Tumor Necrosis Factor Alpha and Insulin-Like Growth Factor 1 Induced Modifications of the Gene Expression Kinetics of Differentiating Skeletal Muscle Cells

Introduction TNF-α levels are increased during muscle wasting and chronic muscle degeneration and regeneration processes, which are characteristic for primary muscle disorders. Pathologically increased TNF-α levels have a negative effect on muscle cell differentiation efficiency, while IGF1 can have a positive effect; therefore, we intended to elucidate the impact of TNF-α and IGF1 on gene expression during the early stages of skeletal muscle cell differentiation. Methodology/Principal Findings This study presents gene expression data of the murine skeletal muscle cells PMI28 during myogenic differentiation or differentiation with TNF-α or IGF1 exposure at 0 h, 4 h, 12 h, 24 h, and 72 h after induction. Our study detected significant coregulation of gene sets involved in myoblast differentiation or in the response to TNF-α. Gene expression data revealed a time- and treatment-dependent regulation of signaling pathways, which are prominent in myogenic differentiation. We identified enrichment of pathways, which have not been specifically linked to myoblast differentiation such as doublecortin-like kinase pathway associations as well as enrichment of specific semaphorin isoforms. Moreover to the best of our knowledge, this is the first description of a specific inverse regulation of the following genes in myoblast differentiation and response to TNF-α: Aknad1, Cmbl, Sepp1, Ndst4, Tecrl, Unc13c, Spats2l, Lix1, Csdc2, Cpa1, Parm1, Serpinb2, Aspn, Fibin, Slc40a1, Nrk, and Mybpc1. We identified a gene subset (Nfkbia, Nfkb2, Mmp9, Mef2c, Gpx, and Pgam2), which is robustly regulated by TNF-α across independent myogenic differentiation studies. Conclusions This is the largest dataset revealing the impact of TNF-α or IGF1 treatment on gene expression kinetics of early in vitro skeletal myoblast differentiation. We identified novel mRNAs, which have not yet been associated with skeletal muscle differentiation or response to TNF-α. Results of this study may facilitate the understanding of transcriptomic networks underlying inhibited muscle differentiation in inflammatory diseases.


Introduction
TNF-α levels are increased during muscle wasting and chronic muscle degeneration and regeneration processes, which are characteristic for primary muscle disorders. Pathologically increased TNF-α levels have a negative effect on muscle cell differentiation efficiency, while IGF1 can have a positive effect; therefore, we intended to elucidate the impact of TNF-α and IGF1 on gene expression during the early stages of skeletal muscle cell differentiation.

Methodology/Principal Findings
This study presents gene expression data of the murine skeletal muscle cells PMI28 during myogenic differentiation or differentiation with TNF-α or IGF1 exposure at 0 h, 4 h, 12 h, 24 h, and 72 h after induction. Our study detected significant coregulation of gene sets involved in myoblast differentiation or in the response to TNF-α. Gene expression data revealed a time-and treatment-dependent regulation of signaling pathways, which are prominent in myogenic differentiation. We identified enrichment of pathways, which have not been specifically linked to myoblast differentiation such as doublecortin-like kinase pathway associations as well as enrichment of specific semaphorin isoforms. Moreover to the best of our knowledge, this is the first description of a specific inverse regulation of the following genes in myoblast differentiation and response to TNF-α: Aknad1, Cmbl, Sepp1, Ndst4, Tecrl, Unc13c, Spats2l, Lix1, Csdc2, Cpa1, Parm1, Serpinb2, Aspn, Fibin, Slc40a1, Nrk, and Mybpc1. We identified a gene subset (Nfkbia, Nfkb2, Mmp9, Mef2c, Gpx, and Pgam2), which is robustly regulated by TNF-α across independent myogenic differentiation studies.

Introduction
Myoblast differentiation is a multistep process, which involves proliferation, exit from the cell cycle, migration, alignment, and fusion into multinucleated myotubes [1,2]. This process is mediated by a cascade of changes in gene expression [3] and is essential for muscle repair. Myoblast differentiation can be promoted by growth factors such as IGF1 [4], but it is impaired by elevated concentrations of inflammatory cytokines such as TNF-α [5][6][7]. IGF1 increases myoblast differentiation via both hyperplasia and hypertrophy [5]; however, the underlying regulatory mechanisms at the transcriptomic level are poorly understood. The inhibitory effect of inflammatory levels of TNF-α on myoblast differentiation and muscle repair is associated with cachectic muscle wasting [8,9] and several chronic diseases or muscular disorders [10][11][12]. Moreover, human aging is associated with muscle inflammation susceptibility [13]. The molecular mechanisms leading to inhibition of mybolast differentiation because of elevated TNF-α concentrations are highly complex, involving modulations at the mRNA level [7] as well as epigenetic implications [3], among others. The molecular signaling pathways mediating the inhibitory effect of TNF-α on myogenic differentiation are not yet completely elucidated. To date, cachectic muscle wasting is an incurable complication [14]; however, several therapeutic strategies are currently being investigated to promote skeletal muscle growth and regeneration [15]. Therefore, the current study addressed the mRNA expression kinetics within the first 24 h up to 72 h of differentiation and concomitant response to IGF1 and TNF-α ( Fig 1A). Kinetic expression data obtained from the current study and pathway association analyses as well as principal component analyses and the self-organizing tree algorithm-based clustering provide valuable insights into the molecular signaling mechanisms, which mediated the effect of TNF-α and IGF1 (Fig 1B).

RNA extraction and quality control
The cells were washed with PBS and lysed in Trizol (Invitrogen, Life Technologies GmbH) to harvest approximately 2 × 10 6 cells per 1.5 mL Trizol. The samples were homogenized by vigorous shaking. A total of 0.3 mL chloroform was added per 1 mL Trizol, and the samples were mixed for 15 s by vigorous shaking. Phase separation was allowed by placing the samples on the bench top for 10 min followed by centrifugation at 12,000 ×g for 25 min at 4°C. The upper aqueous phase was transferred to a fresh tube. A total of 0.75 mL isopropanol per 1 mL Trizol was added, thoroughly mixed, and incubated for 10 min and centrifuged at 12,000 ×g at 4°C to precipitate the RNA. The RNA pellet was washed with 0.5 mL ethanol per 1 mL Trizol and centrifugation at 12,000 ×g for 10 min at 4°C. The supernatant was aspirated and the sediment was air dried for 15 min. Total RNA was dissolved in nuclease-free water and photometrically quantified by NanoDrop 1000 ND-1000 (Peqlab, Erlangen, Germany) measurement. Moreover, approximately 250 ng RNA were analyzed on 1% Agarose gel with a 1-KB marker for overall RNA quality control.

Gene expression profiling by hybridization microarrays
Analysis of gene expression was performed with GeneChip Mouse Gene 1.0 ST Arrays (Affymetrix, Santa Clara, CA, USA) following the manufacturer's instructions. Triplicate samples were analyzed for each time point and treatment. A total of 250 ng total RNA were reverse transcribed using the Ambion WT Expression Kit (Ambion, Life Technologies GmbH, Darmstadt, Germany) including the GeneChip Poly-A RNA Control Kit (Affymetrix) according to the manufacturer's instructions. The cDNA yield and size distribution were determined and the cDNA was then purified, fragmented, labeled and hybridized applying the GeneChip WT Terminal Labeling and Controls Kits (Affymetrix) following the manufacturer's instructions. For washing and staining steps the GeneChip Hybridization, Wash, and Stain Kit (Affymetrix) was used according to the manufacturer's instructions. Fluorescence signals were acquired with the AGCC Scan Control Software. Affymetrix CEL files were read, normalized and summarized using the RMA method [17] as implemented in the Affymetrix apt package. Probe sets were retained if they had at least two "detected above background" present calls in at least one experimental group. GeneChip Mouse Gene 1.0 ST Array data were MIAME [18] compliant and was submitted to the ArrayExpress database (www.ebi.ac.uk/arrayexpress) [19], a publicly available repository consistent with the MIAME guidelines. Data are available with the ArrayExpress accession number E-MTAB-3474.

Reverse transcription of RNA to cDNA for individual expression analysis
Validation of mRNA expression results obtained by microarray profiling was performed by individual reverse transcription, using gene-specific reverse primer and subsequent qPCR analysis. Reverse transcription was performed using 100 ng total RNA for each reaction and the components of the miScript Reverse Transcription Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions.

Statistics
Significant differences between mRNA expressions measured by individual RT-qPCR analyses were determined using a parametric unpaired two-tailed student's t-test. Differential expression of genes measured by microarray analysis was determined with LIMMA (Linear Models for Microarray Data) [23] using a factorial design with treatment and time-point as factors. Pairwise comparisons were extracted for all combinations of consecutive time points for the same treatment and between all treatments at the same time point. We clustered expression profiles of all samples for all probesets that were significantly different (fdr < 0.01 and log2 fold change > 1) in at least one pairwise comparison with the self-organizing tree algorithm (SOTA) method [24]. Dynamic PCA was performed within GenEx Software (MultiD Analyses AB, Gothenburg, Sweden) comparing myoblasts to the other treatment groups. Genes were filtered on the basis of p values to identify the most relevant genes explaining the observations [25]. Hierarchical clustering and heatmap generation was performed using GenEx Software (MultiD Analyses AB).

Bioinformatic analysis of data
The Genomatix Pathway System (GePS) within the Genomatix Software Suite (Genomatix Software GmbH, Munich, Germany) uses pathway data from the Pathway Interaction Database [26]. We applied GePS analysis for identifying significant pathway associations and gene ontology terms of input genes derived from our gene expression profiling data.

Immediate response to differentiation and TNF-α treatment
The effect of myoblast differentiation as well as the response to TNF-α or IGF1 exposure modified gene expression patterns, which resulted in separation of treatment groups by hierarchical clustering (S1 Fig) or principal component analysis (PCA) (Fig 2A, 2D and 2G). Genes which separated by principal component analyses (S1 Table) were analyzed for enrichment of pathway associations ( Table 1). As early as 4 h after the induction of differentiation and treatment, the gene expression pattern of myoblasts, myotubes, and myotubes treated with TNF-α clearly separate by PCA (Fig 2A, 2D and 2G), whereas the effect of IGF1 became clearly distinct after 24 h (Table 1). Principal component analysis showed that the differentiation effect had the strongest impact on the proportion of variance followed by the effect of TNF-α treatment whereas IGF1 treatment had a minor effect (Fig 2B, 2E and 2H). Dynamic PCA identified a subset of 61 genes after 4 h of treatment ( Fig 2C. and Table 2), a subset of 27 genes and two microRNAs after 12-h treatment ( Fig 2F, Table 2) as well as a subset of 19 genes and two microRNAs after 24-h treatment ( Fig 2I, Table 2); in each case, these were sufficient to separate Most of the variance is described by PC 1 followed by PC 2 and PC 3. PC 1 explaines most of the variance of myocyte differentiation while PC 2 represented most of the variance induced by TNF-α whereas PC 3 characterized most of the variance caused by IGF1 treatment. Moreover, the treatment groups by principal components. Furthermore, differential gene expression kinetics revealed dynamic time-specific changes of gene regulation (S2 Table). Moreover, certain genes, which were among the 20 most differentially expressed ones, were regulated immediately after the induction (4 h) as well as during very early (12 h) and early (24 h) myoblast differentiation, such as upregulated Adamts5, Ccdc141, Fibin, and downregulated Serpinb2, Gm12824, Npr3, and Sp7 (S2A Table). TNF-α treatment upregulated several genes immediately after induction (4 h), which were still upregulated 12 h as well as 24 h after induction (S2B Table); these included Ccl2, Ccl7, Nfkbie, Tnfaip3, Nfkbia, Bcl3, Vcam1, Slc2a6, Cxcl10, and Mmp9. IGF1 treatment did not result in differentially expressed genes derived from microarray analysis when compared with nontreated myotubes. However, when IGF1-treated samples were compared with TNF-α, several genes were inversely regulated (S2C Table).

Coregulation of gene sets
Self-organizing tree algorithm (SOTA) analyses of gene expression over time (0 h, 4 h, 12 h, 24 h, and 72 h) revealed significant coregulation of gene sets in response to the induction of differentiation and confirmed the immediate response to TNF-α or IGF1 treatment (Fig 3). Clustered cohorts of the gene expression pattern showed a distinct shift in expression levels as early as 4 h after induction of differentiation and TNF-α or IGF1 treatment (Fig 3). The majority of differentially expressed genes fitted in one of the six clusters as shown in  Table) are shown in the supporting information section. The data collected from the current study suggests that >80% of the differentially expressed genes clustering in cohorts are assigned to three SOTA clusters: cluster A, which includes genes upregulated during very early differentiation ( Fig 3A); cluster B, which represents genes upregulated during late differentiation ( Fig 3B); and cluster C, which visualized cohorts of genes downregulated as early as 4 h after induction of differentiation ( Fig  3C). We examined whether gene expression transcripts with similar regulation also demonstrate related biological implications. Analysis of signal transduction pathway associations and gene ontology annotation class "biological processes" (S4 Table) demonstrate that pathways such as cyclin G1 and semaphorin pathway were enriched in cluster A (early myotubes genes up). Cluster B (late myotube genes up) was enriched for genes with a function in the ryanodine receptor and calcineurin pathway for example, whereas cluster C (early myotube genes down) showed enrichment of genes e.g. involved in dual-specific phosphatase and fibroblast growth factor pathway as well as TGFbeta signaling. Genes with a function in, for example, the cyclindependent kinase inhibitor 2 pathway were enriched in cluster D (TNF induced, suppressed in late myotubes), while cluster E (specifically induced by TNF) overrepresented genes with a function in pathways such as NFkappaB and tumor necrosis factor. Finally, cluster F (late myotubes genes down) was enriched for pathways such as nuclear factor (erythroid derived 2) like 2, tumor protein p 53, and other cell cycle-related pathways.
Specific signaling pathway regulation during myoblast differentiation and TNF-α response Signal transduction pathway associations were enriched in a time-dependent manner (Table 3) for the effect of myoblast differentiation. Immediately after the induction of differentiation results from dynamic principal component analyses (dPCA) (group selection myoblasts) are shown for gene expressions (C) 4 h, (F) 12 h, and (I) 24 h after induction of differentiation and treatment. DPCA identified a minimal subset of genes, which could describe the treatment effects (see Table 2) and separate the effects by principal components. (4-h differentiation), mothers against DPP homolog (Smad) and TGFbeta pathway associations were significantly enriched amongst others (Table 3). During very early differentiation (12-h differentiation), signaling pathways such as mothers against DPP homolog, notch, and semaphorin were enriched (Table 3). After 24-h differentiation (early differentiation) enrichment analysis of pathway associations revealed involvement of cyclin A2, ryanodine receptor, and E2F transcription factor 1 pathway among others (Table 3). For example, pathways related to TGFbeta or SMAD signaling (Table 3) were additionally allocated to SOTA cluster C. However, part of the enriched signal transduction pathway association included genes which were not similarly regulated during myoblast differentiation as the respective pathway associations were not retrieved in SOTA clusters.
In contrast to the effect of myoblast differentiation, the effect of TNF-α treatment on gene expression, and thus pathway enrichment, was approximately constant over time (Table 3). However, slight time-specific enrichment of signal transduction pathway associations were evident as the number of signal transduction pathway associations peaks at 12 h after induction of  differentiation. Signal transduction pathway association analysis of genes regulated by TNF-α during myoblast differentiation revealed that the following pathways enriched at 4 h, 12 h as well as at 24 h included tumor necrosis factor (TNF superfamily, member 2), NFkB, and chemokine (C C motif) ligand 2 (Table 3). Moreover, TNF-α treatment regulated matrix metalloproteinase signaling after 24 h of TNF-α and differentiation stimuli. Genes with a function in TNF-α or cytokine signaling were retrieved in SOTA cluster E (specifically induced by TNF). Genes upregulated by TNF-α after 24-h incubation had a function in the chemokine (CC motif) ligand 2 or matrix metalloproteinase pathway, which are both enriched in SOTA cluster C (early myotube genes down). Furthermore, the effect of IGF1 compared with TNF-α revealed enrichment of similar pathways as observed for the effect of TNF-α compared with the untreated control (Table 3). In summary, enrichment of several pathways was validated across methods (compare Tables 1 and 3). Inter method validated pathways were highlighted in bold (Tables 1 and 3). Pathways which did not match between results from principal component analysis and results from differential gene expression analysis resemble the consequence of different analyses approaches. TNF-α inversely regulated early differentiation-associated genes TNF-α impaired myoblast differentiation; therefore, we aimed to identify differentiation-associated genes inversely regulated by TNF-α. We identified genes counteracted by TNF-α (Table 4) after 24-h treatment; this included several genes that were among the top 20 most upregulated genes during differentiation, such as Cpa1, Aspn, Adamts5, and Fibin. Most of the inversely regulated genes were upregulated during differentiation but downregulated because of TNF-α treatment (Table 4).    Only one gene, Serpinb2, was downregulated during myoblast differentiation but upregulated upon TNF-α stimulus (Table 4). Inversely, regulated genes were indicative of which pathways may be counteracted by TNF-α that lead to the observed phenotypic impairment of differentiation [27]. These genes included Aspn, Adamts5, Trdn, Slc40a1, Capn6, and Ser-pinb2, which are involved in the following enriched pathways: mothers against DPP homolog, TGF beta, matrix metalloproteinase, ryanodine receptor, tumor necrosis factor (TNF superfamily, member 2), NFkB, or TNF (compare Tables 3 and 4).   Fig 4A) and F-box protein 5 (Fbxo5/Emi1) (Fig 4B). Despite the downregulation of Mybl2 mRNA during differentiation, there was no significant regulation of Mybl2 protein as confirmed by western blot analysis (Fig 4C). Selection criteria for genes which were analyzed by RT-qPCR or western blot were based on the integrative analysis of microRNA and mRNA expression data as described by Meyer et al. [28,29].

Discussion
Gene expression kinetics of in vitro myoblast differentiation in the presence of IGF1 or inflammatory levels of TNF-α have not yet been described in detail. Based on microarray data of PMI28 myoblasts, the current study elucidated gene expression kinetics and its networks immediately after induction of differentiation (4 h), during very early (12 h), and early (24 h) differentiation as well as late (72 h) differentiation. Results from the current study indicated significant effects of TNF-α and subtle changes in gene regulation because of IGF1 treatment. Thus, the discussion section focuses on the effects observed for TNF-α treatment on gene expression of differentiating myoblasts.

Immediate response to differentiation as well as TNF-α and coregulation of gene sets
The current study detected significant co-regulation of gene sets as well as an immediate and specific response to TNF-α, which interfered with gene expression regulation during normal differentiation. In summary, the vast majority of genes differentially regulated in myoblast differentiation and response to TNF-α or IGF1 were upregulated during early or late differentiation. Our findings relate to Henningsen et al. [2] who reported that a higher proportion of muscle-released proteins exhibited an increased level of secretion compared with the proteins with a decreased secretion profile during the course of C2C12 differentiation. Moreover, we found that genes with similar relative expression profiles were enriched for genes with similar biological implications indicating significant co-regulation of functionally related gene sets. Genes upregulated during very early and or late myoblast differentiation were associated with muscle cell differentiation, muscle structure development, or muscle contraction, which is in agreement with the observed phenotypic differentiation [27] including withdrawal from the cell cycle, myoblast fusion, and formation of multinucleated myotubes. In harmony with this, we identified accumulation of coexpressed genes belonging to pathways which are upregulated during differentiation or which are positive regulators of differentiation such as cyclin G1 [30], semaphorin [31][32][33][34]2], ryanodine receptor [35,36], calcineurin (protein level: [37], activity level: [38]), and doublecortin like kinase.
We propose that one of the inhibitory effects of TNF-α on myoblast fusion could be associated with NF-kappaB activation and ryanodine receptor regulation. This assumption is based on a study by Valdes et al. [39], which suggested that NF-kappaB activation in skeletal muscle cells is linked to membrane depolarization and depends on sequential activation of calcium release mediated by the ryanodine and by IP(3) receptors [39]. Moreover, RyR1 alters the expression pattern of several proteins involved in calcium homeostasis [40], which regulates calcineurin amongst others. Calcineurin may have therapeutic potential, as Stupka et al. [41] demonstrated that calcineurin is essential for skeletal muscle regeneration in wild type mice or in young mdx mice in which calcineurin stimulation can ameliorate the dystrophic pathology [41]. Moreover, after 24 h of differentiation, pathways including the doublecortin like kinase pathway were enriched. Doublecortin like kinase encodes a microtubule-binding protein. To date, the doublecortin like kinase pathway has not been discussed in the context of myoblast differentiation or response of differentiating myoblasts to TNF-α. We speculate that doublecortin like kinase may play a role in myoblast migration or guidance as it has been known that doublecortin like kinase is associated with interneuron migration [42] and axon guidance [43].
Genes downregulated during late myotube formation. Moreover, signal transduction pathway associations of coregulated clustering genes, which decreased in expression during later differentiation, included nuclear factor (erythroid derived 2)-like two (Nrf2), tumor protein p53, breast cancer 1 early onset (Brca1) as well as the cell division cycle. Consistent with a role of the Nrf2 pathway in myogenic differentiation, it has been reported that Nrf2 protein expression increased during myogenesis and regulated muscle differentiation [65]. Nrf2 promoted muscle regeneration and protected against TWEAK-mediated muscle wasting [66]. However, our data shows down-regulation of Nrf2 signal pathway associations. Furthermore, p53 signal transduction pathway associations were in agreement with the finding that p53 activation was measurable during myoblast differentiation and that p53 had a specific role in this process [67][68][69]. Moreover, Brca1 was involved in cell differentiation, and it has been shown to be upregulated during C2C12 myoblast differentiation [70]. However, our data is contradictory to the findings of Kubista et al. [70] as we detected downregulation of Brca1. In addition, the gene ontology term, cell cycle, was significantly enriched in genes downregulated during later differentiation, which is represented by serin/threonine-protein kinase (Chk1) gene expression for example. Chk1 was associated with several enriched signal transduction pathways, including breast cancer 1 early onset and tumor protein 53. Chk1 activity was associated with regulation of cell cycle and differentiation [71] in other cell types. The known functions of Chk1 are discussed in paragraph "Gene expression profiling results were validated at the mRNA and protein level".
TNF-induced and suppressed genes during late myotube formation. TNFα-induced genes downregulated during late myoblast differentiation were of special interest as they were modulated by TNF-α, and at the same time essential for skeletal muscle cell differentiation. These genes may point to possible therapeutic strategies to ameliorate the inhibitory effect of TNF-α. In harmony with this assumption, we found gene ontology biological process terms enriched, which are associated with regulation of cell proliferation, differentiation, migration, and motility. Interleukin 1 receptor antagonist amongst others was upregulated by TNF-α but downregulated during differentiation, which was in harmony with a known positive effect of IL-1 on myogenic differentiation [72]. Moreover, Cdk6 expression regulation was significantly associated with signal transduction pathway cyclin dependent kinase inhibitor 2. In agreement with this, it has been known that myoblast cell cycle exit and differentiation are mediated in part by down-regulation of cyclin D1 and associated cyclin-dependent kinase (Cdk) activity [73]. Consistent with a role for Cdk4/Cdk6 activity as a regulator of myogenic differentiation, Saab et al. [73] observed that Cdk4/Cdk6 inhibition promoted morphologic changes in myoblasts and enhanced the expression of muscle-specific proteins [73].
TNF-specific induced genes. Genes specifically induced by TNF-α were involved in the immune response and were associated with signal transduction pathway associations such as NF kappa B, TNF-α signaling, chemokine (C C motive) ligand 2, toll like receptor, IL-1, IL-6, and IL-18. More importantly, these pathways have been associated with cell proliferation and differentiation. Inflammatory cytokines such as TNF-α have been known to inhibit myogenic differentiation, in part through sustained NF-kappaB activity [9]. Activated NF-kappaB interfered with the expression of muscle proteins in differentiating myoblasts [9] by inducing loss of MyoD mRNA [74] or interference with the function of MyoD [75]. Moreover, NF-kappaB activates cyclin D1 expression at the transcriptional level, which inhibits myogenesis [76] and regulated cyclin D1 protein D1 stability [77]. In addition, our data revealed that TNF-α exposure increased gene expressions associated with the IL-1 pathway in differentiating myotubes. Grabiec et al. [72] reported that interleukin-1beta stimulated early myogenesis of mouse C2C12 myoblasts, and concluded that IL-1beta was associated with the impact on myogenic regulatory factors [72]. On the other hand, IL-1beta induced Id2 gene expression in vascular smooth muscle cells [78], which could point to an inhibitory effect of IL-1beta in skeletal muscle cells. Furthermore, TNF-α specifically induced expressions were enriched for genes associated with the IL-6 and IL-18 pathways. It has been reported that TNF-α exposure increased IL-6 in skeletal myoblasts [79,80]. IL-6 increased myogenic differentiation [81] and the mRNA expression of myocyte enhancer factor 2D [82], while IL-6 has been known to stimulate myoblast proliferation [83,84]. It has been shown that IL-18 stimulated airway smooth muscle cell proliferation [85] and activated NF-kappaB amongst others [86]. Our data suggested that IL-1, IL-6, and IL-18 pathway associations could be mediators of the inhibitory effect of TNF-α on skeletal muscle differentiation, or may have implications in compensating for anti-myogenic effect of the pathological concentrations of TNF-α levels.
In summary, we confirmed known gene regulations and identified new genes, which have not yet been described, to play a role in mediating the response to TNF-α in skeletal myoblast differentiation. Moreover, we provided kinetic gene expression data of the very early and early differentiation response, which facilitated the understanding of the regulatory networks, leading to impaired myoblast fusion upon pathological concentrations of TNF-α. Coregulated gene sets were enriched for pathways, which have been described in the context of myoblast differentiation. However, our data showed new avenues in the complexity of gene expression kinetics and networks, and pointed to findings contradicting the current literature on first sight. Moreover, we have identified TNF-α-regulated genes in skeletal muscle cell differentiation, which have not been implicated in this process before. An increased understanding of gene expression regulation during skeletal muscle cell differentiation may provide new approaches for the development of strategies to counteract impaired muscle regeneration or muscle wasting.
Specific signaling pathway regulation during myoblast differentiation and TNF-α response Differential gene expression kinetics revealed dynamic, time-specific change of gene regulation as well as genes constantly downregulated immediately subsequent induction and during the course of differentiation, including mothers against dpp homolog signaling or semaphorin signaling associated genes. Our data confirmed the importance of regulating mothers against dpp homolog or Smad protein signaling in myoblast differentiation. Moreover, semaphorins have been linked to muscle regeneration [34]. However, the upregulated isoforms of semaphorins identified within the current study, namely Sema6a, Sema6d, Sema5a, and Sema3c, have not yet been described in myoblast differentiation. The majority of differentially regulated genes were enriched during signal transduction pathway associations in a time-dependent manner. Induction of myoblast differentiation can be characterized as being regulated by genes involved in mothers against DPP homolog and TGFbeta signaling. With Smad proteins being downstream mediators of TGFbeta signaling, our data emphasize the role of TGFbeta and downstream Smad signaling in modulating myoblast differentiation and myotube formation (compare [56]). After 12 h, differentiation genes involved in signaling pathways such as mothers against DPP homolog, notch, semaphorin, cadherin, TGF beta, and fibroblast growth factor were enriched. Thus, we can conclude that TGFbeta signaling was a major regulatory pathways during the first hours (4 h-12 h) of differentiation. After 24 h of differentiation we identified enrichment of pathway associations including notch signaling, cyclin, cyclin dependent kinase, and cycline dependent kinase inhibitor amongst others. Thus, notch signaling can be attributed to the differentiation phase from 12 h to 24 h of differentiation, whereas cell cycle regulation is the predominant theme after 24 h differentiation.
TNF-α treatment specifically upregulates several genes immediately after induction (4 h), which remain upregulated after 12 h as well as 24 h incubation in differentiation medium, and which are related to TNF, NFkB, chemokine ligand, and interleukin pathways in agreement with immune responsive reactions of muscle cells. Moreover, after 24 h, TNF-α treatment regulated genes associated with matrix metalloproteinase which may indicate that TNF-α excerts part of its pro-proliferative functions through modulating the MMP pathway and thus myoblast migration.
However, it has not yet been reported that Adamts5 was one of the 20 most up-regulated genes during differentiation and that its expression was negatively regulated by TNF-α during myogenic differentiation. Furthermore, the described expression regulation or known functions of triadin, Adamts5, Cnr1, Itm2a, and Fzd4 in skeletal muscle differentiation underline the significance of our findings that TNF-α deregulated these genes during myogenic differentiation and reduced fusion capacity of myoblasts. However, our findings regarding Capn6 expression regulation were contradictory to a study by Tonami et al. [100] reporting that Capn6 was a suppressor of skeletal muscle differentiation, and a study by Liu et al. [101] indicating that Capn6 promoted cancer cell proliferation and was positively regulated by the PI3K-Akt signaling pathway. On the other hand, it has been shown that Capn6 expression was suppressed by serum in fibroblast cell culture [100], which would be in harmony with the upregulation of Capn6 upon serum deprivation observed in the current study. Remarkably, this effect was reversed by TNF-α treatment. Other members of the calpain family have been discussed in myocyte differentiation, namely Capn1, which has been reported to play an important role for satellite cell myogenesis [102], and Capn2 / m-calpain, which had been shown to play a role in the control of muscle precursor cell differentiation [103]. Therefore, we hypothesize that Capn6 could be a myofusion marker which may participate in promoting in vitro differentiation of skeletal myoblasts through an unknown physiological mechanism. We identified inversely regulated genes described in the context of skeletal muscle but not in the context of myoblast differentiation or TNF-α response such as Aspn, Fibin, Slc40a1, Nrk, and Mybpc1 (Fig 5). To date, Aspn had been described in the context of congenital muscular corticollis [87], cardiac remodeling [104], or the transition from a hyperplasic myotube-producing phenotype to a hypertrophic growth phenotype in fish [105]. The current study is the first identifying Aspn expression regulation in the context of skeletal myogenic differentiation and its response to TNF-α as well as enrichment in TGFbeta signaling pathway associations in skeletal muscle. Similarly to Aspn, we detected fibin among the top 20 most upregulated genes during myoblast differentiation. It is reported that fibin is expressed in skeletal muscle [89] amongst other tissues. We provided indications for a role of fibin in the regulation of myogenic differentiation and its response to TNF-α. Slc40a1, which encoded ferroportin [106], was hypothesized to influence skeletal muscle iron content [91]. We found that Slca40a1 was associated with the tumor necrosis factor pathway ( Table 3), but the current study is the first describing a role of Slca40a1 in myogenic differentiation.
We observed upregulation of Nrk/Nik related kinase during differentiation and downregulation by TNF-α treatment. Nrk has been known to be expressed in skeletal muscle during mouse embryogenesis [107,93] and may be involved in the regulation of actin cytoskeletal organization in skeletal muscle cells through cofilin phosphorylation and actin polymerization [93]. However, the current study is the first describing Nrk/Nik regulation during in vitro skeletal myoblast differentiation and TNF-α response. Myosin binding protein C (MyBP-C) is expressed in striated muscles where it modulates actomyosin cross-bridges [96] and acts as an adaptor to connect myosin and muscle-type creatine kinase for efficient energy metabolism and homoeostasis [108]. Mutations of sMyBP-C have been causally linked to the development of distal arthrogryposis-1, a severe skeletal muscle disorder [96,109], and lethal congenital contracture syndrome type 4 [109]. We describe Mybpc1 upregulation in the context of in vitro myoblast differentiation, which is counteracted by TNF-α.
Importantly, we identified genes inversely regulated during skeletal myocyte differentiation compared with differentiation under TNF-α stimulus, but have not yet been described in the Novel genes and pathways in skeletal myocyte differentiation and TNF-α response. We identified genes and pathway associations, which have not been described before in skeletal myocyte differentiation or have been reported to have a different regulation than the one observed in the current study. A plus indicates upregulation during differentiation and a minus indicates downregulation. Moreover, we show genes which are inversely regulated by TNF-α, but have not been defined before, to be regulated in skeletal myocyte differentiation and response to TNF-α.
doi:10.1371/journal.pone.0139520.g005 context of skeletal muscle or TNF-α effect on muscle cells. Some differentiation and TNF-αregulated genes have been described in smooth or heart muscle, or muscle progenitor cells, but not in skeletal muscle, including Cpa1 [110], Parm1 [111], and Serpinb2 [112] (Fig 5). Parm1 expression was detected in the muscle progenitor cells of the somites [111]. Thus, the role of Parm1 in muscle differentiation and TNF-α response still needs to be unraveled. Serpinb2 was >4-fold downregulated during differentiation, but upregulated during TNF-α treatment. There has not yet been evidence for the observed effect in skeletal myoblast differentiation. However, in smooth muscle, Jang et al. [112] have reported that plasminogen activator inhibitor-2 protein levels were upregulated by TNF-α. It has been reported that PAI-2 (Serpinb2) is upregulated during cell cycle progression in myoepithelial cells [113]. Moreover PAI reduced the capacity of endothelial cells to lyse fibrin [114] and rPAI-2-expressing sarcoma cells showed inhibited invasion characteristics [115]. We speculate that downregulation of Serbinb2 in myoblast differentiation may facilitate cell migration during the early phase of differentiation.
The current study reveals nine genes regulated by differentiation and TNF-α, which have not yet been described in muscle cells. We describe for the first time a significant specific regulation of the following genes in myoblast differentiation: Aknad1, Sepp1, Ndst4, Tecrl, Cmbl, Unc13c, Spats2l, Lix1, and Csdc2 (Fig 5). On the basis of our expression data, we postulate a biological implication of these genes in myoblast differentiation and responsiveness to TNF-α. Comparatively little is known regarding the function of Aknad1 and Tecrl. Sepp1 is known to be involved in selenium distribution to tissues throughout the body [116]. Ndst4 is involved in N-sulfation of heparin sulfate [117] chains and is downregulated in carcinoma [118], indicating an anti-proliferative or pro-differentiative role. Moreover, a role of Cmbl has not yet been explicitly described in the muscular context. The physiological implications during skeletal muscle cell differentiation and its response to TNF-α remain elusive, but are likely to be of biological significance as Cmbl is one of the top 20 most regulated genes during myoblast differentiation within the current study. We identified significant expression regulation for Unc13c/ Munc13-3 in myoblast differentiation. However, Munc13 has been described to be almost exclusively expressed in the cerebellum, which is a presynaptic protein [119] critical in regulating neurotransmitter release and synaptic plasticity [120]. In Munc13-deficient mice, the distribution, number, size, and shape of synapses, as well as the number of motor neurons they originate from and the maturation state of muscle cells, are profoundly altered [121]. The function of Spats2l in skeletal muscle differentiation and response to TNF-α has not been described earlier. Himes et al. [122] suggest that SPATS2L may be an important regulator of β(2)-adrenergic receptor downregulation. The function of Lix1 expression regulation in skeletal muscle cell differentiation remains elusive. It has been suggested that Lix1 plays a role in radial growth of motor axons observed in feline spinal muscular atrophy [123]. We identified inverse regulation of Csdc2 expression during differentiation and TNF-α treatment relative to control cells. However, Csdc2 has previously been described in the context of decidualization [124], but not skeletal muscle.
Gene expression profiling results were validated on the mRNA and the protein level The expression of selected genes such as the Serine/threonine-protein kinase Chk1 [checkpoint kinase 1 homolog (Schizosaccharomyces pombe)] and F-box protein 5 (Fbxo5/Emi1) were validated on the protein level by western blot analysis. Both, Chk1, and Fbxo5 were downregulated during differentiation and in the IGF1 treatment group. However, when differentiation was induced in the presence of inflammatory levels of TNF-α Chk1 and Fbxo5/Emi1 were upregulated. The observed regulation of Chk1 on the protein level was in harmony with its known function as being part of a "G2 restriction point" that prevented premature cell cycle exit in cells programmed for terminal differentiation [71]. Chk1 functioned as a mitogendependent protein kinase that prevented premature differentiation of trophobast stem cells by suppressing expression of p21 and p57, but not p27, the CDK inhibitor that regulated mitotic cell cycles [71]. In the current study, we described for the first time Chk1 regulation during myoblast differentiation and its deregulation because of TNF-α treatment on the mRNA and protein level. Chk1 had been described in the context of estivating frogs, and it has been postulated that Chk1, amongst others, may contribute to preserving muscle function during metabolic depression and immobility [129]. In addition, we describe for the first time a specific regulation of Fbxo5/Emi1, both at the mRNA and the protein level, during myoblast differentiation and TNF-α treatment. Emi1 has been reported to play a role in somitogenesis [130]. Moreover, Emi1 functioned to promote cyclin A accumulation [131], and is known as a key cell-cycle regulator [132,131,133]. The described functions of Fbxo5/Emi1 were in harmony with the exit of the cell cycle of proliferating myoblasts to differentiate into myotubes. The inhibitory effect of TNF-α on myotube formation was resembled by diminished downregulation of Emi1/Fbxo5. Furthermore, we analyzed the protein level of transcription factor. In contrary to the upregulation of Mybl2 mRNA during proliferation and TNF-α treatment compared with differentiation control, there was no significant regulation of the Mybl2 protein. The latter may be because of protein stability and thus extended half-life of Mybl2 protein, which may mask the downregulation. Moreover, it was shown that Mybl2 was regulated posttranscriptionally [134] and phosphorylation was necessary to activate Mybl2 [135]. Further experiments need to analyze Mybl2 activity to possibly identify correlation with mRNA expression levels. Henningsen et al. [2] found little correlation between mRNA and protein levels for proteins secreted during myoblast differentiation. The latter indicates pronounced regulation by posttranscriptional mechanisms, such as functionality of miRNAs.
New genes in TNF-α response during myoblast differentiation Several of the differentially expressed genes identified in the current study were new in the context of impaired myoblast differentiation because of TNF-α exposure (compare S2B Table with  S5 Table). However, a subset of genes has been confirmed in studies by Bhatnagar et al. [136] (S5A Table) or a study investigating TNF-like weak inducer of apoptosis (TWEAK) treatment by Panguluri et al. [137] (S5B Table). TWEAK mediated skeletal muscle wasting [66]. The common subset of genes between the different studies confirmed the validity of identified genes. In addition, the cross-study validated gene subset pointed to the most prominent genes, which were stably regulated across different murine skeletal myoblast cell lines and different treatment conditions. Bhatnagar et al. [136] analyzed C2C12 myoblasts treated with 10 ng/mL TNF-α, whereas the current study applied 5 ng/mL TNF-α. Genes differentially expressed in the current study as well as by Bhatnagar et al. [136] or Panguluri et al. [137] were enriched in the expression cluster (cluster B) containing genes slightly upregulated during early myotubes and highly upregulated during late myotubes (72 h) but downregulated by TNF-α or in expression cluster E which bore genes specifically induced by TNF-α. Thus, a gene subset can be confirmed across independent studies including genes which were i) involved in myotube differentiation but counteracted by TNF-α treatment or ii) TNFα-induced genes. Of note is that there was a common subset of genes regulated by TWEAK after 96 h of differentiation in C2C12 myotubes according to Panguluri et al. [137] and TNF-α treated PMI28 myotubes in the current study (S5B Table). Thus, TNF-α and TWEAK regulated in part common gene sets (S5C Table). Genes identified across different murine skeletal myoblast cell lines and different TNF-α treatment conditions as well as TWEAK treatment included Nfkbia, Nfkb2, Mmp9, Mef2c, Gpx, and Pgam2. Thus, several genes have not yet been described to play a role in the response to TNF-α in myoblast differentiation (S2B Table), while a small subset of genes had been confirmed by others (S5 Table).

Conclusions
The understanding of the gene expression regulation during skeletal myoblast differentiation and how this is impacted by TNF-α and IGF1 is of significant clinical and therapeutic importance. Results of the current study facilitate the understanding of the regulatory networks leading to impaired myoblast fusion upon pathological concentrations of TNF-α. We confirmed genes and pathways prominent in myogenic differentiation or TNF-α signaling, and identified novel genes and pathways (Fig 5). Several differentiation-relevant genes were inversely regulated by TNF-α treatment. Moreover, our data revealed novel kinetic expression dynamics of genes and pathways during differentiation and TNF-α treatment. Moreover, TNF-α and IGF1 treatment could be characterized by a subset of indicative expression markers, of which some are robust inter-study retrieved markers. Results of the current study may point to possible candidates for new strategies to counteract impaired muscle regeneration in inflammatory myopathies, muscular dystrophies, or cancer cachexia. However, further research at the protein level is required.  Table. Gene ontology and pathway enrichment of genes clustering by self-organizing tree algorithm analysis. Gene ontology biological process and signal transduction pathway associations are given for genes separated in clusters (A) cluster A: upregulated during very early differentiation, (B) cluster B, genes upregulated during later differentiation, (C) cluster C: genes downregulated during very early differentiation, (D) cluster D: TNFα-induced and suppressed in late myotubes, (E) cluster E: genes specifically induced by TNF-α, (F) cluster F: late myotubes genes downregulated, (G) cluster G: upregulated by TNF-α and downregulated in late myotubes, (H) cluster H: upregulated in early myotubes, but downregulated in late myotubes, (I) cluster I. (XLSX) S5 Table. Comparison of the effect of TNF-α with previous studies. Genes differentially expressed in skeletal muscle cells in the current study and previous profiling studies are listed. (A) Genes differentially expressed in a study by Bhatnagar et al. [136] (C2C12, 18 h, 10 ng/mL TNF-α) as well as in the current study (PMI28, 4 h, 12 h, 24 h, 72 h, 5 ng/mL TNF-α) and the SOTA cluster in which the respective gene was retrieved (B) Regulated genes identified in a study by Panguluri et al. [137] (C2C12, 96 h, 10 ng/mL TWEAK) and in the current study (PMI28, 4 h, 12 h, 24 h, 72 h, and 5 ng/mL TNF-α) as well as the SOTA cluster in which the respective gene was retrieved. (C) Genes published by Bhatnagar et al. [136] and Panguluri et al. [137]. Bold genes were detected by Bhatnagar et al. (2010) [136], Panguluri et al. (2009) [137] as well as in the current study. (XLSX)