The combined use of cell lines and microarrays offers a major opportunity to study gene expression patterns and/or dynamics during different physiological and pathological processes. However, the substantial findings generated by the use of pangenomic microarrays have generally been difficult to interpret in terms of the gene regulation controlling biological functions. In this study, we chose to explore the expression dynamics of just one part of the mouse genome, called the 'glycogenome', in the context of myogenesis. For this purpose, we first of all standardized the experimental conditions for the differentiation of C2C12 (a mouse myogenic cell line), and analyzed the expression of myogenic markers by quantitative real-time RT-PCR. The expression of 375 glycogenes was then monitored in differentiating C2C12 cells using quantitative real-time RT-PCR with TLDA (TaqMan Low Density Array, see Methods section). Highly deregulated genes were next clustered as a function of their expression profiles. Their functions were analyzed and used to suggest new roles for glycoconjugates in myogenic differentiation.
The expression of MRF and marker genes is consistent with C2C12 cell differentiation
When cultured in vitro
, C2C12 myoblasts start to differentiate following serum deprivation. The first myotubes appeared 48 hours after serum starvation and a maximum of mature multinucleated cells was obtained after 11 days in the differentiation medium (Figure ). Expression levels of the four MRFs (Mrf4, Myf5, MyoD
) and four marker genes (Csrp3
) known for their involvement in myogenic differentiation [29
], were determined by quantitative real-time RT-PCR at different time points following the induction of C2C12 differentiation.
Figure 1 Time course of C2C12 differentiation. (A) C2C12 cells were placed in differentiation medium when culture reached 80% of confluence (t = 0 h). After 48 hours, the first myotubes, indicated by arrows, had clearly formed and after 11 days (t = 264 h), most (more ...) MyoD, Myf5
genes were expressed throughout C2C12 differentiation (Figure ). MyoD
mRNA levels only changed slightly, regardless of the time elapsing after the start of differentiation. Beyond t = 48 h, the expression of Myf5
decreased more than two-fold and remained down-regulated, while the myogenin
gene was up-regulated (~100-fold). For Mrf4
, transcripts were only detected at t = 192 h. Therefore, the expression profiles of myogenic regulatory factors during C2C12 differentiation were in agreement with their expression patterns (top diagram inset in Figure ) described in the literature [15
Expression of the muscle transcription factors Mef2a
increased as from 6 h of differentiation to reach 60-fold for Mef2a
and 3.4-fold for Mef2d
at the end of the experiment (Figure ). Their expression was in line with their myogenic activator roles [29
]. Interestingly, the increase in Hes6
expression started at t = 6 h of differentiation and reached 6.5-fold after 72 h. As demonstrated elsewhere [31
], this last result argued in favor of Hes6
involvement at the onset of C2C12 differentiation and more generally of the myogenic process. Unlike the Hes
expression was first detected at t = 18 h of differentiation and increased to reach a peak at t = 120 h (Figure ). The expression profile of Csrp3
, encoding the LIM protein, correlated with its activator function of C2C12 differentiation. Indeed, it has been showed that LIM protein is not necessary for myoblast proliferation but plays a key role in upcoming myogenic differentiation [30
]. Thus, the transcriptional expression profiles of both myogenic marker genes and MRFs genes attested to the accurate time course of C2C12 differentiation.
Most glycogenes are expressed during the onset of C2C12 differentiation
The glycogenome refers to all genes involved in glycosylation. It includes ~600 genes and accounts for ~2 percent of the mouse genome. The expression of 375 glycogenes was analyzed during the first 72 h of C2C12 differentiation, when the first myotubes are formed. These 375 glycogenes account for more than 60% of the mouse genes known to be related to glycosylation (Table ). The proteins encoded by these genes belong to glycosyltranferases, glycosidases, lectins, sulfotranferases or proteins involved in sugar metabolism or transport [see Additional file 1
]. Given the known weak expression of most glycogenes, their expression patterns were determined by quantitative real-time RT-PCR using the TLDA technology which allows the simultaneous analysis of 375 selected genes [see Additional file 2
Data summary of Mus musculus glycogene expression during C2C12 differentiation.
Three-quarters of the genes analyzed were expressed (Table ): 276 genes displayed significant quantities of transcripts (Ct ≤ 33) during at least one point of the differentiation time course. Among the 375 glycogenes of this study, 202 genes were also analyzed in Tomczak et al
. study [16
]. The microarray and TLDA approaches gave similar results for 91 genes, 43 were expressed and 48 unexpressed. For the remaining common genes (111), only TLDA revealed significant expression levels. This could be explained by the methodologies employed, insofar as microarray techniques are less precise and sensitive than quantitative real-time RT-PCR [34
Among the genes expressed, 34% had a minimum 2-fold modification of their expression for at least one kinetic time, and 10% displayed a variation of at least 4-fold (Table ). The significant number of glycogenes thus regulated underlined the critical function of glycosylation in this differentiation process. Lectin genes appeared to be regulated preferentially, because only 57% of them were expressed, compared to 73% or more for the other gene families. Within each glycogene family, it is interesting to note that no correlation was observed between the number of genes analyzed and the number of those regulated. Indeed, glycosyltransferase genes accounted for about 40% of analyzed genes and only 11% of them displayed an mRNA variation of more than 4-fold. At the same time, ~50% of lectin and sulfotransferase genes, representing ~27% and ~6% of analyzed genes respectively, were significantly modified in terms of their expression. In addition, no glycogene sub-family, such as fucosyltransferases or sialyltransferases, was preferentially repressed or expressed.
Genes displaying more than 4-fold variation (37 genes) were distributed into four groups according to their glyco-family (Table ). The first group included lectin and sulfotransferase genes (26% of them with significant mRNA variations), the second contained glycosyltransferase and sugar carrier genes (11-16% deregulated), the third included glycosidase genes (only 4% of genes deregulated), and the final group comprised translocase and sugar metabolism genes in which no gene displayed a variation in mRNA expression. Thus, a large proportion of the modifications to glycogene expression that occurred during C2C12 differentiation mainly seemed to affect proteins giving rise to the glycans or lectins required for cell contacts. These results are consistent with the cellular events involved in myotube formation, i.e. cell interactions and fusions.
Among the genes analyzed, 99 were poorly or not expressed. Their corresponding mRNA were not detected (Ct = 40) or not significantly quantified (Ct>33). These genes encoded proteins involved in physiological processes unrelated to myogenesis. For example, Has3
encodes a hyaluronan synthase which is active in hyaluronan/hyaluronic acid synthesis and known to be involved in the inflammatory response [35
], and Icam2
encodes a lectin which mediates adhesive interactions during the immune response.
Nearly half of analyzed glycogenes could be cell homeostasis genes
Among the 276 genes expressed, 181 were invariantly transcribed (Table ). These constitutively expressed genes could be divided into three sets, according to their functions. The first set corresponded to genes involved in cell homeostasis, the second to genes involved in myogenic cell homeostasis and the third to myogenic genes that could probably undergo a late modification to their expression. In this respect, most of the genes encoding proteins involved in N
-glycan precursor synthesis and present on our mouse glycogenome TLDA were homeostasis cell genes and were constitutively expressed. Alg2, Alg3, Alg9, Alg12
(mannosyltransferase genes) and Alg6
(a glucosyltransferase gene), which are responsible for N
-glycan precursor synthesis, were expressed without any significant variations. This was also the case for Dpia3
), an ER chaperone-encoding gene involved in disulfide bond formation [36
]. The second set of genes, although constitutively expressed during the first 72 h of differentiation, could have crucial functions at all stages of myogenesis. The myogenic factor MyoD
, or the sialidase gene Neu3
are representative of this group [37
]. Finally, the expression of the third set of genes may be modified after 72 h of differentiation and be required for later stages of myogenesis. For example, the expression of Pomt1
, encoding an O
-mannosyltransferase which is known to glycosylate the muscle membrane protein α-dystroglycan linking cytoskeleton actin to ECM components, could be tardily up-regulated [38
Glycogenes with significant mRNA variations are sequentially expressed
On the 95 regulated genes, 37 whose expression levels were modified more than 4-fold were retained for further analyses. In order to obtain a global vision of their expression profiles, Principal Component Analysis (PCA) was performed. Its efficiency was excellent since ~89% of information in the data set was recovered on the first ordinate (~70% on component 1 and ~19% on component 2). The localization of each gene in the Figure indicates its expression as a function of differentiation times (6 to 72 h), compared with the precursor state at t = 0 h of differentiation (Figure ). The position of a gene in the same direction as a vector indicates an increase of expression. By contrast, the position of a gene in the opposite direction to a vector means that the gene was down-regulated. Because of their reduced sizes, 12 h and 18 h vectors were only weakly informative.
Figure 2 Expression dynamics of up- and down-regulated glycogenes during the onset of mouse C2C12 differentiation. (A) According to their expression profiles, the 37 glycogenes expressed with more than-4 fold variations were analyzed using a principal component (more ...)
Gene clustering was performed using the Euclidean distances calculated with their coordinates on the first plan of PCA. This clearly highlighted three groups (Figure ). The first contained 12 genes, the second four and the third 21. The myogenic marker Myf5 was classified in cluster 2, MyoD and myogenin in cluster 3 (data not shown); Mrf4 was not clustered since it was not expressed during the first 72 h of differentiation. mRNA levels in the cluster 1 displayed a general tendency to decrease that was more pronounced towards the end of the time course (Figure ). Cluster 2 included genes with a peak mRNA expression at 24 h of differentiation. Genes in cluster 3 had expression profiles opposite to those of cluster 1 because these expressions increased and became more important at the end of the time course (Figure ).
The 37 highly regulated glycogenes were examined according to the activity/function of the enzymes they encode. Only their functions linked to myogenesis were considered (Figure ). Functions unknown or unrelated to myogenesis, such as intracellular transport, were grouped in "other function". In the light of the literature, several functions could be assured by one protein. Genes in cluster 1 encoded proteins mainly involved in cell adhesion and interaction, GAG biosynthesis and signal transduction. The down-regulation of most of them could be required for the early mechanisms of myogenesis, especially for the switch from a proliferative to a quiescent state and then to a differentiated state. The four genes in cluster 2 were mainly involved in glycosphingolipid and GAG biosynthesis (Figure ). These functions suggest early rearrangements of the plasma membrane and ECM, leading to the first fusion events. Among the up-regulated genes in cluster 3, some genes were also involved in glycosphingolipid biosynthesis while the others encoded proteins that were mostly implicated in cell adhesion and interaction and in intracellular biological functions. These functions were consistent with the fusion events leading to myotube formation and maturation beyond 48 h of serum deprivation.
Figure 3 Cell functions in which regulated glycogenes are involved. The function assigned to each gene was extracted from the Kegg Pathway database . Numbers indicate how many genes are concerned for each function, one gene being able to be involved in different (more ...)
With regards the sequential expression of ≥ 4-fold variant glycogenes and the function of encoded proteins, the early differentiation of C2C12 cells seemed mainly to require: (i) the specific expression of molecules involved in cell signaling and a modification to ECM composition, (ii) the expression of CAMs, and (iii) qualitative and/or quantitative modifications to plasma membrane glycoconjugates.
Cell signaling and GAGs sulfation contribute to the initiation of myogenesis
The functions assured by some down-regulated genes in cluster 1 suggested an involvement of cell signaling in myogenic differentiation. The commitment of C2C12 cells to the myogenic or adipogenic lineage is controlled by specific transcription factors. Myogenesis is regulated by MRFs [4
], while adipogenesis is controlled by PPAR-γ and the C/EBP families of transcription factors [39
]. The Olr1
gene encodes a lectin which is activated by PPAR-γ signaling [41
]. The down-regulation of Olr1
is consistent with the commitment of C2C12 to myogenic differentiation. Lfng is an enzyme that elongates O
-fucose on some EGF-like domains of the Notch receptor. It belongs to the Fringe family [42
] and acts as a modulator of the Notch signaling pathway [43
]. It also influences cell fate during embryonic development [44
]. Given the involvement of Notch in the myogenic process [45
down-regulation in differentiating C2C12 cells argues for the involvement of Lfng in myogenic differentiation. Interestingly, among the up-regulated genes in cluster 3, Lgals12
encoded the galectin-12 which is required for adipogenic signaling and adipocyte differentiation [46
]. This gene is indeed weakly expressed at early stages, but its important transcriptional induction beyond 48 h of differentiation suggests, for the first time, its later implication in myogenesis.
GAGs are known to have many biological functions, including cell adhesion, migration and signaling [47
]. Three sulfotransferase genes from cluster 1 (Chst1, Chst2
) are known for GAG sulfation. Chst1 and Chst2 are involved in the sulfation of keratan GAG and Hs3st3b1 in that of heparan GAG. Because Hst3st3b1
is the only gene in cluster 1 which was up-regulated at an early stage (Table ), heparan GAG could become preferentially sulfated. Moreover, the Extl1
gene in cluster 3 encoded a glycosyltransferase that contributes to heparan/heparin sulfate biosynthesis. Thus when C2C12 cells differentiate, they seem to undergo a switch from the sulfation of keratan GAG to the predominant sulfation of heparan GAGs. Such a modification has not previously been reported in myogenesis and it could contribute to the activation of myogenesis. Keratan sulfate GAG may have anti-adhesive properties [48
] that are obviously incompatible with up-coming myoblast fusion events during myogenic differentiation.
Deregulated glycogenes during the onset of C2C12 differentiation.
CAMs, glycosphingolipids and glycoproteins of the C2C12 plasma membrane appeared to be reshaped for cell fusion
Myoblast fusion into myotubes requires cell interactions. Ten highly regulated glycogenes are involved in cell adhesion (Figure ). Among the genes in cluster 1, four encoded lectins (Itga3
) and one a sulfotransferase (Chst10
). These five genes have been described in different developmental processes. For example, Itgα3 associated with Itgβ1 have been shown to mediate the migration of endothelial cells and angiogenesis [49
]. In the present case, the down-regulation of Itgα3
may have been linked to the arrest of myoblast migration and proliferation. In addition, five lectin genes encoding for three integrins (Itgα4, Itgα7 and Itgβ1bp2), one galectin (Lgals7) and Ncam1, belonged to up-regulated genes (Cluster 3). Most of them have important functions in myogenesis: NCAM1 in myoblast fusion [23
], melusin (encoded by Itgb1bp2
) in the maturation and/or organization of muscle cells [50
], and Itgα7 (with Itgβ1) in myogenesis [51
]. Up-regulation of these CAM-encoding genes, combined with the down-regulation of the four genes in cluster 1 mentioned above, also suggests a potential switch of CAM during myogenic differentiation.
Cell fusion obviously requires a modification to the quality and quantity of glycans in plasma membrane glycolipids and glycoproteins. Three genes in cluster 2 (β3GalT1
) encoded glycosyltransferases implicated in glycosphingolipid biosynthesis (Figure ). β
3GalT1 is responsible for synthesizing the precursor of lactoseries glycolipids. Fut4 and Sec1 are involved in the terminal fucosylation of lacto and/or neo-lactoseries glycolipids. Four other genes involved in these different biosyntheses were found in cluster 3 (Figure ). They encoded two other fucosyltransferases, a sialyltransferase and a galactosyltransferase. The sialyltransferase is involved in ganglioside synthesis, while the three other enzymes are required for lacto and/or neo-lactoseries glycolipid biosynthesis. For glycoproteins, three genes in cluster 3 were revealed: Galnt5
-glycan synthesis proteins and St8sia2
a sialyltransferase involved in the biosynthesis of Ncam1 polysialic acid (PSA). The latter bears polysialylated N
-glycans and mucin type O
-glycans on a muscle-specific domain which is involved in myoblast fusion [24
]. The up-regulation of these three genes was in good agreement with the findings of the previous study. Therefore, myoblast fusion may require some glycosphingolipid rearrangements and/or terminal modifications (as fucosylation and sialylation) to glycans of membrane glycoproteins and glycolipids.
Figure 4 Schematic representation of glycosphingolipid synthesis pathways. The enzymes responsible for the main steps of glycosphingolipid biosynthesis are indicated. The expression levels (relative quantity of mRNA) of the corresponding genes are reported. NA: (more ...)
GM3 ganglioside levels increase in differentiating C2C12 cells
In order to confirm some of these membrane glycoconjugate rearrangements, glycolipids were considered for further analyses. According to their metabolic pathways and gene expression patterns, lactosylceramid seemed to be preferentially synthesized when compared to galactosylceramid (Figure ). Indeed, the Ugt8 gene was weakly expressed (Ct>33), while the Ugcg and β4galt6 genes were strongly expressed (Ct<25). Lactosylceramid is the common precursor of four biosynthesis pathways. The expression levels of the analyzed genes implicated in these pathways indicated that some compounds could be preferentially synthesized and/or reshaped. Among these, only GM3 (and its derivatives) could be enhanced because the St3gal5, β3GalT4 and St8sia5 genes were up-regulated (Figure ). In order to test this hypothesis, immuno-cyto-staining was used to analyze GM3 gangliosides on differentiating myoblasts (Figure ). Only a few myoblasts are positively stained at 0 h and 12 h of differentiation. Beyond 24 h, the immunostaining increased, and most of the cells were stained at 48 h and 72 h. This result showed that levels of GM3 indeed increased in the plasma membrane during the onset of C2C12 differentiation. Interestingly, beyond 48-72 h of differentiation, cells with stronger staining were mostly elongated and underwent differentiation, which argues for a role of GM3 in C2C12 differentiation and fusion.
GM3 ganglioside staining in differentiating C2C12 cells. Cells were labeled with an anti-GM3 antibody. This primary antibody was detected by a secondary antibody coupled to FITC. Isotypic controls for each differentiation time point are given.