Introduction

Microtubules (MT) are hollow cylindrical polymers assembled by the non-covalent interaction of α- and β-tubulin heterodimers. MTs are generally nucleated at MT organizing centers (MTOC) by a ɣ-tubulin complex (ɣ-TuC) that acts as a template and stabilizes the so-called MT “minus-end” (Liu et al., 2021; Roostalu and Surrey, 2017; Sanchez and Feldman, 2017; Thawani and Petry, 2021). At the opposite end, the “plus-end” (+end), MTs elongate by the addition of GTP tubulin. The β-tubulin bound GTP is hydrolyzed and stable but transient GDP + Pi intermediates are generated in the region immediately trailing the growing +end. A subsequent Pi release favors MT depolymerization that can be rescued by de novo GTP-tubulin addition (Cleary and Hancock, 2021; Gudimchuk and McIntosh, 2021). As such, MTs alternate periods of growth and shrinkage, a behavior called dynamic instability (Mitchison and Kirschner, 1984). In vivo, a profusion of MT associated proteins (MAPs) regulate MT length and dynamics (Bodakuntla et al., 2019; Goodson and Jonasson, 2018). Besides, MTs are often assembled from multiple tubulin variants or isotypes, and modified by a cohort of post-translational modifications that modulate MT dynamics directly or by influencing the recruitment of MAPs (Janke and Magiera, 2020; Roll-Mecak, 2020). Other MAPs organize MTs into high-ordered edifices either by cross-linking MTs or by connecting them with cellular structures such as membranes, chromosomes or components of the actin cytoskeleton (Bodakuntla et al., 2019; Meiring et al., 2020). Cells fine-tune these mechanisms both in space and in time to give rise to distinct MT architectures with specific function.

In proliferating cells, MT dynamic is crucial for their functions. It allows exploring the cell volume searching for structures to “capture”, like centromeres during mitosis, or to exert pushing or pulling forces required for various cellular processes such as cell migration or morphogenesis (Heald and Khodjakov, 2015; Kirschner and Mitchison, 1986). Yet, the MT cytoskeleton can extensively rearrange and assemble more stable MT structures, notably when cells change fate (Meiring et al., 2020; Röper, 2020). For example, in terminally-differentiated cells, such as epithelial cells, cardiomyocytes or neurons, stable MT networks make up the majority, and ensure critical cellular functions such as cell shape maintenance or long-distance intra-cellular transport (Baas et al., 2016; Muroyama and Lechler, 2017). Defects in MT stabilization are at the origin of many human pathologies, including neuro-degenerative diseases and ciliopathies (Anvarian et al., 2019; Wheway et al., 2018). Effectors responsible for MT stabilization have been actively searched for since the 1980’s. In mammals, the involvement of specific MAPs such as Tau, MAP2, MAP6 and PRC1, or tubulin post-translational modifications have been extensively studied. While their contributions to MT stabilization are central and undisputable, their sole actions do not fully explain the observed levels of MT stability in many types of non-dividing cells (Hahn et al., 2019).

For years, yeast species have been instrumental in deciphering mechanisms that regulate MT dynamics in eukaryotes. While proliferating yeast cells display a dynamic MT network (Winey and Bloom, 2012), proliferation cessation goes with the formation of dramatically stable MT structures in both S. cerevisiae and S. pombe (Laporte and Sagot, 2014; Laporte et al., 2013, 2015). Upon quiescence establishment, S. cerevisiae assemble a bundle composed of stable parallel MTs, hereafter called Q-nMT bundle, for Quiescent-cell nuclear Microtubule bundle, that emanates from the nuclear side of the spindle pole body (SPB), the yeast equivalent of the centrosome. The Q-nMT bundle spans the entire nucleus and relocalizes kinetochores and centromeres that remain attached to its MT +ends (Laporte and Sagot, 2014). Since MTs embedded into the Q-nMT bundle are not all of the same length, they confer to the structure a typical arrow shape. When cells exit quiescence, the Q-nMT bundle depolymerizes and, by pulling the attached centromeres back to the SPB, allows the recovery of the typical Rabl-like configuration of chromosomes found in G1 yeast cells (Jin et al., 1998). The molecular mechanisms underlying the formation of this peculiar stable MT structure and its physiological function(s) are not understood. Yet, cells impaired for Q-nMT bundle formation have a compromised survival rate in quiescence and a drastically reduced fitness upon cell cycle re-entry (Laporte and Sagot, 2014; Laporte et al., 2013, 2015).

Here, we demonstrate that the Q-nMT bundle formation is a multistep process that follows a precise temporal sequence. The first step requires the kinesin-14 Kar3, its regulator Cik1 and the EB1 homolog Bim1, and leads to the formation of a short (≈ 1 µm) and stable bundle that resembles a half mitotic spindle. Importantly, in this first step, MT polymerization and stabilization are coupled. In a second step, additional MTs polymerize from the SPB, elongate, and are zipped to pre-existing ones in a Cin8/kinesin-5 dependant manner. The full Q-nMT bundle stabilization is established in a third and last step that requires the kinesin-5 Kip1. Our observations further indicate that Q-nMT bundle assembly requires not only MT-kinetochore attachment but also inter-kinetochore interactions. Finally, we show that, upon quiescence exit, the Q-nMT bundle disassembles from its +ends, where each MT depolymerizes in coordination with its neighbors, via the action of the depolymerase Kip3, a member of the kinesin-8 family. A complete Q-nMT bundle disassembly is required to authorize cells re-entry into the proliferation cycle. Overall, this study describes for the first time the entire life cycle of a MT structure from the molecular mechanisms involved in its formation and stabilization to its disassembly and further reveals that this atypical quiescent specific structure acts as a sort of “checkpoint” for cell cycle resumption upon quiescence exit.

Results

The Q-nMT bundle formation is a three-step process

To investigate whether the Q-nMT bundle was stable by default, i.e. intrinsically stable, or assembled as a dynamic structure and then stabilized, we quantified Q-nMT bundle length (Fig. 1A) and thickness (Fig. 1B) upon quiescence establishment following carbon source exhaustion. We found that the Q-nMT bundle assembly was a 3-step process. In an initial phase (phase I), MTs elongated from the SPB to reach ≈0.8 µm. The number of these so-called phase I-MTs was about the same as in a mitotic spindle (see inset of Fig. 1B and Sup. Fig. 1A). Importantly, phase I-MTs are resistant to nocodazole (Noc), a MT poison that causes dynamic MT depolymerization, indicating that their stabilization was concomitant with their polymerization (Fig. 1A-B). In a second phase, that started ≈10h after glucose exhaustion, additional MTs emerged from the SPB (phase II-MTs) and elongated along phase-I MTs, approximatively doubling the thickness of the phase I MT bundle (Fig. 1B and Sup Fig. 1A). At this stage, the tip of the newly elongated MTs were instable (Fig. 1A-B). Full Q-nMT bundle stabilization was reached ≈48h after glucose exhaustion (phase III), as Noc treatment affected neither Q-nMT bundle length nor its thickness (Fig. 1A-B and Sup. Fig. 1B). Another way to shed light on the Q-nMT bundle formation steps was to plot Q-nMT bundle width as a function of length for each individual cells before and after Noc treatment (Fig. 1C).

Figure 1:

To confirm the above multistep process, we first followed Nuf2-GFP, a protein that localize at the MT +end. In phase I, Nuf2-GFP signal relocates from SPB to the distal part of elongating Q-nMT bundles (Sup. Fig. 1D). In phase II, when cells were treated with Noc, the Nuf2-GFP signal followed depolymerizing MTs (Fig. 1D), testifying for the instability of phase II MTs. By contrast, the Nuf2-GFP signal remained immobile in phase III, when cells have fully stabilized Q-nMT bundles (Fig. 1D). Second, we measured the γ-tubulin (Tub4) signal at the SPB and found that Tub4 started accumulating at the SPB at the onset of phase I to reach a plateau at the end of phase II (Fig. 1E and Sup. Fig. 1E). Interestingly, the amount of Tub4 at the SPB was proportional to the thickness of the Q-nMT bundle (Sup. Fig. 1F). Finally, in order to confirm the two waves of MT elongation, we developed a strain expressing mTQZ-Tub1 from the endogenous TUB1 promoter and mRuby-TUB1 under the ADH2 promoter, a promoter that is active only after glucose exhaustion. We observed that mRuby-TUB1 was incorporated into the Q-nMT bundle only after phase I (Fig. 1F and Sup. Fig. 1G) confirming the existence of a second wave of MT elongation in phase II. All the above information lead us to propose the model presented in Fig. 1G (see figure legend).

Q-nMT bundle formation is an active process regulated in time

Increasing viscosity of the cytoplasm was shown to dampen MT dynamics (Molines et al., 2022). Upon quiescence establishment, yeast cells experience a fluid phase to solid-like transition (Munder et al., 2016; Joyner et al., 2016) and an acidification (Jacquel et al., 2021) that could contribute to Q-nMT bundle formation. Several evidences indicate that these changes in physico-chemical properties are not involved in Q-nMT formation. First, 4-day-old quiescent cells can simultaneously display both dynamic cytoplasmic MTs (cMT) and a stable Q-nMT bundle (Sup. Fig. 1H and (Laporte et al., 2013)). Second, when we mimicked a fluid phase to solid like transition in proliferating cells by artificially dropping the pH, no stable MTs were observed, while F-actin aggregation was induced (Sup. Fig. 1I) as expected from (Peters et al., 2013). Third, 2-day-old quiescent cells displaying a phase III Q-nMT bundle were able to grow de novo dynamic MTs after a pulsed-chase Noc treatment (Sup. Fig. 1J).

If Q-nMT bundle formation was solely depending on physico-chemical changes, this structure should be able to assemble in late quiescence. By treating cells with Noc upon glucose exhaustion, we managed to prevent Q-nMT bundle formation in early quiescence (Fig. 1H). Strikingly, when the drug was washed out 24 h later, no Q-nMT bundle assembled, even several days after Noc removal (Fig. 1H). We verified that these cells did enter quiescence as they assembled actin bodies (Fig. 1H, right panel), another quiescent cell specific structure (Sagot et al., 2006). This experiment strongly suggests that Q-nMT formation is an active process induced by a transient signal emitted upon glucose exhaustion.

Mutants that cannot assemble Q-nMT bundle have a reduced viability in quiescence and a decreased capacity to form colonies upon quiescence exit (Laporte et al., 2013). To establish a causal link between the absence of Q-nMT bundle and the above phenotypes, we took advantage of our ability to prevent conditionally Q-nMT bundle formation in WT cells using early Noc treatment. We demonstrated that in the absence of Q-nMT bundle, both prototroph and auxotroph WT cells loose viability in quiescence and that survivors had a reduced capacity to generate a progeny upon quiescence exit (Sup. Fig. 1K). This reinforce the idea that the Q-nMT bundle is required for cell survival upon chronological aging.

Tubulin steady state level and Tub3 isoform are critical for Q-nMT bundle formation

We then focused on deciphering the molecular mechanisms involved in Q-nMT bundle formation. Yeast cells express two α-tubulin isotypes, Tub1 and Tub3. In proliferating cells, Tub1 make up the majority (Aiken et al., 2019; Nsamba et al., 2021; Schatz et al., 1986). In 4-day-old cells, both isoforms were embedded into the Q-nMT bundle (Fig. 2A and Sup. Fig. 2A). As Tub3 stabilizes MTs in vitro (Bode et al., 2003), it may be key for Q-nMT bundle formation in quiescence. To test this hypothesis, we first analyzed the phenotype of tub3Δ cells. In phase II, tub3Δ cells displayed short MT bundles that were highly sensitive to Noc. In phase III, no MT bundle could be detected (Fig. 2B). Thus in this mutant, MT bundles assembled during phase I were not stable and ultimately collapsed. Yet, in tub3Δ cells, the steady state amount of α-tubulin is significantly decreased (Sup. Fig. 2B and (Nsamba et al., 2021)). Thus, the amount of α-tubulin, but not Tub3 itself, could be important for Q-nMT bundle formation. Indeed, mutants impaired for α- and β-tubulin folding, such as the prefolding complex mutants pac10Δ, gim3Δ or yke2Δ, or for tubulin heterodimer formation such as pac2Δ, or for β-tubulin folding, such as cin2Δ or cin4Δ, all of which displaying a decreased amount of tubulin, were unable to assemble a Q-nMT bundle (Sup. Fig. 2C-D).

Figure 2:

To identify the role of Tub3 per se, we took advantage of the “Tub1-only” mutant in which the TUB3 gene was replaced by the TUB1 gene at the TUB3 locus. This mutant only expresses Tub1 and displays an α-tubulin level similar to WT (Sup. Fig. 2B and (Nsamba et al., 2021)). We found that “Tub1-only” cells display shorter but stable Q-nMT bundles in phase III (Fig. 2C). Interestingly, “Tub1-only” Q-nMT bundles were thinner than WT Q-nMT bundles (Fig. 2D). Overall, these data indicate Tub3 is involved in MT elongation in phase II.

Kinetochores are critical for phase I but dispensable for the maintenance of Q-nMT bundles

Kinetochore components are enriched at the Q-nMT bundle +end ((Laporte et al., 2013) and Sup Fig. 3A). We thus wondered if kinetochore-MT attachment could play a role in Q-nMT bundle formation. First, we disrupted kinetochore-MT attachment upon phase I using the well-established ndc80-1 allele (Cheeseman et al., 2006; DeLuca et al., 2018; Wigge et al., 1998). As shown in Fig. 3A, no Q-nMT bundles were detected in ndc80-1 cells shifted to 37°C at the onset of quiescence entry. Second, we focused on the chromosomal passenger complex (CPC: Bir1/Survivin, Sli15/INCENP, Nbl1/Borealin and Ipl1/Aurora B), a complex known to regulate kinetochore-MT attachment dynamics (Cairo and Lacefield, 2020). We found that upon quiescence entry, Ipl1 inactivation impeded MT bundle formation (Fig. 3B). While we could not detect Ipl1 and Ndc1 in quiescent cells, we found that Sli15-GFP and Bir1-GFP localized along the Q-nMT bundle with an enrichment at the Q-nMT bundle +end (Fig. 3C). The deletion of Sli15 did not affect Q-nMT bundle (Fig. Sup. 3E). However, in 4-days old bir1Δ cells, Q-nMT bundle length were slightly shorter and not fully stabilized (Fig. 3D). Finally, we investigated the involvement of Bim1, a MT +end binding protein that plays a role in kinetochore MT-end on attachment (Dudziak et al., 2021; Thomas et al., 2016). We have previously shown that Bim1 localizes all along the Q-nMT bundle but that its deletion has no impact on Q-nMT bundle formation even in the presence of Noc ((Laporte et al., 2013) and Fig. 3D). Interestingly, the amount of Bim1 was correlated with tubulin incorporation during Q-nMT bundle formation (Sup Fig. 3D). When we combined bim1Δ with bir1Δ, cells hardly assembled MT structures that vanished after Noc treatment (Fig. 3D). This demonstrates that in the absence of Bim1, Bir1 was critical for Q-nMT bundle stabilization in phase I. A similar phenotype was observed with SLI15 deletion (Sup Fig. 3E). Interestingly, the viability of bim1Δ bir1Δ cells in quiescence was drastically reduced (Sup Fig. 3G). Together, these experiments show that kinetochore-MT attachments are critical for the initiation of Q-nMT bundle assembly.

Figure 3:

We then hypothesized that inter-kinetochore interactions may stabilize the parallel MTs embedded into the Q-nMT bundle. To test this idea, we analyzed MT organization in slk19Δ cells that are defective in kinetochore clustering (Richmond et al., 2013). In quiescence, Slk19 was enriched at both Q-nMT bundle extremities (Fig. 3E). slk19Δ cells displayed shorter and thinner MT structures that were not stable, i.e. sensitive to Noc treatmant (Fig. 3F). Since Bim1 has been involved in kinetochore-kinetochore interaction, we combined slk19Δ with bim1Δ. MT structures detected in slk19Δ bim1Δ cells hardly reached ≈1 µm and vanished upon Noc treatment. In addition, kinetochores localized as a rosette around the SPB (Fig. 3F) and cell viability was strongly compromised (Sup. Fig. 3G). This demonstrates that in the absence of Bim1, Slk19 is strictly essential for MT bundling and stabilization in phase I. Furthermore, since kinetochores are known to be cross-linked by the monopolin complex (Mam1, Lrs4, Hrr25 and Csm1 (Corbett et al., 2010)) we analyzed the phenotype of mam1Δ, lrs4Δ and csm1Δ cells in quiescence. Most of them displayed short MT structures often arranged in a star-like array (Fig. 3G-H, Sup. Fig. 3F). As for Bir1 and Skl19, the combination of bim1 deletion with monopolin disruption worsens MT structure length and stability (Fig. 3H) and the cell viability was strongly compromised (Sup. Fig. 3G). Cells impaired for Spo13, a protein involved in monopolin recruitment to kinetochores (Lee et al., 2004) displayed a similar phenotype (Fig. 3G-H). Altogether, these experiments show that inter-kinetochore interactions are critical for Q-nMT bundle assembly.

Lastly, we investigated if kinetochore-MT interactions were required for Q-nMT bundle stability once it is formed. In 5-days old WT cells, we used a UV-pulsed-laser to break the Q-nMT bundle into two pieces (Fig. 3I). The presence of dynamic cMTs after laser ablation testified for cell viability (Fig. 3I). In most cases, the length of both the released fragment and the fragment attached to the SPB remained constant (Fig. 3I, Sup. Fig. 3H). By contrast, in proliferating cells, dynamic anaphase spindles promptly disassembled following a pulse-laser break (Sup. Fig. 3I). This demonstrates that once Q-nMT bundles are formed, they do not need MT-kinetochore interaction to be maintained. Accordingly, Ndc80 or Ipl1 inactivation in cells that are already in quiescence had no effect on the Q-nMT bundle maintenance (Sup. Fig. 3B-C). Importantly, the above experiment indicates that once formed, Q-nMT stability is established and maintained throughout its length.

Specific kinesins are required for each step of Q-nMT bundle formation

To go further in the deciphering of the molecular mechanism involved in Q-nMT bundle formation, we focused on kinesins. Kar3 is a kinesin that, in complex with its regulator Cik1, can generate parallel MT bundles from a MT organizing center (MTOC) both in vitro and in proliferative cells (Manning et al., 1999; Mieck et al., 2015; Molodtsov et al., 2016). In quiescence, we found that Kar3-3GFP localized at the SPB, but also as dots along the Q-nMT bundle (Fig. 4A). In 4 day-old kar3Δ cells, the majority of the detected MT structures were extremely short compared to WT (Fig. 4B and Sup Fig. 4A). Same results were obtained in cik1Δ but not in vik1Δ cells, which lack the alternate Kar3 regulators (Fig. 4B). Thus, the Kar3/Cik1 complex is required for phase I.

Figure 4:

Cin8 is a Kinesin-5 that cross-links MTs (Bodrug et al., 2020; Pandey et al., 2021; Singh et al., 2018), and as such, could play a role in Q-nMT bundle stabilization. Cin8 was hardly detectable in quiescent cells (Sup. Fig. 4B) and cin8Δ cells assembled Q-nMT bundles that were thinner than in WT cells (Fig. 4C-E). Yet, these thinner bundles became stabilized in phase III (Fig. 4C and Sup. Fig. 4C-D). Overall, these results strongly suggest that Cin8 is required for phase II MT nucleation and/or elongation but not for stabilization in phase III.

Yeast have an additional Kinesin-5 called Kip1 (Fridman et al., 2013). In quiescence, most of the Kip1-GFP signal was observed at the Q-nMT bundle +end (Fig. 4F). In kip1Δ cells, phase I was slightly delayed and Q-nMT bundle were not stabilized in phase III (Fig. 4G-H), yet they were as thick as WT Q-nMT bundles (Sup. Fig. 4E). These data demonstrate that Kip1 is required for Q-nMT bundle stabilization in phase III. Of note, in phase I and II, Q-nMT bundles were slightly longer in cells deleted for the Kinesin-8 Kip3, a MT depolymerase ((Fukuda et al., 2014) Fig. 4 G-H and Sup. Fig. 4G) but were unaffected in cells deleted for Kip2, a kinesin that stabilizes cytoplasmic MTs ((Hibbel et al., 2015) Sup. Fig. 4F-G). Taken together our results demonstrate that each phase of Q-nMT bundle assembly involves specific kinesins.

Q-nMT bundle disassembly conditions SPB duplication/separation

Finally, we questioned the molecular mechanism of Q-nMT bundle disassembly upon quiescence exit. Cycloheximide prevented Q-nMT bundle disassembly indicating that this process requires de novo protein synthesis (Fig. 5A). We then measured Q-nMT bundle thickness upon depolymerization, and found that while Q-nMT bundles shortened, they did not get thinner (Fig. 5B). In agreement, Nuf2-GFP clusters found at the Q-nMT bundle +end moved back to the SPB while the Nuf2-GFP clusters localized in the middle of the structure remained immobile until they are reached by these later (Sup. Fig. 5 A). This shows that all the MT +ends did not start depolymerizing at the same time, and that longer MTs depolymerized first. This finding is in agreement with our model in which MTs are cross-linked all along the Q-nMT bundle length.

Figure 5:

Since Kip3 is the only well-characterized yeast MT depolymerase, we studied Q-nMT bundle disassembly in kip3Δ cells and observed that it was drastically delayed (Fig. 5C), yet its overall depolymerization rate remained unaffected (Sup. Fig. 5B). In fact, we found that in WT cells, Kip3 needs to be resynthesized upon quiescence exit (Fig. 5D). We searched for proteins that could be involved in Q-nMT bundle disassembly together with Kip3. Among proteins required for mitotic spindle disassembly, we found that inactivation of CDC14, CDC15, CDH1, DCC1 and TOR1 had no effect on Q-nMT bundle disassembly, just as with the inactivation of Ipl1 or proteins involved in kinetochore-MT attachment (Sup. Fig. 5 C to F). The sole additional actor we found was She1, a dynein regulator (Bergman et al., 2012), which deletion strongly aggravated the kip3Δ phenotype (Fig. 5C). Thus, Q-nMT bundle disassembly does not rely on the canonical mitotic spindle disassembly machinery but, rather, specifically required Kip3 and She1.

Importantly, when we followed quiescence exit at the individual cell level, Q-nMT bundle depolymerization always preceded SPB duplication/separation (Fig. 5A) and thus, in both WT and in kip3Δ cells (Fig. 5E, F). In fact, when we impeded Q-nMT bundle disassembly using kip3Δ she1Δ, we found that SPB duplication was severely delayed (Fig. 5G-H, Sup. Fig. 5G-H). Importantly, kip3Δ she1Δ cells proliferated just like WT cells (Sup. Fig. 5G) and upon quiescence exit, these mutant cells disassembled actin bodies just as in WT cells (Sup. Fig. 5I), demonstrating that kip3Δ she1Δ cells were not impaired in sensing quiescent exit signals. These observations indicate that Q-nMT bundle has to be disassembled to allow SPB duplication/separation upon quiescence exit and cell cycle re-entry.

Discussion

Our results shed light on the precise temporal sequences that generate a Q-nMT bundle upon quiescence establishment in yeast. From bacteria to human stem cells, modifications of the cell physico-chemical properties are known to accompany quiescence establishment. Among these modifications, a reduction of the cellular volume increases molecular crowding (Joyner et al., 2016) and pH acidification causes changes in the surface charges of macromolecules and viscosity increases (Charruyer and Ghadially, 2018; Jacquel et al., 2021; Munder et al., 2016; Persson et al., 2020). Much evidence points to these modifications as a trigger for the auto-assembly of several types of enzyme-containing granules (Munder et al., 2016; Petrovska et al., 2014; Rabouille and Alberti, 2017), or complex structures such as P-bodies and Proteasome Storage Granules (Peters et al., 2013; Jacquel et al., 2021; Currie et al., 2023). Recently, Molines and colleagues showed that cytoplasmic viscosity modulates MT dynamics in vivo (Molines et al., 2022). Here, we show that Q-nMT bundle formation does not rely on modifications of the physicochemical properties that cells experience upon quiescence establishment (Sup. Fig. 1). In fact, dynamic cMTs and a stable Q-nMT bundle can be observed simultaneously in a quiescent cell (Sup. Fig. 1H), while physicochemical properties evolve in parallel both in the nucleus and in the cytoplasm (Joyner et al., 2016). Importantly, reduction of the cellular volume, pH acidification and increased viscosity are observed within minutes upon glucose starvation (Joyner et al., 2016) while the Q-nMT bundle needs a couple of days to be fully assembled (Fig. 1A). Further, we show that its formation cannot be delayed in time (Fig. 1H) and as such probably depends on an active process induced by a transient signal emitted upon glucose exhaustion. We speculate that the SPB may act as a platform that integrates this signal, and transfers it to the MT machinery in order to assemble a stable parallel structure.

As it is required for the onset of Q-nMT bundle formation (Fig. 3B), Aurora B/Ilp1, a kinase that plays a key role at the MT-kinetochore interface where it senses tension, could be one of the target of the above-mentioned nutritional signal. In parallel, upon quiescence establishment, chromosomes hyper-condensation (Guidi et al., 2015; Rutledge et al., 2015) may modify tension at the kinetochore/MT interface and as such could contribute to the initiation of phase I. Besides, Bim1 has also been involved in MT-kinetochore attachment (Dudziak et al., 2021; Akhmanova and Steinmetz, 2015). Although it deletion alone has no effect on Q-nMT bundle formation, its role become critical for phase I if kinetochore-MT interaction are already destabilized by the absence of other CPC components such as Sli15 and Bir1 (Fig. 3D and Sup. Fig. 3E). Altogether, these observations point to kinetochore-MT interaction as essential for phase I initiation, as confirmed by the absence of Q-nMT bundle in the ndc80-1 mutant (Fig. 3A).

In phase I, SPB-anchored MTs elongate and are concomitantly stabilized (Fig. 1 A-C). This step requires the kinesin-14 Kar3 and its regulator Cik1 (Fig. 4B). During mating (Molodtsov et al., 2016), and in early mitosis, when half-spindles form (Kornakov et al., 2020), Kar3/Cik1, together with Bim1, align and cross-link growing MTs along existing ones, thereby promoting the organization of MTs into parallel bundles. Yet, in mitosis, the kinesin-14/EB1 complex promotes MT dynamics (Kornakov et al., 2020). In phase I, whether the robust MT stabilization depends on the modification of the kinesin-14/EB1 complex properties or is due to additional specific cross-linker(s) remains to be clarified (Fig. 6b).

Figure 6:

In quiescence, deletion of either monopolin or Slk19 leads to the formation of short, shattered and flared MT structures, phenotypes that are worsened by the deletion of Bim1 (Fig. 3F-H and Sup. Fig. 3F). Since Monopolin and Slk19 are involved in kinetochore clustering (Plowman et al., 2019; Rabitsch et al., 2003; Tóth et al., 2000; Lee et al., 2004; Mittal et al., 2019; Movshovich et al., 2008; Richmond et al., 2013; Zeng et al., 1999), we speculate that kinetochore-kinetochore interaction may not only help preventing depolymerization of single MT, but also constrain elongating phase I MTs and facilitate their concomitant cross-linking (Fig. 6c).

With the initiation of phase I, Tub4 starts accumulating at the SPB to allow the second wave of MT nucleation observed in phase II. When phase II MTs elongate, they are concurrently stabilized along pre-existing phase I MTs (Fig. 6d). This step relies on the kinesin-5/Cin8 as Q-nMT bundles assembled in cin8Δ cells are thinner than WT Q-nMT bundles (Fig. 4C and D). Intriguingly, in vitro, the monomeric form of the vertebrate kinesin-5 Eg5 promotes MT nucleation and stabilizes lateral tubulin-tubulin contacts (Chen et al 2019). We thus assume that in the absence of Cin8, either nucleation of phase II MTs cannot occur, or phase II MTs can elongate, but cannot be cross-linked and stabilized to phase I MT while they grow, and as such rapidly vanish. An absence of bundle thickening in phase II is also observed in the “Tub1-only” mutant i.e. in the absence of Tub3. Tubulin isoforms or tubulin PTM can allow the recruitment of specific kinesins and thereby modify MT properties, see for example (Sirajuddin et al., 2014; Peris et al., 2009; Dunn et al., 2008). May Tub3 be important for Cin8 recruitment or function during phase II?

Then, MTs elongate, but their distal parts (>1 µm) are not yet stable (Fig. 1A). The full-length Q-nMT bundle stabilization is reached about 2 days after glucose exhaustion. This step depends on the kinesin-5/Kip1 (Fig. 4 G-H and Sup. Fig. 4E) that is essential to cross-link and stabilize elongating MTs during phase III (Fig. 6e). Accordingly, we observed a Kip1 enrichment at the Q-nMT bundle +ends (Fig. 4F). Thus, in agreement with their ability to form homo-tetramers able to cross-link (Acar et al., 2013; Kapitein et al., 2005; Scholey et al., 2014; Weinger et al., 2011) and stabilize parallel MTs (Kapitein et al., 2005; Shimamoto et al., 2015; Yukawa et al., 2020), the two yeast kinesin 5 are essential for Q-nMT bundle formation. Yet, in quiescence, just as in mitosis, Kip1 and Cin8 have clearly non-equivalent functions (Roostalu et al., 2011; Shapira and Gheber, 2016).

So far, our results indicate that the Q-nMT bundle formation mechanism is different from the one at work during mitotic spindle assembly. Similarly, we show that Q-nMT bundle disassembly relies on a mechanism that does not involves the same pathway than mitotic spindle dismantlement. Indeed, Q-nMT bundle disassembly involves Kip3 (Fig. 5C-D) but is independent of the anaphase-promoting complex and Aurora B (Woodruff et al., 2010). Interestingly, the longest MTs start depolymerizing from their +ends first. Then, when they reach the +ends of shorter MTs, they embark them on depolymerization (Sup. Fig. 5A). This cooperative behavior has been observed in vitro within parallel MT bundles (Laan et al., 2008) and our data indicate that such a comportment can exist in vivo. In telophase, the CPC localizes to the end of the two half-spindles in order to participate in mitotic spindle disassembly (Ibarlucea-Benitez et al., 2018). Intriguingly, in quiescence, CPC is poised at the Q-nMT bundle + end (Fig. 3C). May this CPC localization facilitate Q-nMT bundle rapid disassembly once environmental conditions become favorable, thus ensuring quiescence exit efficiency?

Finally, in the present study, we show that the Q-nMT bundle is required for WT cells survival in quiescence. While we do not yet know why this structure is important to face chronological age, it is clear that the presence of Q-nMT bundle strongly alters the nuclear organization (Laporte et al., 2013; Laporte and Sagot, 2014; Laporte et al., 2016). As chromatin organization influences gene expression, one hypothesis could be that the presence of Q-nMT bundle is required for expression of genes required for cell survival in quiescence. Another attractive hypothesis could be that the Q-nMT bundle is the yeast analog of the primary cilium of mammalian cells. In fact, these two structures formed upon proliferation cessation, they are both composed of highly stable parallel MTs, their formation involve MT motors, and they are both template from the centriole/SPB. More importantly, we show here that the Q-nMT disassembly is needed to authorize nascent SPB separation upon quiescence exit (Fig. 5). Thus, the Q-nMT bundle disassembly may controls reentry into the proliferation cycle, just as the primary cilium does in ciliated mammalian cells (Goto et al., 2017; Kim and Tsiokas, 2011).

Materiel and Methods

Yeast strains, plasmids and growth conditions

All the strains used in this study are isogenic to BY4741 (mat a, leu2Δ0, his3Δ0, ura3Δ0, met15Δ0) or BY4742 (mat alpha, leu2Δ0, his3Δ0, ura3Δ0, lys2Δ0), available from GE Healthcare Dharmacon Inc. (UK), except in Fig. 2, in which S288c strains were used (Nsamba et al., 2021). BY strains carrying GFP fusions were obtained from ThermoFisher Scientific (Waltham, MA, USA). Integrative plasmid pTub1-mTurquoise2-Tub1, pTub1-wasabi-Tub1, pTub1-mRUBY2-Tub1, pTub1-mEOS2-Tub1 were a generous gift of Wei-Lih Lee (Markus et al., 2015). The strain expressing SPC42-mRFP1 is a generous gift from E. O’Shea (Huh et al., 2003). The strain expressing TUB4-mTQZ is generous gift of S. Jaspersen (Burns et al., 2015). Three copies of eGFP in tandem were integrated at the 3’ end of DAD2, TUB3 or KAR3 endogenous loci respectively. Plasmids for expressing Nup2-RFP (p3695) or Bim1-3xGFP (p4587) from the endogenous locus were described in (Laporte et al., 2013).

For Fig. 1F, pHIS3:mTurquoise2-Tub1+3’UTR::LEU2 was integrated at TUB1 locus, then a pRS303-ADH2p-mRuby2-Tub1 was integrated at the HIS3 locus. To generate this plasmid, the ADH2 promoter was amplified from yeast genomic DNA flanked with NotI and SpeI restrictions sites and inserted in pRS303. The mRuby2-Tub1+3’UTR, including the 116-nucleotide intron and 618 nucleotides downstream the stop codon was cloned between SpeI/SalI sites.

Yeast cells were grown in liquid YPDA medium at 30 °C in flasks as described previously in (Sagot et al., 2006) except for experiments using thermosensitive strains, where cells were first grown at 25 °C then shifted at 37 °C for indicated time before imaging.

Experiment in Sup. Fig. 1I was performed as described in (Orij et al., 2009, Peters et al., 2013). In brief, proliferating cells were transfered at an OD_600nm of 0.5 in Hepes buffer (25 mM Hepes, pH 7.4, 200 mM KCl, 1 mM CaCl₂, and 2% dextrose) buffered either at pH 4 or 7.5, in the presence of 100 µM CCCP (Sigma-Aldrich). After 150 min 30 °C shaking, cells were imaged (for mTQZ-Tub1) or fixed and stained using Alexa Fluor 568–phalloidin (Invitrogen) as described (Sagot et al., 2006).

For live cell imaging, 2 μL of the cell culture were spotted onto a glass slide and immediately imaged at room temperature.

For quiescence exit, cells were then incubated 2-3 min in liquid YPD and a 2 µL were spread onto a 2 % agarose microscope pad containing YPD. Individual cells were imaged every hour, up to 6 or 12 h at 21°C. For quiescence exit in presence of cycloheximide (Fig. 5A), cells were pre-incubated 30 min in the presence of the drug prior quiescence exit. Cycloheximide was used at 180 µM (Sigma-Aldrich), Nocodazole was used at 30 µg/mL (7.5 µM) (Sigma-Aldrich).

Fluorescence Microscopy

Cells were observed on a fully-automated Zeiss 200M inverted microscope (Carl Zeiss, Thornwood, NY, USA) equipped with an MS-2000 stage (Applied Scientific Instrumentation, Eugene, OR, USA), a Lambda LS 300 Watt xenon light source (Sutter, Novato, CA, USA), a 100X 1.4NA Plan-Apochromat objective, and a 5 position filter turret. For RFP imaging we used a Cy3 filter (Ex: HQ535/50 nm – Em: HQ610/75 nm – BS: Q565 nm lp). For GFP imaging, we used a FITC filter (excitation, HQ487/25 nm; emission, HQ535/40 nm; beam splitter, Q505 nm lp). For mTurquoise imaging we used a DAPI filter (Ex: 360/40 nm – Em: 460/50 nm– BS: 400 nm). All the filters are from Chroma Technology Corp. Images were acquired using a CoolSnap HQ camera (Roper Scientific, Tucson, AZ, USA). The microscope, camera, and shutters (Uniblitz, Rochester, NY, USA) were controlled by SlideBook software 5. 0. (Intelligent Imaging Innovations, Denver, CO, USA).

For PALM (Fig. 4E) a Nikon Ti-Eclipse equipped with iLas2 TIRF arm, laser diodes (405, 488, 532, 561, 642 nm), a 100x 1.49 oil (TIRF) objective connected to a EMCCD Camera Photometrics Evolve was used.

Immunofluorescence was done as in (Laporte et al. 2015), using AlexaFluor Phalloidin (Invitrogen).

For expansion microscopy (Sup Fig. 2A), spheroplasts were obtained as described in (Laporte et al., 2016). After washes in 10 mg/mL NaBH4 in PEMS for 10 min, spheroplasts were seeded on 12 mm round poly-L-lysine coated coverslips, washed twice for 30 min in 100 μL PEM-BAL (PEM + 1% BSA) and incubated with the primary antibodies (anti-GFP from mouse; Roche, ref. 11814460001, 1/50; anti-alpha-tubulin from rat; YOL1/34, Abcam, ref. Ab6161, 1/100) for 1 hour at 37°C. Cells were then washed three times in PEM-BAL and incubated with the secondary antibodies (Goat anti-mouse AlexaFluor®488 1/200 and Donkey anti-rat AlexaFluor®555; A11029 and A21434 respectively, ThermoFisher Scientific, Waltham, MA) for 45 min at 37 °C. Cells were then washed three times with PBS. Cells were processed for Expansion Microscopy as previously described in (Bahri et al., 2021). Spheroplasts were incubated for 10 min in 0.25 % GA in PBS, washed in PBS three times for 5 min, and processed for gelation. A drop of 100 µL of ExM MS (8.625% (wt/wt) SA, 2.5% (wt/wt) AA, 0.15 % (wt/wt) BIS, 2 M NaCl in 1× PBS) was placed on the chilled Parafilm, and coverslips were put on the drop with cells facing the solution and incubated for 3 min. Then the coverslips were transferred to a 35 μL of ExM MS supplemented with 0.2 % APS and 0.2 % TEMED, with the initiator (APS) added last. Gelation proceeded for 3 min on ice, and samples were incubated at 37 °C in a humidified chamber in the dark for 1 h. Then coverslips with attached gels were transferred into a six-well plate for incubation in 2 mL of digestion buffer (1× TAE buffer, 0.5% Triton X-100, 0.8 M guanidine hydrochloride, pH ∼8.3) supplemented with DAPI (1 µg/mL) for 10-15 min at 37 °C, until the gels detached. Fresh proteinase K at 8 units/mL was then added and samples incubated at 37 °C for 30 min. Finally, gels were removed and placed in 10mL petri dishes filled with ddH2O for expansion. Water was exchanged at least twice every 30 min, and incubated in ddH2O overnight at RT. Gels expanded between 4× and 4.2× according to SA purity. Expanded cells were imaged with a UPlanS Apo 100×/1.4 oil immersion objective in an Olympus IX81 microscope (Olympus, Tokyo, Japan). For Structured illumination microscopy (SIM, Fig. 7D), a ZEISS Lattice Lightsheet 7 was used.

Laser ablation (Fig. 3I) was performed at room temperature with a 100X oil Plan-Apochromat objective lens (NA 1.4) and an Axio-Observed.Z1 microscope (Carl Zeiss) equipped with a spinning disk confocal (Yokogawa), an EMCCD Evolve camera (Photometrics and Roper Scientific) and 491 nm (100 mW; Cobolt calypso) and 561 nm (100 mW; Cobolt Jive) lasers. Images were acquired with Metamorph software (Roper Scientific). Every 5 to 20 s, a Z-series of 0.4 µm steps were acquired. A 355 nm microchip laser (Teem Photonics) with a 21 kHz repetition rate, 0.8 µJ energy/pulse, 2 kW of peak power and 400 ps pulse width, powered with an iLas2 PULSE system (Roper Scientific) was used between 10 to 40 % power with one pulse of a spot length of 100 points. Breakage was considered as successful if a non-alignment between the two remaining Q-n MT bundle fragments was observed.

Z-stacks were deconvolved using the Deconvolution Lab plugin (Fig. 1D; Fig. 3C,E,F Fig. 4A,F). In figures 1F, 4A, B, G, supplemental figure 1H and 3B a faint and fuzzy fluorescence signal is detected in some cell cytoplasm using the GFP filter set. This signal is not GFP but rather due to a non-specific yellow background fluorescence.

Image analysis

Distribution and associated statistics were performed using GraphPad Prism 5 (GraphPad Software, Inc. La Jolla, USA). Unless specified, histograms and scatter dot plots show the mean and the error bars indicate SD. MT length was measured on MAX-projected image using imageJ.

For MT fluorescence intensity measurement, a line scan (i1) of 4 pixel width containing both FP signal and background was drawn along MTs on SUM projection image (3-4 Z-plans) using ImageJ. A line of 8 pixels width (at the same location) was drawn in order to calculate the intensity of the surrounding background (i2). The real intensity (ir) was calculated as follow: int=(i1-((i2×8)-(i1×4))/4). Similar approach was conducted to measure Tub4 fluorescence intensity (Fig. 1E) i.e. (i1) and (i2) were two boxes, the second including the first one. Intensity was calculated as follow: int=[i1-([(i2xi2_surface)-(i1xi1_surface)]/(i2_surface-i1_surface))*i1_surface]. For morphometric Q-nMT bundle property distribution (Fig. 1C and 4B)the mean fluorescence intensity was measured in each individual cell, in a finite area localized adjacent to the SPB (0,4 µm – 3-4 Z stack sum projected) as an estimate of Q-nMT bundle width and plotted as a function of the according measured bundle length.

For Nuf2 fluorescence intensity measurement at SPB (Sup. Fig. 1D), GFP and RFP line scan measurements were done on a 3-4 Z-plans sum projection. The “SPB localization zone” was defined as the length ending at the second pixel after the brightest RFP signal of Spc42-RFP. The remaining Q-nMT bundle zone was defined as the “+end zone”. The total Nuf2-GFP signal detected along the Q-nMT bundle was set to 100 % (sum projection imaged) allowing to determine the percentage of the signal measured at “SPB localization zone” and at “+end zone”

Normalized Q-nMT bundle length (Fig. 3C) was done after line scan intensity measurement. mTQZ-Tub1 intensities slopes were first aligned on their inflexion point. To compare Q-nMT bundle with different lengths, we first sorted Q-nMT bundle <1.8 µm. After an artificial isotropic expansion, we fit all MT structures to length order. Then, corresponding mean intensity of Q-nMT bundle (mTZQ-Tub1) and GFP signal were calculated.

To measure the Q-nMT bundle depolymerization (Fig. 5A and E), individual Q-nMT bundle length were measured over time. The first measurement was set at 100 % of the length, and the following percentages calculated accordingly. A fluorescence drop above or equal to 25 % defines the inflexion point of the slope, and was used to align the different length measurements. The FWHM (Full Width at Half Maximum) was calculated by measuring fluorescence intensity of a line crossing perpendicularly a Q-nMT bundle, from a sum projection. After fitting intensity level with Gaussian distribution and obtained the standard deviation (σ) value, FWHM was calculated using the equation 2√(2ln2)xσ. In Fig. 5B, FWHM was measured at 0.5 µm of the SPB.

Western blots

Western blots were done as described in (Sagot et al., 2006) using anti-GFP antibodies (Roche); anti-Tat1 antibodies (a generous gift from J-P. Javerzat) and antibody against the budding yeast Act1 or Sac6, generous gifts from B. Goode.

Phenotypical analysis of cells without Q-nMT bundle

WT strains were grown to glucose exhaustion (OD ∼6.5), washout with “old YPDA” and pulsed with either DMSO or Nocodazole (30µg/mL) for 24 h. Cells were then washed twice in “old YPDA” before being incubated in the same medium. Dead cell measurement was done using methylene blue staining. The capacity of cells to exit quiescence was scored after cells micro-manipulation (N=4, n>100) as described in (Laporte et al., 2011).

Supplemental data

Supplementary Figure 1 further described the 3 steps of Q-nMT bundle formation. It shows the impact of physico-chemical cell properties on MT stabilization and demonstrate that Q-nMT bundle is required for both cell viability in quiescence and quiescence exit fitness.

Supplementary Figure 2 describes the impact of the alpha tubulin level on Q-nMT bundle assembly.

Supplementary Figure 3 demonstrates that kinetochore-MT interactions are not required for Q-nMT bundle maintenance and that mutants affected for kinetochore-kinetochore interactions have a reduced viability in quiescence.

Supplementary Figure 4 impact of various yeast kinesin deletion on Q-nMT bundle morphometric parameters.

Supplementary Figure 5 Q-nMT bundle disassembly in mutants affected for mitotic spindle dismantlement.

Acknowledgements

The authors would like to thank the Bordeaux Imaging center for the help in super-resolution imaging. We also thank Zeiss for their help with SIM. We express our gratitude to E. O’Shea, S. Jaspersen, J-P. Javerzat, Wei-Lih Lee and B. Goode for sharing reagents. We would like to thank J-P. Javerzat for helpful and constructive discussions about our work. DL, AML, JD and IS were supported by a grant from the ANR-21-CE13-0023-01, a Ligue Contre le Cancer Régionale – Dordogne grant #193366 and the CNRS. MG and EN were supported by a National Science Foundation grant number MCB-1846262. AR was supported by the Conseil Régional d’Aquitaine (#20111301010) and the CNRS.

Author Contributions

D. Laporte did all the experiments described in this manuscript, except Fig. 1F, 1H, Fig. 5 G-H and Sup. Fig. 5G to I that were performed by A. Massoni-Laporte, together with all image deconvolution, and Fig. 4E and Sup. Fig. 2A and 3A that were done by J. Dompierre. A. Royou helped for pulse-laser experiments of Fig. 3I and Sup Fig. 3H-I. C. Lefranc did molecular biology and western blots. D. Mauboules performed yeast genetic for Fig. 2. M. L. Gupta Jr and E. T. Nsamba created the Tub1-only mutant (Fig. 2). L. Gal and M. Schuldiner performed high throughput deletion screen. IS and DL designed and supervised the experiments. IS and DL wrote the manuscript.

Abbreviations

• MT: microtubule

• SPB: spindle pole body

• GTP: guanosine triphosphate

• MTOC: microtubule organizing centers

• ɣ-TuC: ɣ-tubulin complex

• MAPs: microtubule associated proteins

• Q-nMT bundle: quiescence specific nuclear microtubule bundle

• CPC: chromosome passenger complex