Effect of refeeding on growth characteristics
The mean body weight of the trout was 132 g ± 6.0 and the condition factor was 1.6 ± 0.03 before fasting. At the end of the 30-days fasting period the mean body weight decreased to 121 g ± 5.5 and the condition factor to 1.3 ± 0.03. The mean body weight increased to 130 ± 6.3, 144 ± 7.8, 143 ± 6.7 and 183 g ± 14 and the condition factor to 1.4 ± 0.02, 1.5 ± 0.05, 1.5 ± 0.03, 1.6 ± 0.02, 4, 7, 11 and 36 days post refeeding respectively.
Changes in gene expression during a fasting-refeeding schedule: Overview
To screen for genes involved in muscle recovery growth, we undertook a time-course analysis of transcript expression in muscle of trout fasted for one month and then refed for 4, 7, 11 and 36 days. At each time point, eight to nine fish were sampled giving in total 43 separate complex cDNA targets that were hybridized to 43 microarrays (GEO accession number: GSE6841). Unsupervised hierarchical clustering of gene expression patterns from all samples produced a consistent grouping of the samples according to the fish feeding conditions (i.e. fasting and 4, 7, 11 and 36 days post-refeeding) (Fig 1). This validated the experimental design and allowed further analysis. To define those genes whose expression levels were significantly different in muscle from 4, 7, 11 or 36 days refed animals compared to muscle from fasted fish we used SAM analysis [10]. We therefore obtained approximately 2200 genes that were then hierarchically clustered using an average-linkage clustering [11]. This resulted in the formation of four major clusters of genes displaying distinct temporal profiles (Fig. 2). A similar clustering was obtained when using the K-means clustering (not shown). The first cluster was composed of genes with peak expression in muscle from starved fish, the second included genes overexpressed at 4, 7 and 11 days post-refeeding and down-regulated at 36 days post-refeeding (cluster II) the third was composed of genes with a later and more sustained induction (7–36 days post-refeeding) (cluster III) and the fourth (cluster IV) contained genes overexpressed only in 36 day refed trout muscle. These expression profiles and the clusters are available online as a browseable file [12]. The accession number of each spotted clones can be obtained by typing the corresponding uniq_id (clone name) in the nucleotide data base of the NCBI.
Genes with peak expression in starved trout muscle (cluster I)
Cluster I contained approximately 1000 genes showing high expression in the muscle of fasted fish and down-regulation after refeeding. In this cluster were notably identified two major markers of nutrient deprivation: the tuberous sclerosis component 2 (TSC2) an inhibitor of mTOR function and the translational repressor 4E-BP1. The most distinctive feature of cluster I was the presence of a large repertoire of genes involved in the regulation of protein degradation (Fig. 3). These genes participate either in the lysosomal system such as the cysteine protease cathepsins B, D and S or to the ubiquitin-proteasome pathway. In this latter class were several proteasome subunits, proteasome-associated proteins, ubiquitin, several ubiquitin-conjugating enzymes, ubiquitin carboxyl-terminal hydrolases and ubiquitin ligases. Among other genes involved in proteasome-mediated degradation were several COP9 signalosome complex subunits, cullin-3 and the Ariadne-2 protein homolog. Notably we did not observe any significant up-regulation of calpains in the muscle of fasted trout. Cluster I contained many genes involved in catabolic pathways and beta oxidation of fatty acids such as the short, medium and long chains of Acyl-CoA dehydrogenase.
Genes up-regulated 4 to 11 days after refeeding (cluster II)
Cluster II included approximately 550 genes with transient induction 4 to 11 days post refeeding. In this cluster were found more than 40 genes regulating mRNA synthesis, processing and turnover (Fig. 4). Among them were genes encoding small nuclear ribonucleoproteins, transcription initiation factors, RNA helicases of the DDX family, spliceosome-associated proteins such as the NHP2-like protein 1 as well as polyadenylation and export factors.
Cluster II also comprised a large set of genes involved in various aspects of translation (Fig. 5). These genes encode translation initiation factors (including the SUI1 domain containing density-regulated protein DRP), elongation and peptide chain release factors, peptidyl-tRNA hydrolases and several aminoacyl-tRNA synthases. In addition to these genes, a large group (Fig. 6) of genes whose products regulate protein folding and maturation was also found in cluster II. Among these latter were several heat shock proteins (HSP), endoplasmin, several subunits (alpha, beta, gamma, theta, zeta and epsilon) of the chaperonin-containing complex TCP1 and seven peptidyl prolyl cis/trans-isomerases encoding genes (including genes encoding FK506-binding proteins) that are known to catalyze the cis-trans isomerization of prolyl bonds in oligopeptides and various folding states of proteins
Consistent with the induction of genes involved in protein biosynthesis, cluster II included a large number of genes involved in ribosome formation. This group of genes which is presented in Fig. 7 included structural genes encoding ribosomal proteins as well as several genes whose products regulate ribosome biogenesis among which were fibrillarin, nucleolin, Brix domain-containing proteins, PUA domain-containing hypothetical protein MJ1432, RRp5, BMS1 and SAS 10 (something about silencing protein 10). Interestingly, in contrast to ribosomal structural genes which displayed about uniform expression during the 4 to 11 days post refeeding period, most of the genes regulating ribosome biogenesis sub-clustered together (Fig. 7, upper part) peaking at 4 days post-refeeding.
In accordance with an increase in cellular biosynthesis, a large number of genes involved in mitochondrial production of ATP grouped into cluster II (Fig. 8). Among them were several genes which are components of the oxidative phosphorylation system (chains 1, 2, 4 and 5 of the NADH-ubiquinone oxidoreductase, cytochrome c and ubiquinol-cytochrome c reductase complex subunits) as well as several genes of the ATP synthase complexes (alpha, beta and gamma chains). Along with these genes involved in energy production, we observed the induction of several genes important for mitochondrion formation or biogenesis such as TIM10 (a mitochondrial import inner membrane translocase), TOM34 (a mitochondrial import receptor subunit), voltage-dependant anion-selective channel protein 2 and 3 and the metalloprotease AFG3-like protein 1. Some enzymes of the mitochondrial matrix such as pyruvate dehydrogenase isoforms and succinyl-CoA ligase were also found in this cluster. Consistent with a stimulation of mitochondrial biogenesis the gene encoding the mitochondrial single-stranded DNA-binding protein (Mt-SSB) involved in mitochondrial DNA replication was also up-regulated.
Cluster II also contained several genes characteristically expressed during cell proliferation (Fig. 9). These genes are involved in DNA replication (mcm3, mcm6, DNA2-like homolog, DNA topoisomerase I and origin recognition complex subunit 4), progression through the cell cycle (G1/S-specific cyclin D2, CDK5 regulatory subunit, Wee1-like protein kinase, mitogen-activated protein kinase 9, M-phase induced phosphatase 2, pelota homolog and tumor protein D53 homolog), chromosome condensation (histone H2Az, H3, H5A, chromobox protein homolog 3 and regulator of chromosome condensation) or associated with proliferation (protein 2G4).
Among the most differentially regulated genes with miscellaneous functions and belonging to cluster II, we found uridine-cytidine kinase 2, various genes preventing cell apoptosis such as MCL1. Genes whose products mediate cAMP-dependant signalling such as the cAMP-dependent protein kinase beta catalytic subunit were also present. Among the genes belonging to cluster II with unknown functions we found C9orf32, the surfeit locus protein 2 and 4 encoding genes, the genes for the hypothetical protein ZK637.2 and for the hypothetical WD-repeat protein C1A6.02 in chromosome I.
Genes up-regulated 7–36 days after refeeding (cluster III)
Cluster III that contained approximately 480 genes up-regulated 7 to 36 days post refeeding was enriched in genes that encode components of the reticulum and Golgi apparatus such as triadin, and exostosin-2, and proteins involved in transport from the reticulum to the Golgi apparatus including golgi SNAP receptor complex member 1 (Fig. 10). Numerous genes in cluster III function in actin cytoskeletal rearrangements (including dynactin subunit 6, actin-like protein 3, actin-related protein 2/3 complex subunits and ankyrins) and organisation of the sarcomere (skeletal actin, myosins, tropomyosins, troponins, nebulin) (Fig. 11). Also were grouped in cluster III several collagen genes (collagen alpha 1(I), alpha 2(I), alpha 5 (IV) and alpha 1 (V) chains) that participate in the synthesis of the muscle extracellular matrix. Among the genes of cluster III with miscellaneous function were several Ras-related proteins such as Rab-24, Rab-26, and Rab-11b, several members of the glutathione S-transferase family, the bifunctional methylene tetrahydrofolate precursor and the myeloid leukaemia factor 1 (MLF1).
Genes up-regulated in the muscle of 36 days refed trout (cluster IV)
Cluster IV contained fewer (less than 200) genes compared to the other clusters. As observed for cluster III, cluster IV comprised several genes regulating reticulum and Golgi biogenesis and activity such as glycosyltransferases and the protein transport Sec24D (Fig. 10). A distinctive feature of cluster IV was to contain several genes encoding glycolytic enzymes (triose-phosphate isomerase, 6-phosphofructokinase, alpha-enolase and phosphoglycerate kinase). In addition, cluster IV contained several genes involved in lipid biosynthesis such as 24-dehydrocholesterol reductase precursor, ethanolamine kinase, Scavenger receptor class B member 1 and 1-AGP acyltransferase.
Functional categorization of genes contained in clusters I–IV as shown by GoMiner algorithm
To further assess the enrichment of a particular functional class of genes in each cluster we used the GoMiner algorithm [13]. Given that the GO data bases that served as input for GoMiner did not encompass the whole of genes present in our membrane only a subset of genes contained in cluster I–IV was selected for GoMiner analysis. Nevertheless, in agreement with the detailed identification of the genes contained in the different clusters (see above), the GoMiner notably highlighted an enrichment of cluster I for genes regulating protein degradation while cluster II was found to be enriched for genes involved in RNA metabolism (and more particularly rRNA metabolism), ribosome biogenesis, mitochondrion, translation and protein folding (Fig. 12). On the other hand, cluster III was particularly found to be enriched for genes participating in actin cytoskeleton and myofibrillar organisation (Fig. 12). This functional categorization was supported by probability values (Fig. 12) which yielded the measure of the likelihood that a particular biological process was overrepresented in a cluster compared with that expected by random selection from the SAM list.
Transcriptional regulators induced during muscle recovery growth
Some of the genes induced during muscle recovery growth were themselves regulators of transcription (Fig. 13). Among those present in cluster II, were two cyclic AMP-dependent transcription factors closely related to CREB3 and to ATF4 respectively, MEF2a, SF-1/fushitarazu homolog 1 related protein, cell growth regulating nucleolar protein LYAR, apoptosis-antagonizing transcription factor AATF and an unidentified zinc finger protein encoding gene. Cluster III included MTF1, the LIM/homeobox protein Lhx8, the homeobox proteins Hox-C9 and Hox-B1, the LIM domain containing transcription factor LMO4, the Homeodomain only protein Hop, a CREB1 related protein, sox11 as well as four unidentified zinc finger protein encoding genes. In addition to transcriptional regulators that trigger gene-specific expression by binding to sequence of promoters, we found other genes overexpressed during muscle recovery growth which potentially control gene expression by inducing histone modifications (SmyD1) or DNA modification (DNMT1 and DNMT2).
Validation of the microarray gene expression data
The accuracy and reliability of the results obtained with microarrays were tested by quantitative RT-PCR (Q-PCR) of 10 selected genes belonging to different functional classes. The gene expression levels obtained by Real-time PCR were normalized to that of the 18s. A good agreement was observed between microarray and Q-PCR analysis (Fig. 14) with discordant results obtained for only two genes. Thus, differential expression detected by microarray analysis is highly predictive of expression levels measured with an independent methodology such as Q-PCR (80% confirmation).