A novel benzodiazepine derivative that suppresses microtubule dynamics and impairs mitotic progression

Microtubules are composed of α- and β-tubulin heterodimers whose assembly undergoes stochastic phases of growth and shrinkage in a process called dynamic instability (Mitchison and Kirschner, 1984). The microtubule network serves diverse and specific functions in differentiated cells such as scaffold and cargo transport. As the cell proceeds through mitosis, the microtubule network undergoes a dramatic reorganization to build a mitotic spindle used to equally segregate chromatids in the two daughter cells. In vivo, microtubule nucleation and organization, as well as their intrinsic dynamic instability, are regulated by microtubule-associated proteins (MAPs). The influence of MAPs on the dynamic assembly of microtubules contributes to the microtubule assembly patterns required for the proper formation of the mitotic spindle (Desai and Mitchison, 1997; Prosser and Pelletier, 2017).

The dynamic properties of mitotic spindle microtubules are required for fast and amphitelic attachment of sister kinetochores to the opposite spindle poles, an event essential to avoid mis-segregation of chromatids during anaphase and subsequent aneuploidy. In case of erroneous attachments of chromosomes to spindle microtubules, the spindle assembly checkpoint (SAC) remains unsatisfied and inhibits anaphase onset through inhibition of the anaphase promoting complex/cyclosome (APC/C). Cells are then delayed in mitosis until all defective attachments have been corrected (Khodjakov and Rieder, 2009).

The central role of microtubules in orchestrating cell division, together with the property of SAC-dependent mitotic delay, have made microtubules a target of choice for cancer chemotherapeutic agents aiming to block cell division (Janssen and Medema, 2011). Microtubule-binding compounds that block microtubule dynamics, such as Vinka alkaloid and taxanes, have emerged as efficient anti-tumor agents and are currently used in therapeutic treatments (Dumontet and Jordan, 2010). However, their use leads to severe side effects, including neuropathy (Argyriou et al., 2012). Moreover, a variability in the response to the treatment with these compounds in different tumor types is observed (Kavallaris, 2010). The identification of new anti-mitotic drugs with alternative modes of action thus remains a priority for anti-cancer research (Dumontet and Jordan, 2010).

Activation of the AMPA receptor by glutamate has long been known to enhance cell proliferation (Stepulak et al., 2014). A recent screen to identify new anti-proliferative compounds has highlighted the non-competitive AMPA receptor agonist derivate 1-(4-amino-3, 5-dimethylphenyl)-3,5-dihydro-7,8-ethylenedioxy-4 h-2,3-benzodiazepin-4-one (called hereafter 1g) as a potent novel growth inhibitor (Parenti et al., 2016). Upon treatment with 1g, human leukemia T cells accumulate with a 2N DNA content, suggesting an arrest of cells at the G2/M phase of the cell cycle. In the current study, we have thus evaluated the potential of 1g as a novel promising and efficient anti-mitotic drug with a unique effect on microtubule growth.

A previous study has indicated that 1g induces an accumulation of Jurkat cells at the G2/M phase of the cell cycle (Parenti et al., 2016). To confirm the effect of 1g on the proliferation of transformed human epithelial cells, we followed an asynchronous population of HeLa cells by phase contrast time-lapse microscopy for 20 h (Fig. 1A; Movies 1 and 2). Control cells progressively rounded up to enter mitosis, divided and spread back down over this period of time, increasing the total number of cells. In contrast, cells treated with 1g similarly rounded up, but no cell division was observed, suggesting that cells remained arrested in prometaphase and were unable to progress into anaphase and complete their division. This progressive accumulation of rounded cells was observable in a dose-dependent manner with an optimal 1g concentration of 2.5 µM (Fig. 1B,C).

To evaluate whether cells treated with 1g display a cell division defect, we synchronized HeLa cells at the G2/M transition using the CDK1 kinase inhibitor RO3306 (Vassilev, 2006), then released them in the presence of DMSO (control) or 1g and followed their progression through mitosis (Fig. 1D,E). Their mitotic stage was defined by staining the cells for tubulin and DNA. As expected, control cells reached the metaphase–anaphase transition within 45 min and 79% of the cells were in telophase at 90 min post-release. In contrast, only 6% of cells treated with 1 µM 1g had passed the metaphase–anaphase transition by 90 min and none of the dividing cells exposed to 2.5 µM and 5 µM 1g had reached anaphase. Furthermore, dividing cells treated with 2.5 µM and 5 µM 1g remained in prometaphase, displaying an abnormal mitotic spindle (Fig. 1D, bottom right panel). Whereas control cells were able to fully congress their chromosomes to the metaphase plate within 45 min, cells treated with 1g still displayed major congression defects at 90 min (Fig. 1D,F and Fig. 2B,D). This failure to establish a metaphase plate was associated with the detection of the SAC component BUBR1 on the kinetochore in agreement with an activation of the SAC (Fig. 1F). The drastic increase in the mitotic index (Fig. 1C) of the asynchronous cell population after a 16 h exposure to 1g implies that at high drug concentration, cells remain blocked in mitosis (82±5% in 2.5 µM 1g versus 7.4±0.81% in DMSO). Instead, the more modest increase in mitotic index (26.8±0.8% in 1 µM 1g versus 82.0±5.0% in 2.5 µM 1g) when cells were exposed to 1 µM 1g suggests that at suboptimal 1g concentrations, cells can progress through mitosis but with extensive delay. Indeed, anaphase figures could be observed 4 h post-release in cells treated with 1 µM 1g, but not with higher concentrations (data not shown).

To confirm that the accumulation of rounded cells observed in the asynchronous cell population treated with 1g (Fig. 1A) was indeed due to activation of the SAC making the cells unable to exit mitosis, HeLa cells were treated with 1g alone or in combination with the SAC inhibitor AZ3146 (Tipton et al., 2013), and followed by videomicroscopy. In the presence of AZ3146, 1g-treated cells progressively rounded up, but unlike cells treated with 1g alone, proceeded through a defective anaphase-telophase and spread back down. It is interesting to note that these cells displayed no cytosolic sign of cell death throughout the 16 h of the treatment with both inhibitors (Fig. 1G; Movies 3 and 4). These data indicate that 1g impacts cell proliferation via mitotic arrest and not direct cell death. However, prolonged mitotic arrest has been shown to result in subsequent cell death (Brito and Rieder, 2006; Stanton et al., 2011), explaining the previously described apoptotic effect of the 1g compound observed after 48 h of treatment (Parenti et al., 2016). All together, these experiments demonstrate that treatment of proliferating HeLa cells with 1g activates the SAC, arresting the cells in prometaphase.

The disorganized mitotic spindle observed in 1g-treated cells, prompted us to further characterize the dynamics of the mitotic spindle assembly. To that purpose, using spinning-disk microscopy, we imaged live HeLa cells expressing GFP-tubulin after their release from the CDK1 inhibitor-induced G2/M arrest (Fig. 2A,B; Movies 5–7). Control cells treated with DMSO rapidly separated their spindle poles and 100% of the cells formed a bipolar spindle within 15±4 min (Fig. 2C). All cells achieved full chromosome congression into a metaphase plate and progressed into anaphase in an average time of 55±7 min (Fig. 2B,D). When treated with low 1g concentrations (1 µM), 90% of the cells formed a bipolar spindle within 28±13 min. However, at 80 min after nuclear envelope breakdown (NEB), 90% of cells treated with 1 µM 1g still displayed unaligned chromosomes and thus had not satisfied the SAC (arrowheads in Fig. 2B). At higher 1g concentrations (2.5–5 µM), several microtubule nucleation sites could be observed and the microtubule growth from centrosomes was strongly reduced (Fig. 2A, arrowheads). Although clustering of the microtubule asters could be observed, in 70% of the cells exposed to 2.5 µM 1g and in 100% of those treated with 5 µM, three asters or more were present and strong chromosome congression defects were observed at 80 min post NEB (Fig. 2B, arrows, D,E). No bipolar mitotic spindles were formed at those concentrations (Fig. 2C). Visualization of the centrosomes, using HeLa cells expressing the centriolar protein centrin-GFP, confirmed that microtubule nucleation emanated from the two centrosomes. However, the additional small asters observed did not contain centrin-GFP, indicating that 1g did not trigger centrosome amplification (Fig. 2G). In addition, a decrease of 25% (1 µM 1g) and 37% (2 µM 1g) in spindle poles distance compared with control cells was observed (Fig. 2F). The defect in centrosome separation was further accentuated at 5 µM, with a 64% decrease in distance between centrosomes in cells treated with 5 µM 1g compared with DMSO (Fig. 2H). The formation of shorter microtubules observed in the presence of 1g most likely accounts for the reduced centrosome separation, since centrosome separation is a microtubule-dependent process occurring in prophase and prometaphase (Wittmann et al., 2001).

The dynamic properties of microtubules are crucial for the extensive remodeling of interphase arrays of microtubules into a mitotic bipolar spindle (Desai and Mitchison, 1997). We thus investigated whether 1g interferes with microtubule polymerization. We first performed a depolymerization–repolymerization type of assay (Fig. 3A). The microtubule network of interphase cells was depolymerized at 4°C in the presence or absence of 1g. We then followed the dynamics of microtubule regrowth as the samples were returned to 37°C. Long newly nucleated microtubules could be observed 5 min after the temperature switch in control cells, and extensive microtubule repolymerization was present after 10 min. Presence of 1g did not affect cold-induced depolymerization of microtubules, indicating that 1g does not act as a microtubule-stabilizing agent (Fig. 3A, t=0 and G). Indeed, in the presence of stabilizing agents, such as taxan, cold-resistant microtubule bundles can be detected (Stanton et al., 2011). However, 1g markedly slowed down the dynamics of the microtubule network reformation (Fig. 3A, t=5, 10 and 15 min). Only short and fragmented microtubules were observed in 1g-treated cells in the first 10 min, and it required 30 min to polymerize a microtubule network equivalent to the one formed after 15 min in control cells.

The growing microtubule plus-end tracking protein (+TIP) EB1 has been commonly used to image microtubule plus-ends and to quantify their dynamics (Matov et al., 2010). To further investigate and quantify the effect of 1g on microtubule dynamics, we imaged HeLa cells expressing EB1-GFP by spinning disk confocal microscopy at 1 s time intervals before and immediately after addition of 1g or DMSO (Fig. 3B; Movies 8 and 9). In control HeLa cells, measurements of EB1 comet velocity indicated a microtubule growth speed of 16.6±0.83 µm min⁻¹. Whereas addition of DMSO in control cells did not significantly alter microtubule growth speed, 1g treatment resulted in a 25% decrease in mean growth rate (from 14.2±0.3 before to 10.6±0.3 µm min⁻¹ with 2.5 µM 1g) (Fig. 3C,D) without affecting the median growth lifetime. Furthermore, the initial microtubule growth pattern (before treatment, Fig. 3B,C) was recovered shortly after drug removal (washout, Fig. 3B,C) indicating that the action of 1g is reversible (15.3±1.1 µm min⁻¹ after washout) (Fig. 3B,C; Movies 8 and 9). Treatment with 1g did not significantly alter the number of nucleation events (Fig. 3E) or the number or duration of growth pauses (Fig. 3F). In control cells, we could infer a shrinkage rate of 32.9±2.2 µm min⁻¹, which was not statistically different in DMSO-treated cells (34.5±1.7 µm min⁻¹). In the presence of 1g, no shrinkage events were detected. When tracking EB1 comets, shrinkage can only be inferred when followed by significant EB1-labelled microtubule regrowth (Matov et al., 2010). The absence of detectable shrinkage in the presence of 1g could be explained by the fact that the main events occurring were terminal shortenings and shrinkage followed by short or slow growth phases that did not produce detectable EB1 comets within the temporal window analyzed. We thus clearly observed a slowdown in microtubule polymerization in interphase cells. Altogether, these data indicate that 1g is able to promote a fast and reversible inhibition of microtubule growth during interphase and mitosis.

To assess whether the 1g compound directly targets tubulin polymerization, we tested its impact on microtubule self-assembly in vitro. We performed turbidity assays classically used to analyze the effect of drugs or MAPs on microtubule assembly, including microtubule nucleation and elongation (Gallaud et al., 2014) (Fig. 3G). Absorbance at 350 nm, which is directly proportional to the amount of microtubule polymers formed, revealed no significant difference when purified tubulin was incubated in polymerizing buffer at 37°C in the presence of a range of 1g concentrations or DMSO. More specifically, we did not observe any effect of the drug on nucleation (same lag-phases at different concentrations), on elongation (sigmoid) phase, or on the plateau of polymerization (total mass of microtubules assembled). In addition, no significant formation of aggregates was detected when the temperature was shifted back to 0°C as the average absorbance returned to the base line for all samples. The absence of significant alteration in tubulin polymer assembly or on formation of aggregates indicates that 1g does not interfere with microtubule self-assembly directly.

During the past decades, Drosophila melanogaster has emerged as an interesting model to identify new genes required for cell division as well as for cancer research (Gonzalez, 2013). Therefore, we examined the impact of the compound on neural stem cell (neuroblast) division, in the developing brain of Drosophila larvae. Third instar larval brains expressing RFP-α-tubulin and H2A-GFP were dissected and cultured in Schneider medium containing DMSO, or various concentrations of 1g. For comparison, we used in parallel 20 µM taxol and 10 µM nocodazole treatments (Fig. 4A; Fig. S1). Control neuroblasts formed bipolar spindles and started to segregate their chromosomes 5.9±0.4 min after the NEB. As expected, following prolonged mitotic arrest, neuroblasts treated with taxol or nocodazole underwent slippage (Fig. S1D,E). In the presence of 1g, mitosis duration was significantly increased in a dose-dependent manner (6.1±0.6 min, 8.2±0.6 min and 10.9±3.1 min for 5 µM, 10 µM and 20 µM treatment, respectively) (Fig. 4D), suggesting an activation of the SAC. The delay observed was in the range of that observed in neuroblasts depleted of Msps (15.5±0.8 min, n=44) (Fig. S1D), a MAP that has been described to severely disrupt the integrity of the mitotic spindle when downregulated (Cullen et al., 1999). Furthermore, in agreement with SAC activation, 1g-treated neuroblasts exhibited dose-dependent mitotic spindle assembly defects (5 µM: 35.3%, 10 µM: 100% of neuroblasts) (Fig. 4B). Lagging chromatids were frequently observed in 31.3% (n=6/17) and 68.4% (n=13/19) of neuroblasts treated with 10 and 20 µM of 1g, respectively (Fig. 4C). In 79% (n=42) of cells, we observed the presence of tripolar spindles and the formation of two central spindles (Fig. 4A, 1,g 8, 9, 13 min, Fig. 4E,F). The outcome of mitosis was variable. Whereas some cells proceed to double cytokinesis, leading to the formation of three daughter cells (Fig. 4A, circled in 1g panels), in some instances, regression of the initial double cytokinesis furrow could also be observed. In either case, following defective division 1g-treated neuroblast progenies were able to polymerize microtubules from their interphase centrosome (Fig. 4G,H).

We show here that the anti-proliferative effect of 1g in cancer cell lines observed previously (Parenti et al., 2016) is caused by interference with microtubule polymerization and defective mitotic spindle assembly, leading to SAC activation. Of note, the 1g concentrations used in cell treatments (5 µM or less) are 20 times less than the K_d for the AMPA receptor (K_d>100 µM; Micale et al., 2008), indicating that the strong effects of 1g on cell division are independent of the AMPA receptor signaling pathway.

Our in vitro assays with pure tubulin indicate that 1g does not directly target the intrinsic microtubule polymerization. When 1g was used in vitro within cells or in vivo, we did not observe the massive effects on the microtubule cytoskeleton that are triggered with microtubule binding agents such as taxol or nocodazole. Instead, under 1g treatment, the overall architecture of the HeLa cell interphase microtubule network remained intact. Moreover, Drosophila neuroblasts were still able to nucleate de novo microtubules from the daughter centrosome following division. As cells proceed from interphase to mitosis, a 10-fold increase in the turnover rate of microtubules is required for the reorganization of the microtubule network into a mitotic spindle and for the capture of chromosomes (Desai and Mitchison, 1997; Prosser and Pelletier, 2017). This change in microtubule turnover rate implies that a drug that moderately impedes microtubule dynamics in interphase would be expected to trigger more drastic effects on mitotic microtubules. Indeed, the main effect of 1g was observed during cell division in fly brain neuroblasts and in HeLa cells.

In both systems exposed to 1g, SAC activation was observed, resulting either in mitotic delay in fly neuroblasts or in mitotic arrest in mammalian cells. Whereas neuroblasts treated with taxol or nocodazole (which severely impair spindle assembly) remained arrested and underwent mitotic slippage, the mitotic delay in fly neuroblasts treated with 1g, did not exceed three times the duration of cell division. The presence of kinetochores not attached to microtubules of the mitotic spindle causes SAC activation and mitotic arrest. In 1g-treated neuroblasts, the observation of a mitotic delay rather than a mitotic arrest, suggests that despite displaying a tripolar shape, the spindle microtubules still manage to grow and correctly attach kinetochores. However, the altered microtubule growth results in a time delay to perform that task and satisfy the SAC. In contrast, HeLa cells remained arrested following 1g treatment. The difference in genome size could account for this difference in the 1g drug response. Fly cells harbor only four chromosomes, making it easier to achieve kinetochore attachment under defective MT polymerization and thus satisfy the SAC and exit mitosis. By contrast, SAC satisfaction is likely more problematic for HeLa cells that require the correct attachment of 70–82 chromosomes (Landry et al., 2013).

Nevertheless, the mitotic phenotypes are similar between neuroblasts and HeLa cells: the centrosomal microtubule asters are present, but they only nucleate short microtubules. Moreover, additional microtubule asters are detected, which fail to coalesce into a bipolar structure. This inability to form a bipolar spindle suggests that not only is the growth of MT altered, but the structure of the metaphase spindle, such as the organization/bundling of the interpolar microtubules, is also affected. Altogether, our results support the hypothesis that 1g targets a MAP involved in the regulation of microtubule growth during interphase and mitosis, which is essential for mitotic spindle assembly. The similarities in the mitotic phenotypes observed in HeLa cells and fly neural stem cells indicate that the 1g target and its functional motifs are likely to be conserved between humans and flies. Further studies will be needed to identify the direct cellular target of 1g in order to obtain further insight on its mechanism of action.

Microtubule-targeting agents are widely used in chemotherapy, but their lack of specificity for dividing tumor cells is a limitation. Indeed, their toxicity for the neural, immunological and gastric systems, due to their profound effect on interphase microtubules functions as well as tumor resistance, foster a need for development of new agents (Dumontet and Jordan, 2010; Stanton et al., 2011). The targets of choice are cellular microtubule regulators that modulate microtubule dynamics and organization specifically in proliferating cells. As discussed above the 1g compound may comply with the requirement of new pharmaceutical compounds with a more specific mode of action.

HeLa Kyoto cells were grown in Dulbecco's modified Eagle's medium Glutamax (Gibco) supplemented with 10% fetal calf serum (PAA), 100 U/ml penicillin and 100 µg/ml streptomycin. For synchronization experiments, cells were treated for 16 h with 5 µM RO-3306 to arrest cells in G2/M, cells were then washed in complete medium and released for the indicated times. pEGFP-Tub (BD Biosciences), pGFP- EB1 (gift from Philippe Chavrier, Institut Curie, France) and were used to generate tubulin-GFP and EB1-GFP HeLa stable cell lines, respectively.

1g was synthesized by Roberta Ettari (University of Messina, Italy) (Parenti et al., 2016) and its purity verified by ¹H- NMR and ¹³C-NMR. Absence of contaminants was also subsequently confirmed by HPLS-MS-MS. 1g was dissolved in DMSO at a stock concentration of 50 mM. Further dilutions were performed in tissue culture medium. DMSO control treatment is equivalent to the amount of DMSO present in the highest 1g treatment concentration used in the experiment. The final DMSO concentrations in the assays were therefore less or equal to 0.01% in HeLa cell experiments and 0.04% for Drosophila tissue experiments. The Mps1 inhibitor AZ3146 was from Calbiochem, the CDK1 inhibitor RO-3306 was from Merck, taxol and nocodazole were from Sigma-Aldrich.

The following commercial antibodies were used: α-tubulin (clone YL1/2, Millipore; 1:1000), anti-BUBRI (clone 9, BD Biosciences; 1:500), anti-GFP (clone7.1 and 13.1, Roche; 1:1000), anti-phospho histone H3 (clone CMA312, Millipore; 1:1000). DNA was stained with Hoechst 33342 or To-Pro-3 iodide (Invitrogen). For microtubule immunofluorescence staining, cells were grown on glass coverslips and fixed with methanol at −20°C (tubulin). For BubR1 staining, cells were first permeabilized with 0.5% Triton in PHEM (60 mM PIPES, 25 mM HEPES, 10 mM EGTA, 4 mM MgSO₄•7H₂O) for 5 min at room temperature and then fixed with paraformaldehyde in PHEM for 10 min. Antibody staining was then performed as described previously (Benaud et al., 2004). Brain from wandering third instar larvae were dissected and maintained in 100 μl of Schneider medium supplemented with DMSO or 20 μM 1g for 1 h at 25°C before fixation and immunostaining, as described previously (Gallaud et al., 2014).

Flies were maintained under standard conditions at 25°C. w¹¹¹⁸ flies were used as controls for immunostaining experiments. Flies expressing H2A-GFP (Clarkson and Saint, 1999) and the recombinant Insc-Gal4, UAS-mCherry-α-tubulin as well as UAS-ChRFP::Tubulin (BDSC 25774) and Insc-Gal4 (BDSC 8751) were obtained from the Bloomington Drosophila Stock Center (Indiana University, #5941 and #25773, respectively). The flies ubiquitously expressing tubulin tagged with RFP (RFP-tub) was a gift from Renata Basto (Institut Curie, Paris, France). The cell membrane marker Ubi-PH(PLCγ)-GFP (Gervais et al., 2008) was a gift from Antoine Guichet (Institut Jaccques Monod, Paris, France). The Msps RNAi line (ref. GD21982) was purchased from the Vienna Drosophila RNAi Center. The knockdown of Msps in central brain neuroblasts was driven by Insc-Gal4.

For live imaging, Drosophila brains expressing H2A-GFP and mCherry-α-tubulin were dissected in Schneider's Drosophila medium containing 10% FCS. Following 10 min pre-incubation with DMSO or the indicated concentration of chemical compounds, isolated brains were loaded and mounted on stainless steel slides. The preparations were sealed with mineral oil (Sigma-Aldrich) as previously described (Gallaud et al., 2014). HeLa cells were grown in Lab-Tek I chambered coverglasses (Nunc). Bright-field images of asynchronous dividing HeLa cells were acquired every 5 min with a 20× objective on a DMRIBE inverted microscope (Leica) equipped with CO₂ heated incubator chamber and a CoolSNAP ES BW camera (Roper Scientific). Live fluorescent images were acquired on a spinning disk microscope using a Plan Apo 60×/1.4 NA objective on an Eclipse Ti-E microscope (Nikon) equipped with a spinning disk (CSU-X1; Yokogawa), a thermostatic chamber (Life Imaging Service), Z Piezo stage (Marzhauser), and a charge-coupled device camera (CoolSNAP HQ2; Roper Scientific). Drosophila live images were alternatively acquired with a spinning-disk system consisting of a DMi8 microscope (Leica) equipped with a 63×/1.4NA oil objective, a CSU-X1 spinning-disk unit (Yokogawa) and an Evolve EMCCD camera (Photometrics). The microscope was controlled by the Inscoper Imaging Suite and the dedicated software (Inscoper).

For HeLa cells, time-lapse tubulin-GFP images were acquired every 1 min for the 45 min time lapse or every 5 min for the 90 min time lapse, and EB1GFP comets every 0.5 s using Metamorph Software (Universal imaging). For Drosophila neuroblasts, Z-series were acquired every 30 or 60 s. Immunofluorescence images of fixed samples were acquired with an SP5 confocal microscope (Leica) or with an API DeltaVision microscope equipped with a coolSnapHQ camera (Princeton instruments) using the SoftWorX software. Image acquisition was coupled to deconvolution when indicated. Images were processed and measurements performed using Fiji software (http://fiji.sc/). Analysis of EB1 comets was performed using the MATLAB-based open source u-track particle tracking (version 2.0) software (Danuser Lab, UT Southwern Medical Center).

Commercial lyophilized tubulin (PurSolutions, Nashville, USA) was reconstituted at 500 µM in distilled water according to the manufacturer's instructions. Tubulin was diluted at 50 µM in 10% glycerol, 1 mM GTP, 0.02% DMSO in BRB80 buffer (80 mM K-Pipes, 1 mM EGTA, 1 mM MgCl₂, pH 6.8 with KOH), and in the presence of 0 µM, 5 µM or 10 µM 1g. Control samples contained the same amount of DMSO as 1g samples. Suspensions were centrifuged at 33,000 g at 4°C for 5 min before polymerization. Samples were transferred into 100 µl quartz cuvettes (Hellma), and measurements at 350 nm were performed in a UVIKON XS spectrophotometer maintained at 35°C to stimulate tubulin polymerization. After 30 min of polymerization, depolymerization was induced by a cold temperature shift at 4°C to assess the presence of aggregates.

The imaging work was performed on the platform MRic-Photonics (BIOSIT, Université Rennes1). We thank Thibault Courtheoux for his help with the use of plus-tip tracker software and Laurent Richard-Parpaillon for critical discussion.

The peer review history is available online at https://jcs.biologists.org/lookup/doi/10.1242/jcs.239244.reviewer-comments.pdf