Temperature influences the coordinated expression of myogenic regulatory factors during embryonic myogenesis in Atlantic salmon (Salmo salarL.)

Variations in embryonic temperature can produce both time-limited and persistent alterations to skeletal muscle phenotype in teleost fish (reviewed in Johnston, 2006). For example, embryonic incubation temperature changes mitochondrial abundance(Vieira and Johnston, 1992),the timing of expression of developmental-stage specific myofibrillar proteins(Johnston et al., 1997; Johnston et al., 1998) and the number and size distribution of myotomal muscle fibres in larval and juvenile stages (Stickland et al.,1988; Johnston et al.,2000). In amniotes, the myogenic progenitor cells responsible for embryonic and postnatal muscle growth express the transcription factors Pax3/Pax7 (Relaix et al.,2005) and are derived from the dermomyotome(Gros et al., 2005), a transient compartment of the dorsal somite of embryonic stages. In zebrafish,Pax7-expressing cells of the anterior part of the epithelial somite migrate laterally to form a layer of cells external to the myotome (the external cell layer), as the entire somite rotates through 90° from its starting position (Hollway et al.,2007). Self-renewing, undifferentiated Pax7-expressing cells persist in the external cell layer in adult zebrafish and provide myogenic precursors utilised during larval and possibly adult muscle growth(Hollway et al., 2007; Stellabotte et al., 2007; Devoto et al., 2006). Additionally, the external cell layer is the source of quiescent Pax7⁺ muscle progenitors found under the basal lamina of muscle fibres in adult zebrafish (Hollway et al.,2007), reminiscent of the satellite cells that are required for muscle growth and repair in mammals(Mauro, 1961). The anterior somite of the zebrafish also supplies progenitors used in the growth of the dermis and pectoral and dorsal fin muscles, supporting a functional role for this region equivalent to the amniote dermomyotome(Hollway et al., 2007).

Teleosts produce new myotubes throughout larval, juvenile and adult stages,a reflection of the large increase in body size that occurs during ontogeny. For example, in Atlantic salmon (Salmo salar L.) there were ∼5000 fast muscle fibres per myotomal cross-section in hatched embryos and fibre number expanded to around 850 000 by the time recruitment stopped in adult fish (Johnston et al., 2003). The final fibre number in adult salmon (FN_max) can be modified by around 20% according to the temperature experienced during the early life history stages (Johnston et al., 2003). If temperature affects the number of post-embryonic myogenic precursors originating from the external cell layer, this could provide a plausible explanation for later changes in FN_maxin adult fish. The ecological significance of the developmental plasticity of fibre number is unknown, but a higher fibre number would be expected to increase the potential for fast growth(Johnston et al., 2003) at the expense of higher routine maintenance costs(Johnston et al., 2005).

The MRFS are a conserved family of four proteins (myf5, myoD, myoG and MRF4), related by ancient gene duplication(Atchley et al., 1994). The MRFs are potent transcriptional activators of muscle-specific genes, owing to two domains conserved in each family member: the basic region and helix-loop-helix (HLH) domain (Weintraub et al., 1991). The ubiquitously expressed E-proteins share these regions and dimerize to MRFs via the HLH and the resulting complexes then bind via the basic regions to a specific motif (CANNTG, the e-box) conserved in the regulatory region of most muscle genes(Murre et al., 1989; Lassar et al., 1989). The MRFs share partial redundancy and in vitro, can each convert several cell lines to differentiated skeletal muscle(Weintraub et al., 1989). However, each gene has evolved a unique expression pattern and specialist function in initiating or maintaining myogenesis. Mouse double knockouts have shown that myoD and myf5 are critical for myogenic specification, as indicated by a lethal phenotype lacking myoblasts and skeletal muscle(Rudnicki et al., 1993). In contrast, myoG knockout mice have myoblasts, but die from a lack of differentiated muscle (Hasty et al.,1993). MRF4 plays a double role in muscle specification/differentiation since myogenesis occurs normally in myf5:myoD^–/– mice, when MRF4 is not compromised (Kassar-Duchossoy et al.,2004).

Heterochronies in the expression of myoD family members might be expected if they are involved in the developmental plasticity of myogenesis, including effects on the number of muscle fibres formed. To test this hypothesis, the expression of myoD family members has been investigated by in situhybridisation in fish embryos reared at different temperatures. The majority of studies have found no difference in the relative timing or intensity of myoD or myoG expression with respect to somite stage in embryos reared at a range of temperatures [Atlantic cod Gadus morhua(Hall et al., 2003); Atlantic herring Clupea harengus (Temple et al., 2001); common carp Cyprinus carpio(Cole et al., 2004) and Atlantic halibut Hippoglossus hippoglossus(Galloway et al., 2006)]. However, in rainbow trout Oncorhynchus mykiss it was reported that myoD and myoG expression was more intense at the mRNA and protein levels and also more advanced with respect to somite stage in embryos incubated at 12°C versus 4°C(Xie et al., 2001).

MyoD in the rainbow trout was shown to occur as two paralogues,which was thought to reflect the tetraploidization of the salmonid genome(Rescan and Gauvry, 1996). Subsequently two myoD paralogues of lower percentage identity were identified in five Percomorpic teleost fish(Tan and Du, 2002; Galloway et al., 2006; Macqueen and Johnston, 2006; Fernandes et al., 2007). A third myoD paralogue was recently characterised in Atlantic salmon,rainbow trout and brown trout (Salmo trutta) and a phylogenetic analysis showed that each salmonid myoD paralogue was orthologous to a universal teleost myoD gene (named myoD1) and distinct from the second myoD paralogue (named myoD2) found in some fish (Macqueen and Johnston,2006). The three salmonid myoD paralogues, which were named myoD1a/1b/1c, had distinct expression patterns during embryonic development and probably represent a whole genome duplication followed by a more recent local gene duplication event(Macqueen and Johnston, 2006). In the light of these recent discoveries it was thought worthwhile to re-examine potential developmental plasticity of myoD expression with temperature in Atlantic salmon. Furthermore, in order to thoroughly test the hypothesis that developmental plasticity to temperature is associated with heterochronies in MRF expression, it is necessary to extend the study to include the other myoD family members not so far investigated. MRF4is of particular interest since to our knowledge its expression during embryonic development has not previously been described in fish.

Salmo salar (L.) embryos were reared by Akvaforsk (Sunndalsora,Norway) at 2°C, 5°C and 8°C. Embryos were sampled based on the staging system of Gorodilov (Gorodilov,1996), which accounts for the rate of embryonic Atlantic salmon development at different temperatures. The following seven stages were selected: (1) the end of gastrulation, (2) 1–3-somite stage (ss), (3)10–15 ss, (4) 30–40 ss, (5) 45–50 ss, (6) toward the end of segmentation (60–65 ss), (7) post-segmentation (the `eyed stage'). Embryos were fixed in 4% (m/v) paraformaldehyde/PBS and then dehydrated by consecutive washes in increasingly concentrated methanol (until 100% m/v) and stored at –80°C until later use.

Juvenile Atlantic salmon (N=6; mean mass=291±36 g, mean fork length=263±27 mm), obtained from EWOS innovation (Lonningdal,Norway), were sampled for fast muscle, which was dissected from the dorsal epaxial myotome and flash frozen in liquid nitrogen. For total RNA extraction,100 mg of muscle was added to FastRNA Pro Green Beads (MP Biomedicals,Stretton, Cheshire, UK) with 1 ml of Tri Reagent (Sigma, Gillingham, Dorset,UK) and then homogenised with a Fast Prep instrument (MP Biomedicals). Genomic DNA was removed from the RNA sample using the TURBO DNA-free kit (Ambion,Huntingdon, Cambs, UK). RNA quality was confirmed by assessing the integrity of 28S and 18S ribosomal RNA by gel electrophoresis. Digested RNA was quantified using the fluorescent nucleic acid dye Ribogreen (Invitrogen,Paisley, Scotland, UK). First strand cDNA was synthesised using 1 μg of total RNA and a RETROscript kit (Ambion). Genomic DNA was extracted from 50 mg of spleen tissue (Dneasy Tissue Kit, Qiagen, Crawley, W. Sussex, UK). The primers shown in Table 1 were then used to amplify Atlantic salmon full coding sequences of smyoG,smyf5 and s-smlc1, and a partial sMRF4 sequence, using several standard PCR reactions with gDNA (MRFs) and cDNA (MRFs and s-smlc1). Additionally, to obtain the 3′ of the sMRF4gene (plus full coding sequence), a BD Smart™ RACE cDNA amplification kit was used (BD Biosciences, Oxford, Oxon, UK) (primer in Table 1). PCR products were separated using, and isolated from 1.1% (m/v) agarose gels, purified using a QIAquick gel extraction kit (Qiagen) and then ligated into a pCR4-TOPO T/A vector (Invitrogen) before transformation into chemically competent Escherichia coli cells (Invitrogen). At least two clones per gene fragment were then sequenced in sense/antisense directions by the University of Dundee sequencing service.

A consensus nucleotide and amino acid (AA) translation of each gene was constructed from each sequencing result. The identity of putative genes was confirmed against the complete non-redundant NCBI database using BLAST and TBLASTN searches(http://www.ncbi.nlm.nih.gov/blast/). Subsequently, each gene was submitted to the GenBank public database(http://www.ncbi.nlm.nih.gov). The intron–exon structure of each MRF was assessed by aligning cDNA and gDNA sequences in the program Spidey(http://www.ncbi.nlm.nih.gov/IEB/Research/Ostell/Spidey/). For sequence alignments, Clustal X(Thompson et al., 1997) was used with the Gonnet 250 matrix for pairwise comparisons, and Gonnet series parameter for alignments. Amino acid translations of the following mRNA sequences were aligned: tmyoD1a (X75798), tmyoD1b (Z46924), smyoD1a (AJ618978), smyoD1b (AJ557150), smyoD1c(DQ317527), btmyoD1c (DQ366710), smyf5 (DQ452070), tmyf5 (AY751283), sMRF4 (DQ479952), smyoG(DQ294029), tmyoG (Z46912) and amphi-myoD1(AB092415). Maximum likelihood was then performed on this alignment using PHMYL (Guindon and Gascuel,2003) and the WAG model(Whelan and Goldman, 2001),with 500 pseudoreplicate bootstraps. For comparison, a Neighbour Joining (NJ)analysis was performed on the same alignment in Mega 3.1(Kumar et al., 2004) using the JTT model and 1000 bootstrap iterations for branch support. Trees produced by both methods were reconstructed in Mega 3.1

To make DNA templates for RNA probe synthesis, PCR was used with T3/T7 primers (Invitrogen) and as a template, a pCR4-TOPO T/A plasmid (Invitrogen)containing the cDNA products of smyf5, smyoG, s-smlc1 and sMRF4 (Table 1)excluding the sMRF4 RACE product. The smyoD1a/1b/1c probe templates were as described previously (Macqueen and Johnston,2006). Finally, nucleotides 502–1119 of the Atlantic salmon Pax7 gene previously reported(Gotensparre et al., 2006)were amplified, cloned and sequenced as described above (primers in Table 1).

Each cRNA probe was synthesised in sense and antisense directions using T3/T7 RNA polymerases (Roche, Lewes, E. Sussex, UK) with concurrent incorporation of digoxigenin (DIG) or fluorescein (FLU) labelling (both Roche). In situ hybridisation was based on a standard procedure(Jowett, 2001) and all hybridisation and stringency washes were performed at 70°C. Probes were detected with alkaline-phosphotase-conjugated antibodies (Roche) using NBT/BCIP (Roche) for DIG and Fast Red (Invitrogen) for FLU. Different temperature treatments were incubated in each solution for identical time periods. This ensured that differences recorded between temperature groups in the colour development step were attributable to differences in gene expression rather than unequal sample treatment.

All embryos from each temperature treatment and stage were studied using both a DMRB compound microscope and a Leica MZ7.5 binocular microscope (Leica Microsystems Ltd., Milton Keynes, Bucks, UK). When DIC optics was used,embryos were flat-mounted with a coverslip on a clear microscope slide and orientated to a dorsal or lateral perspective. Embryos were staged by counting the somite number and photographs were recorded on a Nikon P4500 camera. Subsequently, representative embryos were mounted in cryomatrix (Thermo Electron Corp., Waltham, MA, USA), orientated and then frozen in isopentane cooled to its freezing point (–159°C) by liquid nitrogen. Serial 18μm cryosections (–20°C) were cut on a Leica cryostat (Leica Microsystems, CM1850). Differences in gene expression patterns between temperature treatments were considered reliable when replicated in each embryo at each stage (N=6). When figures were constructed, representative images of embryos from equivalent somite stages were selected from each temperature treatment. This meant that differences in temperature groups were not considered in developmental windows when embryos could not be accurately staged, i.e. prior to somite formation and after the completion of segmentation.

In salmonids, myoD has a complex evolutionary history and is represented by three paralogues (myoD1a, 1b and 1c)(Macqueen and Johnston, 2006). In the current study, we have also obtained full coding sequences of other Atlantic salmon myoD related genes: myf5 (smyf5), myoG (smyoG) and MRF4 (sMRF4). Using primers designed from tmyf5, a complete coding sequence (cds) of smyf5 (DQ452070) was obtained incorporating 720 bp that translated into a protein of 239 AA. The percentage identity conserved between smyf5 and other vertebrate myf5 orthologues at the respective nucleotide/protein level was 97.9/96.2% with rainbow trout(AY751283), 75.9/73.0% with pufferfish Takifugu rubripes(NM_001032770), 71.7/76.3% with zebrafish Danio rerio (AF253470),59.2/54.8% with frog Xenopus leavis (AJ579311), 60.7/56.3% with chicken Gallus gallus (NM_001030363) and 63.4/54.5% with human Homo sapiens (NP_005593).

Using primers designed from tmyoG (Z46912), a complete cds corresponding to smyoG (DQ294029) was obtained, which was 789 bp long, and translated into an open reading frame (ORF) of 254 AA. The percentage identity conserved between smyoG and other vertebrate myoG orthologues at the respective nucleotide/protein level was 97.9/98.4% with rainbow trout, 76.2/77.6% with pufferfish (AY566282),72.7/73.5% with zebrafish (NM_131006), 61.9/58.1% with frog (NM_001016725),64.1/56.4% with chicken (D90157), and 65.4/53.2% with human (NM_002479).

MRF4 had not previously been cloned in any salmonid fish. For this reason, primers used to amplify MRF4 were initially based on an expressed sequence tag (DN165140) obtained from a TBLASTN search of the salmon genome project(http://www.salmongenome.no/cgi-bin/sgp.cgi#Blast)using the translated D. rerio MRF4 mRNA (NM_001003982) as a probe. A reverse primer was designed from this sequence and was used with a forward primer designed in the start region of MRF4 based on alignments with several vertebrate MRF4 sequences(Table 1) to amplify nucleotides 1–649 of the salmon coding sequence. Finally, a 3′RACE primer was designed at the 3′ of the confirmed sMRF4sequence (Table 1) and this was used in a 3′ RACE reaction to obtain a whole coding sequence for sMRF4 and a complete 3′ untranslated region, with a poly-A tail and one polyadenylation signal (AATAAA) (not shown). SMRF4 shared closest homology to its orthologue in the knifefish Sternopygus macrurus (DQ059552) with 75.0/76.9% nucleotide/AA identity. The percentage identity between sMRF4 and other vertebrate orthologues at the coding nucleotide/protein level was 70.9/71.6% with pufferfish (AY445320),73.9/76.0% with zebrafish, 63.8/59.3% with frog (S84990), 62.8/60.9% with chicken (D10599), and 62.7/62.1% with human (NM_002469).

Fig. 1 shows an alignment of all the known salmonid MRFs with an ancient myoD homologue in the cephalochordate Branchiostoma belcheri. The bHLH domain and cis–his-rich region (just N-terminal to the basic region) are strongly conserved in all salmonid MRFs and with the ancient myoD gene. Additionally,the helix-III domain of myoD1 paralogues (AAs 206–221 of smyoD1a) is most similar to cephalochordate myoD (5/15 substitutions vs smyoD1a) >salmonid-myf5 genes (6/15 substitutions vs smyoD1a)>sMRF4 (8/15 substitutions vs smyoD1a) >salmonid myoG genes(10/15 substitutions vs myoD1a). Additionally a highly conserved motif is present in salmonid myoD paralogues and other vertebrate myoD proteins (not shown), that is not conserved in other MRFs, but partially conserved in amphi-myoD1 (Fig. 1). This motif has not been assigned any function at present. The NH₂- and COOH-terminals are the least conserved regions of the salmonid MRF proteins.

Primers to amplify a complete coding sequence of Atlantic salmon smlc1 (s-smlc1) were designed from the rainbow trout sequence previously reported [EST (BX076946)(Chauvigne et al., 2005)]. The coding sequence of s-smlc1 (DQ916288) was 561 bp that translated into an ORF of 185 AA. The percentage identity conserved between s-smlc1and other vertebrate smlc1 orthologues at the respective nucleotide/protein level is 99.1/99.5% with rainbow trout, 78/81% with the pufferfish Tetraodon nigroviridis (putative: predicted within CAAE01014556), 80/83% with zebrafish (NP_956810), 65/67% with frog (EST:AAI28964), 66/69% with chicken (P02606) and 64/67% with human (NP_002467).

The exon–intron structure of all known Atlantic salmon MRFs is presented in Fig. 2. Common to all vertebrate MRFs, each salmonid myoD family gene is represented as three exons and two introns. For each gene, exon 1 is the largest,incorporating the NH₂-terminal activation domain, basic and HLH motifs, and in vertebrate myoD genes, a highly conserved region that has no assigned function currently. Exon 2 is the smallest for each MRF, and exon 3 incorporates the helix-III domain.

A maximum likelihood analysis was performed on an alignment of all known salmonid myoD family members (Fig. 3) using amphi-myoD1 as an outgroup. A NJ analysis was performed on the same alignment, producing a tree entirely consistent with Fig. 3 (not shown). Salmonid MRFs cluster distinctly into four groups, representing the four myoDfamily members. The three myoD1 proteins clustered together, representing their close paralogy. This is also reflected in their highly conserved genomic organisation (Fig. 2). As expected, smyoD1/smyf5 and smyoG/sMRF4 separately branch from amphioxus myoD(100% and 93% bootstrap support). This is consistent with the established evolutionary scenario where an ancestor myoD homologue (represented in sub-vertebrate taxa as a single gene) duplicated twice to produce the ancestor lineages to myoD/myf5 and myoG/MRF4 and then the individual MRFs(Atchley et al., 1994).

We have recorded the mRNA expression patterns of six MRF genes throughout salmon embryogenesis. To place the expression of each MRF in the context of known muscle fibre differentiation events we also studied the expression of s-smlc1, which is expressed in rainbow trout adaxial cells as they differentiate (Chauvigne et al.,2005), and Pax7, which in zebrafish is expressed in the myogenic precursors of the external layer(Stellabotte et al., 2007; Hollway et al., 2007).

To simplify the expression data we present our findings in two formats. Firstly, a schematic diagram shows the progressive expression of each gene in the most anterior somite of salmon embryos during segmentation and post-segmentation stages of embryogenesis(Fig. 4). This excludes the complexity generated when considering the embryos rostral–caudal axis and associated gradient in expression patterns due to changing somite maturity. Fig. 4 enables the reader to quickly establish the spatio-temporal correlations between the expression patterns of the six MRFs with s-smlc1 and Pax7 in a single maturing somite. Next, we produced a detailed inventory of expression images for smyoD1a, smyf5, smyoG, sMRF4 and s-smlc1 from the 30–45 ss (Fig. 5) and these genes plus Pax7 at the end of segmentation and in a post-segmentation stage (Fig. 6). Only one myoD1 paralogue (1a) is represented in this figure considering the recent detailed description of the expression of smyoD1a/1b/1c(Macqueen and Johnston, 2006). Each of the descriptions has been written for independent use, and thus a degree of overlap exists between them.

The first somite (So-1) is the oldest at any given stage of development and is used here to describe the progression in MRF, s-smlc1 and Pax7 expression during its maturation. When So-1 arose from the unsegmented mesoderm, smyoD1a and smyf5 were, respectively,expressed in the adaxial myoblasts flanking the notochord (A1) and throughout its entire lateral width, excluding the most anterior quarter (B1). Smyf5 was not expressed in So-1 adaxial cells during any period of embryogenesis (B1–6) in contrast to other teleosts studied to date (e.g. Coutelle et al., 2001; Cole et al., 2004). At the point when there were around 20 newer somites caudal to So-1, smyoD1aand smyf5 did not change significantly (A2 and B2). However at this time three other MRFs were turned on in So-1. SmyoD1b expression was similar to smyf5, extending through the entire posterior domain of So-1, including the undifferentiated adaxial cells (C1 vs B1). SmyoD1c had a comparable expression pattern to smyoD1a and 1b: transcripts were detected in the adaxial myoblasts (D2 vs A2) and diffusely in the posterior-lateral region of So-1,although less extensively than smyoD1b/smyf5 (D2 vsC2, B2). At this time, smyoG was also expressed across So-1 but it expression extended further anteriorally than smyoD1b or smyf5 (E2 vs C2/B2).

The first expression of sMRF4 was present in So-1 at the 25 ss in the adaxial cells adjacent to the notochord (F3), just before an identical expression field was recorded for s-smlc1 at the 30 ss marking the onset of adaxial cell differentiation (G3). At this time smyf5expression was downregulated in So-1 (B3), whilst smyoD1b and smyoG extended anteriorally (C3 and E3), coinciding with the first differentiation of fast muscle fibres. In contrast smyoD1a/1c expression spread laterally whilst maintaining a signal to the medial s-smlc1 expressing adaxial cells (A3 and D3). This phase of expression preceded the lateral migration of the adaxial cells that was marked by an inward facing triangular wave of s-smlc1expression throughout the middle of So-1 at the 45 ss (G4). At this point smyoD1a/1c and sMRF4 expression was comparable, but not identical to s-smlc1 and was no longer present in the medial myotome of So-1 (A4, D4 and F4). In contrast, smyoD1b and smyoG were detected in the entire length and width of So-1 at this time (C4 and G4). Additionally, smyf5 expression re-accumulated at the superficial edge of the posterior region of So-1, before the completion of adaxial cell migration (B4).

At the end of segmentation, So-1 had fully acquired the chevron-shaped phenotype, and the adaxial cells had spanned the myotome to form a single layer of slow-fibres, indicated by s-smlc1 expression (G5). Pax7 expression was present external to this layer, presumably marking myogenic progenitors of the external cell layer (H5). At the end of segmentation, smyoD1a, smyoD1c, smyoG and sMRF4 were each expressed throughout the bulk of the myotome of So-1 (A5, D5, E5, F5). Conversely smyf5 expressed was limited to the lateral edge of the myotome, in the posterior domain of So-1 (B5). From the 45 ss-end of segmentation smyoD1b was rapidly downregulated in all but the superficial region of the So-1 myotome (C5).

As So-1 matured further, s-smlc1 and Pax7 expression respectively remained in the single-slow layer and external cell layer (G6 and H6). At this time, each MRF was expressed most strongly in superficial regions of the So-1 myotome, particularly in dorsal and ventral regions and at the level of the horizontal septum (A6-E6). Smyf5 staining was still restricted to the posterior region of So-1, faintly along the whole superficial edge of the myotome, and more strongly in the dorsal-ventral-zones(B6). SmyoD1b expression was very similar to smyf5 in any cross-section, although the staining was present throughout the length of So-1(C6). MyoD1a/1c, sMRF4 and smyoG expression was not restricted to dorsal-ventral regions and each was also present in the deeper fast muscle fibres (A6, D6, E6, F6), although smyoG expression was comparatively fainter in ventral regions of the myotome (E6).

Several MRFs were expressed in the adaxial myoblasts before s-smlc1. SmyoD1a was expressed in a bilateral strip flanking the nascent notochord of some pre-somitic embryos, although more often in adaxial progenitors of the presomitic mesoderm (PSM)/somites from the 0–10 ss and then maintained here in the newest somites/PSM throughout segmentation(Fig. 5Bii,Dii, Fig. 6Bii). Smyf5 was expressed before or contemporaneously to smyoD1a, in two triangular fields of the PSM either side of the notochord, but did not colocalise with smyoD1a in pre-somitic adaxial myoblasts at this stage (not shown). During segmentation, smyf5 was expressed throughout the mid-posterior of the newest somites, and in the anterior PSM, displaying a pattern of interspersed strong and reduced signal where the newest two somites arose(Fig. 5Bi,Di). Expression continued moving down the tailbud, terminating adjacent to the notochord's end(Fig. 5Bi,Di), but unlike other teleosts (e.g. Coutelle et al.,2001; Cole et al.,2004), smyf5 was not expressed in the adaxial myoblasts of the anterior PSM or caudal somites (Fig. 5Bi,Di) until the end of segmentation when a residual PSM remained(Fig. 6Bi).

As somitogenesis progressed, other MRFs were expressed in adaxial myoblasts of somites, but not the PSM as for smyoD1a. At the ∼20 ss smyoD1c colocalised with smyoD1a in somite adaxial cells(see Macqueen and Johnston,2006), but also with smyf5/smyoD1b in the posterior domain of the newest somites (not shown). As somites matured, smyoD1b spread anteriorally to encompass the whole myotome (not shown) whereas smyf5 was initially downregulated and barely detected in the rostral somites at the 30 ss (Fig. 5Ai). SmyoG mRNA was also detected at the 20 ss and was present in the adaxial myoblasts of the final few caudal somites(Fig. 5Biii,Diii, Fig. 6Biii), before rapidly spreading to encompass the whole myotome of more anterior somites(Fig. 5Aiii,Ciii, Fig. 6Biii). The final myoD family member detected before the expression of s-smlc1 was sMRF4 at ∼25 ss, in a faint transient wave of rostral–caudal expression in adaxial cells (30 ss stage, mid-somites shown: Fig. 5Biv).

S-smlc1 marks the differentiation of adaxial cells to slow muscle myocytes, which started in the rostral somites of ∼30 ss embryos and progressed in a caudal direction as newer somites matured(Fig. 5Av). The progression of s-smlc1 expression could be correlated with that of some myoD family members, whereas others seemed independent. For example,at the 25 ss, sMRF4, expression was present in adaxial cells of the rostral somites, immediately before s-smlc1 expression at the 30 ss(not shown and Fig. 5Av) and similarly progressed caudally. However, the rostral–caudal progression of sMRF4 was initially transient, disappearing in more rostral somites as it accumulated in newer somites. The timing of sMRF4preceded s-smlc1, so at the 30 ss, s-smlc1 was maintained in rostral somites (Fig. 5Av), sMRF4 had been downregulated at this site(Fig. 5Aiv), but was expressed in the mid-caudal somites (Fig. 5Biv), prior to s-smlc1 expression here(Fig. 5Bv). In the rostral somites at the 30 ss, smyoD1a/1c transcripts had spread laterally away from the medial somite but this domain still overlapped with s-smlc1 expression in differentiating medial adaxial cells(smyoD1a shown: Fig. 5Aii). As somites matured, the adaxial cells migrated laterally,indicated by a wave of s-smlc1 transcripts in the rostral somites of 45 ss embryos (Fig. 5Cv). By the end of segmentation, this migration was occurring from around the tenth most caudal somite (Fig. 6Bv)and was completed in the rostral somites(Fig. 6Av). During adaxial cell migration, smyoD1a/1c and sMRF4 transcripts moved away from the notochord and at the 45 ss, mRNA of each gene was present in a broad v-shaped domain similar to s-smlc1 expression (e.g. smyoD1a: Fig. 5Cii, sMRF4: Fig. 5Civ). During this time, each of these MRFs remained in the adaxial myoblasts of the caudal somites, co-expressed with smyoG, before s-smlc1expression (e.g. Fig. 5Dii,Diii,Div,Dv, Fig. 6Bii,Biii,Biv).

In contrast to the 30 ss, where smyf5 was downregulated in maturing somites (Fig. 5Ai), by the 45 ss, smyf5 had accumulated in the rear quarter of the rostral somites at the superficial myotome, before the adaxial cells had completed their migration (Fig. 5Ci,Cv). This pattern was maintained, so that at the end of segmentation (65 ss), smyf5 was expressed along the entire outer edge of the myotome at the rear border of the rostral-mid somites(Fig. 6Ai). Smyf5transcripts were present at this site before the adaxial cells had completed migrating, making it unlikely that this domain was limited to the slow layer. Instead, we suggest that smyf5 expression marked the earliest production of muscle fibres sourced from the external cell layer. In support of this, at this time Pax7 was clearly expressed specifically in the external cell layer of the rostral somites outside of the single slow layer(Fig. 6Avi). In more caudal somites, where smyf5 had not reached the myotome border(Fig. 6Bi), Pax7 was distributed throughout the somite, and particularly strongly at the anterior border (Fig. 6Bvi). Thus, the migration of Pax7 mRNA to a position external to the myotome occurred at a similar time as the restriction of smyf5 mRNA to the posterior border of the myotome.

The expression domains of smyoG and smyoD1b from the 30–65 ss also suggest a role for these transcription factors that is independent of adaxial cell migration. For example, both genes were unchangingly present across the width/length of the myotome in all but the most caudal somites at the 20–65 ss, irrespective of the migration state of adaxial cells [smyoG shown: Fig. 5Aiii,Ciii, Fig. 6Aiii, see also myoD1b (fig. 3B,F in Macqueen and Johnston, 2006)]. Additionally, the extension of smyoD1b/smyoG transcripts occurred in an anterior direction during somite maturation; adaxial cells migrated laterally.

After segmentation, when the eyes became pigmented, the fin buds lengthened and all somites developed the chevron shape (the eyed stage), s-smlc1expression was present as a single superficial layer of slow-twitch fibres in rostral-mid (Fig. 6Cv) but not caudal somites (e.g. Fig. 6Dv)and Pax7 was expressed in the external cell layer and dorsal spinal cord along the embryos rostral–caudal axis(Fig. 6Cvi,Dvi). At this time, smyf5-expressing cells were present in the rear portion of all somites, mainly in the dorsal and ventral superficial fast myotome, adjacent to the horizontal septum and more faintly adjacent to the single slow muscle layer (Fig. 6Ci,Di). SmyoD1b was also expressed in similar regions at the superficial myotome, but was maintained along each somites length (not shown). Conversely, smyoD1a/1c and sMRF4 transcripts were detected to a greater or lesser extent throughout the entire myotome (as in Fig. 6Aii,Aiv). As embryos (and somites) matured further, staining for these genes was reduced in the medial myotome but increasingly maintained in more superficial regions of the myotome, particularly in dorsal/ventral regions (smyoD1a: Fig. 6Cii,Dii; sMRF4:Civ,Div). Similarly, smyoG expression was present to a greater or lesser extent throughout the myotome, but as embryos matured, expression was reduced in the medial myotome but maintained at the dorsal (and faintly at the ventral) edge of the myotome and adjacent to the horizontal myoseptum(Fig. 6Ciii,Diii).

Fig. 7 shows the relationship between the rate of somitogenesis and embryonic temperature. Segmentation proceeded from around 750–1700, 425–960 and 250–600 h post-fertilization (h.p.f.) at 2°C, 5°C and 8°C,respectively. A first order linear regression was fitted to data of developmental time (h.p.f.) versus somite number during the linear phase of somitogenesis, which occurs from the 0 ss until the last few somites are added as segmentation is completed(Gorodilov, 1996). Using the regression equation from each plot, it was calculated that somitogenesis proceeded at a respective rate of one somite added each 15 h, 8 h and 5 h at 2°C, 5°C and 8°C.

The expression of smyoD1a, smyoG, smyf5, sMRF4 and s-smlc1 was investigated at three embryonic temperatures (2°C,5°C and 8°C). SmyoD1a and smyoG expression showed no variation between temperature treatments for corresponding somite stages (not shown). In contrast, at several equivalent somite stages, replicated differences (in six embryos per stage) were recorded in the mRNA expression profiles of smyf5, sMRF4 and s-smlc1 with respect to somite stage. The expression pattern of each gene at 5°C was approximately intermediate between that observed at 2°C and 8°C (not shown). In situ hybridisation cannot be used as a quantitative tool for comparative analysis and therefore we only highlight cases in which differences in staining intensity bordered on the presence or absence of transcripts in all embryos examined.

At the 30 ss and 45 ss, smyf5 staining was intense in the newly formed caudal somites, presomitic mesoderm and tailbud at 8°C, but faint at 2°C (45 ss shown, Fig. 8A). In embryos approaching the end of segmentation (with ∼63 out of 65 somites), smyf5 staining had reached somite number 58 at 8°C, but was almost absent from somites 59–63(Fig. 8B). In contrast, at 2°C, an smyf5 mRNA signal was detected in somites 58–63 and within the residual presomitic mesoderm(Fig. 8B, see arrows in corresponding transverse sections). We interpret these results to show that smyf5 expression was retarded with respect to somite stage at 2°C, with staining in the caudal somites and PSM peaking and subsequently retracting earlier at 8°C compared to lower temperatures.

In somites 30–45 of 45 ss embryos, sMRF4 transcripts were detected in the medial somite at both temperatures (not shown), but as somites matured staining was more advanced at 8°C. For example, in somites 20–25, sMRF4 transcripts were starting to migrate laterally away from the notochord at 8°C but not 2°C(Fig. 9B). Furthermore, sMRF4 staining had advanced into somites 1–15 at 8°C, but not 2°C (Fig. 9A, see arrow on transverse sections). Towards the end of segmentation, while the most caudal somites (53–63) had sMRF4 transcripts in adaxial cells at both temperatures (Fig. 9D),in more rostral somites (numbers 43–50) the medial compartment showed a strong sMRF4 signal at 8°C, but was virtually unstained at 2°C (Fig. 9C, see arrowheads on transverse sections). These results indicate that the wave of sMRF4 expression in maturing somites was retarded with respect to somite stage at lower temperatures.

As segmentation reached completion, the most newly formed somite with s-smlc1 expression in the adaxial cells at 2°C was number 52–53, compared to 56–57 at 8°C(Fig. 10A). Thus at an equivalent somite stage, s-smlc1 expression was delayed by 4–5 somites at 2°C (illustrated by blue arrowhead in Fig. 10A: also see arrowhead on transverse sections through equivalent somite number of 2 and 8°C embryos). In more rostral somites, a clear wave of s-smlc1transcripts could be seen migrating laterally away form the notochord between s43-s48 at 8 but not 2°C (not shown). In rostral somites (numbers 1–20) an s-smlc1 signal was detected in the superficial slow layer at 8°C, but not 2°C (not shown). These results suggest that like sMRF4, s-smlc1 expression was strongly retarded at 2°C compared to 8°C.

We studied the coordinated expression of six Atlantic salmon MRFs during embryonic myogenesis. In mice, myoD and myf5 have mainly redundant roles in myogenic specification (Rudnicki et al.,1993). In the zebrafish, myoD and myf5 are the first MRFs expressed in adaxial myoblasts(Weinberg et al., 1996; Coutelle et al., 2001) and while the morpholino-knockdown of either gene has no effect on slow muscle formation, when both are nulled, slow muscle formation is ablated(Hammond et al., 2007). In salmon, myf5 and myoD1a are the first MRFs to be expressed in myogenic precursor cells of the segmental plate followed by myoGas differentiated muscle is formed. Although the order that MRF transcripts appear in salmon myotomes is similar to that described in zebrafish(Coutelle et al., 2001; Weinberg et al., 1996) there are some notable differences in expression patterns, which are probably related to the tetraploid nature of the salmonid genome(Allendorf and Thoorgard,1984). The lineage leading to modern salmonids has undergone two whole genome duplications relative to the common tetrapod ancestor(Jaillon et al., 2004; Allendorf and Thoorgard, 1984). The salmonid-specific genome duplication is thought to have occurred 10–25 million years ago (Allendorf and Thoorgard, 1984). Approximately 50% of the duplicated genes have subsequently been lost from the genome and are represented by a single paralogue (Bailey et al.,1978). For example, only one myoG gene has been described in zebrafish (Weinberg et al.,1996), common carp Cyprinus carpio(Cole et al., 2004), rainbow trout (Delalande and Rescan,1999) and Atlantic salmon (present study). The highly conserved expression pattern of myoG during embryonic myogenesis in these species suggests that myoG is retained as a single gene in salmonids. In other cases,duplicated genes have been retained. For example, salmonid fish have three myoD paralogues (myoD1a/1b/1c), which, based on phylogenetic evidence (see Fig. 3), are thought to have arisen from a single gene orthologous to zebrafish myoD1 via a whole genome and subsequent local duplication(Macqueen and Johnston, 2006). During the segmentation period, smyoD1a and smyoD1c are sequentially expressed in adaxial cells and their progenitors as they differentiate in overlapping domains whereas smyoD1b is expressed exclusively in the lateral somite. The expression patterns of smyoD1a/1c and smyoD1b correspond to two embryonic phases of expression of the single myoD1 gene in zebrafish in the adaxial cells and posterior-lateral somite(Weinberg et al., 1996). These expression domains mark the progenitors of the embryonic slow and fast muscle fibres (Devoto et al., 1996),respectively, which arise earlier than and are distinct from those muscle progenitors arising from the anterior somite(Hollway et al., 2007; Stellabotte et al., 2007). The myogenic precursors of the posterior somite and the adaxial cells are respectively regulated by hedgehog and fgf8 signals(Barresi et al., 2000; Groves et al., 2005). Our working hypothesis is that the role of the teleost myoD1 gene was sub-functionalised in Atlantic salmon according to the model of Force et al.(Force et al., 1999), with each paralogue regulated by a sub-set of the cis-acting elements found in the promoter region(s) of the single myoD1 gene(Macqueen and Johnston, 2006). A second myoD gene, myoD2, has been retained in five percomorphic teleosts (Tan and Du, 2002; Galloway et al., 2006; Macqueen and Johnston, 2006; Fernandes et al., 2007). To date myoD2 paralogues have not been identified in salmonids or zebrafish. Since hundreds of skeletal muscle genes are regulated by myoD(Bean et al., 2005), the presence of multiple paralogues may provide additional levels of control and complexity of expression patterns providing some selective advantage leading to their retention in the genome.

In Atlantic salmon, myf5 was expressed in the posterior domain of recently formed somites, the anterior PSM and tailbud(Fig. 5Bi,Di), in a similar pattern to that described in zebrafish(Coutelle et al., 2001) and common carp (Cole et al.,2004). However, in contrast with the other teleosts studied, smyf5 was not expressed in adaxial cells during most of segmentation(Fig. 5Bi,Di), until a small residual PSM was present at the 65 ss. This finding is consistent with the presence of two possible myf5 paralogues. To examine this further, we designed primers in conserved regions of smyf5 to amplify intron 2. A single band was obtained by PCR using a gDNA template and, despite multiple sequencing, the identical sequence was represented in all clones. Furthermore,a single myf5 orthologue was retrieved when BLAST searches were performed at the salmon genome project(http://www.salmongenome.no/cgi-bin/blast.cgi),TGI (Atlantic salmon/rainbow trout databases at: http://compbio.dfci.harvard.edu/tgi/)and GRASP(http://web.uvic.ca/cbr/grasp/)databases using smyf5 as a probe. An interesting alternative possibility is that the two myf5 genes produced during the tetraploidisation of the salmonid genome became sub-functionalised before one paralogue (expressed in adaxial myoblasts) was lost, perhaps because of the abundance of transcribed myoD1 paralogues in adaxial cells(Macqueen and Johnston, 2006)and known redundancy of myf5/myoD proteins in myogenic specification in vertebrates (Rudnicki et al.,1993; Hammond et al.,2007). If a second Atlantic salmon myf5 paralogue does not fulfil the known role of the single zebrafish myf5 orthologue in adaxial cell specification (Coutelle et al., 2001), then slow muscle development in salmonids is likely to vary significantly to other teleosts. A morpholino-based knock-down of individual salmonid MRFs would be informative in this respect.

No previous MRF4 expression pattern has been described in fish embryos for comparison with our results and nothing is currently known about its regulation. In wild-type mouse embryos, MRF4 is the third myoD family member to be expressed within the hypaxial region of each thoracic somite whilst it is also expressed contemporaneously with myf5 in the undifferentiated dermomyotome (Summerbell et al., 2002). When MRF4 expression was not compromised in myoD/myf5 double null mice, normal myogenesis occurred,indicating that MRF4 can substitute for myf5/myoD in initiating muscle growth(Kassar-Duchossoy et al.,2004). It is unknown whether this dual role of MRF4 also occurs in teleost fish or is specific to mammals. In Atlantic salmon, MRF4 is first expressed in a transient rostral to caudal wave in somitic adaxial progenitors just prior to their differentiation and the expression of s-smlc1,suggesting it acts downstream of smyoD1a. It is interesting to note that that the Helix-III of sMRF4 is more distinct from smyoD1a (8/15 substitutions: Fig. 1), than a comparable alignment of mouse MRF4 vs myoD [6/15 substitutions in MRF4 compared to AA 245–258 of myoD (see Bergstrom and Tapscott, 2001)]. Substituting the helix-III of mouse myoD with the equivalent MRF4 region resulted in a chimera that efficiently activated endogenous muscle-specific genes(Bergstrom and Tapscott, 2001). The equivalent region of mouse myoG (with 8/15 substitutions, i.e. the same as sMRF4) could not replace the original myoD motif. It is possible that the increased number of substitutions in the helix-III of sMRF4 compared to mammalian MRF4 and/or differences in regulatory elements that have arisen during evolution have resulted in a reduced potency for myogenic specification whilst maintaining its role in differentiation. It is currently unknown whether multiple MRF4 paralogues are conserved in salmonids. Again, in silico searches in the same resources as described above for myf5 lead us to retrieve MRF4 cDNA sequences likely originating from the single gene described here (Figs 1, 3).

We have shown that following the end of segmentation, each MRF is expressed in zones of new myotube production that occur at the lateral edge of the fast myotome (stratified hyperplasia), particularly in dorsal and ventral areas and adjacent to the horizontal myoseptum. Smyf5 was present at the superficial edge of the myotome in rostral somites from the 45 ss, initially prior to the completion of adaxial cell migration and was thus independent of the first wave of slow muscle differentiation. It is possible that smyf5 marked the onset of stratified hyperplasia, which began at a similar stage of development in the closely related salmonid, S. trutta, evidenced by myoD/myoG expression(Steinbacher et al., 2007). SmyoD1a/1c, sMRF4 and smyoG expression in the bulk of the myotome was reduced from the end of segmentation onwards, but maintained (or upregulated) at the lateral edge of the fast myotome at either the dorsal and/or ventral extremes and/or adjacent to the horizontal myoseptum. The source of additional embryonic fast muscle fibres is likely to be the external cell layer, which is marked by Pax7 expression(Fig. 4H5,H6 and Fig. 6Avi,Cvi,Dvi) (Hollway,2007; Stellabotte et al.,2007; Steinbacher et al.,2007; Devoto et al.,2006).

We have shown that altering egg incubation temperature produces heterochronies in the expression of some myogenic regulatory factors but not others. Thus, whereas smyoD1a and smyoG expression showed no consistent differences with temperature with respect to developmental-stage,the expression of sMRF4 and smyf5 and the slow muscle differentiation marker s-smlc1 were retarded at 2°C compared to 8°C. Our finding that the relative timing of smyoD1a and smyoG expression was independent of temperature parallels observation in Atlantic cod (Hall et al.,2003), Atlantic herring(Temple et al., 2001), common carp (Cole et al., 2004) and Atlantic halibut (Galloway et al.,2006), but differs from the result reported in rainbow trout(Xie et al., 2001). As a consequence of the heterochronies in sMRF4 and smyf5expression, the ratio of the individual myoD family members at each developmental stage was a function of environmental temperature. It is known that the different MRF proteins vary in their intrinsic abilities to initiate myogenesis or promote muscle differentiation(Bergstrom and Tapscott, 2001; Ishibashi et al., 2005). For example, whilst myf5 and myoD targeted a similar array of genes involved in myogenic specification, myoD was markedly more efficient at inducing muscle differentiation genes (Ishibashi et al.,2005). Functional analysis in mouse has shown that myoD strongly upregulates capn2, a protease required for myoblast–myotube fusion,whereas myoG has a weak effect and myf5 no effect(Dedieu et al., 2003). Using a combination of genome-wide transcriptional factor binding and expression profiling in the mouse a total of 126 genes were identified that bound myoD(Blais et al., 2005). Many of these genes were transcription factors that propagate and amplify signals initiated by the MRFs (Blais et al.,2005). MyoD and myoG occupied 91 and 137 promoters in differentiating myotubes, indicating the MRFs recognise distinct, but overlapping, targets (Blais et al.,2005). Of particular interest was the finding that MRFs bind a set of genes involved in synapse specification and the function of the neuromuscular junction (Blais et al.,2005). In Atlantic herring, embryonic temperature has been shown to produce major changes to the timing of development of neuromuscular junctions in the myotomal and fin muscles(Johnston et al., 1997; Johnston et al., 2001). Herring were reared at 12°C and 5°C until just after hatching and then transferred to a common ambient temperature. The development of dorsal and anal fin ray muscles and their neuromuscular junctions occurred at shorter body lengths in the 12°C-group, resulting in improved fast-start swimming performance relative to the 5°C-group(Johnston et al., 2001).

Morpholino knock-down experiments of myoD and myf5 in the zebrafish resulted in an increase in the number of Pax3/7-expressing external cells on the lateral surface of the somite(Hammond et al., 2007). Since these cells are a source of fast muscle growth throughout post-embryonic zebrafish growth (Hollway et al.,2007; Stellabotte et al.,2007), heterochronies in MRF expression provide a potential mechanism that could explain some of the major changes in muscle phenotype that occur with variations in developmental temperature, including changes in muscle fibre number.

Since this paper was submitted the expression pattern of the zebrafish orthologue of MRF4 has been published(Hinits et al., 2007). This work showed that zebrafish MRF4 has a comparable expression pattern to sMRF4,being initially expressed in differentiated slow muscle precursors of somites after myoD1 (regulated by hedgehog signalling, and ablated by the morpholino antisense knockdown of myf5 and myoD1) and later in differentiated fast muscle fibres subsequent to the expression of myoD1 and myoG in the lateral somite.

This project was supported by EWOS innovation in conjunction with the Integrated Project SEAFOODplus granted by the European Union under contract No. 506359. D.J.M. is supported by a NERC studentship(NER/S/A/2004/12435).