Gastrulation consists of a series of coordinated cell movements that organize the germ layers and establish the major body axes of the embryo. The first coordinated cell movement during zebrafish gastrulation is epiboly, which involves the thinning and spreading of a multilayered cell sheet (Warga and Kimmel, 1990). Before the start of epiboly, the embryo is organized into three layers, all of which undergo epiboly. The deep cells of the blastoderm, which eventually give rise to all embryonic tissues, are covered by a monolayered extra-embryonic epithelium, the enveloping layer (EVL). The EVL and deep cells sit on top of a large yolk cell. At the interface between the yolk and the blastoderm lies the yolk syncytial layer (YSL), an extra-embryonic syncytium to which the EVL is attached.
There is considerable evidence that the yolk cell plays a critical role in teleost epiboly. The yolk contains arrays of microtubules aligned along the animal–vegetal axis, which shorten over the course of epiboly (Strähle and Jesuthasan, 1993; Solnica-Krezel and Driever, 1994). Disruption of these microtubule arrays impairs epiboly in zebrafish and in a related fish species Fundulus (Betchaku and Trinkaus, 1978; Strähle and Jesuthasan, 1993; Solnica-Krezel and Driever, 1994). In Fundulus, removal of the blastoderm from the yolk results in accelerated epiboly of the YSL, suggesting that the YSL normally pulls the blastoderm along with it (Trinkaus, 1951; Betchaku and Trinkaus, 1978), presumably by means of its attachment to marginal EVL cells.
Recent work has highlighted the importance of an intact and properly differentiated EVL for epiboly of the EVL, YSL, and deep cells. EVL cell fate must be properly specified for epiboly to proceed normally (Pei et al., 2007; Sabel et al., 2009). For example, simultaneous morpholino knock-down of several EVL-specific intermediate filament genes results in epiboly delay of the EVL, YSL, and deep cells (Pei et al., 2007). In addition, epithelial cell adhesion molecule (EpCAM), which is expressed exclusively in the EVL during gastrulation, is required for normal epiboly of the EVL and deep cells (Slanchev et al., 2009). Taken together, this work suggests that EVL integrity is critical for epiboly.
Tight junctions (TJs) are one type of junctional complex that links cells in epithelia. Consisting of both transmembrane and cytoplasmic proteins, TJs serve both a barrier function to control the flow of ions between cells (the paracellular pathway) and a fence function to separate apical and basolateral domains of the plasma membrane within cells. The precise function of TJs depends on both cell type and protein composition (Brandner, 2009). In Fundulus, the EVL is attached to the YSL by means of TJs, and breaking this attachment results in retraction of the EVL toward the animal pole (Trinkaus, 1951). The situation is similar in zebrafish, where tight junctions connect marginal EVL cells to the YSL (Köppen et al., 2006). By contrast, EVL cells are connected to each other by TJs, adherens junctions, and desmosomes (Montero et al., 2005; Slanchev et al., 2009). We hypothesized that the regulation of TJs might be a requirement for normal epiboly.
Few studies have addressed the role of TJs in zebrafish morphogenesis. However, in a recent study, the function of the TJ scaffold protein tight junction protein-3/zonula occludens-3 (Tjp-3/Zo-3) was investigated using antisense morpholino oligonucleotides (MOs; Kiener et al., 2008). Despite the fact that Tjp-3/Zo-3 is expressed throughout cleavage, blastula and gastrula stages, no early defects were observed in morpholino injected embryos. At later stages, morphant embryos displayed tail fin defects, pericardial edema, and a lack of blood circulation (Kiener et al., 2008). These phenotypes were ascribed to defects in the EVL permeability barrier, resulting in defective osmoregulation.
We have focused on the claudin gene family. Claudins are the major transmembrane protein component of TJs, and the selectivity of the paracellular pathway is regulated by Claudin interactions on adjacent cells. Epithelia typically express more than one Claudin, raising the potential for both homo- and heterotypic interactions (Findley and Koval, 2009). The C-terminal region of Claudins contain a PDZ binding domain, which allows interactions with PDZ containing proteins, most notably the zonula occludens proteins (ZO-1,2,3), which link TJs to the actin cytoskeleton (Angelow et al., 2008). In addition, localization of Claudins to TJs is mediated by ZO-1 and ZO-2 (Findley and Koval, 2009).
Mammals have 24 claudin genes, several of which have been implicated in human diseases (Brandner, 2009). Of interest, the claudin gene family has expanded within the teleost lineage. Fifty-six claudin genes have been identified in the Fugu genome and phylogenetic analysis indicated that 17 of these genes have no mammalian orthologs and, hence, are unique to the fish lineage (Loh et al., 2004). The fact that some of these 17 genes are also found in the zebrafish genome suggests that expansion of the teleost claudin gene family was an early event in teleost evolution (Loh et al., 2004). To date, 23 zebrafish claudin genes have been annotated (zfin.org) but this number is likely to increase with further analysis of the zebrafish genome. It is currently unclear what contributions the expansion of the gene family may have played in teleost evolution, although it has been suggested to relate to the need for extensive osmoregulation in aquatic organisms (Loh et al., 2004).
No studies have addressed the function of Claudins during zebrafish gastrulation. In Xenopus, one study examined the function of Xclaudin1 (Xcla1), which is maternally and zygotically expressed throughout early development (Brizuela et al., 2001). Overexpression of Xcla1 caused increased cell adhesion, whereas overexpression of a C-terminal truncated version, lacking the PDZ binding domain, resulted in increased cell dispersion, suggesting that cell adhesion was reduced. In zebrafish, Cldn15 was shown to be important for lumen formation in the gut (Bagnat et al., 2007) and a mutation in the claudin j (cldnj) gene produces larvae lacking otoliths in the ears (Hardison et al., 2005).
We have examined the expression of cldnd, cldne, and cldn7 during zebrafish development. As, thus far, our functional analyses have only been informative for cldne, that is our focus here. cldne has no direct mammalian ortholog, although it is most closely related to mammalian Cldn3 and 4 (Kollmar et al., 2001). We found that morpholino knock-down of cldne resulted in defects in both the initiation and progression of epiboly. We propose that the morphant phenotype is the result of weakened attachment between the EVL and YSL.
RESULTS AND DISCUSSION
Expression of cldne in the Early Embryo
To examine the spatial and temporal expression of cldne, we performed whole-mount in situ hybridization on selected stages from cleavage through 1 day postfertilization (dpf). Our results are in agreement with gene expression data available through the Zebrafish Information Network (Thisse et al., 2001; Rauch et al., 2003). Here we extend this previous work, describing new details of the expression pattern of cldne.
During cleavage stages, cldne was maternally expressed in all blastomeres (Fig. 1A), while cldne expression was confined to the EVL during blastula and gastrula stages. At sphere stage (4 hours postfertilization [hpf]), which occurs shortly after the midblastula transition, the EVL becomes lineage restricted (Kimmel et al., 1990). Patchy cldne expression began to appear in a subset of EVL cells at sphere stage and a similar pattern of expression was observed at dome stage (4.3 hpf, Fig. 1B, arrowhead). Expression throughout the EVL was evident by shield stage (Fig. 1C,D), and this pattern was maintained for the duration of gastrulation (Fig. 1E).
A region of more intense hybridization in the EVL became apparent at the margin at approximately 60% epiboly. We postulated that this more intense expression might represent EVL cells at the dorsal margin overlying the dorsal forerunner cells (DFCs). DFCs are derived from EVL cells which ingress to form a tight cluster that later gives rise to Kupffer's vesicle (Oteíza et al., 2008), the teleost organ of left–right asymmetry, analogous to the mammalian node (Essner et al., 2005; Kramer-Zucker et al., 2005). This hypothesis was confirmed by performing double in situ hybridizations for cldne and the DFC marker casanova (cas) on embryos at 80% epiboly. This analysis demonstrated that the more intense cldne expression domain corresponded to dorsal marginal EVL cells (Fig. 1E, inset), and might reflect the fact that there are tight junctions between a subset of DFCs and overlying EVL cells (Oteíza et al., 2008). At 1 dpf, cldne expression was detected in the skin, olfactory placodes, ears, and proctodeum (Fig. 1F), similar to previous reports.
Morpholino Knock-down of cldne Causes Epiboly Defects
To investigate cldne function during gastrulation, two independent and nonoverlapping translation blocking morpholinos (cldne-MO1 and cldne-MO2) were used to target both maternal and zygotic transcripts (see Experimental Procedures for details). Injection of either MO at the one-cell stage caused epiboly delay of the EVL, deep cells and YSL. Because cldne-MO1 was effective at a lower dose than cldne-MO2, it was used in all subsequent experiments (referred to as cldne-MO).
Injection of different amounts of cldne-MO produced dose-dependent epiboly delays. At high doses (4 ng/embryo), most embryos failed to dome and lysed at the animal pole during mid-gastrulation (data not shown). At moderate doses (0.5 ng/embryo) all cldne-MO injected embryos exhibited delayed epiboly (113/113, 100%; Fig. 2). When control embryos had reached 30% epiboly (4.7 hpf), morphant embryos were still at sphere stage (normally at 4 hpf; Fig. 2A,B). By the time that control embryos reached shield stage (6 hpf, Fig. 2C), most morphant embryos had only reached 40% epiboly (5 hpf; (Fig. 2D). The narrowing of the external YSL, which occurs during dome stage and leads to crowding of the YSL nuclei (Solnica-Krezel and Driever, 1994) occurred normally in morphant embryos (Fig. 2E, arrow), suggesting that the microtubule arrays in the yolk functioned normally. The majority of embryos injected with a control cldne mismatch morpholino (cldne-mmMO) were normal (10/60, 17% exhibited epiboly delay, Fig. 2F). To confirm the specificity of the morpholino, RNA rescue experiments were performed using cldne RNA which lacked the morpholino target site. While injection of cldne mRNA had little effect (5/70, 7% had delayed epiboly, Fig. 2G), rescue of the morphant phenotype was observed when cldne RNA and cldne-MO were co-injected (39/59, 66% rescued, Fig. 2H). The cldne morphant embryos continued to progress through epiboly more slowly than the control embryos, such that morphant embryos were at 70% epiboly stage (7.5 hpf) when control embryos had progressed to 90% epiboly (9 hpf; Fig. 2I,J). The majority of morphant embryos failed to complete epiboly and these embryos eventually degenerated and died. Although morphant embryos exhibited an overall developmental delay, involution and dorsal convergence occurred, albeit more slowly (Figs. 2, 3).
Deep Cell Fate Is Unaffected in Morphant Embryos
To assess whether deep cell fates were altered in cldne morphant embryos, we performed whole mount in situ hybridization for no tail (ntl), a mesodermal marker, and cas, an endodermal and DFC marker (Schulte-Merker et al., 1994b; Kikuchi et al., 2001). At shield stage, expression of ntl was normal in morphant embryos (Fig. 3A,B), indicating that mesoderm was properly specified. At 70% epiboly stage, cas expression in both DFCs (Fig. 3C,D, arrowhead) and endoderm (Fig. 3E,F, arrows) was also overtly normal in cldne-MO injected embryos. Thus, mesoderm and endoderm development was not affected in cldne morphant embryos.
EVL Cell Fate Is Normal in Morphant Embryos
We postulated that TJs in the EVL might be disrupted in cldne morphant embryos, leading to a loss of EVL integrity. Previous work demonstrated that EVL differentiation requires cell–cell contact (Sagerström et al., 2005). Thus, if cell–cell contacts were reduced in the EVL of morphant embryos, expression of EVL specific genes might be lost. Therefore, we examined expression of the EVL marker keratin4 (ker4) by in situ hybridization (Fig. 4). Embryos injected with high doses of MO had dramatically reduced expression of the EVL marker ker4 (data not shown), while embryos injected with moderate doses of MO expressed ker4 relatively normally (Fig. 4A–C). However, closer examination of ker4 expression revealed that the EVL margin was jagged and irregular in MO injected embryos (Fig. 4E) when compared with uninjected embryos (Fig. 4D). Thus, although the EVL marker ker4 was expressed in morphant embryos, the organization of the EVL at the margin appeared to be disrupted.
We also examined the expression of the tight junction associated scaffold protein ZO-1 and found that it was expressed in morphant embryos (Fig. 5). Although this suggested that TJs were present in morphant embryos, ZO-1 does not require the presence of Claudins to localize to the membrane (Fanning and Anderson, 2009). It is unlikely that all TJ connections between the EVL margin and the YSL were lost, as in that case the EVL would be expected to retract toward the animal pole. Furthermore, other claudin genes including cldn7, cldnb, and cldnf are also expressed in the EVL during epiboly (zfin.org and our unpublished data). The ZO-1 staining clearly showed that marginal EVL cells had a scalloped appearance in morphant embryos compared with control embryos (compare Fig. 5B and D), suggesting that the EVL was under less tension than in control embryos.
The EVL Margin Is Irregular in cldne Morphant Embryos
To examine the EVL further, phalloidin staining was used to visualize the cortical actin organization of EVL cells. Cortical actin appeared relatively normal in MO injected embryos; however, abnormalities were apparent at the margin (Fig. 6). The first tier of EVL cells (adjacent to the yolk) were often rounder (Fig. 6B,E,F, arrowheads) and had fewer points of contact with their neighbors, resulting in an irregular and jagged appearance (Fig. 6B,C,E,F, arrows), as compared to the margin of control embryos (Fig. 6A,D). The marginal actin band visible in YSL during late epiboly (Cheng et al., 2004), formed normally in morphant embryos (data not shown). These findings suggested that the EVL margin in morphant embryos was under less tension, perhaps as a consequence of reduced adhesion between marginal EVL cells and the YSL. The disruptions in marginal EVL cells were apparent as early as the 30% epiboly stage, indicating that this was likely to be a primary rather than a secondary effect of the morpholino.
SUMMARY AND CONCLUSIONS
We have found that morpholino knock-down of the EVL specific tight junction component cldne results in a highly penetrant epiboly defect. Epiboly initiation and progression of the EVL, deep cells, and YSL were all delayed. The epiboly delay resulted in a general developmental delay with most morphant embryos dying by the end of gastrulation. Deep cell delay may be secondary to the EVL defect, as deep cells are unable to progress past the EVL margin (Köppen et al., 2006). Several cldn genes are expressed in EVL, in addition to adherens junction components. These factors may function redundantly to maintain EVL integrity, thereby explaining why the EVL remained intact in cldne morphant embryos. In the future, simultaneous knock-down of several cldn genes may reveal redundant or cooperative roles for these genes in the EVL. Although additional studies are needed, the primary defect in morphant embryos appears to be at the EVL margin.
Based on work in Fundulus, the prevailing model in zebrafish is that during epiboly the EVL is passively towed vegetally by means of its tight junction attachments to the YSL (Betchaku and Trinkaus, 1978). Severing the EVL–YSL connection in Fundulus, results in rapid retraction of the EVL, indicating that it is under considerable tension during epiboly (Betchaku and Trinkaus, 1978). In cldne morphant embryos, marginal EVL cells were rounder with reduced cell–cell contacts and overall the EVL margin was irregular. These observations suggest that the EVL margin was under less tension in morphant embryos, which likely reflects reduced adhesion to the YSL. We propose that regional variation in the strength of EVL–YSL attachment in morphant embryos could result in slower and uneven advancement of the EVL. Reduced movement of the EVL, would in turn slow epiboly of the deep cells and produce the overall developmental delay that we observed.
Preliminary transmission electron microscopy studies revealed that apical adhesion complexes were present between EVL cells in morphant embryos (data not shown). However, additional work is necessary to examine the EVL–YSL junctions and to determine if the permeability of the EVL is altered in morphant embryos. In sum, our work is the first to show that a tight junction component, Cldne, is important for zebrafish epiboly.
AB embryos were obtained by natural spawning and were staged as described (Kimmel et al., 1995). Animals were treated in accordance with the policies and procedures of the University of Toronto animal care committee.
Microinjections into the yolk of one-cell stage embryos were performed as previously described (Bruce et al., 2003).
Expression Construct and Antisense Morpholino Oligonucleotides
Using the forward primer 5′-CGTTCAACTTCACAAGC-3′ and the reverse primer 5′-GCATCCGTGAGCAGAGG-3′, the full-length open reading frame (630 base pairs) of cldne was PCR amplified from 24 hpf cDNA. The PCR product was ligated into EcoRI/XbaI digested pCS2+ (Rupp et al., 1994) and sequenced. Capped sense mRNA was generated from NotI digested plasmid using the SP6 mMessage mMachine kit (Ambion).
Two antisense morpholino oligonucleotides targeted against cldne were obtained from Gene Tools LLC (Philomath, OR), cldne-MO1: 5′-CCATGTTTGCTTGTTTGTTTGTGGG-3′ (targets bases −21 to +4 of cldne) and cldne-MO2: 5′-CTCGGCACATAGACACCATGTTTGC-3′ (targets bases −7 to +19 of cldne). In addition, a mismatch morpholino version of cldne-MO1 was obtained, with the sequence: 5′-CCATCTTTCCTTCTTTCTTTCTGGG-3′. In BLAST searches against the zebrafish genome, the top hit for both cldne-MO1 and -MO2 was cldne; no other claudin genes were positive hits.
Whole-Mount In Situ Hybridization
Whole-mount in situ hybridization was performed as described (Jowett and Lettice, 1994). pBS-cldne (expressed sequence tag cb84 from zfin.org) was used to synthesize antisense digoxigenin-labeled riboprobe by linearizing with NotI and transcribing with T7. Probes to no tail (Schulte-Merker et al., 1994a), keratin4 (Thisse et al., 2001), and casanova (Kikuchi et al., 2001) were synthesized as previously described.
Phalloidin and ZO-1 Staining
Phalloidin staining using Alexa Fluor 488 phalloidin and ZO-1 (Invitrogen) antibody staining were performed as described (Köppen et al., 2006). Embryos were mounted in 3% methycellulose and photographed on a Zeiss AxioImager Z1 compound microscope using an Orca-ER camera (Hammamatsu) and Openlab software (Perkin Elmer).
Our thanks to Isaac Skromne for suggesting we examine the claudin gene family. For helpful discussions, A.B. thanks Rudi Winklbauer. For comments on the manuscript, we thank Stephanie Lepage, Tamara Smith, and Rudi Winklbauer. For suggestions and protocols, we thank Mathias Köppen, and for reagents, we thank Vicky Prince. For the preliminary TEM studies, we are indebted to Henry Hong. A.B. was supported by NSERC and CFI. H.S. and C.T. were supported by NSERC Undergraduate Summer Research Awards.