Comparative gene expression profiling between human cultured myotubes and skeletal muscle tissue

Skeletal muscle specimens and cultures

Paravertebral muscles biopsies were obtained from five subjects: females of age 12-15 years devoid of neuromuscular disease during surgery for idiopathic scoliosis. Informed consent and approval from the Ethics Committee of the Hospital Sant Joan de Déu (Barcelona) was obtained. Biopsies were embedded into RNAlater for RNA extraction or into cell culture medium. Cultures were prepared through an explant technique [11, 17]. Briefly, SM biopsy pieces were dissected under a stereomicroscope; fat, connective tissue and blood were removed; and the pieces were frozen in DMEM medium with 25% fetal bovine serum (FBS) and 6% DMSO. To set up the culture, first the biopsy pieces were spaced embedded in a semi-solid matrix, composed of 5 ml of DMEM/M-199 medium (3:1) with 37.5% FBS and 1.25 ml of human plasma (Sigma-Aldrich, St. Louis, MO, USA), overspread onto tissue culture plates, which were then incubated for 6 to 8 days to permit fibroblast outgrowth. The biopsy pieces were then removed from the matrix with forceps under the stereomicroscope, dissected and embedded in a matrix, composed of 1 ml of human plasma and 2 ml of 1.5% gelatin, spotted and stuck onto new dishes and overlaid with DMEM/M-199 medium (3:1) with 10% FBS, 10 μg/ml insulin, 4 mM glutamine, 25 ng/ml fibroblast growth factor, and 10 ng/ml epidermal growth factor, for 5 to 7 days to permit myoblast proliferation and migration. After myoblast generation, the biopsy pieces were removed, and eventually re-explanted, and myoblast monolayers were dissociated with trypsin and subcultured. Myogenic cells from the same explant were subjected to limited consecutive subcultures to avoid differences resulting from senescence [17]. Myotubes were derived from confluent myoblast cultures; immediately after initiation of myoblast fusion, medium was replaced by DMEM/M-199 medium (3:1) with 10% FBS and 10 μg/ml insulin, to further stimulate differentiation. Myotubes were used 7 days later. Myotube preparations were labeled B19, B22, B24, B25 and B26.

Immunocytochemistry

Cells grown on coverslips were fixed in 4% paraformaldehyde in PBS for 15 min, then washed in PBS, and incubated for 10 min in PBS containing 50 mM NH₄Cl, 10 min in PBS containing 20 mM glycine, 10 min in PBS containing 0.1% Triton X-100 and 30 min in PBS containing 10% FBS. Subsequently coverslips were incubated for 1 h at room temperature with a rabbit anti-desmin antibody (1:100; AB907 Millipore, Billerica, MA, USA). Primary antibody was detected with an Alexa Fluor-488 goat anti-rabbit antibody (1:500; Molecular Probes, Eugene, OR, USA). Both primary and secondary antibodies were diluted in blocking solution. Nuclei were stained with Hoescht (1 μg/ml) during secondary antibody incubation. After staining, samples were mounted on Mowiol mounting medium and analyzed with a Leica TCS SP2 confocal microscope. Images were processed using Photoshop CS software (Adobe Corp, San Jose, CA, USA). From 200 to 500 nuclei in each of the 5 myotube cultures were analyzed with ImageJ (Rasband WS, ImageJ, National Institutes of Health, Bethesda, MD, USA, http://rsb.info.nih.gov/ij/ and the mean ± SD values of the percentage of nuclei located in desmin-labeled myotubes were calculated.

Gene expression analysis

Ten samples, 5 from cultured SM cells, and 5 from SM biopsies were analyzed by DNA microarrays. Total RNA was extracted from SM biopsies with RNeasy fibrous tissue kit and homogenized using a TissueLyser, and from cultured myotubes with RNeasy kit (Qiagen, Valencia, CA, USA), with the same DNase treatment than in the RNeasy fibrous tissue kit, and homogenized with a Polytron. RNA samples were quantified with the RiboGreen RNA Quantification Kit (Molecular Probes), and monitored with the Agilent 2100 Bioanalyzer, to check high-quality RNA (RNA integrity = 8). We used HumanRef-8 v2.0 Expression BeadChips (Illumina, San Diego, CA, USA), which comprise probes to interrogate 22200 transcripts, based on the curated content of the NCBI Reference Sequence database, release 17.

An aliquot of 150 ng of total RNA was used to produce double-stranded cDNA, followed by transcription in vitro, and by cRNA labeling with biotin using the TotalPrep RNA Amplification kit (Applied Biosystems/Ambion, Austin, TX, USA) recommended by Illumina. This method is based on the protocol developed in J. Eberwine's laboratory [18]. The procedure consists of reverse transcription with an oligo(dT) primer bearing a T7 promoter using a reverse transcriptase engineered to produce high yields of first strand cDNA. The reverse transcriptase catalyzes the synthesis of full-length cDNA which then undergoes second strand synthesis and clean-up to become a template for in vitro transcription (IVT) with T7 RNA Polymerase. The IVT, along with biotin UTP, is used to generate hundreds to thousands of biotinylated, antisense RNA copies of each mRNA. These procedures were performed on an automated system primarily developed for the preparation of samples for Affymetrix arrays [19], which we adapted for the Illumina procedure. It is composed by a Microlab Star liquid handling system (Hamilton, Bonaduz, Switzerland) for reagent and sample pipetting and mixing, coupled with a thermocycler TRobot (Biometra, Goettingen, Germany) for incubations, and a microtiter plate reader SpectraMax M2 (Molecular Devices, Sunnyvale, CA, USA) for nucleic acids quantification with RiboGreen assay. The 10 samples analyzed in the present experiment were processed in a single robotic session. The labeled-cRNAs thus produced were measured with the Agilent 2100 Bioanalyzer to control for length. All cRNA sizes were close to 1200 nt and therefore eligible for hybridization onto the microarrays. Then, 750 ng of labeled-cRNAs was added to the hybridization mix, which contained control oligonucleotides in hybridization buffer. Then, 15 μl of each hybridization mix was dispensed on the BeadArrays. After hybridization (16 h, 58°C), the arrays were washed to remove non-hybridized material and were stained with streptavidin-Cy3, which bound to biotin. Scanning was performed using the BeadArray Reader, which provided intensity values for all transcripts. Bead redundancy allowed the calculation of a detection p-value which served at declaring the transcripts "significantly detected" or not. Signal intensities were extracted and summarized in the BeadStudio software. Data were expressed as absolute intensities, to which we applied a background correction. Background was calculated for each array by the average signal of the negative control probes. All microarray experiments were run at the same time. The Illumina HumanRef-8 BeadChip is a glass slide comprising 8 identical microarrays. We therefore used 2 slides for this experiment. To minimize possible batch effect that can occur between the two glass slides, we randomized the sample distribution in the following order: in one slide A (B19 in vitro), B (B26 in vitro), C (B25 in vivo), D (B24 in vitro), E (B24 in vivo), F (B22 in vivo), G (B26 in vivo) and H (B25 in vitro) and in the other slide A (B19 in vivo) and B (B22 in vitro). "In vitro" stands for cultured myotubes and "in vivo" means SM tissue biopsies. As shown in a recent study [20], randomization on Illumina slides greatly improves the results reducing the number of possible false positive occurrences that would lead to a misinterpretation of the data. Microarray data files have been deposited in GEO Omnibus [http://www.ncbi.nlm.nih.gov/geo/ GEO accession number GSE17503].

QuantiGene Plex 2.0 assay (Panomics/Affymetrix, Fremont, CA, USA) was used for confirmation of the microarray analysis. The advantage of this assay is that no RNA amplification is needed, unlike microarrays or quantitative PCR. A comparison to other gene expression platforms in the frame of the MAQC project demonstrated the sensitivity and accuracy of QuantiGene techniques [21]. This technology uses three sets of pooled oligonucleotides that function to capture, provide binding sites for signal amplification molecules, and stabilize the mRNA of interest. The capture extender (CE) pool contains oligonucleotides that are complementary to the mRNA of interest and to capture probes (CPs) covalently linked to carboxylated fluorescent-encoded micro spheres (Luminex, Austin, TX, USA). The label extender (LE) pool has oligonucleotides that contain sequences complementary to the mRNA of interest and to the amplifier molecules which carry a streptavidin phycoerythrin (SAPE) conjugate. The blocker (BL) pool contains oligonucleotides that are complementary to the mRNA where neither CE nor LE bind and it protects the mRNA by maintaining an intact RNA-DNA hybrid. The micro spheres (or beads) thus prepared are analyzed with a Luminex reader, which utilizes two lasers, the first for bead identification, measuring the fluorescent signal specific to each bead type, and the second for mRNA quantification, determining the fluorescence level of SAPE and therefore the expression level of the transcript. The assay was performed according to the manufacturer's instructions, accessible at: http://panomics.com/downloads/UM13075_RevB_QGP2_080430.pdf. Briefly, the RNAs extracted from tissue and cell samples were deposited on a 96-well plate along with the QuantiGene hybridization solution (containing RNase-free water, lysis mixture, blocking reagent, proteinase K, capture beads and probe set). The hybridization plate was then sealed and placed in a shaking incubator (Vortemp 56 Shaker/Incubator, Labnet International, Inc., Woodbridge, NJ, USA) at 54°C and 600 rpm for overnight hybridization (18-22 h). After centrifugation to remove condensation, the hybridization mixture was transferred to a pre-wetted filter plate, where the mixture was filtered and washed three times in wash buffer. All washing steps were performed with a MultiScreen Vacuum Manifold (Millipore) at low pressure, and after the final wash, the filter plate bottom was blotted to remove any remaining liquid to prevent leakage. The samples were then incubated successively in three working solutions: pre-amplifier and amplifier, biotinylated probes and SAPE. Two washes were performed between each incubation to remove unbound reagents. The plate was covered with aluminum foil during SAPE incubation to prevent photo bleaching. For Luminex measurement, after a calibration of the Luminex, the plate was placed in the instrument. Prior to the measurement, each bead type was identified in the Luminex software in order to specify to which transcripts the beads corresponded. At the start of the measurement, the content of each well was aspirated and pushed towards the lasers. As occurs in a flow cytometer, the beads passed successively in front of the two lasers. Beads were identified, and fluorescent signals were measured. The final outcome was a raw data file containing signal intensity and background level for each gene in each sample. Data were analyzed using the transferrin receptor (TFRC) gene as the reference gene, since its expression was similar within the two examined groups, cultured myotubes and SM biopsies.

For reverse transcription (RT) and real-time PCR, an aliquot of 0.5 μg of total RNA was retro-transcribed (RT) with TaqMan reverse transcription reagents from Applied Biosystems (Foster City, CA, USA) using random hexamers and in a 25 μl total volume reaction. Real-time PCR was performed on 1 μl of the RT reaction mixture with the TaqMan universal PCR master mix and TaqMan Gene Expression Assays using the ABI PRISM 7700 sequence detection system. Probes for PYGM (Hs00194493_m1), UCP3 (Hs00243297_m1), THY1 (Hs00264235_s1) and B2M (Hs99999907_m1) were from Applied Biosystems. The threshold cycle (Ct) values of the reference probe beta-2 microglobulin (B2M) were subtracted from each target probe Ct values (ΔCt), the ΔCt values of the SM biopsies were subtracted from the ΔCt values of myotube cultures. Data were expressed as mean values of 2^-ΔΔCT for upregulated genes and 1/2^-ΔΔCT for downregulated ones and the significance of differences was estimated by a paired t-test.

Statistical analysis and filtering

After the scanning of the microarrays, BeadStudio generated a dendrogram to control for the sample clustering. Subsequent steps of quality control (QC), statistical analysis and filtering of microarray data were carried out with GeneSpring GX software v10.0.2 (Agilent Technologies, Santa Clara, CA, USA). Raw data with background subtracted were loaded from Illumina BeadStudio to GeneSpring GX. This step was followed by quantile normalization and log2 transformation. The samples were then grouped into two conditions which we called "in vitro" and "in vivo", corresponding respectively to the SM cultures and the tissue biopsies. GeneSpring GX was used as the first step for QC of data on all the samples with two unsupervised methods, Pearson correlation and Principal Component Analysis (PCA). The PCA vector space transformation was used to reduce multidimensional data sets to lower dimensions for analysis to assess the behavior of the samples and to identify or confirm possible outliers. Within the quality control step, Genespring GX enables to compute the 3 first principal components. The percentage of total variance that each of these 3 principal components capture is shown in the legend of Additional File 1.

To assess which genes were differentially expressed between the two conditions examined, we performed a corrected paired t-test on all probes. Because of the small number of biological replicates, we used the Benjamini-Hochberg false discovery rate (FDR) method controlling for false positives, in order to select genes with the lowest FDR. This method is less stringent than other corrections such as the Bonferroni, and provides a good balance between discovery of statistically significant genes and limitation of false positive occurrences. A corrected p-value cutoff of 0.01 was used to select the regulated genes with the lowest FDR. A fold change value was also computed by GeneSpring GX to assess the level and the direction of the gene regulation. Fold change was calculated as the absolute ratio of normalized intensities between the mean values of the samples in the two conditions, cultured myotubes and SM tissue biopsy, grouped.

Data from QuantiGene Plex were also loaded into GeneSpring GX software. Originally GeneSpring GX was designed for gene expression microarray experiments and uses the main commercially available arrays (Affymetrix, Illumina, Agilent and CodeLink). However, it is possible to configure a "custom" array. The QuantiGene Plex assay was therefore configured in GeneSpring GX. After normalization to the TFRC control gene, data were log2 transformed, and significance of differential expression was assessed by a paired t-test.

Bioinformatic analysis

Functional annotation and classification analysis was also carried out with GeneSpring GX program, running a Gene Ontology (GO) analysis on the gene set selected with the statistical analysis described above. GeneSpring GX computed a p-value to quantify the significance in the GO analysis. This p-value is the probability that a random subset of x genes drawn from the total set of n genes will have y or more entities containing a given GO term. This probability is described by a standard hypergeometric distribution. GeneSpring GX uses the hypergeometric formula from first principles to compute this probability. Since a large number of hypotheses will be tested, some form of correction is required. GeneSpring GX addresses this issue using the Benjamini-Yekutieli correction, which takes into account the dependency among the GO terms. In the GO analysis done in GeneSpring GX, a low p-value therefore implies that the given GO term is enriched. The Gene Ontology data in GX 10.0.2 is compiled from version 1.2 of OBO (Open Biomedical Ontologies).

Data were interpreted using Ingenuity Pathways Analysis (IPA) (Ingenuity Systems, Redwood City, CA, USA) http://www.ingenuity.com. The list of significantly regulated genes selected by the microarray analysis described above was loaded in IPA with the following criteria: Reference set: HumanRef-8 v2.0; Direct and Indirect relationships included; filtered by species (human), and in a second round, by tissue (skeletal muscle). Then IPA computed the data to generate significant networks of genes that are associated with particular biological functions, diseases, and molecular processes. It showed also the canonical pathways stimulated by the experiment. IPA used the right-tailed Fisher's exact test to extract significant pathways. IPA canonical pathways are well-characterized metabolic and cell-signaling pathways coming from articles, reviews, books, and KEGG Ligand.

Gene cluster analysis was performed with the REEF (REgionally Enriched Features in genomes) program, which scans the genome using a sliding window approach, and calculates the statistical significance of each window using the hypergeometric distribution and the false discovery rate [22].

Genes differentially expressed in cultured versus skeletal muscle tissue

Microarray analyses were performed on the 5 SM cultures and the 5 SM biopsies. We confirmed that myotubes were the prevalent cell type in each of the SM cultures by immunostaining for desmin, a muscle-specific intermediate filament protein [23] (Additional file 1, Figure S1). Most of the Hoescht-stained nuclei were located in desmin-labeled myotubes: B19 (78.3 ± 3.12%), B22 (61.4 ± 6.60%), B24 (61.8 ± 8.36%), B25 (83.6 ± 14.50%) and B26 (79.7 ± 5.05%). The microarray analysis showed that in the cultured myotubes compared to the tissue biopsies 1260 transcripts, which correspond to 1216 nuclear genes, were differentially expressed (selection criteria: absolute fold-change value > 2 with a corrected p-value of p < 0.01). Of these, 583 were downregulated and 633 were upregulated. The complete list of regulated genes is shown in Additional file 2, Table S1. On the basis of the correlation analysis performed, Additional file 3, Figure S2 shows that the samples were clearly separated based on their condition cultured myotubes or SM biopsy. The dendrogram also illustrates that the paired samples were differently distributed in the two conditions. PCA and correlation matrix procedures show that the two experimental conditions were well separated and there were no particular outliers in the experiment. Moreover, they highlight that the variability was higher within the SM culture group (in vitro). Pearson coefficients are also provided; the average Pearson coefficient was 0.97 for the SM biopsies and 0.93 for the SM cultures. PCA, correlation matrix and dendrogram also confirmed that there was no significant batch effect between the two Illumina slides in our experiment. In an attempt to restrict differentially expressed genes to those expressed in SM, microarray data were filtered according to the IPA knowledge-base, so that genes expressed in SM were selected. SM-specificity is shown in Additional file 2, Table S1. IPA knowledge about mRNA expression in tissues is based on two sources: the GNF body atlas [24], and literature findings. Among the downregulated genes, 384 passed this filter along with 318 of the upregulated genes. The filtering served as a means for identifying genes that are expressed in other cell types, present in the SM tissue, within the most regulated genes. For instance, hemoglobin HBB and HBA2 genes, which are mainly expressed in erythroid cells [25], from the most downregulated group and the THY1 gene, which is mainly expressed in human fibroblasts, neurons and endothelial cells [26], from the most upregulated group. Remarkably, the ratios between the normalized signal of the THY1 and desmin gene markers in the SM cultures were in a narrow range: B19 (1.94), B22 (1.12), B24 (1.42), B25 (1.28) and B26 (1.44), again indicating that fibroblast presence did not vary widely within SM cultures. However, the THY1/desmin ratios were from 1.3- to 3.2-fold higher in the SM cultures than in the SM biopsies.

The complete set of differentially expressed genes were classified, separately for downregulated and upregulated genes, according to GO (Additional file 4, Table S2A). With respect to the cellular component, genes that were downregulated in cultured myotubes were mainly associated with cytoplasm and enriched in mitochondria. The over-represented biological processes were metabolism, with the largest group of regulated genes, and muscle-system/contraction. With respect to metabolism, genes were mainly involved in the generation of precursor metabolites and energy, cellular respiration and oxidative phosphorylation; quinone cofactor metabolism was also affected. The most augmented molecular function was oxidoreductase activity. Among the genes that were upregulated in cultured myotubes, the enriched cellular components were cytoplasm (with the largest group associated with endoplasmic reticulum (ER)) and the extracellular matrix. The enriched molecular function here was oligosaccharyl transferase activity. The augmented biological process was modification of an amino acid residue in a protein: N-linked posttranslational glycosylation at asparagine residues. GO annotations of the SM-expressed subset of downregulated genes (Additional file 4, Table S2B) did not differ greatly from those of the whole gene set, except for the occurrence of the following over-represented GO annotations: cytoskeletal protein binding (molecular function); fatty acid beta-oxidation and organic acid metabolism-related terms (metabolic processes); and muscle development (biological process). There was no GO annotation enrichment within the upregulated SM-expressed subset for molecular function and biological process categories (Additional file 4, Table S2B).

Among the most upregulated genes detected in this study (Table 1) are well-known regulatory factors such as CDKN1A/p21, TOP2A and HIF1A. Several genes are related to the extracellular matrix, such as those encoding: the Ca(2+)-binding proteoglycan SPOCK1; isoforms of lysyl hydroxylase (PLOD3, PLOD2) [28], an enzyme that catalyzes the formation of hydroxylysine in collagens; CSGlcA-T, which is involved in the synthesis of chondroitin sulphate as a glucuronyltransferase [29]; the metalloproteinase MMP14; TIMP2, an inhibitor of the metalloproteinase type IV collagenolytic activity [30] and the protease inhibitor SERPINE2 [31]. The NFASC gene, which promotes axon subcellular targeting and synapse formation [32], was another highly induced gene, suggesting stimulation of neuronal signaling. Two genes are related to stress: the gene encoding the putative glutathione peroxidase GPX8, an enzyme involved in protection against oxidative stress [33], and the DNAJC10/ERDJ5 gene. The PTX3 gene is related to inflammation [34]. Other genes are related to sphingomyelin metabolism such as those encoding: GBA, a lysosomal enzyme that catalyzes glucosylceramide breakdown [35], and SGMS2, an enzyme that produces sphingomyelin [36]. Finally, a group of genes encodes proteins that either have an undefined (i.e. TMEM200A), or an unknown function in the SM tissue (KIAA1199, PSD3, CCDC80/URB/SSG1/DRO1 and LPIN2).

Genome mapping

Since members of the same gene family or genes that are coordinately expressed might form clusters on chromosomes, we tested the possibility that coregulated genes showed spatial clustering. We applied the GeneSpring GX program to the entire set of 1216 differentially expressed genes to identify the chromosome and genomic position. Of these genes, 983 had chromosome annotations that were distributed throughout all chromosomes (Figure 1A), although the highest number was on chromosome 1. We applied the REEF program to the whole set of 983 differentially expressed chromosome-annotated genes to search for local clustering. Three gene clusters were recognized using the REEF program (Figure 1B). The cluster on chromosome 7 (Figure 1C) contained the largest number of genes and included five members of the GTPase of the immunity-associated protein (GIMAP) gene family, plus ASB10 and CSGlcA-T. Except the CSGlcA-T gene, all genes in the cluster are highly downregulated. The genes encoding GIMAP4, GIMAP5, ASB10 and CSGlcA-T are expressed in SM (Additional file 2, Table S1). No gene families were detected in clusters on chromosomes 1 and 20.

Bioinformatic analysis and interpretation of microarray expression data

To reveal regulated metabolic and signaling pathways the complete set of differentially expressed genes was analyzed by IPA. According to IPA, 10 signaling and 13 metabolic canonical pathways are regulated in myotubes versus SM tissue (Table 3). The most regulated metabolic pathways are related to mitochondria: the citrate cycle, oxidative phosphorylation and ubiquinone biosynthesis. Consequently, mitochondrial dysfunction is the most regulated signaling pathway. Network analysis of this pathway illustrates known interactions between the identified genes and overall pathway downregulation (Figure 2). Included genes are those encoding: components of the oxidative phosphorylation system (most of which are in complex I, some are components of complexes II, III and IV and cytochrome c CYCS); UCP3, an inner mitochondrial membrane transporter that dissipates the proton gradient [37]; CPT1B, which is associated with the outer mitochondrial membrane and facilitates the mitochondrial import of long chain fatty acids, and ACACB, which converts acetylCoA into malonylCoA to inhibit CPT1B [38]; MAOB, an enzyme involved in the degradation of biogenic amines [39]; the pyruvate dehydrogenase E1-alpha subunit (PDHA); and alpha-ketoglutarate dehydrogenase (KGDH). All these genes are expressed in SM except the NDUFAB1, NDUFA4, COX10 and UQCRB genes according to the IPA knowledge-base. Additional regulated metabolic pathways involve glucose, amino acid and monocarboxylic acid metabolism. Results were similar when the subset of SM-expressed genes was analyzed. However, the pathways One carbon pool by folate and Biosynthesis of steroids were not enriched, whereas the pathways Fatty acid metabolism, Phenylalanine metabolism and Glyoxylate, dicarboxylate metabolism were enriched with -log(p-values) of 2.38, 2.13 and 1.81, respectively.

	Score	Ratio	Number of Focus Genes
Signaling Pathways
Mitochondrial Dysfunction	7.88	17.60%	30
Integrin Signaling	3.52	14.50%	28
Hepatic Fibrosis/Hepatic Stellate Cell Activation	2.38	14.10%	19
Caveolar-mediated Endocytosis	2.02	14.80%	12
Regulation of Actin-based Motility by Rho	1.77	13.00%	12
Actin Cytoskeleton Signaling	1.46	10.40%	23
Apoptosis Signaling	1.35	12.10%	11
Circadian Rhythm Signaling	1.31	15.60%	5
p53 Signaling	1.29	12.60%	11
PTEN Signaling	1.29	11.70%	11
Metabolic Pathways
Citrate Cycle	6.27	20.30%	12
Oxidative Phosphorylation	4.61	15.90%	25
Ubiquinone Biosynthesis	3.66	12.50%	13
Glycolysis/Gluconeogenesis	3.39	12.10%	17
Pyruvate Metabolism	3.15	9.66%	14
Tyrosine Metabolism	3.01	6.45%	12
Butanoate Metabolism	2.61	8.53%	11
One Carbon Pool by Folate	2.51	15.80%	6
Glycosaminoglycan Degradation	1.85	9.84%	6
Biosynthesis of Steroids	1.78	4.72%	6
Alanine and Aspartate Metabolism	1.67	9.30%	8
Propanoate Metabolism	1.54	7.14%	9
Lysine Degradation	1.34	6.25%	9

Canonical pathways that have significant associations with the whole set of regulated transcripts were assessed using IPA software. The scores are shown as -log(p-value) in order to show positive numbers. The p-value was calculated by IPA using Fisher's Exact Test. A threshold value of 1.3 was selected (p-value < 0.05). The first column lists the signaling or metabolic pathways; the second the score value; the third the ratio, which represents the proportion of genes from the dataset over the total number of genes in the pathway; and the fourth the number of focus genes.

Other regulated signaling pathways in the complete set of the differentially expressed genes relate to cell interaction with the extracellular matrix (integrin signaling, hepatic fibrosis), cell communication with its environment (caveolar-mediated endocytosis) and actin-cytoskeleton signaling. In addition, circadian rhythm signaling was altered, with five genes all downregulated. Finally, apoptosis and the signaling of phosphatase and tensin homolog (PTEN) (Figure 3) were regulated. PTEN dephosphorylates the signaling lipid phosphatidylinositol (3,4,5)-trisphosphate and affects cellular processes related to cell proliferation, apoptosis and muscle contractility [40]. All signaling pathways except the p53-pathway were significantly and similarly regulated when the subset of SM-expressed genes was analyzed (data not shown).