Maternal mRNAs of PEM and macho 1, the ascidian muscle determinant, associate and move with a rough endoplasmic reticulum network in the egg cortex

Localization of specific mRNAs is essential for cell polarity, asymmetric division and differentiation in several cells, including yeast, fibroblasts,neurons and oocytes, and in embryos(Bassell et al., 1999; Jansen, 2001; Kloc et al., 2002). In oocytes, eggs and embryos, maternal mRNA determinants that are localized in the egg cortex – oskar and bicoid in Drosophila,VegT and Xdazl in Xenopus, and macho 1 in the ascidians Halocynthia and Cion – specify embryonic axes and cell types (King et al.,1999; Lasko, 1999; Mowry and Cote, 1999; Nishida, 2002; Riechmann and Ephrussi, 2001). It is not well understood how these mRNAs translocate and are anchored to the cortex or how they relocate within the cortex, as they do in oocytes and eggs of ascidians and Xenopus.

Ascidians are urochordates that develop into simple tadpoles of ∼3000 cells (Satoh, 1994). They display a profound degree of autonomous development(Nishida, 1997; Satoh, 2001). Removal and transfer of small fragments of the periphery of the egg show that the formation of three kinds of tissue (tail muscle, endoderm and epidermis)involves peripherally localized maternal determinants that are repositioned in two main phases of reorganisation that occur between fertilization and first cleavage (Nishida, 1997; Roegiers et al., 1999). An essential element of the muscle determinant is a cortically localized maternal mRNA that encodes a zinc-finger transcription factor called macho 1(Nishida and Sawada, 2001). The mRNA for macho 1 is one of a large number of maternal mRNAs(called postplasmic RNAs) that concentrate at the posterior pole of the cleaving embryos. The original, most abundant posteriorly-localized mRNA,called pem (for posterior end mark), was identified in Ciona(Yoshida et al., 1996). This was followed by the discovery of several mRNAs (pem-2-pem-6)with similar localization (Satou and Satoh, 1997). To date, 28 postplasmic mRNAs have been identified in Halocynthia, in the MAGEST maternal mRNA database(http://www.genome.ad.jp/magest/)(Sasakura et al., 2000; Makabe et al., 2001; Nakamura et al., 2003).

The Halocynthia homologue of Ciona pem, HrPEM, plus macho 1 mRNA and some postplasmic mRNAs in Halocynthia are already located before fertilization in an animal-vegetal gradient (classified as type I postplasmic RNAs) (Nishida and Makabe, 1999; Sasakura et al.,2000; Makabe et al.,2001; Nishida and Sawada,2001; Nakamura et al.,2003) (the present study). These type I postplasmic mRNAs then become concentrated in the vegetal cortex after fertilization and relocalize to the posterior cortex before first cleavage (see Fig. 1 for HrPEM and macho 1). Co-ordinated relocalization of cortical and cytoplasmic domains has been well documented in ascidians since 1905, when Conklin described the myoplasm in the egg. This is a pigmented peripheral cytoplasmic domain that is rich in mitochondria and inherited by muscle cells(Jeffery, 1995; Nishida, 1997; Satoh, 2001). The concentration and translocation of the myoplasm occurs between fertilization and first cleavage in two main phases that consists of several subphases of cytoplasmic and cortical reorganizations(Sardet et al., 1989; Chiba et al., 1999; Roegiers et al., 1999). These reorganizations (historically called ooplasmic segregation) are driven by the interplay of microfilaments and/or astral microtubules in the cortex and subcortex. In this way, at least five cortical and cytoplasmic domains are stratified in the vegetal hemisphere by the time meiosis ends(Roegiers et al., 1999). After meiosis, two of the domains, the cortical ER-rich domain and the bulk of the mitochondria-rich subcortical-myoplasm domain, move posteriorly and are distributed equally between the first two blastomeres.

By the 8-cell stage, the bulk of these adjacent cortical and subcortical domains is segregated into the posterior-vegetal blastomere pair(Roegiers et al., 1999). These blastomeres are fated to give rise to 24 primary muscle cells in the tadpole tail (Nishida, 1997). During these early cleavages, postplasmic mRNAs are concentrated in a particular cortical structure, the CAB, that is located in the posterior-vegetal blastomere pair (see Fig. 1D,E)(Hibino et al., 1998; Nishikata et al., 1999). The CAB is a 10-μm-thick cortical specialization at the posterior pole of the cleavage-stage embryo. It is clear that the CAB is a dual-function complex that mediates both unequal cleavages and localization of specific maternal RNAs (Nishida, 2002).

We have shown that it is possible to isolate the cortex of eggs of the ascidian Phallusia mammillata by gently shearing eggs and embryos fixed to a polycationic surface using a stream of isotonic solution that mimicks the cytoplasm of the egg (Sardet et al., 1992). Isolated cortices are constituted of characteristic networks of cER, microfilaments and some microtubules, all of which remain strongly attached to the plasma membranes during the isolation procedure(Sardet et al., 2002). In unfertilized ascidian eggs the cER network is a monolayer distributed along the animal-vegetal gradient (Sardet et al., 1992). It is characterized by the presence of numerous sheets and tubes in the vegetal hemisphere and by a sparse distribution of tubes in the animal hemisphere. After fertilization, this gradient distribution of the cER network increases via microfilament-driven cortical contractions which concentrate the cER into a 2- to 5-μm-thick layer around the contracted vegetal pole (Gualtieri and Sardet,1989; Speksnijder et al.,1993). The vegetal cER network is then displaced towards the posterior pole via microtubule-driven translocations mediated by the sperm aster (Speksnijder et al.,1993). The bulk of the cER network and the myoplasm are distributed equally in the posterior region of the first two blastomeres after amphimixy, mitosis and cleavage (Roegiers et al., 1999). The cER is of particular interest because at low resolution its location and displacement after fertilization seem to correspond closely to the position and relocalizations of type I postplasmic RNAs.

In the present work, we examined the precise localization of HrPEMand macho 1 in eggs, zygotes, embryos and isolated cortices using high-resolution in situ hybridization. Our results showed that HrPEMand macho 1 mRNAs could be isolated with the cortex and that both were associated with the network of rough cER. We propose that some postplasmic mRNAs translocate and relocalize with the cER network in the egg cortex following fertilization and accumulate with cER in the CAB during cleavages. A multimedia BioClip `Polarity inside the egg cortex'(www.bioclips.com/bioclip.html)describes this in more detail.

Adults of the ascidian Halocynthia roretzi were collected near the Asamushi Marine Biological Station, Aomori, Japan, and the Otsuchi Marine Research Center of the Ocean Research Institute, Iwate, Japan. Halocynthia spawns eggs ∼280 μm in diameter. Naturally spawned eggs were fertilized artificially and raised in seawater at 11°C. They start first cleavage 2.3 hours after fertilization and reach the 8-cell-stage at 4.5 hours. Eggs were fertilized with sperm activated by NaOH treatment, and the vitelline membrane was removed chemically either before or after fertilization as described previously(Mita-Miyazawa et al.,1985).

Isolated cortices were prepared using cortex solutions (Buffer X, EMC and CIM buffer) and observed fixed or live as described(Sardet et al., 1992). Devitellinated eggs (15-30), synchronously developing zygotes and embryos were deposited in a Ca²⁺-free medium (EMC solution) onto polylysine-coated coverslips. Unfertilized eggs that were attached, fertilized eggs (20-30 minutes after fertilization) and 8-cell-stage embryos (4-5 hours)were immediately sheared with a gentle stream of Millipore-filtered isotonic cortex solution (Buffer X) to which was added 1 unit μl^–1RNase inhibitor (Toyobo, Osaka, Japan). The cortices were fixed within 1-2 minutes after isolation in a freshly prepared mixture of paraformaldehyde (4%)and glutaraldehyde (0.1%) in CIM buffer. After 30 minutes fixation in a humid chamber at 11°C, cortices were washed in CIM buffer and stored for <6 hours in phosphate-buffered saline (PBS) to which was added 1 unitμl^–1 RNase inhibitor. The ER network in isolated cortices was labelled with carbocyanine dyes, DiIC16(3) (Molecular Probes, Eugene, OR,USA) emulsion or a solution of DiOC6(3) (Molecular Probes) in either CIM buffer or Buffer X as described (Sardet et al., 1992). Each cortex-loaded coverslip was inverted and mounted on a microscope slide using two strips of double-sided sticky tape spacers,then fixed firmly with epoxy glue on the left and right sides to form a perfusion chamber. Several cortices were identified by their position on the coverslip and photographed to record the cER distribution. These cortices were identified again after in situ hybridization to record the distribution of specific mRNAs. Cortices were observed and photographed by fluorescence microscopy (Nikon) and confocal microscopy using a Nipkow disc-based apparatus(Yokogawa, Tokyo, Japan) coupled to a fluorescence microscope (Zeiss).

Specimens were hybridized in situ with digoxigenin (DIG)-labelled HrPEM and macho 1 antisense probes that cover the entire cDNAs (GenBank accession numbers, AB045129 for HrPEM and AB045124 for macho 1). Arginyl-tRNA synthetase was used as a probe for ubiquitously distributed maternal mRNA (GenBank accession number, AV383566). Detection of mRNA in whole-mount specimens was carried out essentially as described previously (Miya et al.,1994).

For cortex specimens, the in situ protocol was carried out without treating fixed samples with detergent such as Tween 20, solvent such as EtOH, and proteolytic enzymes such as proteinase K. Solutions between the coverslip and the microscope slides of the perfusion chamber were changed by drawing solutions from one side with a piece of filter paper whilst perfusing the next solution from the opposite side. The slides were maintained in handmade humid chambers throughout the in situ hybridization protocol. After washing with PBS, samples were prehybridized for 1 hour at 42°C. Then the specimens were allowed to hybridize with the DIG-labelled antisense probe (approximately 0.1 μg ml^–1) overnight at 42°C. The specimens were treated with 20 μg ml^–1 RNase A (Sigma) for 30 minutes at 37°C and washed. The conventional coloring using alkaline-phosphatase conjugated anti-DIG antibody was performed as described(Miya et al., 1994).

The tyramide signal amplification method (TSA Biotin System, NEN Life Science Products, Boston) was used to obtain high-resolution fluorescent images of mRNA localization (Wilkie and Davis, 1998). This technique depends on peroxidase-mediated deposition of biotin-labelled tyramide at the location of the antisense RNA probe labelled with DIG. After washing of hybridized embryos and cortices,samples were treated with an HRP-conjugated anti-DIG antibody Fab fragment(1:2000, Boehringer Mannheim, Mannheim, Germany) overnight at 4°C. The TSA reaction was carried out for 10 minutes according to the supplier's instructions. Samples were then washed with PBS for 30 minutes and reacted with Alexa green 488-conjugated streptavidin (1:100, Molecular Probes, Eugene,OR, USA) for 60 minutes. After washing with PBS for 60 minutes, samples were perfused with 80% glycerol and observed under a fluorescent microscope or a confocal microscope. This method gives high resolution because the peroxidase reaction produces tyramide radicals that react covalently with proteins at the site of HRP activity, which prevents appreciable diffusion of the signal and fluorescently labelling the structures surrounding the target mRNA.

Isolated cortices fixed as described above were treated briefly with OsO₄, fast-frozen and processed for deep-etching and replication as described (Sardet et al.,1992). Replicas were examined in a Hitachi electron microscope.

The fluorescent TSA-biotin system provides a level of detection of hybridized mRNAs with sensitivity close to that of the conventional chromogenic method using the alkaline phosphatase-based substrate-precipitation strategy. Such comparisons are displayed for HrPEM and macho 1 mRNAs localization in 8-cell embryos in Fig. 1D-G. In addition, the use of fluorescence allows high-resolution examination of RNA localization in confocal median (Fig. 1A-C,A1-C1) and tangential sections(Fig. 1A2-C2,A3-C3).

The changing localization of HrPEM and macho 1 maternal mRNAs is shown in Fig. 1. In unfertilized eggs, HrPEM mRNA, the most abundant postplasmic RNA and macho 1 (not shown) formed a thin cortical layer along the animal-vegetal gradient, with a maximum concentration in the vegetal hemisphere (Fig. 1A). High-resolution examination of the surface by confocal microscopy showed that HrPEM and macho 1 are located in a reticulated structure<2 μm thick situated beneath the plasma membrane(Fig. 1A1). The mRNAs became further concentrated in the vegetal pole region as a consequence of the first major phase of ooplasmic segregation, which occurs 20-30 minutes after fertilization (Fig. 1B,B1). After completion of meiosis (40-50 minutes), postplasmic mRNAs moved to the future posterior pole during a second major phase of relocalization (80-100 minutes, Fig. 1C,C1). As a result of these cortical relocalizations, the cortical structure decorated by HrPEM and macho 1 thickened to 5-8 μm(Fig. 1A1-C1). This mRNA-rich cortical structure generally appears reticulate (see particularly the tangential section shown in Fig. 1A2,B2) and sometimes punctate (possibly because the formaldehyde fixation does not preserve the continuity of the structure). Some labelled structure apparently detached from the surface after the second phase of relocalization (Fig. 1C1).

Zygotes of Halocynthia start to cleave at about 140 minutes. The first cleavage partitioned the posteriorly and cortically localized mRNAs,such as HrPEM and macho 1, equally between the first two blastomeres (data not shown). At the 8-cell stage, only posterior-vegetal(B4.1) blastomeres inherited postplasmic mRNAs. These RNAs are concentrated in a tiny area of the posterior region of the B4.1 blastomeres(Fig. 1D-G). This region corresponded to the CAB, the cortical structure that attracts the centrosome and nucleus towards the posterior pole of 8- to 64-cell-stage embryos(Nishikata et al., 1999). Posterior-animal views of the 8-cell embryos displayed the moustache-like shape of the mRNA localization, which corresponds to the characteristic shape of the CAB just before unequal cleavage starts(Fig. 1F,G).

Because the appearance and movements of the reticulated structure decorated by HrPEM and macho 1 reminded us of the cER situated beneath the egg surface (Roegiers et al.,1999; Speksnijder et al.,1993), we isolated cortices from eggs, zygotes and embryos of Halocynthia using the methods previously described for unfertilized eggs of Phallusia (Sardet et al.,1992). We isolated cortices after the first phase of ooplasmic segregation, when postplasmic RNAs are concentrated in the vegetal-pole region(Fig. 2A-D), and from 8-cell embryos where they make a small posterior mark(Fig. 2E-I). Characteristic fields of pancake-like imprints that represent fragments of cortices could be observed on the coverslip (Fig. 2A,C,E). To investigate the localization of mRNAs at the subcellular level, it was essential to preserve the structures of these isolated cortices and, in particular, their membranes (plasma membrane, cER network and occasional membranous organelles). To this end, we omitted the permeabilization, digestion and extractions steps (using Triton X100,proteinase and ethanol) that are generally included after fixation in conventional in situ hybridization procedures.

Initially, we labelled cortices with the lipophilic dye DiIC16(3) to observe and photograph the cER network in isolated cortices. After localizing mRNAs with the chromogenic method (alkaline-phosphatase-generated precipitate)(Fig. 2A,B), we relocated and rephotographed the same cortices. At low resolution there was a good correspondence between the presence of the HrPEM mRNA-rich zone and the DiIC16(3)-labelled area that corresponded to the cER-rich zone in isolated cortices (data not shown). We also compared the chromogenic and fluorescence(Alexa) visualization of HrPEM mRNA in isolated cortices fixed with a mixture of paraformaldehyde and glutaraldehyde to preserve the continuity of the cER network. Consistent labelling of isolated cortices was obtained by both methods (Fig. 2A,C). As seen in Fig. 2B,D, the fluorescent-visualization method provides finer details than the chromogenic method. As in the fertilized egg (Fig. 1,B2) an HrPEM-rich reticulated network could be discerned in the isolated cortex (Fig. 2D). It is clear from both chromogenic and fluorescent methods that the cortical mRNAs are not distributed homogeneously over the entire isolated cortex (Fig. 2A-D),which indicated that the polarized distribution of cortical mRNAs in the egg is preserved in isolated cortical fragments.

Cortices could also be isolated from the 8-cell embryos, which are cube-like. Groups of four, cortical, pancake-like fragments were detected(Fig. 2E,F). In roughly 1 in 6 cases, which represented embryos that fell on the posterior side, isolated cortices retained the characteristic mRNA-rich moustaches(Fig. 2E-I). These moustaches correspond to the postplasmic RNA-rich CAB in Fig. 1D-G. By contrast, pieces of cytoplasm that occasionally adhered to isolated cortices did not hybridize appreciably to HrPEM and macho 1 maternal mRNAs (data not shown).

To verify the specificity of the localization of HrPEM and macho 1 mRNAs in the cortex, we examined whether a ubiquitously distributed maternal mRNA in egg cytoplasm (arginyl-tRNA synthetase;T. Okada and H. Nishida, unpublished) was retained in isolated cortices. The prevalence of this mRNA in eggs is thought to be comparable to that of macho 1, based on the EST data of MAGEST, an EST database of maternal mRNA in Halocynthia (Makabe et al., 2001). When cortices isolated from eggs and 8-cell embryos were examined, the signal from arginyl-tRNA synthetase mRNA is very weak compared with HrPEM and macho 1 mRNAs (see Fig. 2F, in which overexposure was used to increase the background signal), which indicates that noncortical mRNAs are not retained appreciably in isolated cortices.

The cER network can be made visible in live cortices with the lipophilic dyes DiIC16(3). As in Phallusia, the unfertilized egg cortex of Halocynthia was characterized by the presence of a continuous monolayer of cER tubes and sheets, which form a carpet that adheres firmly to the underside of the plasma membrane (Fig. 3A-C). Outside the attached cortices, the inner cytoplasmic tubes of ER that connected with the cER were usually projected onto the glass surface by the stream of cortex solution, forming a characteristic comet tail in the direction of shear (Fig. 4A,C,D,F,G). High-resolution electron microscopy of replicas of cortices generated by the fast-freeze deep-etch technique showed that the tubes and sheets of the cER were studded with particles(Fig. 3C,D). Most particles on the cER had the same size and shape, and were most likely ribosomes(Fig. 3D), as shown for Phallusia cortices (Sardet et al., 1992). Therefore, the cER of Halocynthia cortices is rough ER. However, a few particles were smaller or had elongated shapes,different from those of ribosomes.

When the first phase of ooplasmic segregation was complete (∼20 minutes after fertilization), many isolated cortices showed an area where the cER was highly concentrated (Fig. 2C,D, Fig. 4B). This area corresponds to the region at the vegetal pole of the egg where cER accumulates as a result of the cortical microfilament-driven contraction(Roegiers et al., 1999; Speksnijder et al., 1993). At higher magnification, the edge of the cortex in the zone of ER accumulation was characterized by microvilli (Fig. 4B). This is a hallmark of the vegetal pole region and results from the cortical contraction. As mentioned above, there was a good correspondence between the HrPEM and macho 1 mRNA-rich zone and the vegetal region where ER and microvilli were most concentrated(Fig. 2A-D, Fig. 4B). In this study, we focused mainly on HrPEM localization because it is more abundant than macho 1 mRNA and images obtained using the HrPEM probe had less background. However macho 1 mRNA had similar localization, as shown in Fig. 4G.

The cER network in the unfertilized egg is a monolayer of interconnected tubes and sheets attached to the plasma membrane, so this is the easiest stage at which to analyse colocalization of fluorescently labelled mRNAs and cER. At the highest resolution possible with the fluorescent microscope, mRNAs for HrPEM and macho 1 were distributed in a network that coincided precisely with the cER network(Fig. 4A,E-G). It was also clear that there was no appreciable hybridization of the probes for HrPEM-1 and macho 1 on ER tubes that originated in deeper cytoplasmic regions. Such inner cytoplasmic strands of ER often lie outside the cortex in the comet tails produced during shearing of the eggs, and are in continuity with the cER network (Fig. 4A,D,F). The lack of signal on noncortical ER indicates that the mRNA signal detected on cER is not caused by nonspecific binding of the postplasmic mRNAs to the ER of the whole egg. Further evidence for the specificity of localization of HrPEM and macho 1 on the cER network was provided by the low amount of binding of the sense probes for HrPEM and macho 1, and the antisense probes for the ubiquitously distributed maternal mRNA of arginyl-tRNA synthetase on the cER network (Fig. 2F). To further test the binding of mRNAs for HrPEM and macho 1 to the cER network, we exposed live cortices to hypotonic treatment, which vesiculates the cER network. Previously, we showed that many of the ER vesicles remain attached to the plasma membrane even after hypotonic treatment(Sardet et al., 1992). The HrPEM mRNA hybridization pattern in this case appeared as lines of fluorescently double-labelled spots that coincided with vesiculated ER(Fig. 4H). These observations,therefore, provide strong evidence that cortical postplasmic mRNAs, such as HrPEM and macho 1, are localized on the monolayer of cER that is attached to the cytoplasmic face of the plasma membrane in the unfertilized egg.

We have previously shown that we could strip the cER of particles(ribosomes and others) by treating isolated Phallusia cortices with a solution of puromycin and KCl (Sardet et al., 1992). The puromycin-KCl treatment dramatically decreased the signal for HrPEM RNA on the cER network of unfertilized Halocynthia eggs (Fig. 4A,C).

The colocalization of HrPEM and macho 1 mRNA with the cER is also evident in egg cortices after the cortical contraction triggered by fertilization, at the end of the first phase of ooplasmic segregation, 20 minutes after fertilization (Fig. 4B,D). However because the cER network in the vegetal region of isolated cortices is much thicker and compacted by the cortical contraction(Fig. 1B), precise colocalization is seen only using confocal microscopy(Fig. 4B,D). In these isolated cortices, microvilli at the vegetal pole, which resulted from fertilization-induced cortical contraction, were labelled with the lipophilic dye DiIC16(3) but did not contain significant HrPEM mRNA. This indicates that RNA localization is restricted to the cER network(Fig. 4B).

As described previously (Fig. 2E-I) and shown at higher resolution in Fig. 5A, cortices isolated from mid-8-cell-stage embryos typically contained striking patches of cER accumulation in the shape of moustaches. The two sides of the moustaches correspond to the two CABs present in the cortex of the two B4.1 posterior-vegetal blastomeres (Fig. 5A). Our observation that tight packing of cER filled most of the volume of the CAB that were isolated with Halocynthia cortices fits with the previous detection of many vesicles in the CAB by electron microscopy(Iseto and Nishida, 1999). The CAB of 8-cell-stage Halocynthia embryos is a macroscopic cortical structure ∼4-7 μm thick, 5-10 μm wide and 30-50 μm long. Both DiIC16(3) and DiOC6(3) gave identical labelling patterns that revealed the topography of the ER network in isolated cortices(Fig. 3E,F, Fig. 5F). The ER in the CAB seems to be a continuous network of tightly packed tubes(Fig. 3E,F), which are unfolded occasionally by shearing force (Fig. 5D). The ER network compacted in the CAB was connected to tubes of the cER network that extended as a monolayer to line the plasma membrane outside the CAB (Fig. 3E, Fig. 5C). In some cases, large vesicles were embedded in the ER-rich CAB; we presume that this represents an early stage of compaction of the CAB (Fig. 3E). As expected, the tubular ER network inside and outside the CAB vesiculated when live, 8-cell-stage cortices were perfused with hypo-osmotic solutions (Fig. 3G).

A low-resolution inspection of fields of cortices probed for HrPEMand macho 1 indicated that these transcripts were highly concentrated in the CAB (Fig. 2E,G-I, Fig. 5A,B), as expected from whole-mount specimens of 8-cell-stage embryos labelled for HrPEM and macho 1 (Fig. 1D-G). We also noticed that with HrPEM, which gives a strong signal using fluorescent in situ hybridization, very little mRNA was detected in 8-cell-stage cortices outside the CAB (Fig. 5B,C). The HrPEM mRNA signal decreased with increasing distance from the CAB (Fig. 5B,C). Fig. 5Cshows that HrPEM mRNA formed ER-localized patches that only partially covered the cER network at the periphery of the CAB. Although highly concentrated in the CAB, the macho 1 signal was less intense than that of HrPEM, and labelled patches outside the CAB were less obvious than with HrPEM (Fig. 2I). Although a cER network was also present in animal or anterior blastomeres, it did not bind significant amounts of HrPEM mRNA(Fig. 5A).

At the highest possible magnification of the CAB with confocal microscopy,we observed partial colocalization of ER labelled with DiIC16(3) and of mRNA fluorescently labelled with Alexa green 488. Inside the CAB, the ER network was so tightly compacted that it was difficult to determine the exact degree of colocalization of the ER and mRNAs (Fig. 5F,G). Fortunately, the network of ER tubes and sheets compacted in the CAB occasionally stretched like spaghetti under the shear force applied to prepare cortices (Fig. 2E, Fig. 5D,E). In these`stretched' CABs, the large patches of colocalization of ER and HrPEMmRNA were evident. However, although some mRNA patches coincided with the ER network, others seemed to fill the space between ER sheets and tubes (see green regions indicated by arrowheads in Fig. 5E).

Using conditions that preserved the substructure of the isolated cortices and high-resolution fluorescent in situ hybridization in eggs, zygotes,embryos and their cortices, we showed that two postplasmic RNAs, HrPEM and macho 1, were mainly associated with a polarized network of cER. Control experiments, using sense probes and a probe for a ubiquitously distributed mRNA, indicated that binding to the cER was specific. Moreover, the fact that signals for HrPEM and macho 1 were present only on the cER and not on the inner cytoplasmic ER network indicated that nonspecific binding of probes to ER was unlikely.

The association of HrPEM and macho 1 mRNAs with the cER network, which is present initially as a gradient in the unfertilized egg,provides a simple explanation for the polarization and relocalization of at least some type I postplasmic RNAs, and their ultimate accumulation in the CAB, the structure responsible for unequal cleavages at the 8-64-cell stage. Our observations also indicate that the primary muscle determinant – the cortical maternal mRNA macho 1 – probably reaches its site of localized accumulation in the CAB as a passenger of the cER network. Localization and translocation of cER therefore appear to provide a simple mechanism for the localization of specific cortical mRNAs to a subset of blastomeres.

Microtubules, microfilaments and ER have been implicated in the translocation and anchoring of mRNAs to the cortex in yeast, somatic cells,oocytes and embryos (reviewed in Bassell et al., 1999; Lasko,1999; Jansen,2001; Kloc et al.,2002; Palacios,2002). However, it has been difficult to directly visualize the association between mRNAs and these subcellular structures, because fixation,digestion and permeabilization procedures used for high-resolution in situ localization studies alter the integrity of cellular membranes and subcellular organization. High-resolution localization studies of HrPEM and macho 1 in isolated cortices (an open-cell preparation that needs no permeabilization) provided direct evidence that the rough cER network that is attached to the plasma membrane is a main site of localization for some cortical mRNAs in unfertilized ascidian eggs. We cannot presume, however, that all type I postplasmic RNAs associate with the cER. At present, we do not know how type I postplasmic mRNAs reach their cortical location during oogenesis. In Xenopus oocytes it has been suggested that that localizing mRNAs(such as maternal Veg1 mRNA) might move with the ER to the vegetal cortex during oogenesis (Deshler et al.,1997; Schnapp,1999). It will be interesting to see whether the cER network and cortical mRNAs such as HrPEM and macho 1 reach their cortical location separately or together during oogenesis and maturation, and whether this happens at the same time as the polarized mitochondria-rich myoplasm becomes established as a subcortical basket(Swalla et al., 1991). Another important question is the nature of the localization signal in these postplasmic RNAs. Recently, Sasakura and Makabe(Sasakura and Makabe, 2002)showed that small elements in the 3′-UTR region are sufficient for posterior localization of Halocynthia type I postplasmic mRNAs such as HrWnt-5 and HrPOPK-1, as is the case for localized Drosophila and Xenopus mRNAs(Betley et al., 2002).

For the moment we can only speculate about the nature of the binding of HrPEM and macho 1 to the cER network. We have examined the distribution of cER, microtubules and microfilaments in the isolated cortex(Sardet et al., 1992). The fine distribution of microtubules and microfilaments is very different from that of cER. Therefore, the possibility that these cytoskeletal elements mediate association of cER and mRNAs is unlikely. Sasakura at al.(Sasakura at al., 2000)observed the localization of some type I postplasmic RNAs (HrZf-1,HrPOPK-1 and HrWnt-5) in fertilized eggs treated with cytoskeletal inhibitors. Treatment with cytochalasin and nocodazole (which disassemble microfilaments and microtubules) disturbed the translocation of those postplasmic RNAs as well as the first and second phases of ooplasmic segregation respectively, as observed previously (Sardet, 1989; Speksnijder et al., 1993). However, these inhibitors do not cause postplasmic RNAs to detach from the cortex. This indicates that cytoskeletal elements are required for translocation of cER but not for the association of mRNAs to cER. At present there is no way to disrupt intermediate filaments in ascidian embryos.

One possibility is that these mRNAs bind to ribosomes or other particles present at the surface of the rough cER network. This is indicated by the fact that puromycin-KCl treatment detaches the bulk of HrPEM and macho 1 from the cER network in isolated cortices. This treatment was used originally to detach ribosomes from the rough ER-microsome fraction(Sabatini et al., 1966) and also detaches ribosomes and other particles from the cER in the isolated cortex of Phallusia eggs (Sardet et al., 1992). In Phallusia, the cER network can be labelled with the RNA dye thiazole orange and the staining is abolished by the puromycin-KCl treatment (Sardet et al.,1992). Because rough ER is generally thought to be a site of localization for mRNAs that encode membrane or secreted proteins, it might seem surprising to find that mRNAs such as HrPEM and macho 1, which do not code for such proteins, are also associated with the cER. However, using DNA microarrays for the large-scale identification of secreted and membrane-associated gene products has revealed recently that many mRNAs encoding cytosolic proteins are associated with microsomal membranes. This indicates that the rough ER might be a major site of mRNA localization(Diehn et al., 2000). Localization of specific mRNAs to a cER network has been shown most clearly in rice seeds, in which a specialized subdomain of rough ER is the major localization site for mRNAs that encode prolamin-storage proteins(Choi et al., 2000). It has also been hypothesized that in Xenopus oocytes, developmentally important mRNAs which code for secreted proteins, such as Vg1, are associated with ER during their transport to the cortex, possibly via Vera,the conserved RNA zipcode-binding protein(Deshler et al., 1997; Mowry and Cote, 1999). Although anchoring of Vg1 to the cortex seems to depend principally on the presence of keratin filaments in isolated Xenopus cortices, it might also depend on the cER because Vg1 mRNA is lost from isolated cortices after detergent treatment (Alarcon and Elinson, 2001).

In yeast, fibroblasts and nerve cells, translocation and anchoring of mRNAs involves large irregular particles (0.2-0.7 μm) that, apparently, contain many types of mRNAs, proteins, a large number of ribosomes and, in some cases,ER (reviewed in Jansen, 2001; Kloc et al., 2002). This is also apparently the case in Drosophila oocytes, in which translocation and localization of bicoid mRNA in the cortex seems to be mediated by large particles, in particular by ER-containing `sponge bodies'(Wilsch-Brauninger et al.,1997). A role for intracellular membranes in the localization of Oskar has also been recently proposed(Jansen, 2001; Dollar et al., 2002). In addition, there are several indications that rough ER may play a role in mRNA localization in neuronal dendrites because the translocating particles contain the RNA-binding protein Staufen, which is clearly associated with rough ER in these cells (Kohrmann et al.,1999). Because Staufen is also essential for the MT-dependent localizations of the determinants oskar and bicoid to the poles of the Drosophila oocyte, and for the microfilament-mediated cortical localization of the homeobox-containing transcription factor prospero in neuroblasts (Roegiers and Nung Jang, 2000), it is possible that cortical mRNA localization in Drosophila also relies on binding to the ER and on its translocation.

It is well established that in all ascidians examined, the cortically localized postplasmic RNAs, and the determinants for muscle and endoderm formation, concentrate in the vegetal cortex under the influence of the cortical microfilament-driven contraction that is triggered by the fertilizing sperm (Nishida, 1997; Chiba et al., 1999; Roegiers et al., 1999). We have shown that, in Phallusia, this contraction is a consequence of the Ca²⁺ wave that is triggered at the site of sperm entry, and that it results in the accumulation of cER into a 2-5 μm-thick cortical patch that is located in a microvillus-rich contraction pole in the vegetal hemisphere (Roegiers et al.,1995; Roegiers et al.,1999). In the present work we showed that in the much larger egg of Halocynthia, a cER accumulation also forms in the vegetal pole region and that it is retained in the isolated cortex. This cER accumulation concentrates the postplasmic RNAs in a 5-8 μm-thick layer in the vegetal pole region. If contraction is inhibited by cytochalasin, type I postplasmic RNAs do not concentrate vegetally, but form multiple cortical patches(Sasakura et al., 2000), a behaviour we also noted in the cER network of Phallusia eggs(Speksnijder et al.,1993).

In the Phallusia zygote, the vegetal accumulation of cER moves to a posterior equatorial location after meiosis completion. This second major phase of reorganization is initiated by a displacement of the sperm aster and of its microtubules with respect to the cortex, and is completed by the subsequent microfilament-mediated wave of cortical relaxation initiated in the vegetal pole just before cleavage(Roegiers et al., 1999). We have not yet analysed this posterior translocation of cER in Halocynthia. However, because cortical and cytoplasmic relocalizations and cleavage patterns appear well conserved in ascidian species, we assume that some postplasmic RNAs move posteriorly with the cER accumulation that formed first in the vegetal pole region after fertilization.

Our present work indicates that the extreme concentration of postplasmic RNAs in a small posterior region of the B4.1 posterior-vegetal blastomeres is related to the packing of most of the cER in the unfertilized egg into one tight mass that fills a large part of the volume occupied by the CAB (Figs 3, 5). It has been observed by the electron microscopy that a dense matrix containing vesicles is concentrated in the CAB of Halocynthia (Iseto and Nishida, 1999).

Confocal light microscopy sectioning of either live or fixed cortices isolated at the 8-cell stage showed that the cER formed a tight network in the CAB, and that this network is connected with a cER network that surrounds the CAB and lines the plasma membrane. How compaction of the cER occurs during early cleavages is unknown. Microtubules are, apparently, not involved in the formation of the CAB, whose precursors first appear as a number of small cortical particles at the 2-cell stage(Hibino et al., 1998; Nishikata et al., 1999). Cortical microfilaments or `adhesion complexes' on the surface of the ER might participate in the progressive aggregation of cER into one compact CAB domain.

We observed that the cER near the CAB is covered with patches of HrPEM mRNA, and that the contiguous cER in the same blastomere and the cER network in the animal or anterior blastomeres is devoid of HrPEM and macho 1 mRNA. This argues for the presence of a specific mRNA-rich cER domain in the unfertilized egg that is inherited selectively by posterior-vegetal blastomeres. This cER could also be considered the precursor of the CAB.

Our observations have interesting implications for the localization of some type I postplasmic RNAs in the cortex, and in particular of the muscle determinant macho 1. It is clear that the compaction of the cER network underlying CAB formation represents a way to segregate important cortical mRNAs, such as macho 1, in a small localized area of the cortex in specific blastomeres. By analogy with what is known of the large particles that localize mRNA in many cells, including yeast, neurons and oocytes (reviewed in Jansen,2001; Kloc et al.,2002), we suspect that in the 8-cell-stage ascidian embryo, a translational machinery is concentrated in the mRNA patches that we observed on the cER in and near the CAB. Localization and concentration of the translational machinery and of macho 1 mRNA could be essential for the activation of the translation of the muscle determinant macho 1. Indeed, in order for macho 1 mRNA to be the primary muscle determinant, macho 1 protein must be produced before or at the time the compact CAB is formed, i.e. before the unequal cleavage of the 8-16-cell stage separates the ER-rich CAB from the mitochondria-rich myoplasm domain inherited by primary muscle cell blastomeres.

We wish to thank the staff and members of the Asamushi Marine Biological Station, the Ohtsuchi Marine Research Center and the Misaki Marine Biological Laboratory for their help in collecting and maintaining adult ascidians. We thank Patrick Chang for his supervision of the fast-freeze deep-etching technique, and Christian Rouvière and Mohamed Khamla for their assistance with the imagery. Evelyn Houliston, Janet Chenevert and Patrick Chang provided many useful comments. This work was supported by the HFSP organization (RG-299/98), the `Research for the Future' Program of the JSPS(96L00404) to H.N., and the ARC foundation and ACI program to C.S.