It has been difficult, so far, to detect a hierarchy in the different reorganizations that take place during myogenesis because they occur seemingly simultaneously. Here we have attempted to uncouple them by challenging the cells to differentiate in the presence of drugs that affect microtubules. We find that each of the subcellular systems we examine, i.e. centrosome (MTOC), Golgi complex, and ERES, is able to relocate despite microtubule alterations. We also find that the relocation of the centrosomal proteins paves the way for the relocation of the other systems. Furthermore, centrosomal proteins reorganize more completely than Golgi complex and ERES sites in the absence of a dynamic microtubule network. We conclude that direct transport along dynamic microtubules is not involved in the transfer of centrosomal proteins to the nuclear envelope, but that it plays some role in Golgi complex and ERES reorganization.
Some of the cells that differentiate in microtubule-affecting drugs reach a state of partial organelle redistribution. We could not improve the outcome by washing the drug away. This experiment concentrated on nocodazole treatment, because its effects are known to be reversible within a matter of minutes to hours. Less is known of reversal of the effects of taxol and DW12. Our observations highlight irreversible changes that affect the reciprocal relations between organelles and microtubules during myogenesis.
Our findings support our previously published model 
in which Golgi complex fragmentation and reorganization during myogenesis involve microtubule-dependent retrograde trafficking through the ER, whereas centrosomal proteins move from centrosome to nuclear membrane through a more direct, microtubule-independent path. Our model also predicted that ERES relocate ahead of and guide the Golgi complex. We have not been able to prove this because microtubule manipulations did not uncouple them sufficiently. The observation of a few cases when one reorganized without the other suggests that ERES redistribute independently of the Golgi complex, possibly consolidating around ER areas rich in Golgi complex enzymes, in agreement with Guo and Linstedt 
. ERES fusion and fission have been studied extensively 
but ERES positioning is poorly understood because no other cell types that we know of show specific ERES localization in physiological conditions.
Redistribution of centrosomal proteins is enhanced rather than impeded by treatment with nocodazole and taxol. Since dynein is necessary to bring pericentrin and γ-tubulin to centrosomes 
, nocodazole may free the two proteins from the MTOC, possibly explaining how pericentrin dots come to decorate one end of the short microtubules that remain in the cells after chronic nocodazole treatment (). Taxol similarly enhances redistribution of centrosomal proteins but DW12 treatment, which inhibits GSK3-β, does not. Both taxol and DW12 stabilize microtubules, but the extent and pattern of stabilization differ. In addition, GSK3-β has been indirectly linked to microtubule anchoring at the centrosome through its control of Bicaudal-D, altering the binding of Bicaudal-D to ninein and ninein transport to the centrosome 
In the present work, we have also clarified key aspects of the organelle reorganization in normal cells. We show that MTOC reorganization is dependent on myogenin expression and is upstream of and essential for Golgi complex reorganization. We also show that reorganization is progressive, and that even small patches of centrosomal proteins along the nuclear membrane can support microtubule growth and attract the Golgi complex ( and 
). Pericentrin therefore determines or at least identifies a domain of the nuclear membrane competent for microtubule nucleation, Golgi complex and ERES localization. This correlation persists through muscle maturation since Golgi elements in muscle fibers are associated with concentrations of pericentrin 
at the nuclear poles and in the cytoplasm (Oddoux et al., in preparation). We also noticed that pericentrin forms a full perinuclear shell in myocytes. This possibly occurs by self-assembly as suggested by Srsen et al. 
, and would explain pericentrin's independence from normal microtubule organization.
Quantitative PCR and immunoblotting experiments indicate that relocation of the organelles during myogenesis does not involve transcriptional or translational activation of the genes involved and must therefore be regulated post-translationally. For pericentrin at least we expected significant changes at the protein level. Immunofluorescence gives the impression of a vastly expanded staining when comparing the small centrosome to the full perinuclear shell, but changes during differentiation in the amount of pericentrin and its message are in fact small. We ruled out that differentiation involves a switch from pericentrin to kendrin (pericentrin A and B, respectively) and did not detect pericentrin S, an isoform predicted from RNA analysis to be abundant in muscle 
but never actually reported.
We also revisited microtubule nucleation. Previous studies, including ours 
, have investigated this process only during recovery from microtubule depolymerization. For most cells this is justified, since the microtubule pattern during recovery is similar to that at steady-state, but in myotubes this is not the case. We show, for the first time, that microtubules constantly originate from the nuclear membrane of myotubes at steady-state. This is important because it explains the Golgi complex localization around the nucleus. Nucleation from myotube nuclei is less frequent than from centrosomes, even after taking into account the much larger nuclear surface. Furthermore, we have uncovered a basic difference between myotubes and myoblasts (and, consequently, other proliferating cells) during recovery from microtubule depolymerization. The short microtubules reforming around myotube nuclei appear initially devoid of the plus-tip proteins EB1 and EB3, unlike microtubules reforming from centrosomes. Centrosomes contain EB proteins 
whereas nuclear membranes have not been shown to accumulate them. In myotubes, EB proteins must therefore be recruited from the cytoplasmic pool to the newly forming microtubules along the nuclear membrane. In myoblasts, EB proteins are immediately available to the forming microtubules. EB proteins are considered core components that link other plus-tip proteins to microtubules (see 
). We are not aware of another situation in which newly forming microtubule plus-tips contain neither EB1 nor EB3. We conclude that there are structural and functional differences between centrosomes and unconventional MTOC, a difference relevant to several non-muscle cell types as well.
Time-lapse recordings of EB3-GFP in myotubes give at first an impression of random movement, but there are also repetitive trajectories and crossing points (Movie S2
), suggesting that EB3-GFP labels individual microtubules growing along stationary bundles. Golgi complex elements were occasional but not sole sources of microtubule nucleation at the nuclear envelope, indicating that Golgi elements are positioned at the nuclear membrane because of microtubule nucleation rather than the reverse.
In addition to the traditional nucleation and anchoring observed in myoblasts, other processes may take place in myotubes such as anchoring after nucleation at another site 
and regrowth of microtubules from existing stable tubulin seeds 
. Regrowth from existing seeds seems the most likely explanation for the short Glu-tubulin-containing microtubules depicted in .
Uncoupling subcellular changes taking place during muscle differentiation has allowed us to establish centrosome/MTOC reorganization as the first step of the systems that we monitored and to show that the changes are remarkably resistant to microtubule disruptions. Finding out how pericentrin and other MTOC components in muscle are linked to nuclei is one of the most interesting questions ahead. It seems likely that transmembrane proteins of the nuclear membrane, such as the nesprins (reviewed in 
), which mediate interactions between centrosome and nucleus, must be involved.
Changes in Golgi complex morphology or placement are automatically assumed to affect function negatively. Here we show, however, that mislocalization caused by treatment with a GSK3-β inhibitor barely slows down cargo trafficking, an essential Golgi complex function. Rahkila et al. 
reported that VSV-G is retained in an intracellular compartment after muscle cell differentiation. We did not observe such retention, perhaps because of differences in the experimental conditions: Rahkila et al. 
used full virus infection to express VSV-G in L6 cells, whereas we used cDNA transfection to express VSV-G in C2 myoblasts and myotubes.
Microtubules play structural roles in muscle-specific events such as myoblast elongation, fusion or sarcomere formation 
and they are essential to differentiation 
. They also play a role in processes that are not muscle-specific but are important in muscle, such as macroautophagy 
. The fact that we find only minor defects in cargo trafficking when microtubule organization is altered and the Golgi complex is mislocalized, suggests that cells can tolerate some level of disturbance of microtubules and related organelles. This would be good news for conditions such as Pompe disease, or Duchenne Muscular Dystrophy, in which Golgi complex and microtubules are affected