FoxK1 splice variants show developmental stage-specific plasticity of expression with temperature in the tiger pufferfish

Myogenesis is a complex process involving the commitment of pluripotent stems cells to a myogenic fate, orchestrated by the MyoD family of basic helix-loop-helix transcription factors Myf5, MyoD (also known as Myod1; Mouse Genome Informatics) and MRF4 (also known as Myf6; Mouse Genome Informatics)(Kassar-Duchossoy et al.,2004; Rudnicki and Jaenisch,1995). Muscle progenitor cells (MPCs) represent a self-renewing population, the progeny of which undergo further cell division before exiting the cell cycle and entering a programme of terminal differentiation that involves the upregulation of the MyoD family members Myog and MRF4, leading to the expression of muscle-specific proteins(Andres and Walsh, 1996; Kassar-Duchossoy et al.,2004). In the mouse, proliferating myoblasts withdraw from the cell cycle and fuse to form multinucleated myotubes leading to primary and secondary muscle fibres in the embryonic and foetal stages, respectively(Kelly and Rubinstein, 1994). Although muscle fibre number changes little after birth(Rowe and Goldspink, 1969),the MPCs provide additional myonuclei as the muscle fibres in the growing animal increase in length and diameter(Moss and Leblond, 1971). MPCs are also involved in nuclear turnover in adult muscle and have a role in repair from injury and adaptive responses to exercise (reviewed by Chen and Goldhamer, 2003). The satellite cells found beneath the basal lamina of adult skeletal muscle represent myogenic progenitors that are mitotically quiescent(Schultz, 1996). Quiescent and activated MPCs express paired box protein 7 (Pax7)(Seale et al., 2000), the expression of which is downregulated once the cells express MRF4 and Myog and enter the programme of terminal differentiation(Chen and Goldhamer, 2003).

The FoxK1 protein is a member of the forkhead/winged helix family (Fox) of transcription factors, which constitute a diverse group that display a remarkable diversity and play crucial functions in several biological processes, including development and oncogenesis(Carlsson and Mahlapuu, 2002). All members of the Fox family are characterised by the presence of a forkhead(FH) domain, a 110-residue DNA binding region that consists of threeα-helices and three β-strands flanked by two wing-like loops,resulting in a three-dimensional structure that resembles the wings of a butterfly (Clark et al.,1993). The structure of FoxK1 is rather unusual amongst Fox proteins, in that one of the typical wings is replaced by an 8-residue C-terminal α-helix (Chuang et al.,2002). FoxK1 also contains a forkhead-associated (FHA) domain,which is a phosphopeptide recognition region(Durocher and Jackson, 2002). Since the identification of the homeotic gene forkhead in Drosophila (Weigel et al.,1989) more than 150 members of the Fox family have been found in taxa as diverse as yeast and mammals; this family is now divided into 17 subgroups designated A to Q (Kaestner et al., 2000). Relatively few members of the Fox family have been characterised with respect to their function and target genes. During mouse embryogenesis FoxK1 (also known as Foxk1; Mouse Genome Informatics) was detected transiently in the developing myotome, limb precursors, heart tube and certain regions of the brain (Garry et al., 1997). Knockout mice with a functional null allele at the FoxK1 locus showed impaired satellite cell function, which resulted in delayed and incomplete skeletal muscle regeneration following injury(Garry et al., 2000). The two alternatively spliced isoforms of FoxK1 (FoxK1-α and FoxK1-β) were found to have reciprocal expression patterns during muscle regeneration, indicating that they might exert opposite effects on their target genes (Garry et al.,2000). FoxK1-β expression predominates in quiescent satellite cells, whereas FoxK1-α is the main isoform expressed in proliferating myoblasts derived from activated satellite cells(Garry et al., 2000). FoxK1 is one of the few known markers of quiescent satellite cells in mammalian muscle (Garry et al.,1997).

In teleosts, at least three phases of myogenesis can be distinguished in fast myotomal muscle: an embryonic phase, stratified hyperplasia from distinct germinal zones and mosaic hyperplasia(Johnston, 2006). The majority of muscle fibres are formed by mosaic hyperplasia in the larval, juvenile and early adult stages involving the activation of MPCs throughout the myotome(Rowlerson and Veggetti,2001). The continuation of myotube production in adult fish reflects the large increase in body mass between the larvae and the final body size (Johnston, 2006). Teleosts are ectothermic and the outcome of the myogenic programme is profoundly affected by epigenetic factors, particularly embryonic temperature(Johnston and Hall, 2004). For example, in Atlantic herring (Clupea harengus L.) heterochronic shifts were observed with respect to the rostral to caudal progression of myofibril assembly and the outgrowth of primary motor neurons, which started at earlier somite stages as temperature was increased(Johnston et al., 1995). Embryonic temperature has been shown to influence the number of MPCs and muscle fibre recruitment patterns in herring(Johnston, 1993; Johnston et al., 1998),Atlantic salmon (Salmo salar L.)(Johnston et al., 2000a; Johnston et al., 2000b) and European sea bass (Dicentrarchus labrax L.)(Alami-Durante et al., 2007),indicating that temperature experienced during early development has a lasting influence on adult muscle phenotype. In Atlantic salmon, the maximum fibre number in seawater stages varied by up to 20% according to the temperature regime experienced during early development(Johnston et al., 2003).

The availability of a draft genome sequence of the tiger pufferfish(Takifugu rubripes Temminck and Schlegel)(Aparicio et al., 2002)provides an excellent opportunity for investigating the molecular basis of developmental plasticity of myogenesis in teleosts. The aim of the present study was to characterise the FoxK1 gene and its splice variants in T. rubripes and to test the hypothesis that its expression with respect to embryonic stage was a function of developmental temperature.

Two juvenile tiger puffer fish Takifugu rubripes Temminck and Schlegel, with an approximate mass of 160 g, were bred in captivity at the Fisheries Laboratory (University of Tokyo, Maisaka, Shizuoka Prefecture,Japan). Two wild-caught adult specimens, weighing ∼1.4 kg, were purchased from the local fish market at Maisaka (Shizuoka Prefecture, Japan). Fish were humanely killed according to the British Home Office guidelines by over-anaesthesia in a solution of 0.2 mmol l^–1 3-aminobenzoic acid ethyl ester (Sigma, Gillingham, Dorset, UK) buffered with sodium bicarbonate (Sigma). Samples of fast and slow myotomal muscle, heart, liver,skin, brain and gonads (adult fish only) were dissected and stored in RNAlater(Ambion/Applied Biosystems, Warrington, Lancashire, UK) for subsequent RNA extraction. T. rubripes eggs were purchased from a commercial source(Nisshin Marinetech Co., Yokohama, Japan). The eggs from a single female were fertilised at 17°C using the sperm of two males and, after approximately 4 h, they were transferred to the Fisheries Laboratory. Embryos were split into three temperature groups and incubated at either 15°C, 18°C or 21°C (within ±0.5°C). After hatching, the temperature of all tanks was gradually increased to 18°C and larvae were reared for approximately 2 months. Samples of embryos and larvae collected throughout development were preserved in RNA later.

Tissue samples and eggs (0.1 g) were placed in 1 ml Tri reagent (Sigma) and homogenised with FastRNA Pro Green beads (Qbiogene Inc., Cambridge, UK) using the FastPrep Instrument (Qbiogene) for 40 s at a speed setting of 6.0. Total RNA was isolated according to the manufacturer's protocol. Following DNase treatment (Turbo DNA-free; Ambion) to remove any potential genomic DNA contamination, RNA quality was verified by electrophoresis on a 1% (m/v)agarose (Bioline, London, UK) gel under denaturing conditions. Total RNA was then quantified with the fluorescent nucleic acid stain RiboGreen (Molecular Probes/Invitrogen, Paisley, UK), according to the instructions provided by the manufacturer. First-strand cDNA was synthesised from 1 μg of total RNA using a RETROscript kit (Ambion), according to the recommended method. A 1:1 mixture of random decamers and oligo(dT)₁₈ was used as first-strand primers for cDNA synthesis. Following denaturation of the RNA by incubation at 85°C, the reverse transcription was performed at 50°C. A negative control lacking reverse transcriptase was included.

The protein sequences available for mouse FoxK1 (NP_951031 and NP_034942)were used for TBLASTN similarity searches against the third assembly of the T. rubripes genome (available at http://www.ensembl.org/Fugu_rubripes/index.html),using a BLOSUM80 matrix, a word size of four and a maximum expected value cut-off equal to 1×10^–5. This prediction of the T. rubripes orthologue of FoxK1 was refined manually and specific primers were designed. Sequences of the primers (Invitrogen, Paisley, UK) used for cDNA amplification (cDNA-FoxK1) are shown in Table 1. FoxK1 was amplified by PCR using cDNA obtained from 2-month post-hatch larvae. The 25μl reaction mixtures for PCR amplification contained 1 μl cDNA template,40 nmol of each primer, 0.1 μmol l^–1 dNTPs, 1× PCR buffer (Amersham, Amersham, Buckinghamshire, UK) and 1 i.u. Taq DNA polymerase(Amersham). Amplification reactions were performed on a Genius thermocycler(Techne, Duxford, Cambridgeshire, UK) as follows: initial denaturation at 95°C for 3 min, 35 cycles of denaturation for 30 s at 95°C, annealing at 56°C for 30 s and extension at 72°C for 1 min and one final extension for 10 min at 72°C. PCR products were analysed by electrophoresis on a 1.2% agarose gel in modified Tris–acetate–EDTA buffer (Millipore, Billerica, MA, USA) and extracted from the gel using the Montage gel nebuliser system (Millipore). The purified PCR products were ligated to a pCR4-TOPO T/A vector (Invitrogen),which was then used to transform chemically competent TOP10 Escherichia coli cells (Invitrogen).

Sequencing reactions of the plasmid clones were performed in both directions with T3 or T7 primers and the DNA was sequenced with an ABI PRISM 377 DNA Sequencer (Applied Biosystems, Warrington, UK) at the Dundee Sequencing Service (University of Dundee, UK).

BLASTX similarity searches of the T. rubripes FoxK1 sequences were performed against the complete non-redundant GenBank database(http://www.ncbi.nlm.nih.gov/BLAST/)using the default parameters. Following translation of the T. rubripes FoxK1 nucleotide sequences using DNAman (Lynnon Biosoft, Quebec, Canada),the Conserved Domain Database and Search Service (v2.04) at NCBI was used to identify conserved domains in the predicted protein sequences(Marchler-Bauer and Bryant,2004). The putative FoxK1 proteins from T. rubripes were aligned with their orthologues from mouse (NP_951031, NP_034942) and zebrafish(NP_956196) using ClustalW (Thompson et al., 1994) on the BioEdit sequence alignment editor(Hall, 1999). A sequence identity matrix between these sequences was also obtained with BioEdit(Hall, 1999). For genomic sequence analyses, data were obtained from the current Ensembl assemblies of the T. rubripes, zebrafish and mouse genomes(http://www.ensembl.org/). Genomic organisation of FoxK1 was determined by comparison of cDNA and genomic sequences using the alignment program Spidey(Wheelan et al., 2001). The structure of donor and acceptor splice sites in T. rubripes FoxK1 was analysed using the Splice Site Prediction by Neural Network(Reese et al., 1997). The MartView data mining tool(http://www.ensembl.org/Multi/martview)was used to identify the genes present in a 100 kb region either side of the T. rubripes FoxK1 locus and the corresponding orthologues in mouse and zebrafish.

The three splice variants of T. rubripes FoxK1 were subcloned using the primers listed in Table 1 and the method described above, in order to obtain cDNA clones of suitable size. FoxK1-α, FoxK1-γ and FoxK1-δ DNA templates for probe synthesis were obtained from the corresponding pCR4-TOPO plasmids by PCR using standard M13 primers and the thermocycling conditions described above. T7 and T3 RNA polymerases (Roche,East Sussex, UK) were used to synthesise digoxigenin (DIG)-labelled RNA probes by in vitro transcription, according to the manufacturer's protocol. Sense probes were used as negative controls. Whole-mount in situhybridisation with DIG-labelled probes for FoxK1-α,FoxK1-γ and FoxK1-δ was performed essentially as described previously (Fernandes et al.,2006), with minor modifications. DIG-labelled probes for detection of FoxK1-α shared 44% and 46% identity with those for FoxK1-γ and FoxK1-δ, respectively, whereas FoxK1-γ and FoxK1-δ DIG-labelled probes were 74%identical. These differences in probe sequences should permit the specific detection of each splice variant of FoxK1. For each selected developmental stage, five T. rubripes embryos, reared at 18°C, were used. Optimal permeabilisation of T. rubripes embryos was achieved by incubation at 20°C with 20 μg ml^–1 proteinase K(Roche) for 5, 10 and 15 min for pre-somite, segmentation and post-somitogenesis stages, respectively. Whole embryos and flat-mounted embryos were observed under a binocular microscope (Leica MZ7.5, Milton Keynes, UK) and Leitz DMRB microscope (Leica) with DIC optics, respectively,and images were acquired with a Nikon Coolpix 4500 digital camera (Surrey,UK).

T. rubripes embryos incubated at 15°C, 18°C or 21°C were collected at different developmental stages. Tissues from juvenile and adult stages were collected as previously described. Total RNA extraction and cDNA synthesis were performed as described above. PrimerSelect software(DNAStar Inc., Madison, USA) was used to design specific FoxK1-α,FoxK1-γ and FoxK1-δ primer pairs(Table 1). Quantitative real-time PCR (qPCR) was carried out using an ABI Prism 7000 instrument(Applied Biosystems) with SYBR Green reagents (QuantiTect SYBR Green PCR,Qiagen, Crawley, West Sussex, UK), as recommended by the manufacturer. The 25μl reaction mixtures contained 1 μl cDNA (diluted 1:5), 0.4 μmol l^–1 each primer and 1× QuantiTect SYBR Green PCR master mix. PCR amplification of target genes was performed in duplicate using the following thermal profile: initial activation at 95°C for 15 min followed by 40 cycles of 15 s at 94°C, 30 s at 56°C and 30 s at 72°C. After each run, a dissociation protocol with a gradient from 60°C to 90°C was used to ascertain the specificity of the primers. ROX was used as passive dye for normalisation of SYBR Green fluorescence. RNA polymerase IIwas used as internal standard, since it was a more stable housekeeping gene than 18S rRNA or elongation factor 1α. The primer pair used to amplify 171 bp from the large subunit of RNA polymerase II was:5′-CAGCCCAGATGAACTTAAACGG-3′ (forward) and 5′-CCAGGACACTCTGTCATGTTGC-3′ (reverse). Threshold cycle values(C_T) were determined with the 7000 System Sequence Detection Software (Applied Biosystems) using an arbitrary threshold of 1 and a baseline set between 6 and 15 cycles. Standard curves for each gene were obtained by amplifying fivefold serial dilutions (ranging from 1:5 to 1:625)of a reference mixture containing equal amounts of cDNA from each sample. These standard curves were used to estimate the PCR efficiency of each amplicon, using the Relative Expression Sofware Tool (REST)(Pfaffl et al., 2002). C_T values were converted into relative expression levels according to the mathematical model proposed by Pfaffl(Pfaffl, 2001). Statistical analysis of TFoxK1 expression during development at different incubation temperatures was performed on SPSS 12.0 (SPSS Inc., Chicago, USA),using a general linear model with stage, splice variant and temperature as fixed factors. The Bonferroni test was used for post-hoc multiple comparisons between categories. Differences in tissue distribution of the three TFoxK1 splice variants were investigated by two-way ANOVA with Holm–Sidak post-hoc tests using the SigmaStat statistical package (Systat software, London, UK). In all instances significance levels were set at P<0.05.

Anti-FoxK1 polyclonal antibodies were prepared against a synthetic peptide designed from the putative translation of T. rubripes FoxK1-α. The antibody's epitope was located in a conserved region near the carboxyl terminus with a high antigenic index, as determined from the antigenicity plot of TFoxK1-α(http://bioinformatics.org/JaMBW/3/1/7/index.html). The peptide antigen Tyr-Arg-Tyr-Ser-Gln-Ser-Ala-Pro-Gly-Ser-Pro-Val-Ser-Ala-Gln-Pro-Val-Ile-Met was commercially synthesised and coupled to keyhole limpet haemocyanin. This hapten-carrier conjugate was used to immunise rabbits for antisera production,following a standard immunisation schedule (Cambridge Research Biochemicals,Durham, UK).

Myonuclei expressing FoxK1-α were identified by immunohistochemistry using the polyclonal antibody described above. Transverse sections of fast myotomal muscle from adult T. rubripes were stained for FoxK1-αaccording to the method described by Johnston et al.(Johnston et al., 2004). The primary anti-FoxK1-α antibody was used at a final concentration of 1:1500, whereas the anti-rabbit IgG-biotin conjugate was diluted 1:800.

The best match to mouse FoxK1 was found on scaffold 40 of the FUGU 4.0 T. rubripes genome assembly (P=4.1 e^–152). Two other significant matches with P values of 8.6 e^–144 and 4.0 e^–55 were also identified on scaffolds 443 and 847, respectively, but the predicted transcripts did not correspond to FoxK1. Unexpectedly, PCR amplification using primers designed from the sequence predicted on scaffold 40 yielded three specific products of different sizes. These cDNA sequences had open reading frames of 1676, 1868 and 2105 bp, which shared an overall 79.5% identity. BLASTX searches revealed that the three T. rubripestranscripts were most similar to mouse FoxK1, and were therefore designated TFoxK1. TFoxK1-α (GenBank: AY566278) coded for a putative protein of 558 residues that contains the FH domain typical of Fox proteins (Fig. 1). TFoxK1-γ (GenBank: AY566280) and TFoxK1-δ(GenBank: AY566279) translated into the same 344-residue protein with a partially truncated FH domain (Fig. 1). The first 332 amino acids of TFoxK1-γ and TFoxK1-δwere identical to those of TFoxK1-α, as was the FHA domain(Fig. 1). The C-terminal residues VGPFWLKLNALQ present only in TFoxK1-γ and TFoxK1-δ(Fig. 1) were encoded by the cDNA sequence 5′-GTGGGCCCATTCTGGCTGAAACTTAATGCTTTGCAA-3′. TFoxK1 showed a high degree of global conservation with the other vertebrate FoxK1 proteins (Fig. 1). TFoxK1-α shared 53% identity with its homologues in mouse and zebrafish. Within the FHA domain, TFoxK1-α showed 74% and 84% identity to the mouse and zebrafish sequences, respectively. The FH domain was particularly well conserved, displaying 92% identity between TFoxK1-α and the mouse and zebrafish FoxK1 proteins. Zebrafish and mouse FoxK1 shared 84% and 99% of their residues in the FHA and FH domains, respectively.

FoxK1 genes in T. rubripes and zebrafish spanned 6.4 kb and 30.9 kb, respectively, and the difference in size was attributable to larger introns in the zebrafish gene (Fig. 2). Introns I, II and VIII in zebrafish FoxK1 were particularly long, at 4.7 kb, 10.5 kb and 3.7 kb, respectively. The structure of the TFoxK1 gene (partial coding sequence) consisted of seven exons and six introns, whereas the zebrafish orthologue had nine exons and eight introns, including a 463 bp untranslated region (UTR)(Fig. 2). The complete coding sequence of zebrafish FoxK1 (DFoxK1; also known as foxk1 – Zfin) was slightly larger (1920 bp) and shared 58%identity with TFoxK1. The splice sites of exon–intron boundaries for the first six exons were conserved between TFoxK1 and zebrafish FoxK1. However, no evolutionary conserved intronic regions(i.e. over 60% identity over 100 bp) were detected between the TFoxK1and zebrafish FoxK1 genes. Comparison between the TFoxK1cDNA and genomic sequences clearly revealed that TFoxK1-α,TFoxK1-γ and TFoxK1-δ arise from alternative splicing events. Is it unlikely that the alternative transcripts identified in this study represent RT-PCR artefacts, since identical splice variants were identified in repeated amplifications from independent samples and splice variant-specific primers were used for the real-time PCR assays. The alternative transcript TFoxK1-δ is the largest, for it contains both introns IV and V, whereas the splice variant γ resulted from retention of intron IV (Fig. 2). Most of the donor and acceptor splice sites of TFoxK1had good matches to the consensus sequences and were considered strong splice sites (Table 2). However, donor sites 3 and 4 had significantly lower scores than the mean donor score for all exons (0.86) and the acceptor score for exon 2 was significantly lower than the mean acceptor value of 0.90 (Table 2).

FoxK1 is located on scaffold 40, chromosome 3, and chromosome 5 band G2, of the T. rubripes, zebrafish and mouse genomes,respectively. Comparative mapping of the genes surrounding TFoxK1showed that it lies within a region of conserved synteny between T. rubripes, zebrafish and mouse, thus confirming that it is the true orthologue of FoxK1 in these species(Fig. 3A,B). The genes in the vicinity of TFoxK1 consisted of abcg1 (white protein homologue), arpc1a and arpc1b (actin-related protein 2/3 complex subunits a and b), mmd2 (monocyte to macrophage differentiation factor 2), slipr (scaffolding protein SLIPR), cyp3a (cytochrome P450), bat4 (G5 protein), sdk1(sidekick homologue 1) and two predicted genes coding for hypothetical proteins, herein designated hyp1 and hyp2(Fig. 3). With the exception of bat4, which was located on chromosome 17, the mouse orthologues were present in a 3.3 Mb syntenic region (Fig. 3A). Local gene inversions could be observed in the region surrounding FoxK1 in mouse. Synteny conservation was not as prominent between T. rubripes and zebrafish, since most of the T. rubripes genes had orthologues on zebrafish chromosome 1(Fig. 3B). hyp1 was duplicated in zebrafish and one of its copies was inverted. Only arpc1a and arpc1b were present on the same chromosomal segment as FoxK1, and these genes were inverted in relation to the T. rubripes orthologues. T. rubripes cyp3a and hyp2did not have orthologues in either mouse or zebrafish in this region.

Expression of TFoxK1-α in T. rubripes embryos incubated at 18°C was not detected during gastrulation(Fig. 4A) but it showed a dramatic increase at the early stages of the segmentation period [63 h post-fertilisation (h.p.f.) 2–6 somites; Fig. 4B]. At this stage, TFoxK1-α was expressed in the cephalic region and developing somites. At the 5- to 6-somite stage, TFoxK1-α expression was found in pairs of symmetrical bands on either side of the notochord(Fig. 4B). Lower levels of TFoxK1-α transcripts were also detected in the pre-somitic paraxial mesoderm. Rostrocaudal progression of the staining throughout segmentation was not observed. TFoxK1-α expression was downregulated as segmentation progressed and approximately 13 h later (76 h.p.f., 10- to 12-somite stage) it was limited to the optic vesicles, the developing midbrain and a subset of cells flanking the notochord(Fig. 4C). Approximately midway through somitogenesis (16-somite stage, 86 h.p.f.), TFoxK1-αtranscripts could be detected in the lens and retina of the developing eyes,in the tubular heart and in the rhombo-mesencephalic fissure that marks the boundary between the midbrain and the hindbrain(Fig. 4D,E). TFoxK1-α expression declined rapidly after this stage and by the end of the segmentation period it was no longer detectable. The spatial and temporal distribution of TFoxK1-γ and TFoxK1-δ was also investigated by whole-mount in situ hybridisation. No qualitative differences in the developmental expression patterns of FoxK1-α, TFoxK1-γ and TFoxK1-δwere apparent (data not shown).

The expression levels of the three TFoxK1 splice variants during development of embryos incubated at 15, 18 or 21°C were determined by real-time PCR. Data were represented as ratios in relation to the expression level of TFoxK1-α during mid-gastrulation at 18°C,following normalisation with RNA polymerase II as internal standard. At each developmental stage there were differences in the relative proportions of TFoxK1-α, TFoxK1-γ and TFoxK1-δ(Fig. 5) and the splicing pattern changed throughout development (P<0.001). TFoxK1-δ was the least expressed of the three alternative transcripts at any stage and temperature(Fig. 5C). The effect of temperature on expression of TFoxK1 splice variants depended on the developmental stage (P=0.04). TFoxK1-α and TFoxK1-γ had similar expression patterns and their highest transcript levels at 15°C were observed at the onset of somitogenesis. By contrast, TFoxK1-α and TFoxK1-γ expression at 21°C was higher at the end of the segmentation period and maximal at the hatching stage. Differences in relative amounts of TFoxK1-γ and TFoxK1-α were greater at 21°C than 18°C or 15°C,and particularly striking at later developmental stages(Fig. 5A,B). At the hatching stage, the expression ratios between TFoxK1-γ and TFoxK1-α were 0.9, 1.2 and 2.4 at 15, 18 and 21°C,respectively.

Analysis of variance of qPCR results showed that the three TFoxK1splice variants were differentially expressed in different tissues of T. rubripes, but no significant differences in expression levels were found between juvenile and adult fish, representing growth stages in which myotube production was active or inhibited, respectively(Fernandes et al., 2005). The three TFoxK1 alternative transcripts were expressed in all tissues examined, including cardiac muscle (H) and fast (WM) and slow (RM) myotomal muscle (Fig. 6). TFoxK1-α and TFoxK1-γ were found to be the most abundant splice variants, with expression levels 10- to 20-fold higher than those of TFoxK1-δ. In order to investigate the localisation of TFoxK1-α protein in adult fast myotomal muscle a specific antibody was constructed. TFoxK1-α protein was exclusively expressed in mononuclear cells (arrowheads, Fig. 6inset) corresponding to the MPCs, which represented 3–5% of the total myonuclei (data not shown).

In the present study we have characterised the T. rubripesorthologue of mouse FoxK1, a transcription factor that is crucial for myogenic progenitor cell function in mouse (Garry et al., 2000). To the best of our knowledge, this is the first report of a non-mammalian FoxK1 gene. TFoxK1-α is homologous to mouse FoxK1-α and zebrafish FoxK1-α, and contains the two domains characteristic of the Fox family of proteins: the FHA domain, which is involved in phospho-dependent protein–protein interactions, and the FH region that is essential for DNA binding.

The amino-terminal, proline-rich region of the transcriptional activation domain of mouse FoxK1 (Bassel-Duby et al.,1994) is absent in TFoxK1. Both TFoxK1-γ and TFoxK1-δ code for the same, truncated isoform of TFoxK1, which differs from TFoxK1-α at the 12 carboxyl-terminal residues. The putative translation product of these transcripts contains an incomplete helix H3 in the FH domain and, therefore, it is likely to have a limited DNA binding ability, since helix H3 is crucial for DNA binding(Gajiwala and Burley, 2000)and some of the residues that are involved in its insertion into the major groove of DNA are absent. Hence, it seems that neither TFoxK1-γ nor TFoxK1-δ correspond to mouse FoxK1-β, which is the shorter of two isoforms of FoxK1. Mouse FoxK1-β comprises only 409 amino acid residues,compared with 617 in FoxK1-α, and differs from FoxK1-α by only six residues at its carboxyl terminus. Nevertheless, mouse FoxK1-β binds DNA with high affinity in vitro and can function as a transcriptional repressor in transient transfection assays of C2C12 myogenic cells(Yang et al., 1997). A member of the forkhead/winged helix family of transcription factors has also been recently identified in Xenopus laevis and termed XFoxK1(Pohl and Knochel, 2004). Despite its name, this gene does not correspond to mouse FoxK1; in fact, it seems to be the orthologue of mammalian interleukin-2 enhancer binding factor (FoxK2).

The 6.4 kb TFoxK1 gene is composed of seven exons and six small introns located on scaffold 40 of the FUGU 4.0 T. rubripes genome assembly. A comparison between mouse FoxK1 and TFoxK1 gene structures was not completed owing to extensive gaps in this region of the mouse genome sequence. The human orthologue of FoxK1, consisting of nine exons, has been identified using computer-based searches and mapped to chromosome 7p22.1 (Katoh,2004). Zebrafish FoxK1 was more similar to the mammalian gene, as it also contained nine exons and eight introns. Nonetheless, the splice sites of the first six exon–intron boundaries were conserved between zebrafish FoxK1 and TFoxK1, and they shared 65.8%identity at the nucleotide level within these 6 exons. The introns of zebrafish FoxK1 were substantially larger than the corresponding ones in TFoxK1, particularly intron II, which was five times larger than in pufferfish. The short intronic and intergenic regions are a characteristic feature of the compact genome of the Tetraodontidae family(Aparicio et al., 2002). The TFoxK1 gene was linked to abcg1, arpc1a, arpc1b, mmd2, slipr,cyp3a, bat4 and sdk1 genes on scaffold 40 of the T. rubripes genome assembly. The first two genes found downstream of TFoxK1 are the cytochrome P450 enzyme cyp3a and the HLA-B-associated transcript 4 (bat4), the function of which has yet to be determined. It is possible that these neighbouring genes might share common regulatory elements, since they are in such close proximity. Interestingly, there is a predicted gene (hyp2) found downstream of TFoxK1, which does not have orthologues in mouse or zebrafish and codes for a putative peptide of 437 residues that contains a cytochrome c haeme-binding site and a C2H2-type zinc-finger domain. Synteny analysis revealed that TFoxK1 was the true orthologue of zebrafish and mouse FoxK1. In spite of the longer evolutionary distance between mouse and pufferfish, the nature and order of genes in the TFoxK1chromosomal segment were more similar between these two species than between the two fish species. The murine orthologue of TFoxK1 was present in a 3.3 Mb syntenic area of chromosome 5G2 that seems to have been subjected to local gene inversion. This region contains genes with varied functions,including the cell adhesion protein sidekick 1 (sdk1) that is involved in axonal guidance (Yamagata et al., 2002), the scaffolding protein SLIPR (slipr) that is thought to link the receptor protein tyrosine phosphatase β with its substrates at the plasma membrane (Adamsky et al., 2003), the monocyte to macrophage differentiation factor 2(mmd2) whose function is unknown, and the actin-related protein 2/3 complex subunits a and b (arpc1a and arpc1b), which are involved in actin cytoskeleton organisation and biogenesis(Gournier et al., 2001). The murine orthologues of the white protein homologue (abcg1), cyp3a and the hypothetical genes hyp1 and hyp2 were absent in this segment of chromosome 5G2. Synteny conservation between T. rubripes and zebrafish was disrupted as a result of chromosomal translocations and inversions. Only the arpc1a and arpc1bgenes were found in the vicinity of zebrafish FoxK1 on chromosome 3.

Alternative pre-mRNA splicing plays a crucial role in regulating gene function by generating a large number of mRNA transcripts and protein isoforms from a limited number of genes. The variant transcripts generated by alternative splicing can have changes in coding sequence, premature stop codons or alterations in the 5′ or 3′ UTRs. The effects of alternative splicing vary from a complete loss of function to acquisition of a new one, to complex and subtle changes in protein function(Stamm et al., 2005). Importantly, alternative splicing can also modulate transcript levels by targeting mRNAs for degradation by nonsense-mediated decay (NMD)(Lewis et al., 2003). One of the main types of alternative splicing is the retention of introns that would normally be excised. This seems to be a frequent phenomenon in the human transcriptome, as demonstrated by a large scale analysis of intron retention in 21 106 known human genes, which revealed that 14.8% of these genes exhibited intron retention events, mostly located in UTRs(Galante et al., 2004). TFoxK1 is expressed as three alternatively spliced transcripts, two of which had non-excised introns: TFoxK1-δ contained introns IV and V, whereas TFoxK1-γ only retained intron IV. The poor splicing efficiency of intron IV might be explained, at least in part, by the relatively weak donor site of exon 4 and acceptor site of exon 5, both of which had consensus scores lower than the average scores for all TFoxK1 exons. The strong consensus splice sites of exons 5 (donor)and 6 (acceptor), indicate that retention of intron V was rather unexpected. However, it is well known that splice site selection depends not only on the 5′ and 3′ splice sites but also on branch points and exonic–intronic sequence enhancer and silencer elements(Stamm et al., 2005). Retention of intron IV results in the introduction of a premature stop codon,which would result in a shorter variant of FoxK1. However, it is improbable that this truncated isoform be translated, since TFoxK1-γ and TFoxK1-δ have translation termination sites 760 and 525 nucleotides, respectively, upstream from the 3′-most exon–exon junction. Hence, TFoxK1-γ and TFoxK1-δ splice variants are likely to be degraded by NMD(Conti and Izaurralde, 2005)and this might be a mechanism of regulating TFoxK1 transcript levels with important physiological consequences. Indeed, recent studies of gene expression profiling in NMD-deficient human cells revealed that NMD is a crucial post-transcriptional event that regulates expression of a significant number of transcripts involved in a broad range of cellular processes(Mendell et al., 2004).

No significant differences were observed in the temporal and spatial expression patterns of the three TFoxK1 splice variants, except that the intensity of TFoxK1-δ staining was considerably weaker,indicating that TFoxK1-δ is expressed at lower levels than the two other transcripts. The low abundance of TFoxK1-δ compared with TFoxK1-α or TFoxK1-γ might be related to the presence of strong consensus splice sequences around intron V, which would tend to promote its excision. The earliest detectable expression of TFoxK1-α coincided with the onset of somitogenesis(Fig. 4B) and the activation of the myogenic regulatory factor Myog(Fernandes et al., 2006). In contrast to Myog (Fernandes et al., 2006), TFoxK1-α transcripts were present in the pre-somitic mesoderm during the early stages of segmentation. FoxK1 expression in mouse has been shown to be independent of Myog, since its spatial expression pattern in embryos bearing null mutations in the Myog gene was indistinguishable from that in the wild type (Garry et al.,1997). In the somites, TFoxK1-α was transiently expressed in a rostrocaudal gradient and markedly downregulated as the somites matured. This pattern was broadly similar to that observed in mouse, as determined by immunofluorescence using a polyclonal antibody that did not distinguish between the two mouse FoxK1 isoforms. The murine gene was expressed concomitantly with Myf5 in a rostrocaudal gradient in the somites of the developing myotome, including the cells migrating to the limb buds (Garry et al., 1997). Whilst Myog expression persisted in the somites of T. rubripes embryos until the end of segmentation(Fernandes et al., 2006), TFoxK1-α expression was ephemeral and by mid-segmentation it was limited to immature somites and pre-somitic mesoderm(Fig. 4C). Besides being expressed in the developing skeletal musculature, TFoxK1-α was also expressed in the heart tube, as reported in mouse(Garry et al., 1997). No transcripts were detected in the pectoral fin bud primordia, which contrasts with the expression of Myog in T. rubripes(Fernandes et al., 2006). TFoxK1-α was expressed in the developing nervous system from an early stage of ontogeny in the head rudiment and developing midbrain(Fig. 4B,C). Transcripts of TFoxK1-α were particularly abundant in the rhombo-mesencephalic fissure (Fig. 4D,E), around the region where the cerebellar primordium is formed(Candal et al., 2005). Murine FoxK1 was also expressed in the developing brain(Garry et al., 1997) and in selected cortical and dopaminergic areas of the adult brain, including the piriform cortex and the Purkinje cell layer of the cerebellum(Wijchers et al., 2006). Taken together with our data, these results indicate that FoxK1 may play a conserved role in maintenance of developing and adult neurons in vertebrates. Additionally, TFoxK1-α was expressed in the optic vesicles,lens and retina of T. rubripes embryos, which suggests that TFoxK1 may have a novel function in eye development.

TFoxK1-α was present in skeletal muscle, heart, brain and liver, as has been reported in mouse (Yang et al., 1997). In addition, it was expressed in the skin and gonads of adult fish. The function of TFoxK1 in adult tissues other than skeletal muscle has not been elucidated to date. No significant differences in TFoxK1 transcript levels were found between juvenile and adult fish. Similarly, the muscleblind-like genes mbnl2a and mbnl3 were equally expressed in fast muscle of fish that were actively recruiting fibres by mosaic hyperplasia and in adult fish that had stopped producing new myotubes (Fernandes et al., 2007; Fernandes et al., 2005). Differences in the abundance levels of the three TFoxK1 splice variants were detected in the different tissues of T. rubripes: TFoxK1-δ was expressed at lower levels than TFoxK1-α and TFoxK1-γ in all tissues studied and, in general, the latter two splice variants were transcribed at similar levels. The presence of TFoxK1 in multiple tissues suggests that it might regulate cell cycle progression and transcription, not only in muscle but also in other cell types. The α and β alternatively spliced transcripts of FoxK1 are molecular markers of proliferating and quiescent satellite cells, respectively(Garry et al., 2000; Garry et al., 1997). In T. rubripes, TFoxK1-α was also detected in mononuclear cells in fast myotomal muscle of adult fish, which correspond to myogenic cells(Johnston, 2006). The function of mammalian FoxK1-α as a transcriptional activator or repressor remains uncharacterised and nothing is known about its potential binding partners or the downstream target genes that are regulated by FoxK1-α, other than the cyclin-dependent kinase inhibitor p21^CIP(Hawke et al., 2003).

Variation in the number of muscle fibres and in some cases myogenic progenitor cell numbers with embryonic temperature in the free swimming larval stages is a widespread phenomenon amongst teleosts and has been documented in a range of unrelated species from temperate and tropical environments including cod (Hall and Johnston,2003), Atlantic salmon(Johnston et al., 2000b),European sea bass (Alami-Durante et al.,2007) and zebrafish (Hung-Tai Lee and Ian A. Johnston, unpublished results). The tiger pufferfish also shows plasticity of myogenesis with embryonic temperature, since the number of fast muscle fibres in newly hatched larvae is significantly less at 21°C than 15°C or 18°C (Ian A. Johnston, unpublished results). Developmental plasticity in embryonic myogenesis could potentially involve any one of a number of steps, including commitment of stem cells to the myogenic lineage, the number of divisions of myogenic progenitors prior to cell cycle exit, apoptosis, migration and fusion events. In T. rubripes, the rate of decline of Myog from the formation of the first to the last somite-pair was greater at 21°C than 15°C, with intermediate rates of decrease at 18°C(Fernandes et al., 2006). These results lead to the hypothesis that lower embryonic temperatures delay and prolong muscle differentiation, at least in part involving changes in Myog expression (Fernandes et al., 2006). The real-time PCR analysis of TFoxK1transcripts demonstrated that the expression of this gene also varies with respect to embryonic temperature for equivalent developmental stages. Moreover, the relative amounts of the three TFoxK1 transcripts were affected by incubation temperature, indicating differential regulation of alternative splicing induced by temperature. Since TFoxK1 is likely to be involved in myogenesis and is expressed in MPCs, it is a good candidate gene for playing a key role in the temperature-induced phenotypic plasticity of muscle development observed in T. rubripes.

List of abbreviations

 
• C_T

  threshold cycle

 
• DIG

  digoxigenin

 
• FH

  forkhead

 
• FHA

  forkhead-associated

 
• Fox

  forkhead/winged helix family

 
• MPC

  muscle progenitor cell

 
• Myog

  myogenin

 
• NMD

  nonsense-mediated decay

 
• QPCR

  quantitative real-time PCR

 
• TFoxK1

  T. rubripes forkhead box protein K1

 
• UTR

  untranslated region

We would like to thank Professor Yuzuru Suzuki, Dr Kiyoshi Kikuchi, Dr Hiroaki Suetake and Mr Naoki Misuno at the Graduate School of Agricultural and Life Sciences (University of Tokyo, Japan) for their invaluable help during the collection of T. rubripes embryos. We thank Mrs Marguerite Abercromby (University of St Andrews, UK) for her assistance with the immunohistochemistry experiments. We are grateful to Ms Shelby Steele(University of Ottawa, Canada) for her critical reading of the manuscript. This work was funded by an Environmental Genomics Grant (NER/S/2001/00250)from the Natural Environment Research Council, UK.