Insulin-like growth factor (IGF) signalling and genome-wide transcriptional regulation in fast muscle of zebrafish following a single-satiating meal

Growth hormone is synthesized, stored and secreted by specialized cells in the anterior pituitary and plays a central role in controlling feeding behaviour, cell growth, osmoregulation and reproduction in teleosts (Kawauchi and Sower, 2006). Growth hormone acts directly on muscle through sarcolemmal receptors and indirectly via the production of insulin-like growth factors (IGFs) in the liver and peripheral tissues, which are released into the circulation (Wood et al., 2005). IGFs are also produced by paracrine pathways and are stimulated by amino acid influx into the muscle (Bower and Johnston, 2010a). In mammals, the IGF system comprises ten components: two hormones (IGF-I, IGF-II), two receptors (IGF-IR, IGF-IIR) and six binding proteins (IGFBP1–6) (Duan et al., 2010). IGFBPs have distinct physiological roles in development and regulate IGF release to tissues in association with specific proteases (Duan et al., 2010). Binding of IGF-I to its receptor activates several downstream signalling cascades including the P13K–AKT–TOR and MAP kinase pathways that are well conserved in fish and mammals (Engert et al., 1996; Duan et al., 2010). Activation of P13K–AKT–TOR stimulates a phosphorylation cascade that increases translation and protein synthesis (Engert et al., 1996; Duan et al., 2010) and inhibits protein degradation by the 26S proteasome system (Witt et al., 2005). In the zebrafish (Danio rerio), no fewer than 16 components of the IGF system have been described (Maures et al., 2002; Maures and Duan, 2002; Chen et al., 2004; Zhou et al., 2008; Wang et al., 2009; Zou et al., 2009; Dai et al., 2010). The larger number of IGF components in zebrafish compared with mammals reflects a whole genome duplication (WGD) that occurred at the base of teleost evolution (Jaillon et al., 2004). It is thought 15% of the duplicated genes or paralogues from this basal WGD have been retained in extant species (Jaillon et al., 2004). The distinct patterns of tissue expression and transcriptional regulation of many IGF system paralogues observed in zebrafish (Maures et al., 2002; Maures and Duan, 2002; Chen et al., 2004; Zhou et al., 2008; Wang et al., 2009; Zou et al., 2009; Dai et al., 2010) is consistent with either subfunctionalization or neofunctionalization of these genes.

Fasting-refeeding protocols are commonly used to investigate transcriptional regulation in the IGF-system in teleosts following the transition from catabolic to anabolic states (Chauvigne et al., 2003; Salem et al., 2005; Gabillard et al., 2006; Rescan et al., 2007; Bower et al., 2008). Feeding to satiation after a prolonged fast results in increased feeding intensity relative to continuously fed controls and a period of compensatory or catch-up growth (Nicieza and Metcalfe, 1997). The transcriptional responses observed in such experiments are dependent on the nutritional state of the fish prior to fasting, particularly the extent of fat stores, the duration of the fast, body size and temperature (Johnston et al., 2011). Fish show diurnal rhythms in feeding behaviour and activity driven by central oscillators in the brain that are synchronised by environmental cycles and co-ordinated with peripheral clock genes regulating metabolism (Davie et al., 2009). In aquaculture, meal times entrain biological rhythms and ready physiological systems in anticipation for processing the food (Sanchez et al., 2009). As a consequence, great care should be taken in designing fasting-feeding experiments in order to define all experimental variables including the frequency and timing of feeding in relation to diurnal cycles.

Following the digestion and assimilation of a meal, metabolism changes from an overall catabolic to an anabolic state, utilizing the nutrients from the meal to produce energy and synthesize new molecules, which is characteristic of the postprandial period. To our knowledge there is no study describing the transcriptional changes during and following a postprandial period in fish, with most studies focusing on the changes in metabolic rate (Clark et al., 2010; Vanella et al., 2010) and the plasma level of metabolites following feeding (Eames et al., 2010; Eliason et al., 2010; Wood et al., 2010). We have investigated transcriptional regulation in the fast myotomal muscle of male zebrafish in response to a single satiating meal delivered at first light. Expression of all 16 genes of the IGF system was investigated by quantitative polymerase chain reaction (qPCR) and supplemented with a genome-wide survey of transcript abundance using the Agilent 44K oligonucleotide microarray. Transcript abundance and the phosphorylation of the signalling protein Akt were determined in relation to the presence of food in the gut as a reference point. The single-meal experimental design potentially provides greater temporal resolution for studying transcriptional responses compared with continuous refeeding where early and late events quickly become confounded. The aims of the study were to test the hypothesis that paralogues of IGF-system genes were differentially regulated with feeding, and to discover novel genes associated with the fasting and fed states in skeletal muscle.

The F5 generation of a wild-caught population of zebrafish [Danio rerio (Hamilton 1822)] from Mymensingh, Bangladesh, was used in this study. All fish were adult males aged 9 months. Prior to the single meal experiment the fish were maintained in a single 50 litre tank at 27.6±0.4°C in a 12 h:12 h light:dark photoperiod. The fish were fed bloodworms (Ocean Nutrition™, Essen, Belgium) to satiety twice daily for 1 week. Nitrite (0 p.p.m.), nitrate (10–20 p.p.m.), ammonia (0 p.p.m.) and pH (7.6±0.2) were tested during acclimation and experimental periods using a Freshwater Master Test Kit (Aquarium Pharmaceuticals Inc., Chalfont, PA, USA).

Two replicate experiments were carried out, 3 months apart, with identical environmental conditions and food to account for any tank-to-tank variation in the feeding response. The experimental protocol involved fasting fish for 7 days and then feeding a single meal of bloodworms delivered over a 3 h period, after which any uneaten food was removed from the tank by siphoning. In the first replicate experiment seven fish per time-point were sampled and in the second replicate experiment six fish per time-point were sampled at the following times: –156, –24, 0 h (before the meal) and 0.75, 3, 6, 7.5, 9, 11, 24 and 36 h (after the meal). The average body mass in the first replicate experiment was 0.46±0.02 g and the standard length (from tip of snout to last vertebra) of the fish was 29.8±0.4 mm (N 77), and in the second replicate experiment was 0.53±0.016 g and 32.6±0.3 mm (N 66), respectively (means ± s.e.m.). Fish were killed humanely by an overdose of ethyl 3-aminobenzoate methanesulphonate salt (MS-222; Fluka, St Louis, MO, USA).

Fast skeletal muscle was dissected from the dorsal epaxial myotomes, flash frozen in liquid nitrogen and stored at –80°C prior to total RNA and protein extraction. The digestive tract was dissected and fixed in 4% (m/v) paraformaldehyde for later quantification of intestine content to the nearest milligram. Fixation was necessary to prevent tissue loss during dissection and to achieve an accurate quantification of intestine content. Because the nature of the tissue and food were the same for all samples, any shrinking caused by fixation should be proportional to the amount of material and was not considered in the interpretation of the results. All experiments and animal handling were approved by the Animal Welfare and Ethics Committee, University of St Andrews and conformed to UK Home Office guidelines.

Total protein was extracted from fast skeletal muscle from five randomly selected fish per time-point in each of the independent experiments. Tissue (30 mg) was homogenized in Lysing Matrix D (MP Biomedicals, Qbiogene, Irvine, CA, USA) in a FastPrep® machine (Qbiogene) using 350 μl of 25 mmol l^–1 MES (2-morpholinoethanesulfonic acid monohydrate) pH 6.0 containing 1 mol l^–1 NaCl, 0.25% (m/v) CHAPS, DNA/RNA nuclease (Invitrogen, Carlsbad, CA, USA) and protease inhibitor cocktail (Invitrogen).

Samples (20 μl, containing 20 μg of protein) were added to 6 μl of a solution containing 5 μl of 5× protein loading buffer and 1 μl 20× reducing agent (Fermentas, Vilnius, Lithuania), heated for 5 min at 95°C, loaded into NuPAGE® Novex 4–12% Bis-Tris gels (Invitrogen) and electrophoresed at 120 V. A protein ladder ranging from 10 to 250 kDa (Fermentas) and a reference sample were loaded in all gels to estimate the molecular mass of proteins of interest and to serve as a normalization sample, respectively. Proteins separated by electrophoresis were transferred to a PVDF Immobilon-P Transfer Membrane (Millipore, Billerica, MA, USA) at 25 V for 105 min. Successful protein separation and transfer were confirmed by Ponceau S staining (Sigma, St Louis, MO, USA). PVDF membranes were blocked overnight at 10°C using 5% (m/v) non-fat milk (AppliChem, Darmstadt, Germany) prepared in PBS (Sigma) containing 0.1% (v/v) Tween 20 (Sigma). Blocked membranes were incubated overnight at 10°C with the following primary antibodies (IgGs): P-Akt [1:1000 dilution (v/v), Cell Signaling #4060, Danvers, MA, USA], Akt [1:1000 (v/v), Cell Signaling #2966], actin [1:20,000 (v/v), Sigma A2066] and GAPDH [1:30,000 (v/v), Sigma G9545]. Probed membranes were incubated at 20°C for 1 h with the secondary antibody against mouse or rabbit IgG conjugated to horseradish peroxidase [both from Sigma and used at 1:60,000 (v/v)]. After incubation for 1 min with ECL Western Blotting Detection Reagents (GE Healthcare, Amersham, Buckinghamshire, UK) at room temperature, membranes were exposed to Hyperfilm ECL (GE Healthcare). Experimental variations in the electrophoresis and transfer were normalized using a reference sample common to all membranes. The fold-change in phosphorylation of Akt at each time-point was compared with the samples from –159 h. Further information on preliminary optimisation experiments, the normalisation strategy and tests of P-Akt antibody specificity are given in the Appendix Fig. A1.

Total RNA was extracted by homogenization in Lysing Matrix D (Qbiogene) using 1 ml of TRI reagent (Sigma) in a FastPrep® machine (Qbiogene). The RNA concentration 260/280 and 260/230 ratios were measured using a NanoDrop® spectrophotometer (ThermoFisher Scientific, Loughborough, Leicestershire, UK) and were 1.5–2.0 and >2, respectively. RNA integrity was also checked by agarose gel electrophoresis. A Quantitect Reverse Transcription kit (Qiagen, Hilden, Germany) was used to produce first-strand cDNA from 0.6 μg of total RNA following the manufacturer's instructions.

Microarray experiments were carried out by an Agilent-certified microarray service provider (Microarray Centre, University Health Network, Toronto, Canada) using the Dual-Mode Gene Expression Analysis Platform (Agilent Technologies Inc., Santa Clara, CA, USA) in a 4×44K slide format [Zebrafish (v2) Gene Expression Microarray]. RNA from the 3 h and 6 h time-points were hybridized with the RNA from the 0 h sample to identify differentially regulated genes in 7-day fasted fish following a single satiating meal. The 3 h sample coincided with 50% of maximum gut fullness (Fig. 1). Six phenotypic replicates from each group were used. R, version 2.9n.0, with arrayQualityMetrics_2.2.0 and limma_2.18.0 was used for quality analysis of microarray data. Microarray results were also analysed using GeneSpring®, v7. The intensities of spots among arrays was normalised using the AQuantile method and intensities were log transformed before performing a t-test using the Benjamini and Hochberg method for multiple testing correction (Benjamini and Hochberg, 1995). A list of differentially regulated genes was produced by screening against the following criteria: >twofold change in expression, B-value statistic >0 and an adjusted P-value <0.05. Blast2GO (www.blast2go.org) software was used to analyse the gene ontology (GO) terms of the differentially expressed genes. GO enrichment analysis of the annotated genes was performed with the GOSSIP tool using Blast2GO software. The European Bioinformatics Institute Miame ArrayExpress accession number for the microarray experiment is E-MEXP-2887.

Transcript sequences from the Ensembl database (release 55; www.ensembl.org) were used to design primer pairs for each gene listed in supplementary material Table S1. Primers were designed using NetPrimer (http://www.premierbiosoft.com/netprimer/index.html; Premier BioSoft, Palo Alto, CA, USA). Where possible the primer pairs were designed so that at least one primer spanned an exon–exon boundary (otherwise primers pairs were in different exons), with a T_m close to 60°C. First-strand cDNA from muscle was used as a template to synthesise PCR products for cloning. The PCR reaction mixture contained 1.5 mmol l^–1 MgCl₂, 1× NH₄ buffer, 0.25 μl BioTaq™ DNA polymerase (Bioline, London, UK), 0.2 mmol l^–1 dNTP (Promega, Madison, WI, USA), 1 μl cDNA and 0.5 μmol l^–1 of each primer (Invitrogen) and the following thermal cycling conditions were employed: denaturation for 5 min at 95°C followed by 36 cycles of 30 s at 95°C, 30 s at 60°C and 30 s at 72°C, final elongation for 5 min at 72°C. After agarose gel electrophoresis the products of expected size were extracted and purified using a QIAquick Gel Extraction Kit (Qiagen) and cloned using a StrataClone™ PCR Cloning Kit (Stratagene, La Jolla, CA, USA) according to the manufacturer's instructions. Positive clones were picked after 16 h of growth at 37°C in LB-agar plates containing ampicillin (0.1 mg ml^–1) and transferred to 96-wells plates containing LB broth and ampicillin (0.1 mg ml^–1). After 16 h of growth at 37°C colony-PCR was performed to confirm the sequence of the insert using Big Dye terminator sequencing (Applied Biosystems, Foster City, CA, USA) at the University of Oxford (T3 and T7 primers were used to confirm the sequence in both directions). The clones bearing plasmids with the expected insert were grown in 5 ml LB broth and plasmids were purified using a QIAprep Spin Miniprep Kit (Qiagen). The plasmid concentration, 260/280 and 260/230 ratios were measured using a NanoDrop® instrument and were higher than 1.8 and 2.0, respectively.

All protocols and reporting of qPCR assays adhered to ‘Minimum Information for Publication of Quantitative Real-Time PCR Experiments’ guidelines (Bustin et al., 2009). The qPCR reaction mixture contained 7.5 μl 2× Brilliant II SYBR® QPCR Low ROX Master Mix (Stratagene), 6 μl 40-fold diluted cDNA, 0.25 μmol l^–1 each primer and nuclease-free water (Qiagen) to a final volume of 15 μl in 96-well plates (Stratagene). The reactions were performed in a Stratagene MX3005P machine (initial activation at 95°C for 10 min, followed by 40 cycles of 30 s at 95°C, 30 s at 60°C and 30 s at 72°C) and the fluorescence results, collected after the elongation step (72°C), were recorded using the MxPro software v4.10 (Stratagene). Negative controls were included and run in duplicate, and contained either all components of the reverse transcription mixture, except reverse transcriptase (no reverse transcriptase control) or water instead of a cDNA template (no template control). After the qPCR, a dissociation curve (from 55 to 95°C) was produced to verify the presence of a single peak. The specificity of each qPCR assay was also validated by sequencing transformants for each qPCR product. Absolute copy number of each gene was calculated based on a standard curve of at least six orders of magnitude, prepared with the plasmids. This data was also used to analyse the efficiency of each primer pair (supplementary material Table S1). The threshold of fluorescence (for dRn values) used for determination of the quantification cycle (C_q) was set to 1.0 for all plates to allow for comparison between plates. Comparison between plates was possible after normalization to six samples loaded on all plates.

Six reference genes selected from the literature (ef1a, rpl13a, bactin2, b2m, usp5, tbp and cyp1a) and four selected from the microarray experiment (tomm20b, rpl7a, lman2 and dis3l2) were analysed using Genorm v3.5 (Vandesompele et al., 2002) with M set to <1.5. The two genes with the most stable level of expression across the experimental conditions were ef1a and lman2 (M=0.4). The expression of genes of interest was normalized to the geometric average of the two most stable genes and gene expression was reported as arbitrary units (a.u.).

All data were analysed for normal distribution and equality of variance. Normally distributed data were analysed using ANOVA followed by Tukey's post hoc tests using PASW Statistics 18 (SPSS Inc., Chicago, IL, USA). Kruskal–Wallis non-parametric tests followed by Conover post hoc tests were used for the data that were not normally distributed using BrightStat software (Stricker, 2008). There was no statistically significant difference in the standard length, body mass, intestine content or normalized gene expression between the two replicate experiments (P>0.05). Therefore, the results from both experiments were combined to facilitate their interpretation, resulting in N=13 per time-point (i.e. seven fish from the first replicate experiment plus six from the second replicate). In order to combine the mRNA expression data from both experiments, the results of gene expression from the second replicate experiment were normalized to the average value of seven samples from the first replicate experiment, which were included from the cDNA synthesis step onwards. Correlation of gene expression from both qPCR and microarray experiments was analysed by Spearman's correlation test using PASW Statistics 18 (SPSS Inc.). Clustering of gene expression was performed using PermutMatrix (http://www.lirmm.fr/~caraux/PermutMatrix/EN/index.html).

Fish were continuously fed to satiation and then fasted for 7 days prior to feeding a single meal over 3 h. Three samples were taken during the fast: –156, –24 and 0 h, corresponding to 9 h, 144 h and 168 h after the last food. Only traces of food remained in the gut after 9 h of fasting (–156 h time-point), corresponding to 0.1% of body mass, with only bile observed with more prolonged fasting (–24 and 0 h time-points). The maximum average gut fullness, equivalent to 6.3% of body mass, was observed at 0.75 h (45 min) after food became available, indicating satiety had been reached. Intestinal contents had decreased by 50% 3 h after food was withdrawn and 95% of the food ingested had been assimilated or excreted by 9 h (Fig. 1).

In fast myotomal muscle the protein Akt showed a marked increase in phosphorylation to peak levels within 3 h of the start of feeding and became steadily dephosphorylated as the intestine emptied (Fig. 2). The level of the protein load controls GADPH and actin and the dephosphorylated Akt did not change significantly over the course of the experiment.

The expression of all the paralogues of IGF-system genes, including those of Igf1, Igf2, Igf receptors and Igf binding proteins, was determined by qPCR. In many cases retained paralogues of IGF system components showed distinct patterns of transcriptional regulation over the course of the experiment. Changes in transcript levels could be directly attributed to feeding because marked responses were largely present in only the first of two light:dark cycles following the meal.

igf1a expression was correlated with, but lagged behind gut fullness showing a distinct peak 3 h after the start of feeding (P<0.05; Fig. 3A). The igf1b paralogue was not detected in fast myotomal muscle after 35 cycles of PCR (Fig. 3C). igf2b expression was also increased following feeding, showing peak expression 3 h later than igf1a (Fig. 3B). The igf2a paralogue was 3.7 times more abundant than the igf2b transcripts at 0 h (Fig. 3C) and was constitutively expressed during the experiment with no discernible pattern in relation to feeding (not shown).

The expression of the IGF receptor gene igf1ra increased more than 2.7-fold between 9 h and 168 h of fasting, decreased to its lowest levels 45 min after feeding and then showed variable, though still depressed, expression over the next 36 h (Fig. 4A). igf1rb paralogue expression was inversely related to gut fullness showing a more than 3.3-fold decrease within 45 min of feeding, returning to levels not significantly different from those after fasting for only 12 h (Fig. 4B). igf1rb transcripts were more abundant than igf1ra transcripts over the single meal experiment, with 7.4 more copies at the start of feeding (Fig. 4D). Transcripts of the single igf2r gene showed no consistent changes in expression over the fasting–feeding–fasting cycle (not shown).

The two retained paralogues of the IGF binding protein 1 gene (igfbp1a and igfbp1b) showed similar changes in expression over the experiment, with high levels during prolonged fasting (144–168 h), a marked reduction in transcript abundance within 45 min of the start of feeding and variable, but generally low levels over the following 36 h (Fig. 4C). Expression of igfbp1a was considerably higher than igfbp1b over the experiment, with a 12.4 times greater copy number at 0 h (Fig. 4D). The remaining IGF binding protein genes (igfbp2a, igfbp2b, igfbp3, igfbp5a, igfbp5b, igfbp6a and igfbp6b) showed no consistent change in expression in response to a single satiating meal (not shown).

In order to identify some nutritionally responsive candidate genes for further investigation, whole genome microarray analysis was performed. Total RNA from maximally fasted fish (0 h) was hybridized with RNA from fish sampled 3 h and 6 h after the initiation of feeding. Genes were considered differentially regulated if they showed a more than twofold change in expression, a B value >0, and the difference in expression was significant at P<0.05, following correction for multiple comparisons. The hybridizations of fasting and 6 h fed samples produced a relatively short list of genes that were also represented in the hybridization of fasting and 3 h fed samples and so were not considered further. Fast skeletal muscle from zebrafish fasted for 7 days had 56 upregulated genes (Table 1) and 91 downregulated genes (Table 2) compared with fish fed to satiation over 3 h. 45 of the upregulated genes in fasting and 79 of the upregulated genes with feeding had associated gene ontology (GO) terms. For the genes upregulated with fasting GO term analysis revealed a significant enrichment of terms associated with catabolic processes, ubiquitin ligase activity and positive regulation of endothelial cell differentiation, including the genes btg1 and btg2 that have anti-proliferative properties (Winkler, 2010) (full listing in Table 3). Analysis of the genes upregulated with feeding showed enrichment for GO terms such as unfolded protein binding, protein folding, endoplasmic reticulum lumen, protein maturation, chaperone binding, sarcomerogenesis, myosin filament assembly, collagen biosynthesis and regulation of the JAK-STAT cascade (full listing in Table 3). The gene showing the largest fold change of 32.2 with fasting (Table 1) coded for a novel protein with ∼23% identity to the mammalian orthologue of harbinger transposase derived 1 (harbi1), which is thought to have nuclease activity (Kapitonov and Jurka, 2004). The list of genes upregulated with feeding included many chaperone genes (unc45b, ptges3 and serpinh1), heat shock protein (hsp90a.1, hsp90a.2, hspd1 and hspa5) and heat shock protein-associated (ahsa1, calrl and stip1) genes (Table 2). Feeding was also associated with enrichment of the interleukin-20 receptor binding GO term (Table 3) and increased il34 expression (Table 2).

A selection of candidate nutritionally regulated genes comprising eight genes upregulated with fasting (fbox32, zgc:86757, klf11b, nr1d1, cited2, bbc3, znf653 and hsf2) and six genes upregulated with feeding (odc1, fkbp5, sae1, foxo1a, hsp90a.1 and hsp90a.2) plus some contractile protein genes (mylz2 and tnni2a.4) was further investigated using qPCR. The log fold-change in expression of all the genes assayed by qPCR showed a good correlation with the microarray experiment (R=0.79; P<0.001; N=76; supplementary material Fig. S1). Genes from the IGF pathway and those selected from the microarray experiment formed five major clusters (Fig. 5). Cluster I contained genes upregulated with feeding, with low levels of expression during prolonged fasting (–24 and 0 h) and intermediate levels of expression from 9 to 36 h (Igf1a, hsp90a.1, odc1, foxo1a, igfbp6a, Igf2b, sae1 and fkbp5). Cluster II comprised genes that were downregulated at –159 h and from 0.75 to 9 h, with upregulation during prolonged (–24 and 0 h) and early fasting (12–36 h) (igf1ra, igf2r, igf1rb, fbxo32, zgc:86757, klf11b, bb3 and znf653). Cluster III contained genes that were only upregulated with prolonged fasting (144 and 168 h) and just after the beginning of feeding (45 min), with low expression at other time-points (tnni2a.4, hsp90a.2, nr1d1, igfbp1a, hsf2 and igfbp1b). Cluster IV comprised genes with high expression during prolonged fasting, low expression while the intestine was full (0.75–6 h) and intermediate levels of expression from 7.5 to 36 h (Igf2a, igfbp3 and cited2). Cluster V comprised genes with high expression with prolonged fasting, but with intermediate level of expression during feeding and early fasting (mylz2, igfbp5b, igfbp2b, igfbp5a, igfbp2a and igfbp6b). Genes pairs that showed strong significant correlations (R>0.7, P<0.05, N=126) were considered candidates for coregulated expression. Eight strong positive correlations were found and included hsf2 versus nr1d1 (R>0.83; P<0.001) and fbxo32 (R=0.78; P<0.001); bbc3 versus znf653 (R=0.76; P<0.001) and klf11b (R=0.75; P<0.001); fbxo32 versus MURF1 (R=0.73; P<0.001); and bbc3 versus cited2 (R=0.73; P<0.001) and fbxo32 (R=0.72; P<0.001). The only strong negative correlation found was between odc1 and bbc3 (cluster II).

All eight of the candidate genes that were found to be upregulated with fasting in the microarray experiments were validated by qPCR. The expression of the muscle-specific ubiquitin ligases, fbxo32 (also known as MAFbx/Atrogin-1) and MURF1 (annotated as zgc:86757 in the zebrafish genome assembly) was highly sensitive to nutritional status. Transcripts for fbxo32 and MURF1 increased 13.3 and 2.7-fold, respectively, between 9 h and 144 h of fasting and were downregulated by 55% (fbxo32) and 77% (MURF1) in the 45 min sample after feeding, with lowest levels observed after 6 h (Fig. 6A,B). Expression of both genes started to increase soon after food was cleared from the intestine and mRNA levels were at the 168-h fasting levels by 36 h after the meal (Fig. 6A,B).

Four genes (bbc3, cited2, znf653 and klf11b) had expression patterns that were inversely correlated with gut fullness. Transcript abundances were lowest 45 min to 3 h after feeding and rapidly increased as the gut emptied (Fig. 7A–D). nr1d1 and hsf2 mRNA levels increased ∼130- and ∼21-fold, respectively, between 9 h and 144 h of fasting, and were strongly downregulated with feeding and showed a transient increase in levels 15 h after the intestine was empty (Fig. 7E,F). Three of the selected genes upregulated with feeding in the microarray experiment showed expression patterns related to gut fullness. fkbp5 and odc1 showed a peak of expression co-incident with the presence of food in the intestine, corresponding to a 5.6- and 17.2-fold increase in transcript abundance relative to maximal fasting values, respectively (Fig. 8A,B). sae1 showed a ∼6.5-fold increase in expression 6 to 9 h after the start of the meal and a steady decline to levels not significantly different from fasting values by 36 h (Fig. 8C).

In the absence of information on plasma hormone and amino acid levels in small tropical fishes, transit of food through the gastrointestinal system and phosphorylation of the IGF-pathway signalling protein Akt provide a means of interpreting the transcriptional response to a single satiating meal. In a previous fasting–refeeding study using rainbow trout, Oncorhynchus mykiss, maximum plasma insulin and amino acid levels were recorded 30 min and 2.5 h after feeding and were quickly followed by phosphorylation of several kinases, indicating activation of the TOR signalling pathway (Seiliez et al., 2008). Amino acids and insulin also rapidly increase the level of circulating IGFI in brown trout (Banos et al., 1999). In our experiments with zebrafish, 50% of the food ingested had been processed and eliminated within 3 h after the start of feeding and the gut was empty after 9 h (Fig. 1). There was a significant increase in phosphorylation of Akt within 45 min, with peak levels at 3 h after feeding and dephosphorylation over a similar time course to the elimination of food from the gut (Fig. 2). Changes in the levels of muscle mRNAs also mostly took place with a time course of hours and were generally quicker than described in the literature for larger temperate fish species maintained at lower temperatures (Chauvigne et al., 2003; Bower et al., 2008).

The first aim of this study was to test the hypothesis that retained paralogues of IGF-system genes were differentially regulated following transition from a catabolic to an anabolic state. Fasting has been shown to result in an upregulation of muscle IGF receptors in a number of teleost species and is probably a response to a decrease in circulating IGF hormone levels (Chauvigne et al., 2003; Gabillard et al., 2006; Bower et al., 2008; Bower and Johnston, 2010b). In rainbow trout, prolonged fasting resulted in an upregulation of IGFR-Ia, but not IGFR-Ib (Chauvigne et al., 2003) whereas in zebrafish both IGFR paralogues were upregulated (Fig. 4A,B). However, zebrafish igf1rb increased more rapidly following gut emptying than igf1ra reaching fasting levels >36 h and <25 h, respectively (Fig. 4A,B), indicating differential regulation of IGFI receptor paralogues. Functional characterisation studies of IGFIIR in fish are scarce and its role in the response to nutrient levels is not yet clear. In Atlantic salmon, IGFIIR transcripts were significantly upregulated with prolonged fasting and downregulated after 7 days of refeeding (Bower et al., 2008). In this study there was no discernible pattern of transcriptional regulation of igf2r over the experiment. Muscle IGF hormone transcript levels showed distinct peaks within a few hours of feeding (Fig. 3A,B). In the case of Igf1a peak values were found within 3 h of feeding and had declined to fasting levels before the gut was emptied (<5 h; Fig. 3A). In contrast, Igf2b transcripts reached a peak 6 h after feeding and were not significantly different to fasting levels by 9 h (Fig. 3B). Igf1b transcripts were not detected in fast muscle whereas the Igf2a paralogue was constitutively expressed and showed no consistent change in expression over the fasting-feeding-fasting cycle associated with a single meal. In vitro studies with Atlantic salmon (Salmo salar) myocytes have shown synergistic affects of insulin, IGF-I and amino acids on muscle IGF-I transcript abundance, indicating multiple pathways leading to IGF transcription (Bower and Johnston, 2010a). In mammals, the binding of IGF to the IGF-IR induces its auto-phosphorylation resulting in the activation of several down-stream signal transduction cascades via adaptor molecules such as the insulin receptor substrate 1 (IRS-1), which has multiple tyrosine phosphorylation sites (Duan et al., 2010). IGF-I stimulates growth via effects on protein synthesis (Rommel et al., 2001), myoblast proliferation and differentiation acting through distinct signalling pathways (Ren et al., 2010). The effective concentration of IGF in the muscle is regulated by six IGFBPs, which are degraded by specific proteases to release the hormone to target sites (Wood et al., 2005). Evidence, largely from mammals, indicates that IGFBPs can inhibit and/or potentiate IGF actions depending on the cellular context and/or environmental conditions (Duan et al., 2010). In Atlantic salmon, the transition from maintenance to fast growth was associated with a constitutive upregulation of IGFBP-4, a transient increase in IGFBP-5.1 and a downregulation of IGFBP-2.1 (Bower et al., 2008). The two retained zebrafish paralogues of igfbp1 had similar expression in the single meal experiment with high transcript abundance during fasting, a marked reduction in mRNA levels within 45 min of feeding and variable, but generally low level of expression over the following 36 h (Fig. 4C). Mammalian studies indicate that in addition to its role as a modulator of IGF1 availability, IGFBP-1 has putative IGF-independent biological activities through interaction with cell-surface integrins, with putative direct effects on the P13K–AKT pathway (Wheatcroft and Kearney, 2009). There was no evidence for the transcriptional regulation of igfbp2a, igfbp2b, igfbp3, igfbp5a, igfbp5b, igfbp6a and igfbp6b paralogues with feeding in our experiments. Although experimental context may explain some of these differences in IGFBP expression it is clear that there are lineage-specific differences in IGFBP function and regulation within the teleosts. For example, igfbp4 is not represented in the current Danio rerio genome assembly (http://www.sanger.ac.uk/Projects/D_rerio/) and is probably absent from the zebrafish lineage. Our results at least indicate that caution is needed in inferring similar functions of IGFBPs between teleost lineages and certainly between teleosts and mammals. Overall we conclude that following WGD some of the retained paralogues of IGF-system genes show differential transcriptional regulation with fasting and refeeding in skeletal muscle, but that complex patterns of regulation have evolved between and within lineages.

Microarray experiments provided a snapshot of the fast muscle transcriptome during fasting and at the point IGF transcripts reached their maximum abundance following feeding. The screening criteria used to build lists of differentially regulated genes were apparently robust because all 14 genes tested were validated by qPCR and were well correlated (R=0.79; supplementary material Fig. S1). Fasting in zebrafish (Fig. 6A,B) and Atlantic salmon (Bower et al., 2008; Bower and Johnston, 2010b) is associated with a large increase in the abundance of E3 ubiquitin ligases MURF1 and fbxo32 transcripts. In mammals, the ubiquitin substrate recognition system has been implicated in specific degradation of myoD (Tintignac et al., 2005; Finn and Dice, 2006) and other promyogenic transcription factors (Finn and Dice, 2006). Two substrate recognition components of the ubiquitin ligase system, F-box only protein 25 (fbxo25) and RAB40B, member RAS oncogene family (rab40b), were also highly upregulated with fasting in zebrafish fast muscle (Table 1). Two genes with putative roles in autophagy were found to be upregulated during food deprivation [microtubule-associated protein 1 light chain 3 beta (map1lc3b) and neighbour of BRCA1 gene 1 (nbr1)] (Table 1). It is known that myofibrillar proteins are degraded to provide a source of amino acids for energy metabolism during prolonged fasting in teleosts (Johnston and Goldspink, 1973). However, the relative importance of the ubiquitin-proteasome degradation pathway, autophagy and other classes of proteases in mediating protein breakdown during normal protein turnover and short periods of fasting remains to be established. Cell cycle arrest and apoptosis also seems to be an important response in adapting to periods of limited energy supply in zebrafish, as evidenced by the upregulation of anti-proliferative protein gene transcripts [B-cell translocation gene 1 and 2 (btg1 and btg12), THAP domain-containing protein 1 (thap1) and pro-apoptotic genes (bbc3 and klf11b)] (Table 1).

Transition to an anabolic state 3 h after the meal resulted in major changes in the muscle transcriptome. Feeding was associated with the upregulation of transcripts for chaperone proteins (unc45b, ppig, pdip5, dnaja and stip1) including various heat shock proteins and associated proteins (hsp90a.1, hsp90a.2, hsdp1 and hspa5; Table 2). Molecular chaperones are essential for both the folding and maintenance of newly translated proteins and the degradation of misfolded and destabilized proteins (Zhao and Houry, 2007). In zebrafish, the chaperones Hsp90a and Unc45 are coregulated and involved in the folding of the globular head of myosin during myofibrillargenesis, associating with the Z-line once myofibrillar assembly is completed (Etard et al., 2008). The sequencing of subtractive cDNA libraries from fast skeletal muscle of Atlantic salmon fed either maintenance or satiating rations also revealed that expression of chaperone proteins indicative of unfolded protein response (UPR) pathways such as DNAJ4, HSPA1B, HSP90A and CHAC1 was an early response to increased food intake and growth (Bower and Johnston, 2010b). Accumulation of unfolded proteins can occur when the amounts of newly synthesized proteins exceeds that of the protein folding capacity of the endoplasmic reticulum (Okada et al., 2002). Taken together these results indicate that an increase in protein chaperone gene expression and activation of UPR pathways is a general response of teleost skeletal muscle following the transition from a catabolic to anabolic state.

The largest increase in transcript abundance found with feeding was for the gene encoding the enzyme ornithine decarboxylase 1 (odc1; 11.5-fold). Ornithine decarboxylase, a key metabolic enzyme of the polyamine biosynthesis pathway, also showed an increase in activity after feeding in fasted rat tissues (Moore and Swendseid, 1983). Polyamines have a number of biological functions including cell growth and apoptosis and act in a concentration-dependent manner (Larqué et al., 2007). The anabolic state was associated with enrichment of GO terms for mitochondrial translocase activity, initiation of protein synthesis [e.g. eukaryotic translation initiation factor 4A isoform 1A (eif4a1a), dual-specificity tyrosine-(Y)-phosphorylation regulated kinase 2 (dyrk2)] and mRNA processing, maturation and export protein genes [e.g. DEAD (Asp-Glu-Ala-Asp) box polypeptide 5 (ddx5) and small nuclear ribonucleoprotein 40 (U5) (snrnp40)]. Furthermore, pias4 and sae1, which are involved in the post-translational conjugation of SUMO (small ubiquitin-like modifier) to target proteins, were upregulated with feeding. Sumoylation of proteins may affect their stability, localization and activity (Geiss-Friedlander and Melchior, 2007). In contrast to the ubiquitin-conjugated proteins, sumo-conjugated proteins are not targeted for degradation and indeed sumoylation may function as an antagonist pathway to the ubiquitin-proteasome pathway (Desterro et al., 1998; Hay, 2005).

Feeding was also associated with significant changes in the expression of genes that may function as epigenetic switches coding for proteins that modify chromatin structure and alter the expression of suites of other genes. For example, SET and MYND domain containing 1b (smyd1b) transcripts increased 4.8-fold between the fed and fasted states (Tables 1 and 2). Smyd1 functions as a transcriptional repressor in mouse cardiac muscle (Gottlieb et al., 2002) and is required for normal skeletal muscle development in zebrafish (Tan et al., 2006). In the present study feeding also resulted in a 4.9-fold increase in transcript abundance for the euchromatic histone-lysine N-methyltransferase 1b (ehmt1b). The human orthologue of Ehmt1b was found to be part of the E2F6 complex and is probably involved in the silencing of MYC- and E2F-responsive genes (Ogawa et al., 2002), and is thought to play a role in G0–G1 cell-cycle transition.

The recruitment and hypertrophy of fast myotomal muscles in zebrafish was shown to be typical of teleosts (Johnston et al., 2009). The present study also demonstrates the utility of the zebrafish for mechanistic studies on the regulation of growth signalling pathways in teleost muscle, providing the advantages of a sequenced genome, commercially available molecular resources and low husbandry costs because of the small body size. However, caution should be applied in extrapolating all results from model to aquaculture fish species in the light of evidence for lineage-specific patterns of paralogue retention (Macqueen and Johnston, 2008a; Macqueen and Johnston, 2008b) and IGFBP gene expression (this study). The single meal paradigm also provides an interesting alternative to the use of continuous feeding regimes to investigate transcriptional regulation during the transition from a catabolic to an anabolic state.

Detailed information on protein extraction and western-blotting detection experimental protocols.

Skeletal muscle protein from five random replicates for each time-point was extracted by homogenization of 30 mg of tissue in 350 μl of 25 mmol l^–1 MES pH 6.0 containing 1 mol l^–1 NaCl, 0.25% (m/v) CHAPS, DNA/RNA nuclease (Invitrogen) and protease inhibitor cocktail (Invitrogen) using Lysing Matrix D (Qbiogene) in a FastPrep® machine (Qbiogene).

The optimal protein amount to be used for electrophoresis separation and transfer was empirically determined by applying from 10 to 60 μg of protein of a reference sample in duplicates and analysing the density of the Ponceau S staining (Fig. A1A,B). TotalLab software (Nonlinear Dynamics, Newcastle upon Tyne, Tyne and Wear, UK) was used to analyse the density of bands from Ponceau S staining and western blots. Protein saturation was observed when more than 30 μg of protein was loaded in the gel. The optimal amount was 20 μg considering both Ponceau-S-staining linearity (Fig. A1B) and total protein availability for the study.

The membranes used for optimal protein loading determination included a reference sample treated with calf intestinal alkaline phosphatase (A2356; Sigma) to confirm that the antibody targeted the phosphorylated moiety of the protein of interest (Fig. A1C). Actin intensity was better correlated with Ponceau S staining when compared with GAPDH (Fig. A1C) and was used to normalize differences in protein loading. Dephosphorylation of P-AKT significantly decreased its detection by P-AKT specific antibody (Fig. A1C) whereas no change in detection was observed for actin antibody in the dephosphorylated sample (Fig. A1C), confirming the specificity of the P-AKT antibody to the phosphorylated moiety of AKT.

The samples (20 μl, containing 20 μg of protein) were added to 6 μl of a solution containing 5 μl of 5× protein loading buffer and 1 μl 20× reducing agent (Fermentas), heated for 5 min at 95°C and loaded in NuPAGE® Novex 4–12% Bis-Tris gels (Invitrogen) and run at 120 V. A protein ladder of 10–250 kDa (Fermentas) and a reference sample were loaded in all gels to estimate the molecular mass of proteins of interest and serve as a normalization sample, respectively. Proteins separated by electrophoresis were transferred to a PVDF Immobilon-P Transfer Membrane (Millipore) at 25 V for 105 min. Successful protein separation and transfer were confirmed by Ponceau S staining (Sigma). PVDF membranes were blocked overnight at 10°C using 5% (m/v) nonfat milk (AppliChem) prepared in PBS (Sigma) containing 0.1% (v/v) Tween 20 (Sigma). Blocked membranes were incubated overnight at 10°C with the following primary antibodies (IgGs): P-AKT (1:1000 dilution; Cell Signaling, #4060), AKT (1:1000; Cell Signaling, #2966), actin (1:20,000; Sigma A2066) and GAPDH (1:30,000; Sigma, G9545). Probed membranes were incubated at 20°C for 1 h with the secondary antibody against mouse or rabbit IgG conjugated to horseradish peroxidase (both from Sigma and used at 1:60,000 dilution). Positive reactions were recorded by exposing Hyperfilm ECL (GE Healthcare) to the membranes after incubation for 1 min with ECL Western Blotting Detection Reagents (GE Healthcare) at room temperature. Experimental variations in the electrophoresis and transfer were normalized using a reference sample common to all membranes. The fold-change in phosphorylation of AKT in each time point was compared with the samples from –159 h.

 
• Myotome

  Tissues within the embryonic somites that form the trunk muscles.

 
• Neofunctionalization

  Gain of a new function by one paralogue, while the other retains the original function performed by the ancestral gene.

 
• Paralogue

  One of a pair of genes that arose from a common ancestral gene via an inherited mutation involving duplication of a region in the genome.

 
• Promyogenic transcription factors

  A family of four basic-helix-loop-helix proteins (myoD, myog, myf5 and myf6) that control the expression of hundreds of muscle-specific genes, acting as master regulators of myogenesis.

 
• Subfunctionalization

  The function previously performed by an ancestral gene is partitioned between the paralogues.

 
• Sumoylation

  Post-translational conjugation of proteins with a small ubiquitin-related modifier (SUMO).

 
• Unfolded protein response

  Cellular response to prevent accumulation of unfolded proteins. It mainly involves the processes leading to protein folding, protein export into the endoplasmic reticulum, degradation of misfolded proteins and apoptosis.

 
• Whole genome duplication

  An inheritable mutation involving a doubling of all sets of chromosomes and usually followed by massive loss of genes duplicates through evolutionary time. Occurred twice in a common ancestor of vertebrates, again in a common teleost ancestor and again in several other lineages; many paralogues generated from these events are still retained.

We thank Dr Neil Bower and Dr Daniel MacQueen for helpful suggestions and advice and Dr Vera Vieira-Johnston for assistance with fish husbandry.