Biogenesis of structural intercellular junctions during cleavage in the mouse embryo

In our laboratory, we study the process of epithelial differentiation and polarization in the cleaving mouse embryo. The epithelium in question, the trophectoderm, forms the wall of the blastocyst at about 3.5 days post-fertilization and encloses the ICM (inner cell mass; progenitor of foetus) and blastocoelic cavity. The first developmental function of the nascent trophectoderm is to generate the blastocoel by vectorial transport (Wiley, 1987; Manejwala et al., 1989; Wiley et al., 1990) and to regulate metabolic exchange with the ICM. After further expansion of the blastocoel, the trophectoderm contributes to embryo hatching from the zona pellucida by enzymic secretion (Perona and Wassarman, 1986; Sawada et al., 1992). It then engages, at 4.5-5.5 days post-fertilization, in specific adhesive and invasive cellular interactions with maternal tissue to accomplish implantation (Holmes and Lindenberg, 1988; Lindenberg et al., 1990). After implantation, trophectoderm (trophoblast) cell lineages contribute exclusively to extra-embryonic tissues, notably giant cells, extraembryonic ectoderm, ectopiacental cone and chorionic ectoderm (Gardner and Beddington,1988). Our interest in trophectoderm lies in understanding its derivation from non-epithelial and non-polar blastomeres and how the epithelial phenotype is propagated selectively in this lineage of the blastocyst (Fleming and Johnson, 1988; Fleming, 1992).

There are three main constraints in studying the trophectoderm as a model for epithelial differentiation: preimplantation embryos are expensive to generate, there is only limited material to work with and, because of inherent developmental asynchrony, staging of embryos must be tackled carefully. However, we believe that there are some real advantages. First, the trophectoderm is a native tissue with important consequences for morphogenesis and development, and hence is of clear medical interest. Second, its formation is rather slow (about 24 hours), providing time for analysis of mechanisms. Third, and perhaps most important, the epithelium differentiates de novo after normal cell cycling and ‘housekeeping’ functions have been reestablished in the embryo following release from oocyte meiotic arrest (Pratt, 1989; Fleming, 1992). Thus, it is possible to study gene and protein expression events required for epithelial differentiation, in addition to changes in cell organisation. Recently, we have studied maturation of cellcell adhesion during trophectoderm differentiation, in particular the formation of structural intercellular junctions. Here, we provide a brief review of this work.

Cell-cell adhesion between blastomeres begins at the 8-cell stage when the embryo compacts and each cell polarises along its apicobasal axis, generating a proto-epithelial phenotype. Adhesion is achieved by uvomorulin (120×10³M_r, E-cadherin) intercellular binding (Hyafil et al., 1980; Peyrieras et al., 1983). Uvomorulin is synthesised and is present in the membrane in a non-polar and non-adhesive state from early cleavage; at compaction, it becomes adhesive and progressively basolateral in distribution (i.e. at cell-cell contact sites) (Vestweber et al., 1987). The mechanism causing uvomorulin adhesion at compaction is not known but coincides with the onset of uvomorulin phosphorylation (Sefton et al., 1992). Moreover, treatment of pre-compact embryos with phorbol ester to activate protein kinase C causes premature compaction (Winkel et al., 1990), further suggesting a role for phosphorylation in initiating adhesion.

As adhesion commences, the proto-epithelial organisation generated in 8-cell blastomeres includes changes in the distribution of cytoskeletal elements (actin filaments, microtubules), cytoplasmic organelles (mostly endocytic vesicles) and components of the membrane and underlying cortex (e.g. microvilli, actin-binding proteins) (reviewed by Fleming and Johnson, 1988; Fleming, 1992; Gueth-Hallonet and Maro, 1992). Experimental evidence suggests that certain, as yet undefined, aspects of polarisation in the membrane and cortex (cytocortex) are of fundamental importance in guiding the reorganisation of other cellular structures (Johnson and Maro, 1985, 1986; Fleming et al., 1986; Johnson et al., 1986a; Wiley and Obasaju, 1988). This primary cyto-cortical state appears also to act as a stable ‘memory’ of the polar axis in subsequent cell cycles, since during mitosis aspects of polarity are transiently run down before being reestablished in the next interphase (Johnson et al., 1988). What is the relationship between cell adhesion and polarisation at compaction? If the onset of adhesion is prevented (by, for example, incubation with uvomorulin antibody), polarisation is delayed, disoriented with respect to cell contact patterns, but is not inhibited (Johnson et al., 1986b). This suggests that adhesion may act normally to induce and orientate polarisation but the molecular pathway involved is not known. In the absence of adhesion, however, other non-specific factors may induce polarisation, reflecting the ‘programmed’ state of blastomeres to begin differentiation. ‘Programming’ for polarisation at compaction requires RNA, DNA and protein synthesis in the 2-cell embryo, 24 hours earlier, but apparently not subsequently (Kidder and McLachlin, 1985; Smith and Johnson, 1985; Levy et al., 1986).

The first epithelial-type junctions to form in the embryo are gap and tight junctions during the 8-cell stage, although an analysis of adherens junction formation has not yet been conducted systematically (Reima, 1990). The process of tight junction (zonula occludens) formation will be considered here; gap junction formation and function during preimplantation development have been discussed elsewhere recently (Barron et al., 1989; Bevilacqua et al., 1989; Nishi et al., 1991; Fleming, 1992; Fleming et al., 1992).

Ultrastructural analysis of embryos indicates that tight junctions begin to form at compaction; however, this process is not completed until the late morula stage (16-to 32-cell stage), up to 24 hours later (Ducibella and Anderson, 1975; Ducibella et al., 1975; Magnuson et al., 1977; Pratt, 1985). Once completed, vectorial transport and blas-tocoel accumulation can begin, dependent upon the segregation of Na⁺,K⁺-ATPase to basolateral trophectoderm membranes (Watson and Kidder, 1988). During the tight junction construction period, the apicolateral contact region between outside blastomeres displays sites of apparent membrane fusion and lanthanum exclusion, corresponding with a freeze-fracture morphology of anastomosing strands and grooves in complementary faces. These sites first appear as a discontinuous series along the apicolateral border and gradually extend laterally to become continuous (zonular) during the morula stage.

We have examined the molecular maturation of the tight junction in embryos using antibodies to peripheral membrane (cytoplasmic face) tight junction proteins. For a review of tight junction molecular composition, see Anderson and Stevenson (1991). Since embryos develop asynchronously, to improve accuracy in the relative timing of assembly of different components, we have used natural 8cell couplets (2/8 cells), synchronised from their time of division from single 4-cell blastomeres (1/4 cells). These couplets engage in adhesion and polarisation at compaction in a similar way to blastomeres in intact embryos. The assembly of ZO-1 (225×10³M_r; Stevenson et al., 1986; Anderson et al., 1988) at the junction begins 1-2 hours after compaction initiates and precedes the assembly of cingulin (140×10³ M_r; Citi et al., 1988, 1989) which, on average, assembles about 10 hours later, usually during the 16-cell stage (Fleming et aL, 1989, 1993). For both proteins, assembly is at first punctate before gradually becoming zonular. Thus, in double-labelled 2/8 couplets recently compacted, ZO-1 but not cingulin is detectable at the tight junction site (Fig. 1 A,B). Our results imply, therefore, that junction formation is progressive, not only in morphological terms but also at the molecular level. Double immunogold analysis of ZO-1 and cingulin localisation at the tight junction in chick and rat epithelia indicate that ZO-1 is positioned closer to the membrane than cingulin (Stevenson et al.,1989). Assuming these proteins occupy similar relative positions in the mouse embryo junction, our results are also consistent with molecular maturation proceeding in the membrane-to-cytoplasm direction (Fleming et al., 1992).

What factors might influence the sequential nature of tight junction protein assembly during cleavage? To investigate this, we have studied the ontogeny of expression of junction components at the protein level. Immunoblot analysis indicates detectable expression of ZO-1 from the late 4-cell stage, some 5-6 hours before ZO-1 incorporation into junctions is evident cytochemically (Fleming et al., 1989). The biosynthesis of cingulin is more complex than that of ZO-1 because this protein is expressed from both the maternal and embryonic genomes (Javed et al., 1993). Thus, immunoblot and immunoprecipitation data show synthesis of cingulin, but not ZO-1, during oogenesis. Maternal cingulin has a short metabolic half-life, ceases synthesis during the 2-cell stage when maternal transcripts degrade globally, and localises in the oocyte cortex where it may participate in the oocyte-cumulus cell interaction (Javed et al., 1993; Fleming et al., 1993). Biochemical and embryo manipulation experiments indicate that this maternal pool does not contribute to tight junction formation as trophectoderm differentiates. However, the cortical binding site for maternal cingulin persists during cleavage (on embryo outer surface) long after the maternal expression programme has ceased, and presumably embryonic cingulin replaces maternal cingulin at this binding site. Cortical cin-gulin is finally degraded by endocytic turnover of the embryo outer surface in late morulae and early blastocysts (Fleming et al., 1993), resulting in cingulin cytoplasmic foci (Fig. 1C,D). Synthesis of cingulin from the embryonic genome is detectable at trace levels during early cleavage (about ten-fold less than in the egg) but a significant increase occurs at compaction, and continues to increase during later cleavage (Javed et al., 1993). The significant enhancement in cingulin synthesis from compaction at the 8-cell stage precedes by approx 5-10 hours the detection of cingulin assembly at the tight junction. This surge in cingulin synthesis is clearly later than the onset of detectable ZO-1 protein expression in the late 4-cell embryo. These differing expression patterns for ZO-1 and embryonic cingulin during cleavage are consistent with, and may control, the sequential membrane assembly pattern at the tight junction. Future work will investigate whether tight junction protein expression is regulated by sequential transcription and/or translation.

What is the relationship between the onset of tight junction formation at compaction and the concurrent differentiation of blastomeres into a proto-epithelial phenotype? Different experimental situations employing synchronised 2/8 cell couplets cultured during the fourth cell cycle have been used to investigate this question (Fig. 2A; Fleming et al., 1989). If uvomorulin adhesion is inhibited (anti-uvo-morulin antibody; calcium removal), ZO-1 membrane assembly is both delayed and distributed randomly (i.e. not at apicolateral contact site), while microvillus polarisation is similarly delayed and oriented randomly, as mentioned earlier (Fig. 2A). Thus, adhesion is necessary both for normal tight junction assembly and for regulating the timing and orientation of cell polarity. If 2/8 couplets are treated with cycloheximide from the time of their division from 1/4 cells, adhesion and microvillus polarisation occur as normal at compaction but ZO-1 membrane assembly is inhibited. Also, if newly formed 1/8 blastomeres are contacted with newly formed 1/4 blastomeres, these heterogeneous couplets will subsequently adhere together with the 1/8 cell polarising apparently normally, but ZO-1 assembly in the 1/8 cell is delayed and randomly distributed (Fig. 2A). These two latter treatments divorce tight junction formation from epithelial polarisation and intercellular adhesion in the embryo (Fleming et al., 1989). Junction formation, but not polarisation or adhesion, requires (i) proximate translation and (ii) adjacent blastomeres to be equally competent to assemble ZO-1 at the correct membrane site. The capacity to inhibit or perturb normal tight junction formation at compaction without apparently affecting epithelial polarisation or cell-cell adhesion argues against the tight junction having a fundamental role in the establishment of a proto-epithelial phenotype in the embryo, data consistent with other systems (e.g. McNeill et aL, 1990).

Why does the tight junction develop only in the tro-phectoderm lineage of the blastocyst? To investigate this question, we have compared cingulin synthesis in trophec-toderm and ICM tissues and found that, as blastocyst expansion progresses, the level of cingulin synthesis declines in the ICM and increases in trophectoderm such that the difference between these two tissues reaches fifteen-fold (Javed et al., 1993). Differential translation of tight junction proteins may therefore provide a biosynthetic mechanism to explain tight junction tissue specificity. How might blastomeres perceive their tissue type and modulate their expression pattern accordingly? Trophectoderm and ICM lineages derive from differentiative divisions of certain polarised 8- and 16-cell blastomeres (Johnson and Ziomek, 1981; Pedersen et al., 1986; Fleming, 1987). Daughter cells (at 16-or 32-cell stage) inheriting the apical microvillus domain (plus the cytocortical ‘memory’ of epithelial polarity, see earlier) remain on the outside of the embryo and continue to differentiate into trophectoderm. Daughter cells not inheriting this domain occupy an internal position in the embryo and differentiate into non-polar ICM cells (Fig. 2B). Polarised 8- and 16-cell blastomeres may also divide conservatively such that both daughters inherit part of the apical domain and continue to differentiate into trophectoderm (Fig. 2B). In newly formed 16- and 32-cell couplets following either differentiative or conservative cleavage, ZO-1 protein is present in both cells and soon associates with the apicolateral region of cell-cell contact between them (Fig. 2B; Fleming and Hay, 1991). Thus, tight junction tissue specificity cannot be explained by differential inheritance of junctional proteins and their membrane binding site (cf. the apical microvillus domain). Following differentiative cleavage, non-polar cells gradually become enveloped by polar cells until all contact-free membrane is lost. Once this has occurred, membrane-associated ZO-1 in non-polar cells is rapidly run down whilst at polanpolar contacts, apicolateral ZO-1 assembly is maintained (Fig. 2B). If contact-free membrane is regained by non-polar cells (e.g. by isolating the ICM), ZO-1 membrane assembly is re-established within a few hours. This assembly process does not require transcription but does require protein synthesis (Fleming and Hay, 1991). These results suggest the following model for tight junction tissue specificity which is currently being tested in our laboratory. Cell position, and hence tissue type, is intepreted by cell-cell contact patterns which regulate tight junction assembly competence. In the absence of membrane assembly (ICM lineage), expression is down-regulated at the protein level but appropriate transcripts are retained. These putative transcripts would allow for rapid formation of a new tight junction network once contact asymmetry is provided. This mechanism, if substantiated, is perhaps required for rapid delamination of primary endoderm by the ICM at the time of implantation, or for replacing damaged trophectoderm cells (Fleming and Hay, 1991).

Desmosome junctions, characterised by their disc-shaped, membrane-associated plaques and inserted cytokeratin filaments, have been reviewed recently at cellular (Garrod and Collins, 1992) and molecular (Buxton and Magee, 1992; Legan et al., 1992) levels. Desmosomes form relatively late in the process of trophectoderm differentiation, at the 32cell stage once blastocoel formation has begun (Ducibella et al., 1975; Magnuson et al., 1977; Jackson et al., 1980; Fleming et al., 1991). In contrast to the tight junction, the first desmosomes do not appear to form gradually, but rather display a mature morphology and molecular composition. Thus ultrastructurally, nascent desmosomes, although rather small, contain membrane-associated cytoplasmic plaques with inserted cytokeratin filaments and an intercellular adhesive domain possessing a dense midline (Fleming et al., 1991). These early desmosomes show immunoreactivity for the major desmosomal proteins, desmoplakin (dp 1+2) and plakoglobin (dp3), and glycoproteins, desmoglein (dgl) and desmocollins (dg2+3). In time-course assays, all of these components first assemble at basolat-eral membranes from the 32-cell stage and, with the exception of plakoglobin, only after blastocoel fluid accumulation has initiated. Plakoglobin may assemble slightly earlier in the same cell cycle, before fluid accumulation begins. The close temporal relationship between the assembly of desmosomes and the start of blastocoel formation suggests that desmosomes may have an important role in blastocyst expansion, perhaps to stabilise the trophectoderm layer as the cavity enlarges.

What mechanisms may control the timing of desmosome formation in the embryo? We have monitored the pattern of synthesis of desmosomal constituents and found that plakoglobin and desmoplakin are both synthesised significantly in advance of their time of membrane assembly at the 32-cell stage. However, synthesis of desmoglein and desmocollins is not detectable before the 32-cell stage, suggesting that their synthesis may be regulatory for desmosome formation at cavitation (Fleming et al., 1991). Currently, we are investigating the time at which transcription of desmocollins begins using the reverse transcriptase-PCR technique. Our preliminary data indicate that desmocollin transcripts are first expressed from the embryonic genome in the 16-cell morula (J. Collins and T. Fleming, unpublished), an event that may therefore regulate desmocollins synthesis and the timing of desmosome assembly. Finally, the close link between desmosome formation in trophectoderm and the start of blastocoel accumulation suggests that other non-biosynthetic processes may ‘fine-tune’ assembly timing. Cavitation, like compaction, is an integrated differentiation event. Tight junction formation is completed, apical and basolateral membrane domains become segregated, Na⁺,K⁺-ATPase localises on basal membranes and vectorial transport initiates (e.g. DiZio and Tasca, 1977; Kaye et al., 1982; Watson and Kidder, 1988; Watson et al., 1990). Under these circumstances, it is plausible that newly synthesised desmosomal glycoproteins, restricted to baso-lateral membranes by the ‘fence’ activity of the tight junction, would rapidly accumulate to a threshold level appropriate for desmosome assembly.

Programmes of expression of junctional constituents in the embryo, in combination with cell-cell interactions, appear to control the timing, characteristics and tissue specificity of assembly of structural intercellular junctions. Biogenesis of the tight junction network is gradual (approximately 24 hours), both in morphological and molecular terms. In contrast, the first desmosomes are constructed, with apparent full molecular complexity, over a relatively short time period. The slower maturation of the tight junction may reflect in part its larger, zonular configuration, requiring a sequential pattern of expression of constituents to achieve efficient assembly. In contrast, the formation of smaller, disc-shaped desmosomes may be controlled by a different mechanism, the later expression of specific constituents (desmosomal glycoproteins) that act as limiting factors for complete assembly. Both of these junction types, however, appear to function for the first time when the embryo accumulates blastocoelic fluid. Their varied patterns of biogenesis may therefore explain why they begin to form at different times during trophectoderm differentiation and perhaps why compaction must occur a day in advance of cavitation.

We are grateful for the financial assistance provided by The Wellcome Trust and the Wessex Medical Trust for research in our laboratory. We are also grateful for essential supplies of antibodies to adhesion and junctional proteins, particularly from Drs B. Stevenson (University of Alberta), S. Citi (Cornell University Medical College), D. Garrod (University of Manchester) and M. Takeichi (Kyoto University).