocytes are activated as a result of compromised skin barrier and enhanced infection, we compared wild-type and mutant embryos kept under semi-sterile conditions or after incubation in water contaminated with bacteria. Under semi-sterile conditions, wild-type and mutant embryos displayed identical patterns of leukocyte distribution, revealed via leukocyte-specific-plastin (lcp1), in situ hybridization at 48 hpf (Figure S2A, S2B). However, after bacterial challenge for 2 hours, the skin of epcam mutants contained much more inate immune cells (Figure S2D) than in challenged wild-type controls (Figure S2C) or un-challenged mutant siblings (Figure S2B). In sum, these data suggest that loss of EpCAM primarily affects epithelial properties of the enveloping cell layer, whereas dysmorphology of the basal layer, hyper-proliferation, apoptosis, infection and inflammation are later or secondary consequences.

14The EVL-specific defects of MZepcam mutants during segmentations stages prompted us to carry out more thorough analyses of this cell lineage during earlier development. During epiboly, cells throughout the EVL flatten out and increase their apical and basal surfaces,, while marginal EVL cells constrict, involving actin and myosin 2 that is localized in the yolk cytoplasm along the margin of the EVL. In MZepcam mutant embryos, this constriction occurred at positions slightly closer to the animal pole than in wild-type embryos, resulting in an extrusion of the vegetal-most part of the yolk at mid gastrula stages (Figure 5A and 5B) and, sometimes, in embryonic death. This suggests that loss of EpCAM might specifically affect the vegetal-wards spreading of the EVL, but not the actin-myosin-dependent constriction of marginal EVL cells. Compromised EVL epiboly as a result of reduced constrictions at the EVL margin has recently been described for embryos after knock-down of msn1, the zebrafish ortholog of the Drosophila Ste20-like kinase Misshapen that is required for actin/myosin 2 recruitment, or after treatment with the specific myosin 2 inhibitor blebbistatin. In contrast, phalloidin stainings of the actin cytoskeleton revealed normal constrictions of marginal EVL cells in MZepcam mutants (Figure S3). However, MZepcam mutants did display a significant reduction in yolk coverage by the EVL at late gastrula stages (77±4.9% (n = 9) versus 89±4.7% (n = 10) in wild-type embryos of the same age; Figure 5C, 5D, and 5G), and a corresponding reduction in the average surface of individual EVL cells (Figure 5I, 5J, and 5K). Epiboly of deep cells was similarly delayed in MZepcam mutants (72±4.8% (n = 9) versus 83±3.1% (n = 10) in wild type embryos; Figure 5C, 5D, and 5G), while the distance between the marginal borders of the EVL and the deep cell layer was as in wild-type embryos (Figure 5E, 5F, and 5H). Furthermore, and most strikingly, whereas the lateral sides of marginal and, to a lower extent, equatorial wild-type EVL cells appeared ruffled, indicating the presence of (basal; see below) cellular protrusions, MZepcam mutant cells lacked these ruffles (Figures 5I and 5J and Figure 6A–6G). Ruffles could be at least partly restored by injection of synthetic epcam mRNA into mutant embryos (Figure 6G–6I). However, no rescue was obtained upon epcam re-introduction into single EVL cells (Figure 6J; n = 0/13), suggesting that EpCAM does not act in a strictly cell-autonomous manner, but is most likely also required in the neighboring cells to allow proper ruffle formation. Together, these data indicate that EpCAM is required for processes of epithelial morphogenesis driving the cell shape changes and spreading of the EVL during zebrafish epiboly.

15To further investigate the cellular basis of the epithelial defects of MZepcam mutants, we performed immunohistochemistry and transmission electron microscopy (TEM). Consistent with the ruffles visualized via phalloidin stainings (see above; Figure 5I), TEM revealed protrusions at the basal side of wild-type EVL cells (Figure 7A–7D). These protrusions started to form during early gastrulation stages (Figure 7A), increased in length during gastrulation (Figure 7B–7D), and shortened again during segmentation stages (Figure 7F). They were in tight physical contact with the basal side of neighboring EVL cells, and even formed in most marginal EVL regions that are devoid of underlying deep cells (Figure 7B). In agreement with the altered actin pattern (Figure 5J), these basal protrusions were much broader and shorter in MZepcam mutants (Figure 7E). TEM sections further revealed that the closely sealed apical junctional complexes (AJCs) of EVL cells were basally extended in MZepcam mutants compared to wild-type controls (Figure 7G–7M). This difference was apparent throughout all investigated developmental stages from mid gastrulation through day 5 of development. However, desmosomes appeared morphologically unaltered (Figure 7G and 7H; and data not shown).

16In addition, consistent with the basal extension of TJs, immunohistochemistry revealed an increase in the staining intensity for Tight junction protein 1 (Tjp1/ZO1) in the lateral membranes of MZepcam mutant EVL cells (Figure 8A and 8B). Supporting results were obtained in EpCAM gain-of-function studies in madine-darby canine kidney (MDCK) cells, which upon transfection with GFP-EpCAM displayed reduced membranous signals for the TJ components Tjp1 and Occludin, whereas desmoplakin stainings appeared unaltered (Figure S4). However, opposite alterations were observed for Ecad and its cytoplasmic binding partners α- and β-catenin (Figure 8C–8H), all of which displayed reduced membranous staining in MZepcam mutant EVL cells. In addition, perinuclear Ecad staining in the Golgi apparatus (Figure 8I) was strongly reduced in mutant cells (Figure 8D) compared to wild-type controls (Figure 8C). Together, these data indicate that EpCAM promotes the presence of cadherin-catenin complexes in the basolateral domain of EVL cells, whereas it opposes the presence of TJ proteins. In line with these differential effects, EpCAM was co-localized with Ecad in lateral (Figure 8J) as well as basal membranes (Figure 8M) of wild-type EVL cells, whereas Tjp1 was localized apical of the EpCAM domain (Figure 8K and 8L), consistent with results obtained in cultured epithelial cells. To investigate whether compromised basal protrusion formation in MZepcam mutants might be caused by this gain of TJPs or loss of E-cadherin, we inactivated zebrafish tjp1–3, via MO injection, and re-introduced E-cadherin by injecting different mRNAs or plasmid DNAs encoding mouse E-cadherin (see Materials and Methods). However, neither of the treatments led to a significant restoration of ruffle formation (Figure S5; and data not shown), suggesting that the molecular effects of EpCAM might be more complex.

17Complementing the failed rescue experiments described above, we next analyzed whether the defects of MZepcam mutants get enhanced upon concomitant loss of the (remaining) E-cadherin. For this purpose, different amounts of ecad MO were injected into MZepcam mutant embryos. In contrast to epcam, ecad is expressed both in the EVL and in the deep cells. Strong ecad morphants and ecad null mutants display defects during epiboly movements of deep cells (see Introduction), whereas the EVL appeared normal (Figure 9C and 9G). In striking contrast, and unlike uninjected MZepcam controls (Figure 9B and 9F), MZepcam mutants injected with high amounts of ecad MO displayed severely compromised cell-cell adhesion of EVL cells at early gastrula stages (5.5 hpf; Figure 9H), leading to embryo lysis at mid gastrulation (Figure 9D). This effect was strictly layer-autonomous and independent of E-cadherin in the underlying deep layer, as indicated by the disrupted morphology of the EVL in genetic mosaics with a mutant EVL and wild-type deep layers (Figure 9J), but a wild-type morphology in the opposite combination (Figure 9I). This indicates that, while dispensable per se, EpCAM and Ecad together are required for intercellular adhesion within the EVL.

18As described above, EpCAM and Ecad are not only co-localized at the lateral, but also at the basal membranes of EVL cells, facing the underlying deep cells (Figure 8M). Previous studies have shown that Ecad is primarily required to mediate the anchorage of deep cells to the inner surface of the EVL during radial intercalation processes driving the epiboly of deep cells (see Introduction). To test whether EpCAM from EVL cells might also be involved in this anchorage, we carried out genetic interaction studies, injecting ineffectively low amounts of ecad MO into wild-type or MZepcam mutant embryos. In mid gastrula wild-type embryos (80% epiboly stage) injected with such low amounts of ecad MO, epiboly of deep cells was normal (Figure 10A; n = 25/25; compare with Figure 5C). In contrast, and unlike the very moderate deep cell epiboly defects of un-injected MZepcam mutants (Figure 5D), mutants injected with the same low amounts of ecad MO displayed an arrest of deep cell epiboly at the equator of the embryo (Figure 10B; n = 27/29), similar to the defects of embryos injected with highest amounts of ecad MO (data not shown; n = 20/21). Time-lapse recording at early gastrulation stages further revealed crucial differences in the behavior of deep cells. In un-injected embryos or in wild-type embryos injected with low amounts of ecad MO, radially intercalating deep cells remained in the exterior layer of the deep layers, most likely stably attached to the inner surface of the EVL (Figure 10E and data not shown; n = 6/6; 3 videos), whereas in MZepcam mutants injected with low amounts of ecad MO, cells usually moved back into more interior layers of the deep layers (Figure 10F; n = 4/5; 3 videos), similar to the situation after complete knock-down of ecad (Figure 10G; n = 5/5; 2 videos). Interestingly, this effect on deep cell behavior again seemed to primarily depend on EpCAM and Ecad function within the overlying EVL. Thus, in genetic mosaics, deep MZepcam mutant cells injected with low amounts of ecad MO displayed the same epiboly behavior like their wild-type neighbors, which was compromised when the host and the EVL were mutant (Figure 10C; n = 5/5), but normal when the host and the EVL were wild-type (Figure 10D; n = 6/6). Together, this suggests that EpCAM from the EVL supports the function of Ecad to drive epiboly of deep cells.

19EpCAM is a well-established carcinoma marker that in addition to its diagnostic value for rapidly growing tumors of epithelial origin is used as a potential target for immunotherapy,,,. Most of the current studies dedicated to the function of EpCAM have used in vitro systems, revealing a rather pleiotropic nature of EpCAM's roles during tumor progression and normal epithelial development (reviewed in,). Initially proposed to be a cell-cell adhesion molecule,,, more recent studies have pointed to additional and, to some extent, seemingly contrary roles of EpCAM in diverse processes such as signaling, cell migration, proliferation and differentiation. Some of these data have led to the notion of a “double-face” of EpCAM function in promoting both adhesion/tissue integrity and cell motility/morphogenesis/metastasis,. Here, using zebrafish genetics, we present in vivo loss-of-function data largely in line with this notion, and discuss how the double-face effect might be achieved on the cellular level.

20Apart from zebrafish, no EpCAM mutants have been described in any other organism. Two mutant zebrafish alleles were isolated, both of which are most likely EpCAM nulls (Figure 1 and Figure 2). Maternal-zygotic (MZ) mutants lacking both maternally and zygotically supplied epcam gene products display early epiboly and later skin defects, whereas epiboly seems to be normal in zygotic mutants and epcam morphants. The weaker phenotype of zygotic mutants suggests that maternally supplied epcam gene products, either mRNA or protein, are sufficient to drive epiboly. The lack of epiboly defects in epcam morphants further points to maternal EpCAM protein. Translational start site morpholinos as used here target both maternally and zygotically provided mRNA, but not maternal protein, suggesting that the weaker phenotype of epcam morphants compared to MZepcam mutants is due to the presence of maternally provided EpCAM protein. EpCAM-specific antibodies will be required to test this notion.

21In addition to the epiboly and skin defects that are the focus of this work, epcam zebrafish mutants display compromised otolith formation in the developing inner ears (Figure 1 and ). The cellular and molecular basis of this phenotype is unclear, however, similar otolith defects have been described for zebrafish mutants in the tight junction component Claudinj. Another site of prominent zygotic zebrafish epcam expression are the neuromasts of the lateral line (Figure 2). According to a previous report, loss of EpCAM function by morpholino injection leads to defects in neuromast deposition during the posterior-wards migration of the lateral line primordium. Although nicely in line with the concept of a role of EpCAM during epithelial morphogenesis, our analysis of MZepcam mutants could not confirm this phenotypic trait (Figure S6). Indeed, Villablanca et al. had to inject highest amounts of morpholinos to obtain the phenotype, suggesting that it might have been an unspecific effect.

22During gastrulation, epcam is exclusively expressed in cells of the enveloping layer (EVL; see Introduction). Nevertheless, MZepcam mutants display moderate defects during epiboly movements of both the EVL and the underlying deep cell layers (Figure 5). Our chimeric analyses in combination with E-cadherin inactivation indicate that the deep cell defects are secondary consequences of failed EpCAM function in the EVL, whereas the EVL defects themselves are layer autonomous (see below). During their spreading over the yolk, EVL cells normally change their shape and flatten out along the radial axis to increase their horizontal size. The cellular mechanisms underlying this transition are largely unknown. Here, we show that they involve basal protrusive activity of EVL cells, and that this activity is severely compromised in MZepcam mutants, leading to an overall reduction in the average horizontal size of mutant EVL cells (Figure 5). Recently, a similar impairment of EVL epiboly has been described for MZpou5f1 mutants. However, in this case, epiboly defects are accompanied by increased, rather than reduced protrusive activity of EVL cells. This indicates that both gain and loss of protrusive activity can compromise epiboly. Future experiments have to address the functional connection between pou5fl and epcam. In MZepcam mutants, the lack of basal activity is accompanied by a basal extension of apical junctional complexes (AJCs) (Figure 7), and by increased levels of TJ components in apico-basal membranes, whereas levels of E-cadherin (Ecad) and catenins are reduced (Figure 8). These effects are consistent with the co-localization of EpCAM and Ecad in the basolateral membranes of wild-type EVL cells, and their exclusion from the apical Tjp1 domain (Figure 8). Together, this suggests that EpCAM pushes the molecular composition of apico-basal membranes from TJ towards AJ components. In addition to the membrane, mutants displayed reduced perinuclear Ecad staining in the Golgi apparatus, suggesting that EpCAM also affects de novo synthesized Ecad protein. Currently, we cannot distinguish which of the observed phenotypic traits are primary, and which secondary. However, neither knockdown of tjp1–3 nor re-introduction of Ecad led to an alleviation of the mutant phenotype or a restoration of basal protrusions, suggesting that EpCAM might have multiple targets and interaction partners, rather than acting via a single mediator. Generally, protrusion formation is driven by rearrangements of the cortical cytoskeleton that are coordinated with local modulations in cellular adhesiveness. Interestingly, according to our TEM studies, the protrusions of EVL cells primarily attach to the basal membrane of adjacent other EVL cells, rather than to underlying deep cells or extracellular matrix components, while protrusions remain much shorter and broader in MZepcam mutants (Figure 7). In this light, it is tempting to speculate that the thinning of EVL cells underlying EVL epiboly might be driven by the “crawling” of basal protrusions on the basal surface of adjacent EVL cells, and that EpCAM might be particularly involved in modulating cell-cell adhesiveness and cortical tension in basolateral domains of EVL cells (Figure 11A and 11B).

23In contrast to cell culture studies identifying EpCAM as a functional antagonist of Ecad,, we found that zebrafish EpCAM and Ecad tightly interact and enhance each other's effects to promote EVL integrity as well as deep cell epiboly. While each of them per se was dispensable for EVL integrity, combined loss of both EpCAM and Ecad led to severe and layer-autonomous EVL disassembly during early gastrulation stages (Figure 9). This indicates that EpCAM and Ecad play essential, yet redundant roles to mediate proper cell-cell adhesion among EVL cells (Figure 11D). Similarly, partial inactivation of Ecad, which had no effect in a wild-type background, led to a complete arrest of deep cell epiboly in MZepcam mutants (Figure 10A and 10B). Interestingly, also here, the effect depended purely on EpCAM and Ecad function in the EVL (Figure 10C and 10D). On the cellular level, the epiboly arrest was accompanied by a failure of intercalating deep cells to remain in the external deep layer directly underneath the EVL (Figure 10E–10G), pointing to defects in the adhesion between EVL and underlying deep cells, thus, across different layers (Figure 11C). In this light, both the tissue integrity defects caused by loss of EpCAM function in the background of complete ecad inactivation, and the tissue morphogenesis defects caused by loss of EpCAM function in the background of partial ecad inactivation, seem to result from reduced intercellular adhesion. But what are the reasons for the exclusive effects on epiboly, but not EVL-EVL adhesion, in the ecad hypomorphic background? Differential contributions of Ecad and EpCAM to EVL-EVL versus EVL-deep cell binding could be one factor. EVL-EVL cell adhesion involves apical junctional complexes and desmosomes, and might therefore be less dependent on “free” basolaterally localized EpCAM and Ecad than EVL-deep cell adhesion, which lacks such junctions. Alternatively or in addition, the deep cell epiboly arrest might be due to combined effects on cell adhesiveness and its dynamic regulation. During gastrulation, deep cells undergo massive spatial rearrangements in addition to radial intercalations, one of the driving forces of epiboly. For instance, they simultaneously move from ventrolateral into dorsal regions of the embryo to form the embryonic body axis. Therefore, neighborships between EVL and deep cells need to change rather rapidly. In this light, it is feasible to speculate that EpCAM might also be involved in regulating the dynamic dis- and re-assembly of EVL-deep cell contacts to allow proper morphogenesis. Cadherin contacts can be regulated at multiple levels, and membrane localization of Ecad has recently been shown to be dynamically regulated via endocytosis and recycling during gastrulation movements of deep cells in zebrafish and frog embryos,,. Furthermore, the cytoplasmic domain of EpCAM contains an NPXY internalization motif, which could possibly trigger the concomitant endocytosis of Ecad, comparable to the recently revealed role of the transmembrane protein FLRT3 in Xenopus embryos.

24In sum, we propose that EpCAM and Ecad play rather similar, and partially redundant roles during zebrafish gastrulation. Furthermore, they seem to mutually influence each other, with a positive effect of EpCAM on membranous Ecad levels and, possibly, Ecad synthesis and recycling. Subject to dynamic regulation, cadherins are well known for their multiple, and seemingly contrary effects, not only promoting cell-cell adhesiveness and tissue integrity, but also cellular plasticity and cellular “grip” during morphogenesis. Similar mechanisms might underlie EpCAM's reported “double face” function also revealed in this work. Future biochemical studies will be necessary to elucidate the molecular basis of the EpCAM – E-cadherin partnership. Since they play largely redundant roles, they do not necessarily have to physically interact at all. They could for instance mediate different modes of adhesion, cadherins in connection with the highly structured actin cytoskeleton and EpCAM more independently of the cytoskeleton,. Consistent with this notion, our immunolocalization studies indicate a salt-and-pepper like distribution of EpCAM and Ecad, rather than complete co-localization in the basolateral membrane of EVL cells (Figure 8J). However, it is interesting to note that in the context of nuclear signaling, complex formation between the intracellular domain of EpCAM and β-catenin has been observed. Furthermore, more indirect mechanisms might be at play, involving adapter or signaling proteins.

25Similar effects of EpCAM might also account for the later defects in the mutant skin. From mid-segmentation onwards (14 hpf), the outer EVL, now also called periderm, is juxtaposed against a single layer of basal keratinocytes, which derive from the ventral ectoderm, a subpopulation of deep cells initially located in ventral-animal regions of the pregastrula embryo. Both layers display epcam expression, and both display compromised epithelial integrity in epcam mutants. However, the defects in the periderm are already apparent when the basal layer is still normal (16 hpf; Figure 3C–3E). This later periderm defects are very similar to those caused by combined loss of EpCAM and E-cadherin during gastrulation. It remains unclear, however, why in contrast to the early stages and despite its maintained expression, E-cadherin fails to compensate for loss of EpCAM after skin formation. Furthermore, consistent with the aforementioned effect of EpCAM from the EVL on epiboly movements of the underlying deep cells, our chimera analyses reveal that defects in the basal layer are due to a non-autonomous effect from the EVL (Figure 3H and 3I). Also, similar to the situation during epiboly, epcam mutants display a persistent basal extension of AJCs in EVL cells during such later stages (Figure 7J and 7L). Although the role of tight junctions in cell adhesion is still disputable, it is generally accepted that their impact compared to adherence junctions and desmosomes is minor,. In this light, and in light of the observed negative effect of the MZepcam mutation on membranous Ecad levels and cell-cell adhesiveness during gastrulation, we assume that the epithelial defects of the mutant periderm are due to reduced, rather than enhanced cell-cell adhesiveness. In addition, defects might be enhanced by reduced epithelial plasticity. Consistent with this notion, initial epithelial lesions in the periderm of epcam mutants were most prominent on the tailbud (Figure 3B) and the head, regions that undergo massive morphogenesis and that are exposed to highest mechanical stress. Very similar shedding of skin cells was previously described for hai1 mutants, which display partial epithelial-mesenchymal transitions of basal keratinocytes. However, no mesenchymal-like behavior was observed in time-lapse recordings of basal keratinocytes or EVL cells of epcam mutants (K.S. and M.H., unpublished observations). In sum, these data suggest that similar to the defects during epiboly, the later skin defects of epcam mutants are due to compromised intercellular adhesion and cellular plasticity of epithelial cells.

26In light of the reported roles of EpCAM in multiple other cellular processes, such as cell proliferation and cell differentiation, we also investigated whether and when skin cells of MZepcam mutants develop corresponding defects. However, in contrast to other reported zebrafish skin mutants, EVL and basal cells showed normal levels of terminal differentiation markers (Keratin, ATPases; Figure S7). Also, although EVL and basal cells displayed an up to 50% increase in proliferation, this hyper-proliferation only became apparent several hours after the epithelial defects (Figure 4B). Furthermore, aggregates of both EVL and basal cells displayed similar rates of BrdU incorporation like regions remote of the aggregates (Figure 4C). An interesting side outcome of the BrdU incorporation studies, also confirmed by time-lapse recordings (K.S. and M.H., unpublished data), was the demonstration that EVL cells divide at all. This had not been shown before. Rather, it was widely believed that the EVL grows by cell shape changes, and that it later sloughs off, being replaced by cells deriving from basal cells. However, our data suggest that the growth of the two larval zebrafish skin layers might involve layer-autonomous, horizontal cell divisions, rather than or in addition to vertical growth/replacement, the typical concept of stratified epithelia.

27In conclusion, our studies indicate that loss of EpCAM in the developing zebrafish skin primarily leads to compromised epithelial plasticity and adhesiveness, whereas hyper-proliferation is a secondary consequence, possibly due to the loss of contact inhibition. Similarly, the higher susceptibility to bacterial infections and enhanced inflammation of epcam mutants (Figure S2, Video S1, Video S2) are most likely secondary consequences of compromised skin integrity.

28In contrast to embryonic and larval stages, however, we found EpCAM to be largely dispensable after metamorphosis, when the skin has become multi-layered. The reason for this later dispensability remains unclear. Functional redundancy with other genes could be one explanation. In mammals, an EpCAM homologue called Trop2/Tasctd2 exists, which shows approximately 50% sequence identity to EpCAM, and which has most likely evolved via a retrotranspositional event. In contrast, searches of zebrafish databases failed to identify further EpCAM-related sequences (M.H. and K.S., unpublished data), suggesting that zebrafish epcam is a single gene. Interestingly, adult zebrafish epcam mutants even displayed normal cutaneous wound healing (K. S. and M.H., unpublished data), indicating that despite its reported elevated expression during epithelial regeneration of the mammalian liver and kidney,, EpCAM is not required for epithelial morphogenesis that takes place during zebrafish skin repair. Future experiments have to reveal whether epcam mutants are less susceptible to epithelial tumor formation, which would reinforce its suitability as a target for anti-carcinoma therapies.

29The EpCAM alleles hi2836 and hi2151 were isolated during an insertional mutagenesis screen. Unless noted otherwise, the hi2151 allele was used. hi2151 mutants were obtained from heterozygous (Z+/−) or homozygous parents (MZ−/−). The Tg(krt4: egfp)gz7 and Tg(βactin:hras-egfp) (allele vu119) transgenic lines have been described previously,.

30The full-length coding region of zebrafish EpCAM cDNA (GenBank accession number NM_213175) was amplified via RT-PCR and cloned into pCRII-TOPO (Invitrogen) or upstream of eGFP into pCS2+ plasmid. To generate pCS2-Ecad-HA, the full-length coding region of mouse E-cadherin was amplified and cloned upstream of HA into pCS2+. For antisense probe synthesis, pCRII-EpCAM was linearized with XbaI and transcribed with SP6 RNA polymerase. For sense RNA synthesis for microinjection, pCS2-EpCAM-eGFP, pCS2-Ecad-HA and pCS2-mCherry (with farnesylation signal; generous gift from Erez Raz) were linearized with NotI and transcribed using SP6 Message Machine kit (Ambion); mGFP mRNA and GM130-GFP mRNA were generated as described. For rescue experiments, EpCAM-eGFP or Ecad-HA mRNA were injected at 150, 350 (EpCAM) or 100 µg/ml (Ecad), respectively (1.5 nl per embryo), pCS2-EpCAM-eGFP, pCDNA3.1-Ecad-GFP, pCS2-Ecad-HA, or pL31NU-Ecad-Venus DNA at 350 µg/ml (EpCAM) or 100 µg/ml (Ecad), respectively.

31In situ hybridizations were performed as previously described, using probes for EpCAM and leukocyte-specific-plastin (lcp1). Whole-mount immunostaining was carried out as described, using fluorescently-labeled secondary antibodies or the Vectastain ABC kit (Axxora) for enzymatic detection. Antibodies, dilutions used and sources were as follows: anti-p63 (1∶200, Santa Cruz), anti-GFP (1∶400, Invitrogen), anti-RFP (1∶200; antibodies-online GmbH, ABIN132020), anti-E-cadherin (1∶200, BD Biosciences), anti-ZO-1 (1∶200, Zymed), Alexa-Fluor-546 goat anti-mouse (1∶400, Invitrogen) and Alexa-Fluor-488 goat anti-rabbit (1∶400, Invitrogen). Tested anti EpCAM antibodies: A-20 (Santa Cruz), 1144 (Epitomics), HO-3 and C-215. For paraffin sectioning (8 or 16 µm), stained embryos were dehydrated in ethanol series and clearing in toluene, the specimens were infiltrated with paraffin, embedded, and sectioned. Epidermal cell proliferation was assessed by BrdU incorporation followed by combined anti-p63 and anti-BrdU immunostaining as described. Apoptotic cells were visualized by TUNEL staining using in-situ cell death detection kit (Roche). Phalloidin stainings of cortical actin cytokeleton were carried out with Alexa 488 or Alexa 594 Phalloidin (1∶200, Molecular Probes), as described.

32epcam MO (5′- GTGCAGAGACTTTCCGGCCATATTT-3′) was obtained from Gene Tools (Philomath, OR) and diluted in Danieau's buffer. 1.5 nl of a 200 µM MO solution were injected per embryo at the 1 cell stage. The ecad MO was as described. For complete knock-down, 1.5 nl of a 200 µM solution was injected, for synergistic enhancement studies, 1.5 nl of a 30 µM solution. The tjp1, tjp2 and tjp3 MOs (generous gifts from Matthias Köppen and Carl-Philipp Heisenberg; ) were injected alone or together at 50 to 200 µM each); the msn1 MO was as described and injected at 200 µM.

33Clusters of mGFP-labeled basal keratinocytes were obtained by homotopic transplantation of approximately 50 non-neural ectodermal cells from Tg(βactin:hras-egfp) transgenic donor embryos into non-transgenic hosts at 6 hpf. Recipients were fixed at 36 hpf, subjected to anti-GFP and anti-p63 immuno-fluorescence staining, mounted in 1.5% low melting agarose and analyzed by confocal microscopy. For localization studies of EpCAM-eGFP on the basal side of EVL cells (Figure 8), deep cells of mCherry mRNA-injected cells were transplanted at the sphere stage into EpCAM-eGFP mRNA-injected hosts. Embryos were fixed at the 90% epiboly stage for anti-RFP and anti-GFP immunohistochemistry, sectioned, and analyzed via confocal microscopy. To distinguish whether for proper EVL adhesion, EpCAM and Ecad are required in the EVL cells themselves or in underlying deep cells (Figure 9), chimeric embryos were generated by transplanting deep cells at the sphere stage from mCherry-labelled wild-type or MZepcam donors into unlabelled MZepcam hosts injected with high amounts of ecad MO, or vice versa, followed by fixation at the 90% epiboly stage for anti-Cherry and phalloidin staining. To investigate whether for deep cell epiboly, Epcam and Ecad are required in the deep cells or in the EVL (Figure 10), cells from mCherry mRNA-injected wild-type or MZepcam mutant donors were transplanted next to cells from mGFP mRNA + low ecad MO-injected MZepcam mutant donors into unlabelled MZepcam hosts or MZepcam hosts injected with low ecad MO. Embryos were fixed at the 90% epiboly stage, and processed via anti-RFP and anti-GFP immunohistochemistry.

34For time-lapse in vivo imaging, embryos were mounted and recorded at a Zeiss Axiophot with Nomarski optics and a Hamamadzu Orca camera, as previously described. 20-minutes videos were taken at lateral marginal regions, starting at shield stage (5.5 hpf), with 30 sec intervals. For Figure 10, single images form the time-lapse video recordings were imported into Adobe Photoshop, and single cells were pseudo-colored to aid the presentation.

35Fluorescent images were taken with a Zeiss Confocal microscope (LSM510 META); Transmission light microscopy was performed on a Zeiss Axiophot or Leica MZ-8 stereomicroscope. For transmission electron microscopy, wild-type and mutant fish were fixed with 2.5% glutaraldehyde in PBS for 30 minutes each at ambient temperature and then on ice. After washing with PBS, the larvae were post-fixed with 1% osmium tetroxide in 100 mM phosphate buffer pH 7.2 for 1 hour on ice, washed with H20, stained with 1% aqueous uranylacetate for 1 hour, dehydrated in a graded series of ethanol and finally embedded in Epon. Ultrathin sections were stained with uranyl acetate and lead citrate and viewed in Philips CM10 electron microscope.