Induced early expression of mrf4 but not myog rescues myogenesis in the myod/myf5 double-morphant zebrafish embryo

Determination of skeletal muscle in vertebrates depends upon three members of a family of four b-HLH transcription factors, known as muscle regulatory factors (MRFs): Myf5, Mrf4, Myod and myogenin (Myog). Skeletal muscle histogenesis is a multi-step process, from precursor determination to patterning, fusion and activation of muscle-specific genes, and MRFs act at multiple steps in this process where they exert both overlapping and distinct functions. In the mouse, Myf5 and Myod (official gene symbol Myod1) function in a large part redundantly in myoblast determination, so that deletion of one gene or the other does not significantly affect muscle development (Braun et al., 1992; Rudnicki et al., 1992), but deletion of both genes eliminates the skeletal muscle lineage (Rudnicki et al., 1993). Recently, it was demonstrated that Mrf4 is also involved in mouse muscle determination. Kassar-Duchossoy et al. (Kassar-Duchossoy et al., 2004) have shown that the Myf5/Myod double mutant mice are in fact partial triple mutants, because the deletion of the Myf5 locus also compromised the genetically linked Mrf4 gene expression (Kassar-Duchossoy et al., 2004). Indeed, in mutant embryos where Mrf4 expression is preserved, embryonic myogenesis takes place in the absence of Myf5 and Myod, even though muscle rapidly degenerates in the foetal stage of development. These findings indicate that both Myf5 and Mrf4 act upstream of Myod to direct cells into the myogenic lineage (Kassar-Duchossoy et al., 2004). This is in agreement with previous expression data, which shows that Mrf4 is transiently expressed during somitogenesis and later during fiber maturation (Bober et al., 1991; Hinterberger et al., 1991). The mouse myogenin gene (Myog) instead acts genetically downstream of Myf5 and Myod to switch on muscle differentiation genes: in its absence, myoblasts are properly specified and positioned but there is a severe deficiency of muscle fibers (Hasty et al., 1993; Nabeshima et al., 1993). Mrf4 is not essential for later muscle development (Braun and Arnold, 1995; Patapoutian et al., 1995; Zhang et al., 1995), however mice Mrf4 and Myod double mutants are phenotypically similar to Myog mutants, indicating that Mrf4 and Myod play redundant roles in the activation of the differentiation program (Rawls et al., 1998).

In organisms other than mouse, the diverse roles of MRFs have been less extensively studied. In contrast to that observed in the mouse, initiation of myf5 and myod (official symbol myod1) expression is presomitic in zebrafish, Xenopus and chick embryos (Della Gaspera et al., 2006; Hopwood et al., 1991; Jennings, 1992). Zebrafish myf5 and myod are temporally and spatially expressed in largely overlapping patterns in adaxial cells and posteriorly in newly formed somites; however, myf5 alone is expressed in the posterior presomitic mesoderm whereas myod expression appears in older somites (Coutelle et al., 2001; Weinberg et al., 1996). It has been reported that either myf5 or myod is sufficient to promote slow muscle formation from adaxial cells, and that myod is required for fast muscle differentiation (Groves et al., 2005; Hammond et al., 2007). Downregulation of both Myf5 and Myod proteins abolishes slow muscle in early embryos (Hammond et al., 2007), whereas Myod but not Myf5 cooperates with Pbx homeodomain proteins to promote fast muscle differentiation (Maves et al., 2007). A possible role for zebrafish mrf4 in muscle development has not yet been addressed, even though its pattern of expression has been described recently (Hinits et al., 2007).

Here, we report that, at variance with Myf5 and Myod, Mrf4 does not control early myogenesis in zebrafish; however, if heterochronically expressed, it is able to drive normal muscle differentiation in their absence via the selective activation of myod; mrf4 does not naturally compensate for the absence of myf5 and myod, as observed in the mouse, because its expression is late. By contrast, myogenin (gene: myog), the fourth MRF, is unable to rescue complete myogenesis in myf5/myod double morphants. Moreover, we observe that in embryos in which morpholino-mediated inhibition is incomplete, some muscle forms with a highly disorganised pattern, whereas in the complete absence of the early myotome, later myogenesis is abolished, underlining a crucial role of the myotome in zebrafish.

To characterise the MRFs in zebrafish, we injected morpholinos against each gene alone, and in combination, into one- to two-cell-stage embryos. Whereas all the single morphants appeared normal at 24 hours post fertilisation (h.p.f.) (supplementary material Fig. S1), 90% (69/77) of the myf5/myod double morphants were immobile (Fig. 1A,B). In these embryos, skeletal myosin was either strongly reduced (25/69) or completely abolished (44/69) as revealed by antibody staining (Fig. 1C-E'). The morpholinos were designed to bind an exact sequence around the start codon of the mRNA of each gene and are therefore highly unlikely to cross react with one of the other MRF genes because their sequence similarity is negligible in this region. This result thus indicates that Myf5 and Myod are the only MRFs required for the induction of skeletal muscle in zebrafish and it also suggests a different function for Mrf4 in the fish compared with its role in the mouse.

We performed in situ hybridisation and real-time PCR to investigate whether mrf4 is expressed in zebrafish muscle precursors. We could not detect zebrafish mrf4 as early as myod or myf5 expression, which can be visualised by in situ hybridisation from 70-80% epiboly onwards; mrf4 expression was detected during early somitogenesis (from the 5-somite stage), similar to myog expression (Fig. 2, and data not shown). This pattern of expression of mrf4 in the zebrafish is in agreement with a recent report (Hinits et al., 2007) and is difficult to reconcile with a possible role in muscle cell determination. Conversely, mrf4 probably participates in muscle differentiation because its expression coincides with the earliest time point when we can detect myosin protein in differentiating muscle cells.

The expression pattern of mrf4 and myog suggests that they might be targets of Myf5 and/or Myod. We thus investigated the expression of the two genes in myf5 and myod single morphants as well as in myf5/myod double morphants and found their expression to be normal in single morphants but strongly reduced or abolished in more than 85% of myf5/myod double morphants (52/60) (Fig. 3), as shown for skeletal myosin. These data are in agreement with recently published results of Maves et al. (Maves et al., 2007), who found that myog, desmin, smyhc1 and mylz2 are not expressed in myf5/myod double morphants. Thus Myod and Myf5 act redundantly for activation of mrf4 and myog.

Although mrf4 does not appear to regulate muscle determination in zebrafish, we wondered whether this depends upon the different time of expression or rather upon some structural difference to mouse Mrf4. Thus we injected myf5 and myod morpholinos together with different amounts of mrf4 mRNA (20-100 pg) in the early fish embryo. Results showed that the majority of these fish were moving at 24 h.p.f. (supplementary material Movies 1-3; Table 1), and myog and myosin expression was rescued (Fig. 4). myog expression was normal in more than 50% of the rescued embryos (Fig. 4A) and only slightly decreased in the remaining embryos. Myosin staining revealed predominantly U-shaped somites, but was strongly positive in all rescued embryos (Fig. 4B,B'). By contrast, co-injection of either 100 pg or 200 pg, of myog mRNA did not activate neither mobility nor strong myosin expression (Table 1; Fig. 4C,C'). Semi-thin transverse sections revealed apoptotic and highly disorganised muscle in double morphants and quasi intact muscle in rescued embryos (Fig. 5). Electron microscopy of these sections showed organised sarcomeric structures in control and rescued embryos, which were either not present or highly disorganised in double morphants. For example, Fig. 5E shows longitudinally and transversally oriented sarcomeres in the same section.

To understand whether rescued myogenesis is comparable to the normal process, we examined the onset of gene expression of myog, pax3, pax7 and slow and fast myosin at their respective developmental stages. Except for myog, whose expression increased, and in some cases was more broadly expressed, in mrf4-rescued embryos, all other genes appeared to be expressed normally (Fig. 6). This result demonstrates that zebrafish mrf4 is able to act as a muscle determination gene in the early embryo, just like mouse Mrf4, but cannot do so during zebrafish development probably because of its late onset of expression.

To understand the molecular mechanisms underlying muscle rescue by mrf4, we examined the expression of Myod protein. The Myf5 antibody (Santa Cruz) recognises Myod in zebrafish, as reported by Hammond (Hammond et al., 2007) and confirmed by us here. Myod single morphants do not express Myod protein as expected, but myf5 single morphants do (Fig. 7). Myod protein in double morphants is absent, as in myod single morphants, but is normal in rescued embryos compared with uninjected embryos (Fig. 7). We further performed real-time quantitative PCR experiments to compare the expression levels of mrf4, myod and myf5 (which cannot be detected by antibody in the zebrafish) mRNAs in embryos injected with both myf5 and myod morpholinos, in rescued embryos, and in embryos injected with two different concentrations of mrf4 mRNA alone. In all cases gene expression levels were normalised to their levels in uninjected embryos. Interestingly, both myf5 and myod mRNAs were more than tenfold upregulated in double-morphant embryos, whereas mrf4 was reduced to one fifth of its normal expression level (Fig. 8), as previously demonstrated by the strong reduction of signal for mrf4 revealed by in situ hybridisation (Fig. 3). Injection of 80 pg of mrf4 mRNA together with myf5 and myod morpholinos resulted in a more than 100-fold upregulation of mrf4 mRNA at midsomitogenesis, confirming the good quality of the injected transcript, and a nearly 50-fold upregulation of myod mRNA, whereas the level of myf5 was unchanged compared with that in the double morphants (Fig. 8). Injection of 50 pg or 20 pg mrf4 mRNA alone increased myf5 mRNA only marginally, but upregulated myod mRNA levels over 30-fold (Fig. 8). Injection of 20 pg of mrf4 mRNA was sufficient to maintain mobility (Table 1), but no elevated levels of this mRNA could be detected by real-time PCR at midsomitogenesis compared with non-injected embryos (Fig. 8), indicating that physiological levels of mrf4 mRNA are sufficient to induce at least functional fast muscle development in myf5/myod double morphants.

Taken together, these data indicate that mrf4 rescues myogenesis in double-morphant fish via the activation of myod. This is the first evidence that mrf4 is able to activate myod in vivo. Moreover, increased levels of myod morpholino (1 pmole instead of 0.5 pmole) compromised the rescue (data not shown), further supporting the finding that mrf4 rescue is via activation of myod, and that myod is necessary for the rescue.

Morpholinos have been reported to successfully block protein expression for at least 3 days in the developing zebrafish (Nasevicius and Ekker, 2000), during which they are progressively diluted and cleared from the tissue. This observation offers the unique opportunity to investigate how muscle development proceeds in a vertebrate embryo, after the transient repression of myf5 and myod. Conditional mutants have not yet been studied in the mouse. We thus investigated whether the double-morphant embryos, with either strongly reduced or absent muscle, would recover over time, after the inhibition of protein synthesis is released. We followed double-morphant larvae for 8 days and found that after 5 d.p.f. they were still unable to move. However, at 7 d.p.f., the situation changed when some larvae started to tremble, and later to swim, whereas the majority of the double morphants was still immobile. The immobile larvae remained largely devoid of skeletal myosin (Fig. 9B,F) and they died the following day, probably due to the inability to ingest food. In the larvae that did regain motility, skeletal myosin was occasionally present but skeletal muscle was highly disorganised at 3 d.p.f. compared with control or mrf4-rescued larvae (Fig. 9, compare C to A,D). Muscle in 8-day-old recovered larvae was better organised but still was not comparable with uninjected or mrf4-rescued larvae (Fig. 9G,E,H), indicating that swimming does not require a perfect muscle organisation. These results were confirmed by coinjecting myf5/myod morpholinos into embryos of the α-actin-GFP transgenic zebrafish line (Higashijima et al., 1997). By following each embryo separately over a time course of 6 days, we confirmed that embryos devoid of GFP signal at 24 h.p.f. did not express GFP later on (8/8) (Fig. 10D-F), whereas those that faintly expressed GFP at 24 h.p.f. did increase the signal over time (2/2) (Fig. 10G-I), and even if the actin-GFP revealed disorganised somites, the embryos regained some mobility.

Here we show that zebrafish myogenesis is entirely dependent on Myf5 and Myod because no muscle is formed in the absence of these proteins. Even though previous work had shown a block of adaxial myogenesis (Hammond et al., 2007), complete inhibition of subsequent fast myogenesis had not been documented before. At variance with the mouse, mrf4 is expressed as late as myog, at the onset of muscle differentiation, and therefore has no role in zebrafish muscle specification. However, when prematurely expressed, zebrafish mrf4 is able to drive myogenesis via the activation of Myod, whereas myog cannot do so. The ability of mrf4 to rescue myogenesis is probably due to the activity of the third α-helix in the C-terminus of the protein, as described by Bergstrom and Tapscot (Bergstrom and Tapscott, 2001). These authors demonstrated that the third α-helix, conserved in all four MRFs, has evolved distinct functions in Myod and myogenin. Whereas in Myod it appears to be a domain critical for the efficient initiation of skeletal muscle gene expression, in myogenin it rather acts as a general transcription activation domain. They further showed that the C-terminal domain of Mrf4 can substitute for the domain in Myod, but the same domain of myogenin cannot. Also, either Mrf4 or Myod is required together with myogenin to mediate terminal muscle cell differentiation, because mutation of mrf4 and myod result in a severe skeletal muscle deficiency, despite normal expression of myog (Rawls et al., 1998). Myogenin is indeed the only MRF that failed to induce muscle-specific RNAs when ectopically expressed in non-muscle cells (Roy et al., 2002). Taken together, these data strongly indicate that Myf5, Mrf4 and Myod, which are lineage specification factors in the mouse, possess a greater intrinsic ability to initiate the expression of silent genes than myogenin, which rather acts as a differentiation factor. The third α-helices in zebrafish Mrf4 and myogenin protein are strongly conserved, containing only one and two amino acid changes, respectively, compared with the same domain in the mouse. It is therefore probably due to the different function of this domain that mrf4 is able to rescue myogenesis in double-morphant zebrafish but myog is not.

It remains unknown why zebrafish and mouse mrf4 have a similar molecular function and yet are expressed at different periods in the two organisms. It is also not clear when, in the course of vertebrate evolution, the expression of mrf4 changed, i.e. when the gene acquired regulatory sequences able to respond to myogenic inducing factors in muscle progenitors. Also, in Xenopus embryos, mrf4 is expressed late, but interestingly here it clearly precedes myogenin expression (Della Gaspera et al., 2006; Hopwood et al., 1989; Hopwood et al., 1991; Jennings, 1992; Nicolas et al., 1998), providing yet another relative expression pattern of the MRFs. However, no functional assays have been performed to elucidate the role of mrf4 or myog in this context.

We show here that mrf4 is able to activate Myod expression, despite the presence of the myod morpholino. Our real-time PCR results demonstrate that mrf4 rescue leads to a 30- to 50-fold increase in myod mRNA, an amount that is probably sufficient to titrate out the amount of morpholino in the picomolar range and to explain the histochemical detection of the Myod protein in the nuclei of the rescued embryos. Following coinjection of myod/myf5 morpholinos, both myod and myf5 mRNAs are upregulated and this could be attributed to a compensation effect due to morpholino-mediated downregulation of Myf5 and Myod. Importantly, only myod mRNA levels are further increased by additional injection of mrf4 mRNA, whereas myf5 levels remain unchanged. Remarkably, even injection amounts of mrf4 mRNA that cannot be detected by quantitative real-time PCR at levels higher than those detected in uninjected control embryos result in an >30-fold activation of myod mRNA, indicating that mrf4 is a potent activator of myod. Increasing the amount of myod morpholino prevented rescue by mrf4, indicating that myod activation is necessary for mrf4 rescue.

In addition, we also show by in situ hybridisation studies that myog mRNA is increased in early mrf4-rescued embryos, which could be either a direct or an indirect activation via myod. Most likely, both the direct activity of mrf4 together with that of induced myod drive myogenesis in the rescued embryos.

Also, in the mouse, Mrf4 might have the ability to activate Myod because the Myf5 single mutant does express Myod but the Myf5/Mrf4 double mutant does not (Kassar-Duchossoy et al., 2004). Additionally, we demonstrate here that zebrafish myod can activate mrf4 in myf5 morphant zebrafish, indicating a positive-feedback loop between these two genes in zebrafish, which has not been reported in any species so far. For a schematic overview of the muscle gene interactions see Fig. 11.

Thanks to the transient nature of morpholino inhibition, we have investigated the development of skeletal musculature in the absence of an anatomically defined myotome. In the mouse, ablation of Myf5 and Mrf4 delays the appearance of a myotome for over 2 days, whereas in Myod mutant embryos the myotome appears normal, probably due to the early expression of Myf5 compared with Myod. However, even in the absence of an early myotome, skeletal myogenesis proceeds normally in the mouse, with only minor defects in the epaxial musculature (Kablar et al., 1997). By contrast, myotome absence precludes muscle patterning in the zebrafish. The minority of embryos that assemble fewer and disorganised muscle fibres compared with control embryos are probably those in which inhibition by morpholinos was not complete, and where some muscle had initially formed. Surprisingly, these embryos recover the ability to swim. In both immobile and motile surviving embryos, we found bundles of sarcomeres perpendicularly oriented to other bundles in the same cytoplasm of the few residual muscle cells. This observation suggests that an ordered pattern of MRF expression is also required to drive correct sarcomerogenesis in the embryonic muscle. Thus it appears that the myotome is crucial for further muscle development in zebrafish, consistent with the notion that further muscle development uses the myotome as a template and the adult muscle anatomy remains morphologically unchanged. By contrast, the large remodelling of muscle patterning that occurs in tetrapods probably developed upon later morphogenetic signals, so that the patterning role of the myotome was progressively diminished. Conditional Myf5 and Myod ablation in the mouse, and morpholino approaches in other classes of vertebrates, might further address these issues in the future.

Breeding wild-type fish of the AB strain were maintained at 28°C on a 14 hours light/10 hours dark cycle. Embryos were collected by natural spawning, staged according to Kimmel (Kimmel et al., 1995), and raised at 28°C in fish water (Instant Ocean, 0.1% methylene blue) in Petri dishes (Haffter and Nüsslein-Volhard, 1996).

Whole-mount in situ hybridisation, WISH, was carried out as described (Thisse et al., 1993) on embryos fixed for 2 hours in 4% paraformaldehyde in PBS, then rinsed with PBS-Tween (PBT), dehydrated in 100% methanol and stored at –20°C until processed (Jowett and Lettice, 1994). Probes were transcribed with T7 polymerase for antisense and SP6 polymerase for sense probes, and in vitro labeled with digoxigenin (Roche). Primers for PCR probe templates for mrf4 and myog are: mrf4_forward, 5′-ATTTAGGTGACACTATAGTTTTCAATGATTTGCGTTATCTT-3′; mrf4_reverse, 5′-TAATACGACTCACTATAGGGGAAGACTGCTGGACTCTGAAGAC-3′; myogenin_forward, 5′-ATTTAGGTGACACTATAGGATAATTTCTTCCAGTCCAGAATCA-3′; myog_reverse, 5′-TAATACGACTCACTATAGGGCTGTCCACTATAGACGTCAGAGACC-3′. Myod probe was transcribed from a plasmid (kindly provided by Steve Wilson, University College London, UK) after linearisation with BamHI.

For immunohistochemistry, embryos were fixed for 2 hours in fish fix (4% paraformaldehyde, 0.15 mM CaCl2, 4% sucrose, 0.1 M phosphate buffer pH 7.3) or for 10 minutes with a mix of 50% methanol and 50% acetone, washed several times in PBT and blocked in 10% donkey serum in PBT for 1 hour at room temperature. Primary antibody incubation was overnight at 4°C, followed by several washes in PBT and incubation of secondary antibody for 1 hour at room temperature. Nuclei were stained with Hoechst 33342. Primary antibodies are A4.1025 (anti-human all myosin) and EB165 (anti-chicken fast myosin heavy chain) purchased from Developmental Studies Hybridoma Bank (mouse hybridoma cells were grown in our lab and medium was collected and diluted 1:30 for antibody staining). Myod antibody is rabbit anti-Myf5 C-20 from Santa Cruz, and was diluted 1:100. Secondary antibodies are TRITC- or FITC-conjugated donkey anti-rabbit or anti-mouse from Molecular Probes, diluted 1:500. Images of embryos and sections were acquired using a fluorescence microscope equipped with a digital camera. Images were processed using the Adobe Photoshop software.

myod mRNA was transcribed from a plasmid kindly provided by Steve Wilson. mrf4 (AY335193), myog (AF202639) and myf5 (AF270789) cDNAs were cloned by us in the pCS2⁺ expression plasmid after amplification of the genes from embryonic cDNA. mrf4 and myf5 were amplified with primers containing the EcoRI (in forward primer) and XhoI (in reverse primer) restriction sites, whereas myog primers have a BamHI site in the forward primer and a XhoI site in the reverse primer. All cloned plasmids were verified by DNA sequencing. Synthetic capped mgn and mrf4 mRNA was injected repeatedly (n>3) at 20, 50, 80, 100 and 200 pg per embryo. Injections were carried out on 1- to 2-cell-stage embryos (with Eppendorf FemtoJet Micromanipulator 5171); the dye tracer rhodamine dextran was co-injected as a control. To repress mrf4 mRNA translation we designed an ATG-targeting morpholino (Gene Tools, LLC): mrf4-MO 5′-CGTTGGTCTCAAACAGGTCCATCAT-3′. To repress myf5 we designed two myf5 morpholinos against the ATG region and got similar results with both: myf5 MO 5′-TACGTCCATGATTGGTTTGGTGTTG-3′; myf5B-MO 5′-GATCTGGGATGTGGAGAATACGTCC-3′. We could further rescue the myf5/myod double morphants by coinjection of myf5 or myod mRNA. To repress myod mRNA translation we designed an ATG-targeting morpholino: myod-MO 5′-ATATCCGACAACTCCATCTTTTTTG-3′; and as negative controls we injected 0.5 pmole of a 5 bp mismatch morpholino against myod 5′-ATtTCCcACAAgTCCATgTTTTaTG-3′ that did result in an abnormal phenotype, or a standard control morpholino oligonucleotide (stdr-MO), a human β-thalassemia-specific morpholino that has not been reported to have other targets or generate any phenotypes in any known test system except human β-thalessemic hematopoietic cells. 0.5 pmole of myod and mrf4 morpholinos and 0.25 pmole of myf5 morpholino were injected in 1× Danieau buffer (pH 7.6) as suggested by Nasevicius and Ekker (Nasevicius and Ekker, 2000). (0.5 pmole morpholino correspond to approximately 4 ng.) All the morpholinos we injected have already been used by others and have been tested for their specificity (Chen and Tsai, 2002; Hammond et al., 2007; Lin et al., 2006; Wang et al., 2008).

Total RNA was isolated from embryos at indicated developmental stages (1-2, 6-8 and 15 somites). Reverse transcriptions (RTs) were performed using 2 μg DNase-treated (DNA-free™, Ambion) total RNA in presence of random hexamers (Invitrogen™) and SuperScript II reverse transcriptase (Invitrogen™). Real-time PCRs were carried out in a total volume of 10 μl containing 1× iQ SYBR Green Super Mix (Bio-Rad) using 0.5 μl of the RT reaction. PCRs were performed using the Mx3000P Real Time Detection System (Stratagene). For normalisation purposes, 18S ribosomal RNA or elongation factor 1 alpha (ef1alpha) mRNA was amplified in parallel with the gene of interest. The following primers were used: myf5_sense, 5′-GAATAGCTACAACTTTGACG-3′; myf5_antisense, 5′-GTAAACTGGTCTGTTGTTTG-3′; mrf4_sense, 5′-ACAACCTGAAGGAAAACCAT-3′;mrf4_antisense, 5′-TCTTCAGTGGAAATGCTGTC-3′; myog_sense, 5′-TCTGAAGAGGAGCACATTGA-3′; myog_antisense, 5′-AGCCCTGATCACTAGAGGA-3′; 18S_sense: 5′-ACCTCACTAAACCATCCAATC-3′ and 18S_antisense, 5′-AGGAATTCCCAGTAAGCGCA-3′; ef1alpha_sense, 5′-CAAGGAAGTCAGCGCATACA-3′; ef1alpha_antisense, 5′-TCTTCCATCCCTTGAACCAG-3′. All primer pairs are located in different exons. To calculate the fold increase in mRNA level of the gene of interest, normalised to the mRNA level of the housekeeping gene, the following equation was used: 2^–ΔΔCT, where ΔΔC_T=(C_T,Target – C_T,ef1alpha)_{Condition x} – (C_T,Target – C_{T, ef1alpha})_{Condition 0} (Livak and Schmittgen, 2001; Pfaffl, 2001). Condition x corresponds to the morpholino and mRNA injections and condition 0 to untreated embryos. Targets are the mrf4, myf5 and myod genes. All samples were run in triplicate and s.d. was calculated.

24 h.p.f. whole zebrafish embryos were manually dechorionated and fixed overnight at 4°C with 1.5% glutaraldehyde and 4% paraformaldehyde in 0.1 M sodium cacodylate buffer pH 7.3. They were rinsed in the same buffer and postfixed for 1 hour in sodium-cacodylate-buffered 1% osmium tetroxide. The samples were then dehydrated in a graded ethanol series, transitioned to propylene oxide and embedded in Epon 812-Araldite. Sections were obtained using a Reichert Ultracut E. 0.5 μm sections were stained with gentian violet and photographed with a digital camera. Thin sections were cut at 70 nm and placed onto copper grids, stained with 2% aqueous uranyl acetate and lead citrate and analysed under a Jeol 100 SX electron microscope. Cryosections were performed on embryos following antibody staining. Embryos were embedded in 5% sucrose and 1.5% agarose, frozen in OCT and cut into 12-m-thin transverse sections on a Leica cryostat.