Who Needs Microtubules? Myogenic Reorganization of MTOC, Golgi Complex and ER Exit Sites Persists Despite Lack of Normal Microtubule Tracks

A wave of structural reorganization involving centrosomes, microtubules, Golgi complex and ER exit sites takes place early during skeletal muscle differentiation and completely remodels the secretory pathway. The mechanism of these changes and their functional implications are still poorly understood, in large part because all changes occur seemingly simultaneously. In an effort to uncouple the reorganizations, we have used taxol, nocodazole, and the specific GSK3-β inhibitor DW12, to disrupt the dynamic microtubule network of differentiating cultures of the mouse skeletal muscle cell line C2. Despite strong effects on microtubules, cell shape and cell fusion, none of the treatments prevented early differentiation. Redistribution of centrosomal proteins, conditional on differentiation, was in fact increased by taxol and nocodazole and normal in DW12. Redistributions of Golgi complex and ER exit sites were incomplete but remained tightly linked under all circumstances, and conditional on centrosomal reorganization. We were therefore able to uncouple microtubule reorganization from the other events and to determine that centrosomal proteins lead the reorganization hierarchy. In addition, we have gained new insight into structural and functional aspects of the reorganization of microtubule nucleation during myogenesis.


Introduction
The Golgi complex is traditionally thought of as a single organelle per cell, and represented as a stack of flattened cisternae next to the nucleus. However, alternative organizations are found in skeletal [1,2,3] and cardiac [4] muscle, in osteoclasts [5,6], plant cells [7], yeasts [8], polarized endothelial cells [9] and Drosophila embryos [10,11]. The Golgi complex organization is also altered during mitosis [12,13,14], in apoptotic cells [15], in diseases such as Amyotrophic Lateral Sclerosis [16,17], and in animal models of diseases such as Duchenne Muscular Dystrophy [18,19,20] and Pompe Disease [21,22,23]. Understanding how the Golgi complex transitions between different morphologies should help us assess the consequences of these modifications.
Skeletal muscle cell cultures are particularly informative since their Golgi complex transitions from a classic to an alternative fragmented organization during differentiation. This reorganization coincides with changes of the microtubule-organizing center (MTOC), from a classic centrosome to a combination of perinuclear belt and centrosomal remnants, and with remodeling of the microtubule network [1,2,24]. The reorganization of the Golgi complex also coincides with that of the ER exit sites (ERES) [1,2,24,25]. The muscle Golgi complex continues to be remodeled during in vivo myogenesis to form a fiber type-dependent network of hundreds of small stacks of cisternae, closely associated with ERES and positioned throughout the fibers [3,26]. These small Golgi complex elements are associated with a three-dimensional microtubule lattice [26,27].
Reciprocal relations between the MTOC, microtubules, Golgi and ERES make the matter more complicated: centrosomes nucleate microtubules but, conversely, microtubules are involved in keeping the essential centrosomal proteins pericentrin and ctubulin in the centrosome [28,29]. Similarly, microtubules are necessary for the integrity of the Golgi complex [30], which is positioned near the centrosome by minus-end directed microtubule motors [31]. However, it is now accepted that the Golgi complex itself is involved in microtubule nucleation [32,33]. Finally, ERES themselves are positioned along microtubules through interaction of their COPII coat proteins with dynactin [34]. Golgi complex elements in differentiated muscle cells can thus interact with microtubules directly, or indirectly through the ERES. In light of all these potential interdependencies, the hierarchy of the microtubule-Golgi complex changes during muscle differentiation is far from clear.
Previously we have shown similarities between Golgi complex changes during myogenesis and during microtubule depolymerization [24,25]: in both cases, the Golgi complex becomes fragmented, and the resulting elements are positioned at ERES. However, Golgi fragments produced by microtubule depolymerization do not form a perinuclear belt, but are dispersed through the cytoplasm [35]. These results suggested that microtubules might be dispensable for some but not all steps of Golgi complex reorganization during myogenesis. To test this notion and to clarify the interdependence of the several reorganizations (MTOC, microtubules, Golgi complex, ERES), we decided to uncouple them by using microtubule-altering drugs.
The results presented here show that altering or removing microtubules does not prevent myogenic reorganizations. Centrosomal proteins provide the platform for the positioning of the Golgi complex and ERES at the nuclear envelope and emerge as the key players. We also present evidence that the reorganizations proceed by progressive modification of existing structures rather than by demolition and de novo rebuilding.
Finally, we used EB3-GFP to compare microtubule nucleation in myoblasts and myotubes, at steady-state and after microtubule depolymerization. These experiments show that nucleation occurs constantly from the nuclear membrane of myotubes. They also reveal differences between centrosome and nuclear membrane as to MTOC organization.

Centrosome, Golgi complex and ERES reorganizations resist microtubule alterations
In order to assess the participation of microtubules in the reorganization of MTOC, Golgi complex and ERES during muscle development, we challenged cultures of the mouse skeletal muscle cell line C2 [36] to differentiate in the presence of microtubule-altering drugs. Besides taxol and nocodazole, whose cell-wide microtubule stabilization and destabilization, respectively, are well known, we also used a specific inhibitor of GSK3-b, DW12 [37]. Inhibition of GSK3-b affects cell migration in the wound model by preventing the preferential stabilization of microtubules in the direction of cell migration [38,39,40]. We have found that DW12 (like lithium chloride, a much less specific inhibitor of GSK3-b) stabilizes microtubules preferentially around the nuclei.
Our aim was to disturb the microtubule network before differentiation starts in the unsynchronized C2 cultures, while keeping cells differentiation-competent. Therefore, the drugs were added 4 hours after plating the cells in growth media (GM). For each drug, we tested a range of concentrations. We found that 50 ng/ml of nocodazole sufficed to stop cell growth ( Figure S1A), but we used 200 ng/ml, the minimum concentration needed to remove all large microtubules in 48 hours. A similar test for taxol and DW12 concentrations led us to choose 50 nM and 100 nM respectively. After 48 hours in GM in the presence of the drugs, the desired effects on microtubules were indeed achieved ( Figure  S1B). Myogenin, the marker used to assess differentiation, was expressed in fewer than 15% of the cells ( Figure S1C). Differentiation was initiated by switching the cultures to fusion medium (FM), and the effects of the treatments were assessed 24 hours later. By that time, control cultures' expression of myogenin had increased and small myotubes were present, indicating that differentiation was well under way. Figure S2 shows the state of the microtubules at the end of the treatments. Glu-tubulin, an indicator of microtubule stabilization, was present with different patterns in taxol and DW12-treated cells but also in very short microtubules left in nocodazole-treated cells. We did not consider the inhibition of fusion by nocodazole and taxol ( Figure S2) a hindrance, since the events we are monitoring precede fusion [24,41]. All quantitations were done on myocytes (mononucleated, differentiated cells).
We assessed the treatment effects by double-staining for myogenin versus pericentrin, GM130, or Sec31, to probe for reorganization of centrosome, Golgi complex, and ERES respectively and distinguish differentiated from undifferentiated cells. Results for pericentrin are shown in Figure 1. All cultures, regardless of treatment, showed increased myogenin expression. Furthermore, a large fraction of the myogenin-positive cells showed the perinuclear pericentrin staining typical of normal differentiation, even when cell and nuclear shape were abnormal. In contrast, none of the myogenin-negative cells showed reorganized pericentrin. Similarly, the Golgi complex and ERES were redistributed to the nuclear envelope of many of the myogenin-positive myocytes, regardless of drug treatment. Quantitation of these observations (Table 1) confirms that a normal microtubule (re)organization is not necessary for that of the other systems. In fact, pericentrin reorganization appeared specifically enhanced by taxol and nocodazole. Thus, pericentrin, Golgi complex and ERES may follow different pathways when reorganizing. We observed GM130 reorganization only in cells with reorganized pericentrin (Figures 2 and S3), thus pericentrin reorganization appears upstream of the Golgi complex reorganization.
We observed three levels of Golgi complex reorganization: (I) completely reorganized, i.e. fragmented into elements positioned along the nuclear envelope (Figure 2A  ( Figure 2C, F). Nocodazole reduced the size of Golgi elements, in agreement with its known effects, but taxol and DW12 did not. DW12 treatments frequently resulted in pericentrin and Golgi elements at the two poles of the nuclei ( Figure 2G), a rare distribution in control cells. This polar distribution was even found in cells in which most of the Golgi complex failed to fragment ( Figure 2H). Quantitation (Table 2) shows that Golgi complex redistribution is not as complete as pericentrin redistribution in treated cells. Golgi complex reorganization is thus more affected by each treatment than pericentrin reorganization.
We previously proposed that ERES determine Golgi complex localization [24] and we therefore monitored any uncoupling of the two. ERES and Golgi complex redistributions were very closely linked ( Figure 2I-N) and this relationship was not significantly affected by microtubule alterations (Table 2). Even abnormally localized Golgi complex was surrounded by ERES (arrows in Figure 2I-K). Only occasional myocytes showed GM130 reorganization with a poor or non-existent Sec31 reorganization. Therefore it appears that Golgi elements do not need microtubules to be associated with and localized next to ERES. Although we routinely used 200 ng/ml of nocodazole, we also tested concentrations up to 2000 ng/ml and pre-treatment with cold, with similar results ( Figure S4).
Thus we find that the redistribution of centrosome, Golgi complex, and ERES only takes place in myogenin-positive cells and that pericentrin reorganization is upstream of the other components. The redistributions can take place in the absence of dynamic microtubules indicating that these are not essential. However, microtubule manipulations reduce the occurrence of fully normal reorganizations, and this affects the Golgi complex and the ERES more than pericentrin.

Trafficking of cargo is intact in cells with mislocalized Golgi complex
As the drug treatments affected both the structure and positioning of Golgi complex and the microtubules normally needed for Golgi complex function, we next assessed their functional consequences, i.e. cargo traffic ( Figure 3). We concentrated on DW12-treated cells, as these had both recognizable Golgi complexes and microtubule networks (requirements for secretory traffic), even if altered. We used the model cargo VSV-   [42,43,44,45]. VSV-G accumulated in the ER at 40uC and was released in both control cells and DW12-treated cells after shifting to 32uC. It reached the Golgi complex after 10 minutes at 32uC and the plasma membrane after 30 minutes. After one hour, the majority of the VSV-G was on the plasma membrane in control and DW12-treated cultures ( Figure 3A). The timing of trafficking was very similar to published reports for COS-7 cells [46,47]. Quantitation showed remarkably little difference between controls and DW12-treated undifferentiated cells, indicating that mislocalized Golgi complexes fully participate in VSV-G traffic ( Figure 3B). The only observed effect of DW12 treatment was a slight delay in clearing VSV-G from the Golgi complex in differentiated cells. These results indicate that even mislocalized Golgi complexes linked to abnormally stabilized microtubules can support normal cargo trafficking.
Developmental stage, not treatment length. predicts success of microtubule recovery after nocodazole treatment It is conceivable that the Golgi complex reorganization would normalize if nocodazole were washed out after differentiation, assuming that microtubules then would regain the distribution normally found in differentiated muscle cells. We thus removed nocodazole after the usual three-day treatment, and assessed recovery of microtubules ( Figure 4). Microtubule regrowth was fast: a complex network was already visible after three minutes ( Figure 4A). The presence of these microtubules however did not improve reorganization of the pericentrin or the Golgi complex. The fraction of cells with perinuclear pericentrin was not modified and Golgi complex reorganization was only partially improved ( Figure 4B, C). However, the reformed microtubule network differed from both myoblast and myotube normal networks. It contained numerous nodes as if it resulted from nucleation on dispersed sites. Indeed, in cells fixed just before nocodazole was washed out, we observed short residual microtubules, each connected to a pericentrin dot, and each containing Glu-tubulin ( Figures S2 and 4D). It looks therefore as if nocodazole washout lets microtubules rapidly reform by elongation from these short residual microtubules. The dispersal of pericentrin to these many short microtubules was found in all nocodazole-treated myoblasts (myogenin-positive or not), but not in untreated cells or in cells treated with taxol or DW12. After 6 hours of recovery from nocodazole, the microtubule network did not show further changes. To assess whether the abnormal recovery is simply due to the length of the treatment with nocodazole (72 hours), we subjected normally differentiated myotubes to the same 72-hour nocodazole treatment. At the end of this treatment, the myotubes had lost their elongated shape ( Figure 5A, D). But when nocodazole was washed out, recovery of the cell shape and of a normal-looking microtubule network was complete in 24 hours ( Figure 5I). Nocodazole also caused all but small pieces of microtubules to disappear ( Figure 5E, F). But unlike in cells differentiating in nocodazole, many short microtubules and some Glu-tubulin staining were concentrated around the nuclei ( Figure 5B, H). Perhaps more importantly, pericentrin ( Figure 5C) and Golgi elements ( Figure 5G) remained positioned around the myotube nuclei. Thus, when differentiation occurs in nocodazole, some irreversible changes affect microtubules. These changes are likely responsible for the incomplete or imperfect redistribution of centrosomal and Golgi complex proteins.
During normal differentiation, pericentrin relocates gradually and provides a platform for microtubule nucleation, Golgi complex and ERES Some rapid cellular events that depend on microtubules, i.e. mitosis, involve a breakdown of existing structures (the normal microtubule network) and de novo buildup of another one (the mitotic spindle). The events are regulated by a cascade of signaling events but are not regulated at the transcriptional level. Myogenesis, in contrast, is largely regulated at the transcription level. However, the specific events we are interested in appear gradual. To find out which scenario is involved in organelle redistribution, we searched for cells with partial reorganization of the organelles in untreated cultures ( Figure 6A, B). We did find, early in differentiation, elongated cells with a partial nuclear cap of pericentrin emanating from the centrosome ( Figure 6C). These cells represent a constant fraction of the myogenin-positive population ( Figure 6E, F), as expected from cells entering and leaving a transitional state at similar rates in unsynchronized cultures. In contrast, a steady increase in the number of cells with the partial phenotype would indicate failed progression of differentiation. Even partial nuclear pericentrin caps supported microtubule nucleation ( Figure 6D).
Golgi complex reorganization appeared progressive as well ( Figure 7A). However, pericentrin expanded more rapidly to a larger domain than the Golgi complex. Furthermore, 3D reconstruction revealed that pericentrin does not form a belt as usually referred to, but a complete shell around nuclei ( Figure 7A, C). GM130 was found in different patterns, from partial belts ( Figure 7B) to full shells ( Figure 7C). The latter became predominant only in multinucleated cells (see quantitation, Figure 7D).
To determine if ERES guide Golgi complex reorganization, as postulated in our model [24], we searched for mismatch between ERES and Golgi complex in transitioning cells. A partial GM130 belt was found as frequently with as without perinuclear ERES clusters ( Figure 7E) but perinuclear ERES clusters without a perinuclear Golgi complex belt were not observed. Quantitative PCR ( Figure 8A) demonstrated that centrosome and Golgi complex mRNA changes are two to three orders of magnitude smaller than changes in myogenin mRNA. In addition, we found no differentiation-related shift between the isoforms of pericentrin (A and B or kendrin [48]). At the protein level as well, changes in centrosomal and Golgi complex proteins are modest compared to those for myogenin ( Figure 8B, C). We conclude that myogenic reorganization of the centrosome and Golgi complex involves a gradual transformation and relocation of the organelles.

Microtubules constantly nucleate from myotube nuclei
We have proposed a model [24], which explains the presence of the Golgi complex around the myotube nuclei by the coalescence of the microtubule nucleation sites and of the ERES. Newly formed pre-Golgi complex vesicles that emerge from the ERES and normally move retrogradely to microtubule minus ends, stay put since they are already at the microtubule minus ends. But this model hinges on microtubules constantly nucleating from the nuclear envelope area [24,49] and this has not been shown in unperturbed myotubes. We and others have previously identified the nuclear membrane area of myotubes as a bona fide microtubule nucleation site only during recovery from complete depolymerization [49,50]. The high density of microtubules in unperturbed cells hampers tracing their origin.
We now use a GFP chimera of the microtubule plus-end protein EB3 to observe microtubule dynamics in live muscle cells ( Figure 9). In myoblasts, EB3-GFP puncta move outward in the well-known fountain-like pattern, at the maximum frequency of 32 puncta per 100 seconds per centrosome (range: 14-32 per 100 seconds; Figure 9A,B, Movie S1). In myotubes, in contrast ( Figure 9A, B, Movie S2), EB3-GFP puncta have no concerted movement, in agreement with the published behavior of EB1 [51]. Rather, EB3-GFP puncta move toward the plasma membrane and toward the myotube ends, often crossing paths. We were able to track EB3 puncta originating from myotube nuclei ( Figure 9C). The maximum number of puncta observed per 100 seconds per nucleus was 10 (range: 1-10 per 100 seconds, Figure 9B). Thus there is steady-state microtubule nucleation at the myotube nuclear envelope. Trajectory analysis showed that some cytosolic spots repeatedly produce EB3 puncta. These spots probably represent centrosomal remnants, nucleation-competent cytoplasmic elements containing a centriole and centrosomal proteins [49].
To determine whether steady-state nucleation at the nuclear membrane is associated with Golgi elements, EB3-GFP and the Golgi marker GalT-mCherry were co-transfected. GalT-mCherry faithfully targets the Golgi complex in myotubes, as verified by anti-giantin staining ( Figure 10A). If Golgi elements nucleate microtubules, EB3-GFP trajectories should start from GalT-mCherry-labeled Golgi stacks. As shown in Figure 10B-D and Movie S3, this was not the case. Microtubule nucleation was observed from areas not labeled with Galt-mCherry.
The number of EB3 puncta originating from myotube nuclei in unperturbed cells appeared modest when compared to the density of small microtubules observed after recovery from complete microtubule depolymerization [49,50]. To understand this difference, we imaged EB3-GFP expressing cells on a confocal microscope at 37uC throughout nocodazole treatment and recovery. Unlike myoblast centrosomes which produce a burst of EB comets immediately following washout ( Figure 11A, Movies S4 and S5), myotube nuclear membranes, in the same experimental conditions, do not show comets emerging from the nuclear rim ( Figure 11B, Movies S6 and S7). Moreover, comets could only be found after several minutes of recovery,and in the cell periphery. Quantitation of EB fluorescence showed that the myoblast centrosomal area becomes transiently brighter after washout while myotube nuclear areas do not ( Figure 11C, D). To rule out that the lack of EB comets in myotubes is due to the GFP tags on the EB proteins, we examined untransfected cells fixed 5 minutes after nocodazole washout and stained with anti-EB1 or -EB3. We found abundant EB1 or EB3 around the centrosomes of myoblasts, but practically none at myotube nuclear rims (Figure 11 E). Since EB1 is normally in constant exchange between microtubules and cytosol, we wanted to rule out that this exchange, possibly accelerated in myotubes, explains the absence of EB staining during microtubule recovery. A pre-fixation protocol combining a brief detergent extraction with taxol to stabilize EB proteins and preserve the accurate state of the in vivo microtubule assembly (see Methods and [52]) allowed us to visualize small newly nucleating microtubules and EB1 simultaneously, 5 minutes after nocodazole washout ( Figure 11F). Both myotubes and myoblasts show many small microtubules ( Figure 11F middle panel). A weak signal along the nuclear membrane of the myotubes indicates the presence of a small amount of EB1, but the staining remains much weaker than in myoblasts.
Thus we have demonstrated that nucleation of microtubules from the nuclear membrane of myotubes is constant, and we have uncovered an unexpected lack of EB proteins on small microtubules emerging at the earliest times after removal of nocodazole.

Discussion
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 [24] 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 [53]. ERES fusion and fission have been studied extensively [54] 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 c-tubulin to centrosomes [28,29,55], 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  Figure 4A). Taxol similarly enhances redistribution of centrosomal proteins but DW12 treatment, which inhibits GSK3-b, does not. Both taxol and DW12 stabilize microtubules, but the extent and pattern of stabilization differ. In addition, GSK3-b 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 [56].
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 ( Figure 5 and [50]). 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 [25,26] 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. [43], 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 [48,57] but never actually reported. We also revisited microtubule nucleation. Previous studies, including ours [1,49,50], 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 [58] whereas nuclear mem-branes 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 [59]). 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 nonmuscle 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 [33] and regrowth of microtubules from existing stable tubulin seeds [60]. Regrowth from existing seeds seems the most likely explanation for the short Glu-tubulin-containing microtubules depicted in Figure 4D.
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 [61]), 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-b inhibitor barely slows down cargo trafficking, an essential Golgi complex function. Rahkila et al. [43] 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. [43] 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 [62,63,64] and they are essential to differentiation [65,66]. They also play a role in processes that are not muscle-specific but are important in muscle, such as macroautophagy [22,23]. 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 [18,19].

Antibodies and Reagents
Mouse anti-GM130 antibodies were from Transduction Labs (Lexington, KY), anti-giantin from Covance (Denver, PA), anti-  . Timelapse recordings of EB1-or EB3-GFP highlight differences between myoblasts and myotubes after washout from acute nocodazole treatment. Myoblasts and myotubes expressing EB1-GFP were imaged before, during, and after treatment with nocodazole (1 hour, 5 mg/ml). (A) In myoblasts, EB1-GFP comets are present before treatment (A, Pre), absent during nocodazole treatment (not shown) and at the start of washout (A, t = 1 min). Over the next several minutes, EB1-GFP accumulates in the centrosomal region (A, t = 5.5-8 min). When the return of EB1-GFP comets throughout the cytosol indicates the restart of steady-state microtubule growth, the level of EB1-GFP at the centrosome returns to its initial value (A, t = 13 min). (B) There is no such increase in the EB1-GFP concentration at myotube nuclear rims, despite the return of EB1-GFP comets in the cytosol at 6.5 minutes (see detail of boxed area), indicating that microtubule growth has commenced. (C) Analysis of the level of EB1-GFP in the areas outlined in red in the images shown in A (black) and B (red). (D) Quantitation of EB1-GFP and EB3-GFP levels after nocodazole washout in myoblast centrosomes and myotube nuclear rims. (E) To exclude that the results in B are due to problems with the GFP-tagged constructs, the same treatment was carried out on untransfected cultures. Five minutes after nocodazole washout, cells were fixed and stained for EB1 and EB3. Again, neither accumulates at myotube nuclear rims (arrowheads), while both are prominent at myoblast centrosomes (arrows). (F) When a microtubule stabilization protocol combining detergent extraction and taxol is used before fixation, only very weak EB1 staining is found near myotube nuclear rims (arrowheads), despite the presence of many short microtubules in this location. Bright staining is present at centrosomal regions (arrows) and on small microtubules in myoblasts' cytoplasm. (E, F: dashed lines outline myotubes). Bars = 5 mM (main panels), 1 mM (detail). doi:10.1371/journal.pone.0029057.g011 and rabbit anti-Sec31/p137 (Dr. F. Gorelick, Yale University School of Medicine, New Haven, CT). Alexa488-and Alexa568conjugated secondary antibodies were purchased from Invitrogen (Carlsbad, CA). HRP-conjugated anti-mouse and anti-rabbit IgG were from Bio-Rad (Hercules, CA) and HRP-conjugated antichicken IgY, together with BlockHenII, were from Aves Labs, Inc. (Tigard, OR). Hoechst 33342, nocodazole and taxol were from Sigma. DW12 was a kind gift from Dr. E. Meggers (University of Pennsylvania, Philadelphia, PA).

Cell culture and drug treatments
Mouse C2 skeletal muscle cells were cultured in growth medium (GM) and differentiated in fusion medium (FM) as described [24]. Drugs were added 4 hours after plating to achieve 200 ng/ml (0.66 mM) nocodazole or 50 nM taxol. In some experiments cells were placed on ice for 30 minutes after the addition of nocodazole and then warmed up to at 37uC in the drug. Because nocodazole and taxol inhibit cell proliferation, dishes for these treatments received twice the control cell number. For nocodazole washout, cells were fixed or rinsed three times in PBS, and further cultured in drug-free media at 37uC for the times indicated before fixation. The drug DW12 was used at 100 nM. This organo-ruthenium compound specifically inhibits GSK3-b and, like the broad inhibitor lithium chloride, leads to local microtubule stabilization [37,67,68].

Immunofluorescence and microscopy
Cells were fixed in 100% methanol at 220uC for 6 minutes, blocked in PBS containing 1% bovine serum albumin, 2% horse serum and 3% normal goat serum for one hour, incubated with primary antibodies for 1-2 hours, and with secondary antibodies for 45 minutes (all at room temperature), counterstained with Hoechst 33342, and mounted in Vectashield (Vector Laboratories, Burlingame, CA) or in home-made mounting media consisting of p-phenylenediamine in 80% glycerol. Wide-field immunofluorescence images were collected on a Leica DMR, using a 636 1.32 N.A. oil immersion objective lens and a Hammamatsu C4742-95 camera driven by IVision software (BioVision Technologies, Exton, PA). Confocal images were collected either on a Zeiss 510 LSM using a 636 1.4 N.A. oil immersion objective lens and multi-track configurations, or on a Leica SP5 confocal microscope with a 636 1.4 N.A. oil immersion objective lens, using sequential scanning. The figure legends specify whether images are single optical planes or projections of stacks of optical sections throughout the cell (z-stacks). Images were exported in 8-bit TIF format and linearly adjusted using Adobe Photoshop.

Quantitation of reorganization
To assess the effect of a range of nocodazole concentrations on subcellular organization, cells were grown for 48 hours in GM and differentiated in FM for 24 hours with the indicated concentration of nocodazole. Images were collected every two hours using an IncuCyte image system (on a kind loan from Essen Instruments, Ann Arbor, MI). Cell confluence was plotted using the manufacturer's software. Since C2 cells kept in microtubuledisrupting drugs fuse poorly, we analyzed differentiated but unfused cells (myocytes), even in control cultures. For Tables 1 and 2, we scored 600 myogenin-positive cells per time-point for the presence of perinuclear staining (belt), by observation on a wide-field microscope. Cells with the majority of the marker in a perinuclear belt were classified as reorganized. Cells with the marker divided between perinuclear belts and additional locations were classified as partially reorganized. Cells without perinuclear belts were classified as not reorganized. In the next step, cells with reorganized pericentrin were selected and analyzed for Golgi complex morphology. Finally, cells with reorganized Golgi complex were analyzed for reorganization of ERES. Data are expressed as means 6 s.d. An unpaired Student's t-test was used for statistical analysis (Excel software). For analysis of reorganization after nocodazole washout, 400 cells in two independent experiments were analyzed for the presence of perinuclear pericentrin or GM130. The analysis of pericentrin reorganization in control cultures (Figures 5 and 6) was done in 200 myogeninpositive and 200 myogenin-negative cells at each indicated time point. Additionally, confocal z-stacks were collected of cells with reorganized pericentrin (one day in FM), and in those, the morphology of the Golgi complex was classified using 3D renderings in Imaris software (Bitplane, Saint Paul, MN).

Microtubule Nucleation Assay
Cells were incubated in 5 mg/ml nocodazole on ice for 30 minutes to sever microtubules and then at 37uC for 30 minutes to depolymerize microtubules completely. To regrow microtubules, cells were quickly rinsed twice in warm PBS and incubated for the indicated times in GM or FM at 37uC. For better visualization of the regrowing microtubules, cells were extracted at room temperature for one minute in PHEM buffer (60 mM piperazine-N,N-bis-2-ethanesulfonic acid, 25 mM HEPES, 10 mM EGTA, 2 mM MgCl 2 , pH 6.9) containing 1% Triton X-100 and rinsed in PHEM buffer for 20 seconds before fixation. For visualization of both EB proteins and small microtubules right after nocodazole washout, cells were incubated for one minute in 4 mM taxol in a buffer containing 0.5% Nonidet P-40, 20 mM Tris HCl, 140 mM NaCl, 1 mM MgCl 2 , 2 mM EGTA (pH 6.8 [52]) at RT before fixation.

Cell homogenates and Western analysis
Cells growing on Petri dishes were trypsinized, collected by centrifugation, washed twice, and taken up in sample buffer (National Diagnostics, Atlanta, GA). The samples were assayed for protein concentration (Pharmacia 2d Quant kit, Peapack, NJ); proteins were separated on pre-cast tris acetate or bis-tris gradient gels (Invitrogen) at 100 V for one hour or at 200 V for 35 minutes and transferred to nitrocellulose membranes at 51 V for 4 hours or at 20 V overnight at 4uC. The membranes were blocked in 5% milk powder in TBS for one hour (except for 2% FBS and 1% NGS in TBS for a-mannosidase II, or 7.5% BlockHen, 5% NGS in PBS for anti-kendrin), incubated with primary antibodies for two hours and with HRP-conjugated secondary antibodies for one hour at room temperature. The peroxidase activity was revealed with Pierce Supersignal West (Rockford, IL) and captured on Kodak BioMax MR film (Rochester, NY). Films were scanned on a UMAX scanner (Dallas, TX) and band intensities were measured with ImageJ software (written by W.S. Rasband, National Institutes of Health, and available at http://rsb.info. nih.gov/ij/). Molecular weights for bands on pericentrin films were estimated using HiMark Prestained HMW protein standard (Invitrogen). Polynomial curve-fitting of the relative migration distances (Rf) was done in Kaleidagraph software (Synergy Software, Reading, PA).

Secretory cargo trafficking assay
Cells were plated on coverglass and DW12 (100 nM) was added after 4 hours. After another 16 hours, the cells were transfected with 1.2 mg of VSV-G ts045 -YFP cDNA using FuGENE transfection reagent (Roche, Indianapolis, IN) and incubated for 24 hours at 37uC. The cultures were then switched to 40uC. Control cultures grow and fuse normally at 40uC. Some cultures were kept at 40uC for one day, then incubated for various times at 32uC for cargo release, and fixed (undifferentiated cultures). Others were cultured in FM at 40uC for two days, then incubated for various times at 32uC, and fixed (differentiated cultures). The location of pools of VSV-G was analyzed in 100 cells per time-point, using 5 possible categories: ER, Golgi (identified by anti-GM130 staining), plasma membrane, and a combination of these.

EB3-GFP imaging
Cells in MatTek dishes (MatTek, Ashland, MA) were transfected one day after plating with one mg of EB3-GFP cDNA (a kind gift from Dr. A. Akhmanova, Erasmus Medical Center, Rotterdam, the Netherlands) using FuGENE transfection reagent. Undifferentiated cells were imaged one day post-transfection, differentiated cells after two days in FM. Image series of living cells were collected at 37uC on a Zeiss 510 LSM, using a 636 1.4 NA objective, at one frame/second. Files were exported in 8-bit TIF format, and images were processed by Kalman filtering in ImageJ. EB3-GFP tracks were analyzed in the nuclear area of two myoblasts (one nucleus each) and three myotubes (with combined total of eight nuclei). In stacks of 100 frames ( = 100 seconds), EB3 puncta arriving at the centrosome (arrow in undifferentiated cell) or at a myotube nucleus were counted as N a and EB3 puncta leaving from it were counted as N l . Each dot in the graph represents the net number of puncta (N l 2N a ).
Analysis of EB1-GFP or EB3-GFP at centrosomes and nuclear rims after nocodazole washout Cells in 35 mm MatTek dishes were imaged on a Zeiss 510 Meta microscope at 37uC at one frame/second, then nocodazole was added as a 26 stock (5 mg/ml final concentration) and the imaging was slowed to 1 frame/5 min. After 1 hour, the dish was washed 3 times. Imaging was restarted at 1 frame/second as soon as the original focus level was restored. Image series were exported as TIF files. For illustrations, images were inverted and non-linear adjustments for brightness were made in Adobe Photoshop. To show details of comets after the washout, images for the separate panels were further enhanced in a non-linear manner. Intensity changes in centrosomal regions (circular) and at nuclear rims (freehand shape outward about 2 mm from the edge) were quantified with ImageJ on unprocessed files. Normalized values (to the first frame after washout) were plotted. The maximal increases measured for myoblast centrosomes and myotube nuclei were plotted using Kaleidagraph. Considerable changes in the position of the myotube nuclei, necessitating refocusing, made the myotube curves less smooth than the myoblast curves. Figure S1 Effects of microtubule-affecting drugs on C2 cultures. (A) Optimization test for nocodazole concentration; cultures were imaged every 2 hours. Cell proliferation is arrested from 50 ng/ml on. (B) State of microtubules (a-tubulin, green) and microtubule stabilization (Glu-tubulin, red) before the switch to FM to initiate differentiation. (C) Myogenin expression in control and treated cultures after 2 days in GM, before the switch to FM to initiate differentiation. Wide-field images. Bar: 10 mm. (TIF) Figure S2 Cell shape and state of the microtubules in C2 cells differentiated in microtubule-disrupting drugs. Control cultures show many elongated cells and multinucleated myotubes with longitudinal microtubules (a-tubulin, green). Microtubule stabilization is reflected by Glu-tubulin staining (red, arrowheads), which increases in differentiated cells. In cultures treated with taxol (50 nM), the cells do not fuse but are large and elongated with high levels of Glu-tubulin. DW12 (100 nM) has no global effect on the microtubule network but, in most cells, increases Glu-tubulin next to the nucleus, and inhibits fusion. Nocodazole (200 ng/ml) prevents both elongation and fusion. Only very short microtubules are left, which partly stain for Glu-tubulin. Images are confocal z-stack projections.