A stable microtubule array drives fission yeast polarity reestablishment upon quiescence exit

Microtubules (MT) are hollow cylindrical polymers that are essential in all eukaryotic cells. They are formed by the non-covalent interaction of α- and β-tubulin heterodimers. MTs are nucleated at MT organizing centers (MTOCs) by a γ-tubulin–containing complex that acts as a MT template. As a MTOC stays attached to MT, it stabilizes the so-called MT minus end (Kollman et al., 2011). At the opposite end, the plus end, MTs elongate by the addition of GTP tubulin. During assembly, the β-tubulin–bound GTP is hydrolyzed and a stable GDP+Pi tubulin intermediate is generated. The eventual Pi release causes a change in tubulin–tubulin interaction that favors MT depolymerization (catastrophe) that can be rescued by de novo GTP tubulin addition (Carlier and Pantaloni, 1981; Nogales and Wang, 2006; Alushin et al., 2014; Mitchison, 2014). In fact, MT plus ends are thought to be protected from catastrophe by a stabilizing GTP tubulin “cap” (Mitchison and Kirschner, 1984; Horio and Murata, 2014). Thus, MTs alternate periods of growth and shrinkage, a behavior called dynamic instability (Mitchison and Kirschner, 1984). In vivo, this dynamic is regulated by plethora of MT-associated proteins (MAPs) that modulate MT length by influencing the polymerization or depolymerization rates and/or the catastrophe or rescue frequencies (Wade, 2009; van der Vaart et al., 2009; Horio and Murata, 2014). Finally, other MAPs are involved in MT organization into diverse structures such as asters, spindles, or bundles, each of which is required for a particular cellular process. One of the key questions is therefore how cells can spatially and temporally control the assembly of distinct MT-containing structures to elicit specific cellular functions.

Fission yeast has proven to be a powerful model to study how different MT structures are constructed all along the cell cycle (Hagan, 1998). Depending on the cell cycle stage, Schizosaccharomyces pombe displays three different classes of MTOCs. In mitosis, the spindle pole body (SPB), a yeast MTOC equivalent to the centrosome, nucleates MTs that form the mitotic spindle required for chromosome segregation. After mitosis completion, an equatorial MTOC appears as a ring in the cell center and nucleates a post-anaphase array of MTs involved in cell septation during cytokinesis. After cell division, the equatorial MTOC breaks down to form multiple interphase MTOCs that localize all around the nuclear membrane. The SPB stays linked to the cytoplasmic face of the nuclear envelope and can possibly nucleate interphase MT. Thus, in interphase, fission yeast cells display three to five long MT bundles that extend along the long axis of the cell (Hagan and Petersen, 2000; Sawin and Tran, 2006).

Interphase MT bundles are composed of about four dynamic antiparallel MTs (Drummond and Cross, 2000). The formation of these structures relies on the Mto1/2 complex that recruits and activates the γ-tubulin–containing complex on the cytoplasmic face of the SPB, at nuclear membrane–associated interphase MTOCs, and onto preexisting MTs, thus allowing the nucleation of new MTs (Venkatram et al., 2004, 2005; Janson et al., 2005; Samejima et al., 2005, 2010; Zimmerman and Chang, 2005; Lynch et al., 2014). In the latter case, the newly nucleated MT is pulled toward the minus end of the underneath MT template via the motor activity of the kinesin-14 Klp2, until Ase1, a homodimeric MT bundling protein, acts as a brake (Carazo-Salas et al., 2005; Carazo-Salas and Nurse, 2007; Janson et al., 2007; Braun et al., 2009, 2011). The MT length is regulated by several MT plus end tracking proteins including Mal3 (the EB1 homologue) and Tip1 (the Clip170 homologue), two proteins that favor MT rescue (Beinhauer et al., 1997; Brunner and Nurse, 2000; Busch and Brunner, 2004; Höög et al., 2013), Alp14 (a TOG orthologue) that has a MT polymerase activity (Al-Bassam et al., 2012), and the kinesins-8 Klp5 and Klp6 that promote catastrophe (Unsworth et al., 2008; Tischer et al., 2009; Erent et al., 2012). This complex interplay between diverse MAPs results in a typical interphase MT bundle organization, such as within the bundle the overlapping stable MT minus ends are found near the nucleus whereas the dynamic plus ends are extending toward the cell tips (see model in Fig. 8 A; Drummond and Cross, 2000; Piel and Tran, 2009; Hachet et al., 2012). This dynamic arrangement enables the MT plus ends to produce a balance of pushing forces that position the nucleus at the center of the cell (Tran et al., 2001) and will ultimately lead to a symmetrical cell division (Daga and Chang, 2005).

Interphase MT bundles have an additional physiological function. Just after division, cells initiate growth at the preexisting cell end, the old end. Dynamic interphase MTs play an active role in this process. Indeed, by touching the cell cortex, growing MT plus ends deliver polarity factors to cell extremity. One of these factors, the Kelch repeat–containing protein Tea1, docks onto its membrane-associated receptor, the prenylated protein Mod5 (Snaith and Sawin, 2003). Once at the cell tip, Tea1 recruits the Formin For3 and its associated partners, such as Bud6, enabling the nucleation of actin filaments that will be subsequently assembled into actin cables. Actin cables will then serve as tracks for the myosin V–dependent polarized transport of all the building blocks required for cell growth. Hence, in interphase, MTs establish the site of polarized growth and mutations that impinge on interphase MT bundle organization or dynamics consequently affect cell shape (Chang and Martin, 2009; Martin, 2009; Hachet et al., 2012).

Like other eukaryotes, upon nutrient exhaustion, S. pombe cells stop growth and proliferation and enter a temporary non-dividing state called quiescence (Sajiki et al., 2009). While quiescence may encompass very different physiological situations depending on the nature of the exhausted nutrient (Costello et al., 1986), it is clear that fission yeast quiescence establishment is accompanied by profound modifications of the cell transcriptome, proteome (Marguerat et al., 2012), and metabolome (Sajiki et al., 2013) and by the acquisition of a variety of features including an enhanced resistance to harmful external conditions or the accumulation of storage molecules (Wei et al., 1993; Su et al., 1996). We have previously shown that upon carbon source exhaustion, the fission yeast proteasome exits the nucleus and forms a specific cytoplasmic structure called proteasome storage granule (Laporte et al., 2008). Such relocalization has also been reported in budding yeast (Laporte et al., 2008). In addition, in Saccharomyces cerevisiae, actin filaments collapse into actin bodies (Sagot et al., 2006), and various chaperones (Tapia and Morano, 2010; Liu et al., 2012), enzymes (Narayanaswamy et al., 2009; Noree et al., 2010), or kinases (Shah et al., 2014) are relocalized upon quiescence entry. Some of these reorganizations have been observed in metazoans (Noree et al., 2010), suggesting that they may be conserved among eukaryotes.

In this study, we show that fission yeast cytoskeletons are completely reorganized upon quiescence entry. Indeed, we primarily observed that the MT cytoskeleton is drastically reshaped as quiescent cells assemble a large SPB-associated MT bundle composed of astonishingly stable antiparallel MTs. Importantly, we also report that quiescence establishment comes along with a massive rearrangement of the actin cytoskeleton and the loss of cell polarity. We finally demonstrate that quiescent cell MT elongation is strictly required for cell polarity reestablishment upon quiescence exit.

Upon carbon source exhaustion, fission yeast cells leave the cell cycle and enter quiescence from interphase (Bostock, 1970; Costello et al., 1986). In these conditions, we have analyzed MT organization in wild-type (WT) cells expressing the α-tubulin 2 (Atb2) fused to GFP. As expected, proliferating interphase cells displayed two to five long cytoplasmic bundles composed of 4 ± 1 MTs (Fig. 1, A and B; Höög et al., 2007). Strikingly, we observed that after carbon exhaustion, the number of MT bundles progressively decreased (Fig. 1 A). Four days after carbon exhaustion, the majority of the cells displayed a single MT bundle (Fig. 1 A and Fig. S1 A) that we named Q-MT bundle, standing for quiescent cell MT bundle. This unique MT bundle was composed of more than 15 MTs (Fig. 1 B) that were not necessarily of the same length, as exemplified by the arrow shape of the bundle extremities (Fig. S1 B). In fact, in ∼15% of the cells, the Q-MT bundle could display internal thickness variations (Fig. S1 C). Importantly, by imaging cells coexpressing GFP-Atb2 with the SPB-associated protein Sfi1 fused to CFP and the nuclear membrane protein Cut11 fused to RFP, we observed that the Q-MT bundle was generally associated with the SPB (>70% of the cells; Fig. 1, C, D, and G), even when quiescence was prolonged (Fig. S1 D). Yet, MT bundles not associated with the SPB displayed the same length and intensity than the SBP-associated ones (Fig. S1, E and F). EM analysis of quiescent WT cells showed that within the Q-MT bundle MTs were regularly spaced (Fig. 1, D–G; and Fig. S1 G). Finally, the Q-MT bundle organization and localization within the cell were confirmed by 3D models constructed using serial section electron tomograms (Fig. 1 G).

In interphase cells, cytoplasmic MTs are known to be extremely dynamic as they rapidly alternate periods of growth and shrinkage (Fig. 2 A; Drummond and Cross, 2000). They are therefore very sensitive to drugs that inhibit MT polymerization and cause dynamic MT disassembly, such as methyl benzimidazol-2-yl-carbamate (MBC) or thiabendazole (TBZ; Fig. 2 B; Sawin and Nurse, 1998; Sawin and Snaith, 2004). In contrast, we showed that in quiescent cells, the Q-MT bundle had a constant length (Fig. 2 A) and resisted against massive amounts of drugs (Fig. 2 B) or a cold treatment known to cause the disassembly of dynamic MTs (Fig. S2, A and B). Importantly, no fluorescence was recovered after GFP-Atb2 photobleaching, demonstrating that within the Q-MT bundle, not only MTs were stable but also there was no MT sliding (Fig. 2 C). Of note, like cells entering quiescence upon carbon source exhaustion, upon nitrogen starvation, cells also stabilized a unique MT bundle (Fig. S2 C).

Tubulin binds GTP and can hydrolyze this nucleotide. The influence of the tubulin nucleotide status on MT dynamics in vivo is still a mater of debate (Kueh and Mitchison, 2009). We have analyzed the GTP concentration variation upon entry into quiescence after carbon exhaustion using high pressure ionic chromatography and observed a drastic decrease of the intracellular pool of this nucleotide (Fig. 3 A). Mycophenolic acid (MPA) is a drug that specifically inhibits inosine-5′-monophosphate dehydrogenase, the enzyme that catalyzes the conversion of inosine-5′-monophosphate into guanosine-5′-monophosphate (Allison and Eugui, 2000), the GTP precursor. As in other organisms (Qiu et al., 2000), MPA treatment of proliferating S. pombe cells caused GTP depletion (Fig. 3 B). The GTP drop was reversed by addition of guanine in the medium, demonstrating the specificity of MPA treatment (Escobar-Henriques et al., 2001). Interestingly, MPA treatment did not significantly affect MT growth or shrinkage rates (Fig. 3 C) nor the MT catastrophe and rescue frequencies (not depicted), excluding the possibility that MT stabilization would be an immediate consequence of GTP depletion upon quiescence entry.

Kinetic analysis of MT stabilization upon entry into quiescence after carbon exhaustion revealed that as early as 6 h after glucose depletion the MT bundle linked to the SPB was more stable, i.e., resistant to the MT-destabilizing drug benomyl, than other MT bundles present in the cell (Fig. 3 D). The stabilization of the SPB-associated MT bundle fit with a non-random stabilization prediction for both early (Fig. 3 E) and late quiescence time points (Fig. S2 D). Collectively, these results indicated that the SPB, rather than the GTP intracellular concentration, played a major role in MT stabilization in quiescence.

To get insights into the mechanism of Q-MT bundle formation, we looked for MAPs colocalizing with this structure. First, we found that astonishingly, in quiescent cells, Mto1 and Mto2 were no longer found as dots around the nuclear envelope but rather localized exclusively onto the Q-MT bundle (Fig. 4 A and Fig. S3 A). Line scan analyses revealed that Mto1/2 localized as dots that most likely mark the MT ends (Fig. 4 B and Fig. S3 A), in agreement with Mto1/2 being a MT minus end binding complex (Sawin et al., 2004; Venkatram et al., 2004, 2005; Samejima et al., 2005; Janson et al., 2007). Second, the regular spacing between MTs within the Q-MT bundle observed by EM (Fig. 1 E) strongly suggested the presence of a MT bundling protein, and, accordingly, we found that Ase1 localized all along the Q-MT bundle (Fig. 4 C). In S. pombe, the CLASP family member Cls1 (Peg1) has been involved in MT stabilization and is known to localize to the overlapping zone of interphase MT bundle (Bratman and Chang, 2007). Similarly, in quiescent cells, we found that Cls1 localized with the denser region of the Q-MT bundle (Fig. 4 C). Finally, the MT plus end tracking proteins Alp7 (the TACC protein orthologue) and Alp14 and Dis1 (two TOG-related proteins) localized as dots that probably correspond to MT extremities within the Q-MT bundle (Fig. 4 C), just like the MT minus end–directed kinesin-14 Klp2, although fewer and fainter Klp2 dots were detected (Fig. 4 C). Of note, Mal3, the EB1 homologue, could not be detected in quiescent cells (unpublished data).

We then analyzed the effect of MAP encoding gene deletion on the Q-MT bundle shape, stability, and formation. None of the deletions tested resulted in the absence of MT bundle (Fig. 4 D) or its destabilization (Fig. S3 C). However, in some mutants, whereas the steady-state level of total α-tubulin remained constant (Fig. S3 B), the Q-MT bundle was either both thicker and longer (klp5Δklp6Δ, alp7Δ, and dis1Δ), just thinner (ase1Δ), or both shorter and thinner (mal3Δ and alp14Δ; Fig. 4, E and F). These results indicated that MAPs involved in interphase MT length regulation are also involved in shaping the Q-MT bundle in quiescence. Besides, mutants with an altered Q-MT shape did not display mortality in quiescence (Fig. S3 D) nor a cell shape defect upon quiescence exit (Fig. S3 E), with the exception of alp14Δ cells (see the following paragraph and Fig. 7 C).

In our experimental conditions, glucose exhaustion is responsible for cell entry into quiescence (Fig. S4 A). When quiescent cells were refed with rich medium, the Q-MT bundle rapidly elongated from both extremities, a proof of the MT antiparallel arrangement within the Q-MT bundle (Fig. 5 A and Fig. S4, B and C). In ∼5 min, dynamic MTs polymerizing from the Q-MT bundle touched the cell poles. Of note, tracking SPB movement upon quiescence exit clearly demonstrated that the Q-MT bundle was attached to the SPB (Fig. S4 E). The MT plus end–tracking protein Alp14, although immobile onto the Q-MT bundle in quiescence (Fig. S4 F), was found at the tip of elongating MTs upon quiescence exit (Figs. 5 B and S4 F). In contrast, the minus end–associated protein Mto2 stayed immobile on the elongating Q-MT bundle (Fig. S4 G). Later (>15 min), new MT bundles were assembled de novo at the nuclear periphery. Interestingly, as for other quiescent cell-specific structures (Laporte et al., 2011, 2013), the sole addition of glucose onto quiescent cells trigged both Q-MT bundle elongation and assembly of new interphase MT bundles at the nuclear periphery (Fig. 5 A). Importantly, these two MT rearrangements occurred even if de novo protein synthesis was inhibited by cycloheximide (CHX; Fig. 5 A). Yet, as expected, cells treated with CHX did not elongate upon refeeding (Fig. S4 D). This experiment demonstrated that the MT cytoskeleton remodeling upon quiescence exit did not need de novo tubulin synthesis.

We then wonder whether, upon quiescence exit, Q-MT bundle elongation was caused by the polymerization of free cytoplasmic tubulin or if it required the depolymerization of the MTs embedded into the Q-MT bundle. Cell treatment with both CHX and TBZ revealed that within the first 30 min after cell refeeding neither the fluorescence nor the length of the quiescent cell Q-MT bundle was affected (Fig. 5 C), indicating that within this time frame the Q-MT bundle did not depolymerized. This was confirmed by the absence of variation of the Q-MT bundle fluorescence intensity for at least 20 min after cell refeeding in the presence of CHX alone (Fig. S4 H). Obviously, a TBZ treatment prevented de novo interphase MT bundle formation at the nuclear membrane. As a control, we showed that inhibiting MT polymerization upon quiescence exit did not influence Q-MT bundle stability. Indeed when quiescent cells were refed with new medium in the presence of CHX, and afterward incubated 10 min with MBC (Fig. S4 I), Q-MT bundle length and fluorescence intensity stayed constant for at least 20 min, whereas MT dynamic was clearly recovered. In fact, Q-MT bundles start to disassemble only 1 h after cells refeeding (Fig. S4 I). Collectively these data demonstrated that a free tubulin pool was present in quiescent cell cytoplasm and that MT elongation upon quiescence exit was caused by free tubulin polymerization at the Q-MT bundle extremities.

In actively dividing S.pombe, MTs deposit polarity factors, such as Tea1, to the cell tip to allow polarized growth (Chang and Martin, 2009; Martin, 2009; Piel and Tran, 2009; Hachet et al., 2012). In quiescent S. pombe cells, Tea1 was no longer detected at the cell tips but was rather localized onto the Q-MT bundle (Fig. 6 A). Tip1, a MT plus end tracking protein of the CLIP170 family, was also lost from cell extremities but couldn’t be detected in quiescent cells (Fig. 6 A). We made a similar observation for active Cdc42 (CRIB domain), the exocyst components Exo70 and Sec8 (Fig. 6 B), the myosin V myo52, the polarity factor Pob1 (Fig. S5 A), and the formin For3 (not depicted). Interestingly, in quiescent S. pombe, no actin cable or patch were found. Instead, we observed a big cytoplasmic F-actin–containing structure (Fig. 6 C). This structure displayed all the characteristics described for actin bodies in quiescent S. cerevisiae (Sagot et al., 2006): they do not have a particular size or shape (Fig. 6 C), they are resistant to Latrunculin A (not depicted), and they contain a specific set of actin binding proteins such as fimbrin (Fim1; Fig. 6 C), capping protein (Acp2; Fig. 6 C), and drebrin (Aap1; Fig. S5 A), but not Bud6, Vrp1, Crn1, Arp5, or End4 (Fig. S5 A). These observations indicated that upon glucose exhaustion–induced quiescence entry, S. pombe cells lose their polarity and entirely reorganize their actin cytoskeleton.

Upon quiescence exit, we found that cells rapidly repolarized. Indeed, within minutes, Tea1 returned to the cell tips (Fig. 7 A) as did Tip1 and Pob1 (Fig. S5 B). In parallel, actin bodies quickly disassembled and actin patches reassembled at the cells extremities (Fig. 7 A, Fim1, Acp2, and App1; and Fig. S5 B, Bud6). Of note, Tea1 repolarization could occur even if de novo protein synthesis was inhibited (Fig. S5 C) but, as expected, it was impaired in a mod5Δ mutant (Snaith and Sawin, 2003; Fig. S5 D).

To decipher the potential role of the Q-MT bundle in cell repolarization upon quiescence exit, cells exiting quiescence were treated with MBC, a drug that prevents MT bundle elongation and de novo interphase MT bundle formation at the nuclear periphery (Figs. 5 C and S5 C). In these conditions, we found that repolarization of Tea1 and the actin components Fim1, Acp2, and App1 was impaired in >60% of the cells and drastically delayed in the remaining 40% (Fig. 7 A). At later time points, these polarity defects translated into major cell shape defects (Fig. 7 B) that were not observed in the control experiments done with proliferating cells. This strongly suggested that Q-MT bundle elongation and the return to dynamic MT was required for polarity reestablishment upon quiescence exit.

Interestingly, we found that quiescent alp14Δ cells were incapable of Q-MT elongation and interphase bundle reassembly upon refeeding (Fig. 7, C and D). This later defect was not the result of an impaired relocalization of Mto1/2 onto the nuclear membrane (Fig. S5 E). This demonstrated that Alp14 was required for Q-MT bundle elongation. Further, as MBC treated cells, alp14Δ cells displayed a strong cell shape defect upon quiescence exit (Fig. 7 C), another proof of the involvement of the Q-MT bundle elongation in cell repolarization upon quiescence exit.

To definitively address this point, we used mto1Δ, a mutant known to be unable to nucleate cytoplamic MTs (Sawin et al., 2004). Interestingly, after 5 d, mto1Δ cells do not assemble a bona fide Q-MT bundle but instead display a unique long and thin MT-containing structure composed of MTs that are sensitive to a cold treatment (Fig. 7, E and F). As mto1Δ cells are unable to nucleate cytoplasmic MTs, a shift back to 30°C after a cold treatment did not allow de novo MT polymerization (Fig. 7 E and not depicted). Using this mutant, we were able to get rid of MTs in quiescence and address the phenotype of cells exiting quiescence in the absence of MTs. As shown in Fig. 7 (E and F), the vast majority of the cold-treated mto1Δ cells exiting quiescence without MT exhibit strong polarity and shape defects (see also Fig. S5 F). Collectively, these results demonstrated that the Q-MT bundle elongation and the recovery of dynamic MTs are involved in polarity reestablishment upon quiescence exit.

In this study we have shown that when fission yeast cells enter quiescence after carbon source exhaustion, they remodel their MT cytoskeleton into a unique MT bundle (Q-MT bundle). This Q-MT bundle is composed of more than 15 stable and regularly spaced antiparallel MTs. Importantly, this structure is associated with the SPB, a MTOC that seems to be involved in Q-MT bundle MT stabilization (Figs. 1, 2, and 3). We have also observed that the actin cytoskeleton is drastically reshaped as cells disassemble actin patches and cables and assemble actin bodies (Fig. 6), stable F-actin–containing structures (Sagot et al., 2006). In parallel, polarity markers, such as Cdc42, the exocyst components, and Tea1, are lost from cell tips (Fig. 6). All these rearrangements result in a loss of cell polarity in quiescence. When cells exit quiescence, the Q-MT bundle immediately elongates from both extremities and actin bodies dissociate (Figs. 5 and 7). Within an hour, a typical interphase organization is restored, i.e., three to four thin and dynamic MT bundles extend along the long axis of the cell, actin patches and cables assemble at cell tips, and cells start to grow in a polarized manner.

Upon quiescence establishment, the SBP-associated MT bundle rapidly becomes more stable compared with other MT bundles present in the cell cytoplasm (Fig. 3 D, t + 6 h). Within <2 d after carbon exhaustion, the SPB-associated MT bundle is fully resistant to treatments that destabilize dynamic MTs. This stabilization will be discussed below. As a second step toward Q-MT bundle formation, we propose that the binding of Mto1/2 onto the SPB-associated MTs allows the nucleation of new MTs (Fig. 8 A, step 2). The displacement of the equilibrium toward de novo MT nucleation by Mto1/2-activated γ-tubulin onto the stable SPB-associated MT bundle may be favored by the concomitant loss of Mto1/2 affinity for the nuclear envelope (Fig. 4 A). As a result, new MTs would pile up onto the SPB-associated MT bundle that progressively thicken, helped by MT bundling activities, including the one of Ase1 (Loïodice et al., 2005; Yamashita et al., 2005) and Clasp (Bratman and Chang, 2007; Fig. 8 A, steps 3 and 4). Interestingly, MAPs that regulate the thickness and the length of interphase MT bundles also influence the Q-MT bundle shape (Fig. 4). Indeed, both in interphase and in quiescence, MT bundles are shorter in a mal3Δ mutant (Beinhauer et al., 1997; Fig. 4, D to F) or longer in a klp5Δklp6Δ mutant (West et al., 2001; Fig. 4, D to F), in agreement with the molecular activities of these proteins that, respectively, promote MT rescue or catastrophe (Unsworth et al., 2008; Tischer et al., 2009; Erent et al., 2012). Alp14 has a MT polymerase activity in vitro (Al-Bassam et al., 2012) and, accordingly, alp14Δ cells display both shorter interphase MT bundles (Al-Bassam et al., 2012) and shorter Q-MT bundles (Fig. 4, D–F). Furthermore, just like interphase MT bundles in proliferating alp7Δ cells (Zheng et al., 2006), the Q-MT bundle is longer in quiescent alp7Δ cells (Fig. 4, D–F). Finally, consistent with Ase1 bundling activity (Loïodice et al., 2005), we found that the Q-MT bundle is thinner in quiescent ase1Δ cells (Fig. 4, D–F). The Q-MT bundle formation may therefore hypothetically rely on a self-assembling mechanisms similar to the one proposed for interphase MT bundles (Carazo-Salas and Nurse, 2006; Daga et al., 2006; Carazo-Salas and Nurse, 2007; Janson et al., 2007; Subramanian and Kapoor, 2012).

During proliferation, nearly 100% of the cells display SPB-associated MTs. In quiescence, ∼75% of the Q-MT bundles are associated with the SPB (Figs. 1 C and S1 D). We have never directly observed a Q-MT bundle “detaching” from the SBP. Therefore, whether free Q-MT bundles result from SPB detachment events or if their stabilization occurred independently of the SPB remain an open question. Yet, whether they are SPB associated or not, Q-MT bundles have an apparent constant length and contain MTs that are insensitive to treatments that destabilize dynamic MTs (Figs. 2, S1 [E and F], and S2 [A and B]).

The molecular events leading to MT stabilization within the Q-MT bundle are still unknown. We can speculate that an enhanced MT cross-linking activity specifically associated with the Q-MT bundle may prevent its depolymerization. Upon quiescence entry, the SPB-associated bundle may recruit proteins that will either modify the bundling activity of known MT cross-linkers, such as Ase1, or load quiescence-specific MT bundling proteins that remain to be identified.

However, an increased MT bundling activity does not explain why MTs apparently do not elongate. We cannot rule out that residual polymerization/depolymerization cycles occur at the MT plus ends. In this scenario, MT elongation would be undetectable using classic fluorescence microscopy because the equilibrium between growth and shrinkage would be displaced toward depolymerization, for example, through the activation of a MT depolymerase. An alternative possibility could be that upon quiescence entry, a MT plus end “capper” would be recruited specifically onto the SPB-associated Q-MT bundle. Although no MT-capping activity has been identified to date, we can envision that a MT plus end tracker, such as a TOG domain–containing protein, for example, could lock MT plus ends, thereby preventing tubulin addition. Finally, the absence of detectable MT elongation in quiescent cells could result from the lack of free polymerizable tubulin. However, we have clearly shown that a free tubulin pool is present in the quiescent cell cytoplasm. Indeed, upon cell refeeding, the Q-MT bundle elongates, even if the de novo protein synthesis is inhibited (Fig. 5, A and C). This elongation occurs without any detectable Q-MT bundle depolymerization (Figs. 5 C and S4, H and I). Could the quiescent cell–free tubulin pool be unable to polymerize? We have shown that quiescence entry is associated with a drastic decrease of the GTP intracellular pool. Yet, an artificial GTP concentration drop does not influence MT dynamic (Fig. 3,A–C), at least within early time points. Further, in vitro experiments have demonstrated that GDP tubulin can polymerize given that few GTP tubulin dimers are present (Valiron et al., 2010). As GTP seems not to be involved, we can speculate that in quiescent cells, free tubulin cannot add on MT ends either because of a posttranslational modification or a tubulin dimer sequestering protein.

Whatever the mechanism responsible for the absence of MT dynamics in quiescence, Q-MT bundle elongation is rapidly triggered by glucose upon quiescence exit. This swift event, which does not depend on de novo protein synthesis, probably relies on signaling cascades involving nutrient catabolism (Laporte et al., 2011). Interestingly, we have shown that Q-MT bundle elongation requires Alp14 (Fig. 7, C and D), a protein displaying MT polymerase activity (Al-Bassam et al., 2012). In interphase, Alp14 is not strictly required for MT elongation, probably because its absence is compensated by a protein with a similar activity. In quiescence, this compensating activity may be inhibited or absent. Consequently, Alp14 would be needed for feeding the elongating Q-MT bundle plus ends upon quiescence exit. This is in agreement with the observation that Alp14 stays poised onto the Q-MT bundle plus ends in quiescence and follows MT elongation upon quiescence exit (Fig. 5 B).

Importantly, we have demonstrated that Q-MT bundle elongation and the recovery of MT dynamics is necessary for an efficient quiescence exit (Figs. 7 and 8 B). Indeed, cells in which the Q-MT bundle elongation is inhibited either pharmacologically or by the deletion of Alp14 exhibit profound cell shape defects. In these cells, polarity markers relocalize to unspecific cortical regions, leading to a random actin repolarization and ultimately to a branched or triangular cell shape (Fig. 8 B). This abnormal cell repolarization upon quiescence exit is a functional proof of the cell polarity loss in quiescence. Further, it explains the long standing observation that mutants impaired for MT dynamic or polarity factors, such as tea1Δ mutant, exhibit more T-shaped cells after a starvation period than during proliferation (Browning et al., 2000; Snaith and Sawin, 2003; Sawin and Snaith, 2004; Fig. 7).

We have previously shown that when budding yeast cells enter quiescence after glucose exhaustion, they assemble a stable MT bundle emanating from the SPB (Laporte et al., 2013). Yet, in contrast to the fission yeast cytoplasmic Q-MT bundle, the S. cerevisiae stable MT array is nuclear and composed of parallel MTs (Laporte et al., 2013; Laporte and Sagot, 2014). We do not know what the rationales for these structural differences are, and a wide range of speculations based on various physiological differences between these two species can be drawn. It is intriguing that in both yeasts SPB-associated MTs are stabilized. Could the SPB be a platform that integrates nutritional signals and translates them into MT cytoskeleton reorganizations?

All quiescent cells need to preserve their ability to proliferate and must produce a healthy progeny (Coller, 2011; O’Farrell, 2011). For microorganisms competing for an environmental niche, quiescence exit must be efficient to guarantee the prevalence of the species. Upon quiescence entry, we have shown that S. pombe and S. cerevisiae cells reorganize both their MT and their actin cytoskeletons to form stable structures (Sagot et al., 2006; Laporte et al., 2013; this study). Whether these reorganizations are just a passive consequence or actively participate in the process of quiescence establishment is a critical question (Daignan-Fornier and Sagot, 2011a,b). Actin and tubulin are very abundant proteins required for cell growth and division. The degradation or the damaging of these proteins in quiescence would require their resynthesis to allow proliferation resumption. This step would need a large amount of amino acids and energy and would require some time that may negatively impinge on the swift reentry into the proliferation cycle. Protein damage might to be a major cause of cell death during a long-term nonproliferative state. Might actin and tubulin be better protected from external damages when packed up into stable polymers rather than as free monomers in the cytoplasm? To definitively address the requirement for actin bodies and stable MT-containing bundles for yeast cell survival in quiescence, conditional mutants or experimental procedures to transiently disassemble these structures in quiescence need to be found.

Nevertheless, here, we have uncovered a physiological function for a quiescent cell–specific structure, as we have shown that fission yeast Q-MT bundle elongation is mandatory for an efficient quiescence exit. As such, this structure clearly participates in the fitness of S. pombe, a single-celled eukaryote in competition with numerous other species. Moreover, our study sheds light on the conservation of the reorganization and stabilization of both the actin and the MT cytoskeletons upon quiescence establishment in two phylogenetically distant unicellular eukaryote species. One obvious direction for future studies is to address this conservation in multicellular organisms.

Table S1 lists the S. pombe strains used in this study. Strains were provided by F. Chang (Columbia University College of Physicians and Surgeons, New York, NY), J.P. Javerzat (Centre National de la Recherche Scientifique, Bordeaux, France), P. Nurse (The Francis Crick Institute, London, UK), A. Paoletti (Institut Curie, Paris, France), P. Perez (Universidad de Salamanca, Salamanca, Spain), K. Sawin (University of Edinburgh, Edinburgh, UK), V. Sirotkin (State University of New York Upstate Medical University, Syracuse, NY), T. Toda (London Research Institute, London, UK), P. Tran (Institut Curie, Paris, France), and J.Q Wu (The Ohio State University, Columbus, OH). All tagged genes are under the control of their endogenous promoters. For all experiments, yeast cells were grown in liquid YPDA medium at 30°C as described previously (Laporte et al., 2008) except for Fig. S1 A and Fig. 3 B for which strains were grown in YE5S+A or in synthetic dextrose medium (2% [wt/vol] glucose, 0.17% [wt/vol]) nitrogen base, and 0.5% [wt/vol] ammonium sulfate] with 0.2% (wt/vol) casamino acids (Difco Laboratories) supplemented with tryptophan and uracil (SD casa WU), respectively. For nitrogen starvation (Fig. S2 C), proliferating cells in EMM2 were washed five times with water, and then transferred in EMM2-N medium at the same cell concentration. For cold treatment, cells were shifted at 4°C for the indicated time, and then shifted back to 30°C and imaged. For drug treatments, cells were incubated with 0.5 mg/ml MBC (Sigma-Aldrich), 0.2 mg/ml benomyl (Sigma-Aldrich), 1 mg/ml TBZ (Sigma-Aldrich), or 180 µM CHX (Sigma-Aldrich) for the indicated time. These concentrations are 10 times higher than the one commonly used with proliferating cells. For quiescence exit in the presence of drugs, cells were preincubated 1 h in the presence of the drug before quiescence exit. For quiescence exit in the sole presence of glucose, cells were washed twice with water, inoculated at OD_600nm = 2 in a solution containing 2% glucose, and incubated at 30°C. Cell viability was addressed after cell incubation for 5 min in a solution containing 0.2% of methylene blue (Sigma-Aldrich) and 2% sodium citrate (Sigma-Aldrich), pH 7. In this study, all experiments were at least duplicated and for each time point, unless specified, >200 cells were scored.

High pressure ionic chromatography experiments were done as described previously (Hürlimann et al., 2011). In brief, cellular extracts were prepared by an ethanol extraction. Cells were harvested by rapid filtration, immediately dropped into 5 ml ethanol/10 mM Hepes, pH 7.2 (4/1 vol/vol), and incubated at 80°C for 3 min. Samples were evaporated using a rotavapor apparatus. Residues were resuspended in water and insoluble particles were eliminated by centrifugation. Separation of metabolites was performed on a CarboPac PA1 column, linked to a chromatography workstation (ICS3000; Dionex), using a sodium hydroxide/sodium acetate (50–800 mM) gradient. Nucleotide derivatives were detected with a UV diode array detector (Ultimate 3000RS; Dionex). MPA (Sigma-Aldrich) was used at the final concentration of 100 µg/ml. Guanine was used at the final concentration of 0.3 mM in 0.1 M sodium hydroxide.

For immunofluorescence experiments, cells were harvested by rapid filtration and fixed by washing the filter with methanol at −20°C. Cells were further processed for immunofluorescence as described previously (Mata and Nurse, 1997). Anti-tubulin staining used the TAT1 monoclonal tissue culture supernatant (gift of K. Gull, Sir William Dunn School of Pathology, Oxford, UK; Woods et al., 1989) at 1:15 followed by Alexa Fluor 488–goat anti–mouse secondary antibody (Molecular Probes) at 1:500.

For actin phalloidin staining, cells were fixed 1 h with freshly made paraformaldehyde (3.8% final [Sigma-Aldrich] in PEM [0.1 M Pipes, 1 mM EGTA, and 1 mM MgSO₄, pH 6.9, with 5 N sodium hydroxide]), vortexed every 10 min for 1 h at 70°C, and then centrifuged 5 min at 3,000 rpm. Cells were then washed twice with PEM and resuspended in 1.5 ml PEM + 1% Triton X-100. Proliferating and quiescent cells were incubated for 30 s and 3 min, respectively. Samples were then washed twice in 1.5 ml PEM and resuspended for 24 h at 4°C in PEM containing 1/10 V of Alexa Fluor Phalloidin (Invitrogen). Finally, cells were washed twice, resuspended in a mounting solution containing 70% glycerol and 5 mg/l paraphenylenediamine, and imaged at room temperature as described previously (Sagot et al., 2006).

For MT behavior after cold experiment (Fig. S2 A), samples were fixed 1 min with formaldehyde (3.8% final; Sigma-Aldrich), washed, resuspended in a mounting solution as above, and imaged.

Cells were observed in a fully automated inverted microscope (Axiovert 200M; Carl Zeiss) equipped with a stage (MS-2000; Applied Scientific Instrumentation), a 300-W xenon light source (Lambda LS; Sutter Instrument), a 100× 1.4 NA Plan-Apochromat objective, and a five-position filter turret. For GFP imaging, we used a FITC filter (excitation, HQ487/25; emission, HQ535/40; beam splitter, Q505lp). For RFP imaging, we used a Cy3 filter (excitation, HQ535/50; emission, HQ610/75; beam splitter, Q565lp). For CFP imaging, we used a CFP filter (excitation, HQ436/20; emission, HQ480/40; beam splitter, 455dclp). For YFP imaging, we used a YFP filter (excitation, HQ500/20; emission, HQ535/30; beam splitter, Q515lp). All the filters were purchased from Chroma Technology Corp. Images were acquired using a CoolSnap HQ camera (Roper Scientific). The microscope, camera, and shutters (Uniblitz) were controlled by SlideBook software 5.0 (Intelligent Imaging Innovations). Images are, unless specified, 2D maximal projection of Z-stacks performed using a 0.3-µm step. For live cell imaging, 2 µl of the cell culture were spotted onto a glass slide and immediately imaged at room temperature.

For fluorescence intensity measurement (Figs. 1 B, 4 F, S1 F, and S4 I), a line scan (i1) of n pixel width (three to eight, depending on the binning acquisitions parameters) containing both GFP signal and background was drawn along MTs using ImageJ software. A line of 2n pixel width at the same location was drawn to calculate the intensity of the surrounding background (i2). The real intensity (ir) was calculated as follow: ib = (i2 × Area i2) − (i1 × Area i1) and ir = i1 − [ib/(Area i2 − Area i)]. To calculate the number of MTs per structure, individual MT fluorescence was determined by three methods: the fluorescence intensity was measured in proliferation cell on either one MT growing or on the nonoverlapping or overlapping region of the spindle, described to contain 11 MTs (Ding et al., 1993). According to these measurements for a single MT, fluorescence intensity measured upon quiescence entry was converted in MT numbers.

To measure MT length (Figs. 4 E and S1 E), positions of the MT-containing structure extremities were followed over time and compared with a fixed point in the cell using MTrackJ plugin (ImageJ). Euclidian distances (D) between two positions over time were calculated as D = √[(x2 − x1)² + (y2 − y1)²]. In time-lapse series (Fig. 2 A), the first measured length was set to zero. The same protocol was applied to measure MT dynamics (Fig. 3 C) and SPB movement in quiescence and upon exit (Fig. S4 E). To measure MT dynamics, four z-frames were acquired every 4 s on proliferating cells treated or not with MPA (2 h). After using a rolling average filter (i.e, the MT extremity fix-point length [w] is the mean of w − 1, w, and w + 1 values), the length was plotted over time and growth/shrinkage speed was calculated over a 300-s observation period.

FRAP experiment (Fig. 3 C) was done as previously described (Laporte et al., 2013). Experiments were done on an inverted microscope (DMI 6000; Leica) equipped with a spinning disk confocal head (CSU-X1; Yokogawa Electric Corporation), a QuantEM camera (Photometrics), and a scanner FRAP system (Roper Scientific). The diode lasers used were at 408 and 491 nm. The objective used was an HCX PL APO CS 100× oil with 1.4 NA. The Z-stacks were done with an objective scanner (Piezo P721.LLQ; Physik Instrumente). This system was controlled by MetaMorph software (Molecular Devices). Recovery from photobleaching was monitored in time-lapse mode. Fluorescence recovery was corrected for background noise and continuous photobleaching using the ImageJ software. Fluorescence was then normalized using a rolling average filter (i.e, the fluorescence intensity y is the mean intensity of y − 1, y, and y + 1). The fluorescence was then normalized to the one measured before bleach (Vavylonis et al., 2008).

For line scan analysis (Fig. 4, B and C; and Fig. S3 A), the real measured intensities (ir) were set to percent, with 100% being the maximum calculated along the line. Slopes were manually aligned to zero, with zero being when ir > 0. The same protocol was applied to compare the different slopes after line scan analysis (Fig. S1, B and C).

Mander’s colocalization coefficients (Fig. 4 A) were calculated using ImageJ software as described in Jimenez et al. (2014). In brief, Z-stacks were first maximum projected and cells were individually selected in both GFP and RFP channels. The same threshold was applied for each channel and signals were binarized (i.e., pixel < threshold = 0 and pixel > threshold = 255). The percentages of GFP pixel (Mto2-GFP signal) that colocalize with RFP pixel (RFP-Atb2 signal) were determined using the JACoP plugin (Bolte and Cordelières, 2006).

For Q-MT bundle fluorescence measurements upon quiescence exit (Fig. S4 H), a line scan of 8-pixel width was drawn along Q-MT bundle using ImageJ software. The fluorescence intensities were measured in function of time and were corrected for photobleaching. Photobleaching was measured on Q-MT bundles in quiescence images with the same acquisitions conditions.

To calculate the theoretical stabilization of MT upon quiescence entry, we started with the mean number (N) of MT bundles observed for cells before treatment (i.e., for 6 h after carbon exhaustion [Fig. 3 E] the mean number of MT bundle per cell N was 2.7 and for 42 h [Fig. S2 D] N was 3.3). We noted n_i-single, the initial percentage of cells with only a single MT bundle. We noted n_multi-before and n_multi-after, the percentage of cells with more than one MT bundle observed before and after benomyl treatment, respectively. For each time point, Δ_n-multi is equal to n_multi-before − n_multi-after. For the random stabilization model, we considered that 1/N of the MT bundle is stabilized, regardless of the association with the SPB. Thus, for each time point, the prediction of the percentage of cells that should display a single MT bundle associated for the SPB is (1/N × Δ_n-multi) + n_i-single. For the nonrandom stabilization, we considered that only the MT bundle associated with the SPB is stabilized. Thus, for each time point, the prediction of the percentage of cells that should display a single MT bundle associated for the SPB is (Δ_n-multi) + n_i-single.

Freezing, freeze substitution (Fig. 1, D and E), and immunostaining (Fig. S1 G) were performed as described previously (Laporte et al., 2013). In brief, cells were washed with glycine, and then with fetal calf serum, and incubated with polyclonal rabbit anti-Tat1 (the Trypanosoma brucei α-tubulin) antibodies diluted 1:250 (a gift of K. Gull; Woods et al., 1989). After a wash with Tris-buffered saline containing 0.1% bovine serum albumin, cells were incubated for 45 min at room temperature with anti–mouse IgG conjugated to 10-nm gold particles (BioCell Laboratories, Inc.). For freeze substitution, yeast were deposited on a copper grid (400 mesh) coated with Formvar. Grids were immersed in liquid propane held at −180°C by liquid nitrogen and then transferred in a 4% osmium tetroxide solution in dry acetone at −82°C for 72 h. Grids were then shifted to room temperature and washed three times with dry acetone. Cell were stained with 1% uranyl acetate and then washed once with dry acetone. Samples were gradually unfiltered with araldite (Fluka). Ultra-thin sections were contrasted with lead citrate and observed in an electron microscope (80 kV; 7650; Hitachi) at the EM facility of the Bordeaux Imaging Center.

For tomography analysis (Fig. 1 G), 150-nm-thick sections were placed on formvar-coated 150-mesh copper grids and poststained 1 min with lead citrate 2% in water. Images were acquired using a Spirit Twin (FEI Tecnai) operated at 120 kV. Images were recorded with a 4k Eagle charge coupled device camera system (FEI Tecnai) at a nominal magnification of 30,000 (sample D7) or 26,000 (sample C10) using a –65 to 65 tilt for Cell 1 or a −65 to 39 for Cell 2 with 2° increments. Tomography tilt series were taken using the batch tomography software (FEI Tecnai). They were then aligned, cropped, and binned using the ETomo program from the IMOD suite. The 3D tomography reconstructed visualization was done using Amira Resolve RT 5.2.0.

For MT distance measurements between two MTs (Fig. 1 F), the centroid of each MT was determined and the Euclidian distance was determined as D = {√[(x2 − x1)² + (y2 − y1)²]}. This distance was then subtracted for the two MTs’ radiuses (∼22 nm). Gaussian fit on these values was then determined using GraphPad Prism (GraphPad Software).

Glucose concentration was measured using the d-Glucose/d-Fructose UV test kit (Roche). Ade13 polyclonal antibodies were raised in rabbit using the purified full-length S. cerevisiae protein (Takara Bio Inc.).

Fig. S1 describes the shape of the Q-MT bundle and its detection by EM using immunogold labeling with anti-tubulin antibodies. Fig. S2 analyzes the Q-MT bundle resistance to a cold treatment and its stabilization with time in quiescence after carbon or nitrogen exhaustion. Fig. S3 provides the localization of Mto1-GFP onto the Q-MT bundle. It shows Western blot quantifying the tubulin steady-state amounts, the Q-MT bundle resistance to benomyl, and the viability of MAP mutants in quiescence. Fig. S4 shows MT elongation at Q-MT bundle extremities and Q-MT bundle stability upon quiescence exit. It also shows Q-MT bundle anchoring at the SPB and Alp14-GFP localization during the quiescence exit process. Fig. S5 displays polarity marker localization upon quiescence entry and exit in WT and mutant backgrounds or in the presence of drugs. Table S1 lists the S. pombe strains used in this study.

We express our profound gratitude to J.P. Javerzat for precious comments and great discussions about our work and his help in writing this manuscript. We are grateful to M. Gupta, A. Paoletti, D. Pellman, and P. Tran for helpful discussions. We are thankful to F. Chang, J.P. Javerzat, P. Nurse, A. Paoletti, P. Perez, K. Sawin, V. Sirotkin, T. Toda, P. Tran, and J.-Q. Wu for providing yeast strains. We thank A. Lebaudy and M. Revilla Guarinos for their help in preliminary experiments. FRAP was done at the Bordeaux Imaging Center facility with the help of Christel Poujol.

This work was supported by the Université Bordeaux and a Young Investigator grant from the Agence Nationale pour la Recherche (JC08 310804 to I. Sagot).

The authors declare no competing financial interests.